CN100466965C - Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management - Google Patents
Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management Download PDFInfo
- Publication number
- CN100466965C CN100466965C CNB038240092A CN03824009A CN100466965C CN 100466965 C CN100466965 C CN 100466965C CN B038240092 A CNB038240092 A CN B038240092A CN 03824009 A CN03824009 A CN 03824009A CN 100466965 C CN100466965 C CN 100466965C
- Authority
- CN
- China
- Prior art keywords
- hba
- data
- smbg
- reading
- rhi1
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Images
Abstract
A method, system, and computer program product related to the maintenance of optimal control of diabetes, and is directed to predicting the long-term exposure to hyperglycemia, and the long-term and short-term risks of severe or moderate hypoglycemia in diabetics, based on blood blucose readings collected by a self-monitoring blood glucose device. The method, system, and computer program product pertain directly to the enhancement of existing home blood glucose monitoring devices, by introducing an intelligent data interpretation component capable of predicting both HbA1c and periods of increased risk of hypoglycemia, and to the enhancement of emerging continuous monitoring devices by the same features. With these predictions the diabetic can take steps to prevent the adverse consequences associated with hyperglycemia and hypoglycemia.
Description
Related application
Present patent application is the domestic applications stage of the international patent application No.PCT/US2003/025053 that files an application in August, 2003, it requires to obtain the priority of following patent application according to 35U.S.C. clause 119 (e), be U.S. Provisional Patent Application series No.60/402,976, it was filed an application on August 13rd, 2002, title is " being used to handle self-monitoring of blood glucose (SMBG) thereby the method for data raising diabetics self management; system and computer program ", and No.60/478,377, it was filed an application on June 13rd, 2003, title is " being used to handle self-monitoring of blood glucose (SMBG) thereby the method for data raising diabetics self management; system and computer program ", and this paper quotes above-mentioned three disclosed whole disclosures as a reference.
Patent application relates to international application No.PCT/US01/09884, it files an application (patent application Nos.WO 01/72208A2 March 29 calendar year 2001, WO 01/72208A3), title is " by method, system and the computer program of self-monitoring data assessment diabetes glycemic control ", with U.S. Patent application series No.:10/240,228, it filed an application in JIUYUE in 2002 on the 26th, title is " by method, system and the computer program of self-monitoring data assessment diabetes glycemic control ", and this paper quotes its whole disclosure as a reference.
Technical field
Native system relates generally to the glycemic control of diabetic individual, more particularly, relates to a kind of computer based system and method, is used for evaluation prediction glycosylated hemoglobin (HbA
1cAnd HbA
1) and the danger of hypoglycemia takes place.
Background technology
A large amount of researchs confirm repeatedly, the most effectual way that prevents the diabetes long-term complications is to take the insulin strengthening treatment that blood glucose (BG) level was strict controlled within normal range, these researchs comprise diabetes control and complication experiment (DCCT), and (see DCCT research association: the diabetes intensive treatment is to the generation and Influence and Development (The Effect Of Intensive Treatment Of Diabetes On The DevelopmentAnd Progresion Of Long-Term Complications Of Insulin-DependentDiabetes Mellitus) the .New England Journal ofMedicine of insulin dependent diabetes mellitus (IDDM) long-term complications, 329:978-986,1993)), Stockholm diabetes intervention study (is seen Reichard P, Phil M: mortality rate in the diabetes intervention study of Stockholm during the long-term intensive treatment of traditional insulin and treatment side effect (Mortality and TreatmentSide Effects Druing Long-term Intensified Conventional InsulinTreatment in the Stockholm Diabetes Intervention Study) .Diabetes, 3:313-317,1994)) and the perspective diabetes study of Britain (see the perspective diabetes study meeting of Britain: strengthen the influence (Effect of Intensive Blood Glucose Control With Metformin OnComplications In Patients With Type 2 Diabetess) (UKPDS 34) of glycemic control for type 2 diabetes mellitus patient complication with metformin.
Lancet, 352:837-853,1998).
Yet, identical research has also confirmed some side effect of insulin strengthening treatment, the most serious is the danger that can increase frequent severe hypoglycemia disease (SH), this is that a kind of Heal Thyself that can't carry out (is seen DCCT research association: epidemiology (Epidemiology of Severe Hypoglycemia In The Diabetes Control andComplications Trial) the .American Journal of Medicine of severe hypoglycemia disease in diabetes control and the complication experiment with the neuroglycopenia incident that could recover that needs outside help, 90:450-459,1991, with DCCT research association: hypoglycemia (the Hypoglycemia in the Diabetes Control and ComplicationsTrial) .Diabetes during diabetes control and complication are tested, 46:271-286,1997).Because SH can cause contingency, stupor even death,, patient and Kang feel lose heart to proceeding intensive treatment so protecting personnel.The result, it is to improve a major obstacle of glycemic control (Cryer PE: hypoglycemia is limiting factor (Hypoglycemia is the Limiting Factor inthe Management Of Diabetes) the .Diabetes Metab ResRev of diabetes management that hypoglycemia is identified as, 15:42-46,1999).
Therefore, diabetics is faced with a lifelong optimization problem, promptly in the danger that keeps blood glucose is not increased again in the strict control hypoglycemia.The main challenge relevant with this problem is to produce a kind of glycemic control and the hypoglycemia danger thereof of assess patient simultaneously, and the better way method that can use in surroundings.
It is well known that over 20 years that glycosylated hemoglobin is the sign of the individual glycemic control of diabetes (1 type or 2 types).A large amount of research worker has been carried out research to this relation and has been found, glycosylated hemoglobin has reflected the average BG level of patient in past two months basically.Because in most of glycosuria patient bodies, the BG level has sizable fluctuation in a period of time, so advise whole glycemic control and HbA
1cBetween actual association can only known patient is in the situation of stable glycemic control in a long term under, observe.
Average BG level and HbA in preceding 5 weeks set up in such patient's early stage research
1cBetween the relation of almost determining, and it is that 0.98 correlation coefficient (is seen Aaby Svendsen P that this curve linear relationship has produced size, Lauritzen T, Soegard U, glycosylated hemoglobin in Nerup J (1982) .1 type (insulin-dependent) diabetes and stable state average blood sugar concentration (Glycosylated Hemoglobin and Steady-State Mean Blood GlucoseConcentration in Type 1 (Insulin-Dependent) Diabetes).
Diabetologia,
23, 403-405).1993, DCCT reached a conclusion, HbA
1cBe golden standard glycosylated hemoglobin chemical examination (gold-standard glycosylated hemoglobin assay) " rationally recommending ", and DCCT has determined previous average BG and HbA
1cBetween linear relationship (see Santiago JV (1993). from diabetes control lesson (Lessonsfrom the Diabetes Control and Complications Trial) with the complication experiment,
Diabetes,
42, 1549-1554).
The guilding principle that has proposed shows, 7% HbA
1cCorresponding to the average BG of 8.3mM (150mg/dl), 9% HbA
1cCorresponding to the average BG of 1.7mM (210mg/dl), and HbA
1cIncrease by 1% corresponding to average BG increase 1.7mM (30mg/dl, 2).DCCT also advises, because directly measure average BG and unrealistic, so can use single simple test, i.e. HbA
1c, the glycemic control of assess patient.Yet research clearly shows HbA
1cTo hypoglycemia and insensitive.
Really, all can not obtain the directly dangerous reliable prediction value of patient SH by any data.DCCT reaches a conclusion, and has only future of about 8% SH can be by known parameter, for example SH history, low HbA
1cWith hypoglycemia stupor (unawareness), forecast is come out.One piece of recent summary describes the current clinical condition of this problem in detail, and, prevention SH (sees Bolli for providing patient and health thereof to protect the selection that personnel can obtain, GB: how in type 1 diabetes reinforcement and nonreinforcement treatment, to improve hypoglycemia problem (How To Ameliorate ThePreblem of Hypoglycemia In Intensive As Well As NonintensiveTreatment Of Type I Diabetes) .Diabetes Care, 22, Supplement2:B43-B52,1999).
Modern household BG monitor provides by self-monitoring BG (SMBG) and has carried out the device that frequent BG measures.Yet the problem of SMBG is, by the data and the HbA of BG monitor collection
1cAnd shortage contact between the hypoglycemia.In other words, also there is not reliable method to estimate HbA at present according to the SMBG reading
1cCan (see Bremer T and Gough DA: blood glucose be predicted out by preceding value with the imminent hypoglycemia of identification? the initiation of data (Is blood glucose predictable from previous values? A solicitationfor data) .Diabetes 48:445-451,1999).
Therefore, an object of the present invention is by proposing three kinds of differences but compatible algorithm provides the contact of this disappearance, three kinds of algorithms are used for by SMBG data estimation HbA
1cWith hypoglycemia danger, with the long-term danger of long-term and the short-term danger and the hyperglycemia of prediction hypoglycemia.
The inventor formerly once reported, the daily acquisition of SMBG data and estimation HbA
1cAnd the reason that lacks contact between the hypoglycemia danger is, the accurate method of the data collection that in diabetes study, uses and clinical assessment seldom obtain diabetes special with mathematics on the support of statistical disposition of precision.
For needs to the statistical analysis that can take the special distribution of BG into account, the inventor has proposed a kind of symmetry transformation to blood sugar measured scope and (has seen Kovatchev BP, Cox DJ, Gonder-Frederick LA and WL Clarke (1997). the symmetrization of blood sugar measured scope and application thereof (Symmetization of the Blood Glucose Measurement Scaleand Its Applications)
Diabetes Care,
20, 1655-1658), its operation is as follows.The BG level is that unit is measured with mg/dl in the U.S., is unit with mmol/L (or mM) in other most countries.The direct relation of two kinds of scales is 18mg/dl=1mM.The whole BG value scope that provides in most of lists of references is 1.1-33.3mM, and thinks that it has covered whole observed values practically.Recommendation according to DCCT (is seen DCCT research association (1993), the diabetes intensive treatment is to the generation and Influence and Development (TheEffect Of Intensive Treatment Of Diabetes On The Development AndProgresion Of Long-Term Complications Of Insulin-DependentDiabetes Mellitus) the .New England Journal ofMedicine of insulin dependent diabetes mellitus (IDDM) complication, 329:978-986,1993)), it is 3.9-10mM that diabetes patient's target BG value scope---is also referred to as the euglycemia scope---, then occur hypoglycemia when being lower than 3.9mM when BG drops to, then occur hyperglycemia when above when BG is elevated to 10mM.Unfortunately, this scope is numerically also asymmetric---and hyperglycemia scope (10-33.3mM) is wider than hypoglycemia scope (1.1-3.9Mm), and euglycemia scope (3.9-10mM) is not or not the center of this scope.The inventor has revised this symmetry by introducing a kind of conversion f (BG), and wherein f (BG) is that the domain of definition is the continuous function on the BG scope [1.1,33.3], has two-parameter analytical form:
f(BG，α，β)＝[(ln(BG))
α-β]，α，β>0
It satisfies hypothesis;
A1:f (33.3, α, β)=-f (1.1, α, β) and
A2:f(10.0，α，β)＝-f(3.9，α，β)。
Then, f (.) thus multiply by the 3rd scale parameter will be separately fixed at through the minimum and the maximum of the BG scope of conversion
With
These numerical value are very convenient, because the random variable with standard normal distribution is in the interval
In have 99.8% numerical value.Measure if BG is unit with mmol/L, when then carrying out numerical solution according to hypothesis A1 and A2, (parameter β) is α=1.026 to function f for BG, α, β=1.861, scale parameter γ=1.794.Measure if BG is unit with mg/dl, the parameter that then calculates is α=1.084, β=5.381, γ=1.509.
Therefore, when with mmol/L being the measurement BG of unit, symmetry transformation is f (BG)=1.794[(ln (BG))
1.026-1.861], when with mg/dl being the measurement BG of unit, symmetry transformation is f (BG)=1.509[(ln (BG))
1.084-5.381].
According to symmetry transformation f (.), the inventor has introduced low BG index---and one is used for (seeing Cox DJ by the new measured value of SMBG reading estimation hypoglycemia danger, Kovatchev BP, Julian DM, Gonder-Frederick LA, Polonsky WH, Schlundt DG, severe hypoglycemia disease frequency among the Clarke WL:IDDM can be predicted (Frequency of Severe Hypoglycemia In IDDM Can BePredicted From Self-Monitoring Blood Glu cose Data) .Journal ofClinical Endocrinology and Metabolism by the self-monitoring of blood glucose data, 79:1659-1662,1994, with Kovatchev BP, Cox DJ, Gonder-Frederick LA, Young-Hyman D, Schlundt D, severe hypoglycemia disease is dangerous among the Clarke WL.IDDM adult estimates: the effectiveness of low-glycemic (Assessment of Risk for Severe HypoglycemiaAmong Adults With IDDM:Validation of the Low Blood GlucoseIndex), Diabetes Care 21:1870-1875,1998).Given a series of SMBG data, when f (BG)＜0, low BG Index for Calculation is 10.f (BG)
2Meansigma methods, otherwise be 0.Also proposed high BG index, according to calculating and get with low BG index symmetrical manner, but this index is not also found its practical use.
In regression model, use low BG index, the inventor can be according to SH 6 SH incident 40% the variations in the middle of the month subsequently of historical and SMBG data interpretation, afterwards this prediction has been brought up to 46% and (seen Kovatchev BP, Straume M, Farhi LS, Cox DJ: estimate the blood glucose rate of transformation and with severe hypoglycaemia relation (Estimating the Speed ofBlood Glucose Transitions and its Relationship With SevereHypoglycemia), Diabetes, 48:Supplement 1, A363,1999).
In addition, the inventor has also proposed some about HbA
1c(see Kovatchev BP with the data of SMBG, Cox DJ, Straume M, the relation of Farhi LS. self-monitoring of blood glucose curve and glycosylated hemoglobin (Association of Self-monitoring Blood GlucoseProfiles with Glycosylated Hemoglobin)
In:Methods in Enzymology, Vol.321:Numerical Computer Methods, Part C, Mechael Johnson and Ludvig Brand, Eds., Academic Press, NY; 2000).
These achievements in research become the part of theoretical background of the present invention.For this theory is committed to practice, added the theoretical parameter of several keys, will describe in the part below.Especially, propose three kinds of methods and be used to estimate HbA
1c, hypoglycemia long-term and short-term dangerous.The proposition of these methods is based on, but is not limited only to, and 867 diabetic individual are surpassed 300,000 SMBG readings, severe hypoglycemia disease record and HbA
1cResult's labor.
Therefore, the inventor attempts to improve the aforementioned limitations relevant with traditional method, and the better way method is provided whereby, thereby can be used in the glycemic control of while assess patient and the danger of hypoglycemia thereof, and can use in its surroundings.
Summary of the invention
The present invention includes a kind of data analysing method and computer based system, be used for estimating two kinds of most important component: HbA of diabetes glycemic control simultaneously by the SMBG data of daily collection
1cWith hypoglycemia danger.Purpose for this document, BG self-monitoring (SMBG) is defined as being used for determining the method for blood glucose under the natural conditions of diabetics, and the method that the SMBG equipment that comprises current common storage 200-250 BG reading uses, and the method for the continuous detecting technology of producing in the future use.By providing this extensive definition of SMBG, the present invention directly is devoted to a kind ofly can predict HbA simultaneously by introducing
1cImprove the performance of (but being not limited only to) existing domestic blood sugar monitoring equipment with the intelligent data interpretation assembly of hypoglycemia high-risk period, and improve the performance of the continuous monitoring equipment of production in the future by identical parts.
One aspect of the present invention comprises a kind of method, system and computer program, is used for by in predetermined period, for example about 4-6 week, the SMBG data estimation HbA of collection
1cIn one embodiment, the invention provides a kind of computerized method and system, be used for HbA according to the BG data estimation patient who in predetermined period, collects
1cThis method (or system or computer usable medium) comprises the HbA according to the BG data estimation patient who collects in first predetermined period
1cThis method comprises: use as the predetermined sequence mathematical formulae of giving a definition and prepare to be used to estimate HbA
1cData: the data pretreatment; Use at least one the estimation HbA in four predetermined formulas
1cWith the effectiveness of estimating by the checking of sample choice criteria.
Another aspect of the present invention comprises a kind of method, system and computer program, is used to estimate the long-term probability of severe hypoglycemia disease (SH).This method is used in the predetermined period, 4-6 week for example, the SMBG reading, and the SH danger in about subsequently 6 months of the prediction.In one embodiment, the invention provides a kind of computerized method and system, be used for long-term probability according to the BG data estimation severe hypoglycemia disease (SH) of in predetermined period, collecting.This method (or system or computer usable medium) comprises according to the BG data estimation patient severe hypoglycemia disease (SH) of collecting in predetermined lasting time or the long-term probability of moderate hypoglycemia (MH).This method comprises: according to collected BG data computation LBGI; With the number that utilizes predetermined mathematical formulae estimation SH incident in future according to the LBGI that calculates.
Another aspect of the present invention comprises a kind of method, system and computer program, is used to discern the hypoglycemia high-risk period of (perhaps other chosen periods) in 24 hours.It is dangerous realization of short-term of calculating hypoglycemia by the SMBG reading that priority of use was collected in preceding 24 hours.In one embodiment, the invention provides a kind of computerized method and system, be used for short-term danger according to the BG data estimation severe hypoglycemia disease (SH) of in predetermined period, collecting.This method (or system or computer usable medium) comprises the short-term probability according to the BG data estimation patient severe hypoglycemia disease (SH) of collecting in predetermined lasting time.This method comprises: according to collected BG data computation scale value; And be the low BG dangerous values (RLO) of each BG data computation.
An aspect of the embodiment of the invention comprises according to the BG data estimation patient HbA that collects in first predetermined lasting time
1cMethod (perhaps interchangeable computer program).This method comprises that the mathematical formulae preparation of using predetermined sequence is used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
An aspect of the embodiment of the invention comprises the BG data estimation patient HbA that a kind of basis is collected in first predetermined lasting time
1cSystem.This system comprises can operate database component and processor that is used to keep discerning described BG data, and wherein this processor is programmed and is used to use the mathematical formulae of predetermined sequence to prepare to be used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
An aspect of the embodiment of the invention comprises the BG data estimation patient HbA that a kind of basis is collected in first predetermined lasting time
1cSystem.This system comprises: BG obtains mechanism, and it is used for obtaining the BG data from the patient; Can operate the database component that is used to keep discerning described BG data; With a processor.This processor is programmed and is used to use the mathematical formulae of predetermined sequence to prepare to be used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
An aspect of the embodiment of the invention comprises a kind of previous HbA that do not need
1cInformation according to the BG data estimation patient HbA that in first predetermined lasting time, collects
1cMethod (perhaps alternative computer program).This method comprises with the mathematical formulae of predetermined sequence prepares to be used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
An aspect of the embodiment of the invention comprises a kind of previous HbA that do not need
1cInformation according to the BG data estimation patient HbA that in first predetermined lasting time, collects
1cSystem.This system comprises can operate database component and processor that is used to keep discerning described BG data.This processor is programmed and is used to use the mathematical formulae of predetermined sequence to prepare to be used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
An aspect of the embodiment of the invention comprises a kind of previous HbA that do not need
1cInformation according to the BG data estimation patient HbA that in first predetermined lasting time, collects
1cSystem.This system comprises: BG obtains mechanism, and it is used for obtaining the BG data from the patient; Can operate the database component that is used to keep discerning described BG data; With a processor.This processor is programmed and is used to use the mathematical formulae of predetermined sequence to prepare to be used to estimate HbA
1cData.This mathematical formulae is defined as: the data pretreatment; By sample choice criteria checking BG data sample; If with sample effectively then estimate HbA
1c
These aspects of the present invention, and other aspects of discussing in full of this document, thus can combine the continuous information that relevant diabetics glycemic control is provided, and improve supervision to hypoglycemia danger.
Herein, these and other target of the present invention with and advantage and characteristics from following explanation, accompanying drawing and claim, will become more apparent.
Description of drawings
Read the explanation of following preferred embodiment together by connection with figures and will have more fully aforementioned and other purposes, characteristics and advantage of the present invention and the present invention self and understand, wherein:
Fig. 1 provides 15 being carried out SMBG by each of the danger level scope of the low BG index definition of example No.1 with illustrating moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia has taken place in 1 month after estimating.
Fig. 2 provides 15 being carried out SMBG by each of the danger level scope of the low BG index definition of example No.1 with illustrating moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia has taken place in 3 months after estimating.
Fig. 3 provides 15 being carried out SMBG by each of the danger level scope of the low BG index definition of example No.1 with illustrating moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia has taken place in 6 months after estimating.
Fig. 4 provides with illustrating and 15 has been undertaken by each of the danger level scope of the low BG index definition of example No.1 SMBG takes place 2 times in 3 months after estimating or more times moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia.
Fig. 5 provides with illustrating and 15 has been undertaken by each of the danger level scope of the low BG index definition of example No.1 SMBG takes place 2 times in 6 months after estimating or more times moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia.
Fig. 6 is the functional block diagram that is used to realize computer system of the present invention.
Fig. 7-the 9th, communicates to connect the schematic block diagram with the alternative modification of system at associative processor of the present invention.
Figure 10 provides with illustrating and 15 has been undertaken by each of the danger level scope of the low BG index definition of example No.1 SMBG takes place 3 times in 6 months after estimating or more times moderate (dotted line) and the seriously experience and the theoretical probability of (solid line) hypoglycemia.
Figure 11 has shown the residual analysis of this model with illustrating, demonstrates close with the normal distribution of example No.1 training data group 1 residual error.
Figure 12 has shown the residual analysis of this model with illustrating, demonstrates close with the normal distribution of example No.1 residual error.
Figure 13 has shown the statistic evidence that the normal probability paper Fig. 1 by example No.1 provides with illustrating.
Figure 14 provides hit rate and the ratio R of representing with percentage rate among the example No.1 with illustrating
UdBetween level and smooth dependency.
Figure 15 provides the dependency between prediction period and corresponding hit rate among the example No.1 with illustrating.
Figure 16 (A)-(B) diagram ground provides the danger in the remarkable hypoglycemia of T1DM predicted by LBGI 1 month, ANOVA (the F=7.2 of each dangerous group severe hypoglycemia disease event number in example No.2, p＜0.001) and ANOVA (F=13.9, p＜0.001) of each dangerous group moderate hypoglycemia event number.
Figure 17 (A)-(B) diagram ground provides the danger in the remarkable hypoglycemia of T1DM predicted by LBGI 3 months, ANOVA (the F=9.2 of each dangerous group severe hypoglycemia disease event number in example No.2, p＜0.001) and ANOVA (F=14.7, p＜0.001) of each dangerous group moderate hypoglycemia event number.
Figure 18 (A)-(B) diagram ground provides the danger in the remarkable hypoglycemia of T2DM predicted by LBGI 1 month, ANOVA (the F=6.0 of each dangerous group severe hypoglycemia disease event number in example No.2, p＜0.001) and ANOVA (F=25.1, p＜0.001) of each dangerous group moderate hypoglycemia event number.
Figure 19 (A)-(B) diagram ground provides the danger in the remarkable hypoglycemia of T2DM predicted by LBGI 3 months, ANOVA (the F=5.3 of each dangerous group severe hypoglycemia disease event number in example No.2, p＜0.01) and ANOVA (F=20.1, p＜0.001) of each dangerous group moderate hypoglycemia event number.
The specific embodiment
The invention enables, but be not limited only to, might produce the exact method that is used to estimate the diabetics glycemic control, and the firmware and the software coding that use when being included in the key component of calculating this method.Be used to estimate HbA
1c, the long-term probability of SH and the danger of hypoglycemia short-term inventive method also can be verified according to collected mass data, and will discuss in the back of this paper.At last, the scheme of these methods can be combined into structuring demonstration or matrix.
I. estimate HbA
1c
One aspect of the present invention comprises a kind of method, system and computer software, is used for by in predetermined period 4-6 week for example, the SMBG data estimation HbA of collection
1cIn one embodiment, the invention provides the method and system of a kind of computerization (perhaps other types), be used for HbA according to the BG data estimation patient who in predetermined lasting time, collects
1cThis method comprises the HbA according to the BG data estimation patient who collects in first predetermined lasting time
1c, this method comprises: utilize the mathematical formulae of predetermined sequence to prepare to be used to estimate HbA
1cData.This mathematical formulae is defined as: the pretreatment of data; Use at least one the estimation HbA in four predetermined formulas
1cWith the effectiveness of estimating by the checking of sample choice criteria.First predetermined lasting time can be about 60 days, and perhaps selectively, the first predetermined lasting time scope is about 45 days to about 75 days, perhaps about 45 days to about 90 days, and perhaps according to expectation.Each patient's data pretreatment comprises: convert blood plasma BG to whole blood BG mg/dl; To be the unit that BG that unit is measured converts mmol/l to mg/dl; With calculating low-glycemic (RLO1) and hyperglycemic index (RHI1).The predetermined mathematical formula that each patient's data pretreatment is used as given a definition: convert blood plasma BG to whole blood BG mg/dl by BG=PLASBG (mg/dl)/1.12; To convert mmol/l unit to the BG that mg/dl measures by BGMM=BG/18; With calculating low-glycemic (RLO1) and hyperglycemic index (RHI1).The predetermined mathematical formula that the pretreatment of data is further used as given a definition: Scale=[ln (BG)]
1.0845-5.381, wherein BG is that unit is measured with mg/dl; Risk1=22.765 (Scale)
2, RiskLO=Risk1 wherein, if therefore there is the danger of LBGI in (BG is less than about 112.5), otherwise RiskLO=0; RiskHI=Risk1, if therefore there is the danger of HGBI in (BG is greater than about 112.5), otherwise RiskHI=0; Each patient's of BGMM1=average BGMM; Each patient's of RLO1=average RsikLO; Each patient's of RHI1=average RiskHI; L06=is only to the average RiskLO that night, reading calculated, if there is not reading then default at night; N06, N12, N24 are the percentage rate of SMBG reading in each interval; The sum of SMBG reading in NC1=first predetermined lasting time; The natural law that has the SMBG reading in NDAYS=first predetermined lasting time.The percentage rate that SMBG reads in N06, N12, the following interval of N24 difference, i.e. approximately 0-6:59, about 7-12:59 and approximately 18-23:59, perhaps other expectation percentage rate and interval number.
This method comprises that further according to the high BG index of patient that calculates with the predetermined mathematical formula be the group assignment.This formula can be defined as: if (if RHI1≤about 5.25 or RHI1 〉=about 16), then assignment group=0; If (RHI1〉if about 5.25 and RHI1＜about 7.0), assignment group=1 then; (if if RHI1 〉=about 7.0 and RHI1＜about 8.5), then assignment group=2; With (if if RHI1 〉=about 8.5 and RHI1＜about 16), then assignment group=3.
Then, this method may further include the predetermined mathematical formula that uses as give a definition and provides estimation:
E0=0.55555*BGMM1+2.95; E1=0.50567*BGMM1+0.074*L06+2.69; E2=0.5555*BGMM1-0.074*L06+2.96; E3=0.44000*BGMM1+0.035*L06+3.65; And if (group=1), EST2=E1 then, if perhaps (group=2) then EST2=E2, if perhaps (group=3) then EST2=E3, otherwise EST2=E0.
This method comprises that the predetermined mathematical formula that uses as give a definition further revises estimating: if (default (L06)), EST2=E0 is (if RLO1≤about 0.5 and RHI1≤about 2.0), then EST2=E0-0.25; (if RLO1≤about 2.5 and RHI1〉about 26), then EST2=E0-1.5*RLO1; And if ((RLO1/RHI1)≤about 0.25 and L06〉about 1.3) then EST2=EST2-0.08.
HbA according to the BG data estimation patient who in first predetermined lasting time, collects
1cCan be by using at least one the estimation HbA in four predetermined mathematical formula
1cAnd realize that four formula are defined as follows:
E) HbA
1c=as EST2 defined above or as revising above;
F) HbA
1c=0.809098*BGMM1+0.064540*RLO1-0.151673*RHI1+1.873325, wherein BGMM1 is the meansigma methods (mmol/l) of BG, and RLO1 is low BG index, and RHI1 is high BG index;
G) HbA
1c=0.682742*HBA0+0.054377*RHI1+1.553277, wherein HBA0 is adopt in about second predetermined period before estimating previous with reference to HbA
1cReading, wherein RHI1 is high BG index; Perhaps
H) HbA
1c=0.41046*BGMM+4.0775, wherein BGMM1 is the meansigma methods (mmol/l) of BG.Second predetermined lasting time can be about 3 months; About 2.5 months to about 3.5 months; Perhaps about 2.5 months to 6 months; Perhaps according to expectation.
Have only when the first predetermined lasting time sample satisfies in following four standards at least one, just use HbA
1cThe effectiveness that the sample choice criteria checking of estimating is estimated:
B) test frequency standard, wherein the first predetermined lasting time sample mean is tested greatly every day at least
About 1.5 to about 2.5 times; B) can select the test frequency standard, wherein the average frequency of predetermined lasting time sample reading in the 3rd predetermined sampling period is about 1.8 reading/skies (the perhaps average frequencies of other expectations);
E) data randomization standard-1 wherein has only as ratio (RLO1/RHI1) 〉=about 0.005 the time
Just verify and demonstration HbA
1cEstimate that wherein RLO1 is low BG index, RHI1 is high BG index; Perhaps
F) data randomization standard wherein has only and just verifies when ratio (NO6 〉=about 3%) and show HbA
1cEstimate, and wherein NO6 is the meansigma methods of reading at night.The 3rd predetermined lasting time can be at least 35 days, and scope is from about 35 days to about 40 days, perhaps from about 35 days to approximately the same with first predetermined lasting time long, perhaps according to expectation.
II. the long-term probability of severe hypoglycemia disease (SH)
Another aspect of the present invention comprises a kind of method, system and computer program, is used to estimate the long-term probability of severe hypoglycemia disease (SH).This method is used predetermined period, about 4-6 week for example, the SMBG reading and the danger of prediction SH in about 6 months subsequently.In one embodiment, the invention provides a kind of computerized (perhaps other types) method and system, be used for long-term probability according to the BG data estimation patient severe hypoglycemia disease (SH) of in predetermined lasting time, collecting.Long-term probability method according to BG data estimation patient severe hypoglycemia disease (SH) of collecting in predetermined lasting time or moderate hypoglycemia (MH) comprises: according to the BG data computation LBGI that collects; With the number that utilizes predetermined mathematical formulae estimation SH incident in future according to the LBGI that calculates.The calculating of LBGI is by at time point t
1, t
2..., t
nGather a series of BG reading x
1, x
2..., x
nThe definition of mathematics ground:
We have defined predetermined risk range (risk category) (RCAT), whereby the numerical range of each risk range (RCAT) expression LBGI; And give in the described risk range (RCAT) at least one with the LBGI assignment.Risk range (RCAT) is defined as follows:
Scope 15, wherein said LBGI is greater than about 6.5.
Then, be respectively each described named peril minus exception scope (RCAT) and limit the probability that selected number SH incident takes place.Utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited to the probability that selected number SH incident takes place in next first predetermined lasting time: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-4.19, b ≈ 1.75 (a and/or b can be other expected values).First predetermined lasting time can be about 1 month, and scope is from 0.5 month to about 1.5 months, perhaps scope from about 0.5 month by about 3 months, perhaps according to expectation.
In addition, utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited to the probability that selected number SH incident takes place in next second predetermined lasting time: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-3.28, b ≈ 1.50 (a and/or b can be other expected values).Second predetermined lasting time can be about 3 months, and scope is from about 2 months to about 4 months, perhaps from about 3 months by about 6 months, perhaps according to expectation.
Further, utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited to the probability that selected number SH incident takes place in next the 3rd predetermined lasting time: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-3.06, b ≈ 1.45 (a and/or b can be other time values).The 3rd predetermined lasting time can be about 6 months, and scope is from about 5 months to about 7 months, perhaps from about 3 months by about 9 months, perhaps according to expectation.
Selectively, utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited in next first predetermined period that (scope is about 1 month, approximately 0.5-1.5 month, approximately 0.5-3 month, perhaps according to expectation) probability of selected number MH incident takes place: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-1.58, b ≈ 1.05 (a and/or b can be other expected values).
Selectively, utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited in next second predetermined lasting time that (scope is about 3 months, approximately 2-4 month, approximately 3-6 month, perhaps according to expectation) probability of selected number MH incident takes place: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-1.37, b ≈ 1.14 (a and/or b can be other expected values).
Selectively, utilize following formula to be respectively each described named peril minus exception scope (RCAT) and be limited to that (scope is about 6 months in next the 3rd predetermined lasting time, approximately 5-7 month, approximately 3-9 month, perhaps according to expectation) probability of selected number MH incident takes place: F (x)=1-exp (a.x
b), x〉0, otherwise be 0, wherein: a ≈-1.37, b ≈ 1.35 (a and/or b can be other issue values).
And, specified the patient at the classification of risks that remarkable hypoglycemia takes place in the future.This classification is defined as follows: minimal risk, and wherein said LBGI is less than about 1.25; Low dangerous, wherein said LBGI is about 1.25-about 2.50; Poor risk, wherein said LBGI are between about 2.50-about 5; And high-risk, wherein said LBGI is greater than about 5 (also can realize other classification range according to expectation).
III. severe hypoglycemia disease (SH) short-term probability
Another aspect of the present invention comprises a kind of be used to discern (perhaps other chosen periods) hypoglycemia in 24 hours method, system and the computer program of high-risk period.The dangerous realization of short-term that this SMBG reading of collecting in preceding 24 hours by priority of use calculates hypoglycemia.In one embodiment, the invention provides a kind of Computerized method and system, be used for short-term danger according to the BG data estimation patient severe hypoglycemia disease (SH) of in predetermined lasting time, collecting.Method according to BG data estimation patient severe hypoglycemia disease (SH) the short-term danger of collecting in predetermined lasting time comprises: according to the BG data computation scale value of described collection; Be the low BG dangerous values (RLO) of each BG data computation.The calculating of RLO (BG) by mathematical definition is: Scale=[ln (BG)]
1.0845-5.381, wherein BG is that unit is measured with mg/dl; Risk=22.765 (Scale)
2, (if BG is less than about 112.5), then: RLO (BG)=Risk, otherwise RLO (BG)=0.Selectively, the calculating of RLO (BG) by mathematics be defined as: Scale=[ln (BG)]
1.026-1.861, wherein BG is that unit is measured with mmol/l; Risk=32.184 (Scale)
2, (if BG is less than about 112.5), then: RLO (BG)=Rsik, otherwise RLO (BG)=0.
Can calculate LBGI according to collected BG data.The calculating of LBGI is by at time point t
1, t
2..., t
nGather a series of BG reading x
1, x
2..., x
nThe definition of mathematics ground:
Can calculate interim LBGI according to the BG data of having collected.The calculating of interim LBGI by mathematics be defined as:
LBGI (1)=RLO (x1); RLO2 (1)=0; LBGI (j)=((j-1)/j) * LBGI (j-1)+(1/j) * RLO (xj); And RLO2 (j)=((j-1)/j) * RLO2 (j-1)+(1/j) * RLO (xj)-LBGI (j))
2
The mathematical formulae of the enough following definition of SBGI energy is calculated:
Then, the invention provides the authentication (qualification) to imminent short-term SH and report to the police.To authenticate and report to the police in the following cases, if: (LBGI (150) 〉=2.5 and (LBGI (50) 〉=(1.5*LBGI (150) and SBGI (50) 〉=SBGI (150)), then confirm or send described warning, perhaps RLO 〉=(LBGI (150)+1.5* (SBGI (150)), then confirm or send described warning, otherwise do not need authentication or warning is provided.
Then selectively, the invention provides authentication or warning to imminent short-term SH.To authenticate and report to the police in the following cases, if: (LBGI (n) 〉=α and LBGI (n) ge (β)), then confirm or send described warning, and/or (RLO (n) 〉=(LBGI (n)+γ * SBGI (n))), then confirm or send described warning, otherwise do not need authentication or warning is provided, wherein α, β and γ are threshold parameters.
Threshold parameter α, β and γ are defined as α ≈ 5, β ≈ 7.5, γ ≈ 1.5.Other possible parameter combinations have been provided in the form below.This numerical value can be close with value given below, also can be any intermediate combination of numerical value in the following form.
| α | β | γ | α | β | γ |
| 6.4 | 8.2 | 1.5 | 5.0 | 7.5 | 1.3 |
| 6.0 | 7.5 | 1.5 | 4.9 | 7.0 | 1.2 |
| 5.5 | 7.5 | 1.5 | 4.9 | 7.0 | 1.2 |
IV. example system
Method of the present invention can be realized with hardware, software or its combination, and can in one or more computer systems or other processing systems, realize, for example PDA(Personal Digital Assistant) is perhaps directly realized in having the blood glucose self-monitoring equipment (SMBG stores meter) of enough storages and disposal ability.In an example embodiment, the present invention is the software of operation on general purpose computer 900 as shown in Figure 6.Computer system 600 comprises one or more processors, and for example processor 604.Processor 604 connects communication infrastructure equipment 606 (for example, communication bus, crossbar (cross-over bar) or networks).Computer system 600 can comprise display interface 602, and it transmits chart, text and other data from communication infrastructure equipment 606 (the perhaps frame buffer that never shows) and is used for showing at display unit 630.
Computer system 600 also comprises main storage 608, random-access memory (ram) preferably, and can also comprise a second-level storage 610.Second-level storage 610 can comprise, for example, hard drive 612 and/or removable memory driver 614, it represents disk drive, magnetic tape drive, disc drives, flash memory etc.Removable storing driver 614 in a well-known manner from and/or read and/or write to removable memory module 618.Removable memory module 618 is represented floppy disk, tape, CD etc., and it is read and write by removable storing driver 614.Can recognize that removable memory module 618 comprises the available storage medium of computer, wherein stores computer software and/or data.
Can select among the embodiment, second-level storage 610 can comprise other device, is used for allowing computer program or other instructions to be loaded into computer system 600.This class device can comprise, for example, and removable memory module 622 and interface 620.The example of this removable memory module/interface comprises a program casket (cartridge) and casket interface (for example seeing) in video game device, removable storage chip (for example ROM, PROM, EPROM or EEPROM) and associated slots, with other removable memory module 622 and interface, it allows slot and data to transfer to computer system 600 from removable memory module 62.
Computer system 600 can also comprise communication interface 624.Communication interface 624 allows software and data to transmit between computer system 600 and external equipment.The example of communication interface 624 comprises a modem, a network interface (for example Ethernet card), a COM1 (for example series connection or parallel connection etc.), PCMCIA slot and card, a modem etc.The form that software that transmits by communication interface 624 and data are signal 628, it can be the signal that electronics, electromagnetism, light or other can be received by communication interface 624.Signal 628 offers communication interface 624 by communication path (passage just) 626.Passage 626 carries signal 628, and can be realized with lead or cable, optical fiber, telephone wire, portable phone connection, RF connection, infrared connection and other communication ports.
In this document, term " computer program medium " and " computer usable medium " are used for usually referring to following medium, and for example removable storing driver 614 is installed in hard disk and signal 628 in the hard drive 612.These computer programs are to be used to provide software to arrive the device of computer system 600.The present invention includes this computer program.
Computer program (being also referred to as computer control logic) is stored in main storage 608 and/or the second-level storage 610.Computer program can also be received by communication interface 624.The feature of the present invention that this computer program can make computer system 600 carry out as hereinafter discuss when carrying out.Especially, computer program makes processor 604 can carry out function of the present invention when being performed.Therefore, this computer program has been represented the controller of computer system 600.
Realizing that with software software can be stored in the computer program, and is loaded in the computer system 600 with removable memory driver 614, hard drive 612 or communication interface 624 in one embodiment of the present of invention.Control logic (software) makes processor 604 carry out function of the present invention as mentioned below when being carried out by processor 604.
In another embodiment, the present invention mainly realizes with example, in hardware, and it for example uses nextport hardware component NextPort such as device-specific integrated circuit (application specific integrated circuit) (ASIC).Thereby carrying out the hardware state machine and carry out the function of explanation here, is conspicuous for those skilled in the relevant art.
In another embodiment, the present invention uses the combination of hardware and software to be realized.
In an example software embodiment of the present invention, above-mentioned method realizes with the SPSS control language, but also can be enough other program realize, for example, but be not limited only to the program that C++ program language or other those skilled in the art can obtain.
Fig. 7-9 has shown the block diagram of representing alternate embodiment of the present invention.With reference to figure 7, shown the block diagram of representative system 710, it comprises that mainly the glucose meter 728 that is used by patient 712 is used for record, especially, the blood glucose of insulin dose reading and measurement (" BG ") level.The data that obtain by glucose meter 728 preferably by suitable communicate to connect 714 or data modem unit 732 transfer to treating stations or chip, for example personal computer 740, PDA are perhaps by portable phone or by suitable Internet port.For example, the data of being stored can be stored in the glucose meter 728 and by the appropriate interface cable and be directly downloaded to personal computer, send to treating stations by the Internet then.Example is ONE TOUCH monitoring system or the meter of being produced by the LifeScan company limited, itself and IN TOUCH softwarecompatible, and comprise interface cable so that download to personal computer.
Blood-glucose meter is general in the industry, and mainly comprises and anyly can play the equipment that BG obtains body function.BG meter or obtain mechanism, equipment, instrument or system and comprise the various traditional methods that are used to each Test extraction blood sample (for example by the thorn finger), and comprise that device that use is read concentration of glucose by motor or claorimetric method determines the level of glucose.Recently, developed the multiple method that hemafecia can be determined blood analyte concentration that need not to get.For example authorize the U.S. Patent No. 5,267,152 (this paper quotes as a reference) of Yang etc., a kind of non-intruding technology of utilizing nearly IR radiation diffuse-reflectance laser spectrum measuring blood concentration has been described.Similarly near IR spectroscopy equipment is in the U.S. Patent No. 5,086,229 of authorizing Rosenthal etc. and authorize in the U.S. Patent No. 4,975,581 of Robinson etc. illustrated (this paper quotes as a reference) arranged.
Authorize the U.S. Patent No. 5 of Stanley, 139,023 (this paper quotes as a reference) illustrated a kind of percutaneous blood sugar monitor, and it relies on permeability booster (for example bile salts) to make things convenient for glucose to move along the Concentraton gradient percutaneous of setting up between interstitial fluid and receiver media.The U.S. Patent No. 5,036,861 (this paper quotes as a reference) of authorizing Sembrowich has illustrated a kind of passive glucose monitor, and it pastes by skin and collects perspiration, wherein uses the preparation of cholinomimetic energy to stimulate the sweat gland secretion perspiration.Similarly sweat collection equipment has explanation (this paper quotes as a reference) in the U.S. Patent No. 5,076,273 of authorizing Schoendorfer and the U.S. Patent No. 5,140,985 of authorizing Schroeder.
In addition, the U.S. Patent No. 5,279,543 (this paper quotes as a reference) of authorizing Glikfeld has illustrated and has used ionotherapy non-invasively material to be sampled in the container of skin surface by skin.Glikfeld has explained, thus biosensor or the special electrode coupling monitoring blood glucose of glucose that this sampling process can be special with glucose.And, the open No.WO 96/00110 (this paper quotes as a reference) in the world that authorizes Tamada has illustrated a kind of iontophoresis apparatus that is used for the percutaneous carbon dioside monitoring target substance, wherein the iontophoresis electrode is used for analyte is moved into catcher, and uses the target analytes in the biosensor exploration vessel.At last, the U.S. Patent No. 6,144,869 (this paper quotes as a reference) of authorizing Berner has illustrated a kind of sampling system that is used to measure the concentration of existing analyte.
Further, BG meter or obtain structure and can also comprise inherent conduit and the sampling of subcutaneous tissue liquid.
Computer or PDA 740 comprise software and hardware, and they are according to the diabetics data of predetermined flow sequence (describing in detail as top) processing, analysis and interpretation oneself record and to produce suitable data interpretation output necessary.Preferably, the result according to data analysis of being carried out by the patient data of computer 740 storages and interpretation is shown by the printer generation form that links to each other with personal computer 740.Selectively, the result of data interpretation program can directly be presented on the video display unit that links to each other with computer 740.
The block diagram that Fig. 8 shows has been represented the selected embodiment with diabetes-management system, and it is the device 810 by patient's operation, its shell (housing) thus the preferably enough compact device 810 that makes can enoughly hand and carry by the patient.On the surface of shell 816, be mounted with a bar guider that is used to receive the test strip for blood-sugar (not shown).Test strip is used to receive the blood sample from the patient.This device comprises a microprocessor 822 and the memorizer 824 that links to each other with microprocessor.Thereby microprocessor 822 is designed to carry out the computer program that is stored in the memorizer 824 and carries out various calculating and control function, describes in detail as top.Keypad 816 links to each other with microprocessor 822 by standard keypad decoder 826.Display 814 links to each other with microprocessor 822 by display driver 830.Microprocessor 822 is communicated by letter with display driver 830 by interface, and display driver 830 is revised and refresh display 814 under the control of microprocessor 822.Speaker 854 also is connected microprocessor 822 with clock 856.Speaker 854 is worked under the control of microprocessor 822, thereby sends the voice that can listen, makes the patient in the future possible hypoglycemia vigilance.Clock 856 provides current date and time to microprocessor 822.
Memorizer 824 is the parameter value of blood glucose value, insulin dose value, insulin type and microprocessor 822 uses of store patient 812 also, and blood glucose value, supplementation with insulin dosage and carbohydrate replenish in order to calculate in the future.Each blood glucose value and insulin dose value all are stored in the memorizer 824 with corresponding date and time, and memorizer 824 is nonvolatile memory preferably, for example electric erasable read only memorizer (EEPROM).
Device 810 also comprises a blood-glucose meter 828 that links to each other with microprocessor 822.Glucose meter 828 is designed to measure the blood sample that is received on the test strip for blood-sugar, and produces the blood glucose value that blood sample is measured.As previously mentioned, this glucose meter is known in this area.The type of glucose meter 828 preferably produces the digital value that directly outputs to microprocessor 822.Selectively, the type of blood-glucose meter 828 can be to produce the analogue value.Select among the embodiment at this, blood-glucose meter 828 links to each other with microprocessor 822 by the analogue-to-digital converters (not shown).
Device 810 further comprises an input/output end port 834, preferably, comprises the port of a series of connection microprocessors 822.Port 834 preferably, connects modem 832 by standard RS232 interface by interface.Modem 832 is used for by communication network 836 at device 810 and personal computer 840, and perhaps health is protected personnel's computer 838, between set up communication.The special technology that connects electronic equipment by connection cord is known in this area.It is " bluetooth " technical communication that another kind can be selected example.
Selectively, the block diagram that Fig. 9 shows has been represented the selected embodiment with diabetes-management system, it is the device 910 by patient operation, with shown in Figure 8 similar, its shell (housing) thus the preferably enough compact device 910 that makes can enoughly hand and carry by the patient.For example, separate type or can differ from and unload the formula glucose meter or BG obtains mechanism/module 928.Self-monitoring equipment has been arranged already, and it is computational algorithm 1,2,3 and need not to send the data to other things to patient's display result directly.The example of this equipment is the ULTRA SMART that the Lifescan company limited is produced, Therasense, Alameda, Milpitas, CA and FREESTYLE TRACKER that CA produces.
Therefore, here Shuo Ming embodiment can for example realize on the Internet at data communication network, make any processor and the computer of any remote location can both obtain this evaluation, estimation and information, as the Fig. 6-9 and/or the U.S. Patent No. 5 of authorizing Wood, 851, describe in 186, this paper quotes its content as a reference.Selectively, the patient of remote location can send to the BG data center health and protect personnel or infirmary or different remote location.
In a word, the present invention proposes a kind of computerized (perhaps non-computerization) data analysing method and system, be used for estimating simultaneously two most important component: HbA of diabetic individual glycemic control
1cWith hypoglycemia danger.Although this method is only used daily SMBG data, except other, three groups of outputs are provided.
The potentiality of the inventive method, system and computer program have been to provide following advantage, but are not limited only to this.At first, the present invention has improved the performance of existing domestic BG monitoring equipment by the content of carrying out and show below: 1) estimate HbA
1cScope, 2) estimate the probability and 3 of SH in 6 months subsequently) estimate the short-term danger (just, 24 hours from now on) of hypoglycemia.The latter can comprise warning, for example reports to the police, and shows to be about to take place the hypoglycemia incident.These three parts can also be combined, thereby the continuous information of relevant diabetic individual glycemic control is provided, and then improve the monitoring to hypoglycemia danger.
As additional advantage, the present invention has improved the performance of existing SMBG data retrieval software or hardware.Almost the manufacturing commercial city of each domestic BG monitoring equipment produces this software or hardware, and patient and Kang protect personnel and use its interpretation SMBG data usually.Method and system of the present invention can directly merge in the existing domestic blood glucose monitor, perhaps can predict HbA simultaneously by introducing
1cImprove the performance of SMBG data retrieval software with the data interpretation assembly of hypoglycemia high-risk period.
The another one advantage is that the present invention can assess the accuracy of domestic BG monitoring equipment in low and high BG scope and in the whole numerical range at BG simultaneously.
And, another advantage, the present invention can assess the effectiveness of various diabetotherapies.
Further again,, diabetics do not make the dangerous optimization problem that increases of its hypoglycemia because being faced with throughout one's life when keeping strict glycemic control, so the present invention has relaxed this relevant issues by using its better way method, just, the present invention's glycemic control of assess patient and danger of hypoglycemia thereof simultaneously can be applied under patient's daily condition simultaneously.
In addition, the present invention is by proposing the contact that three kinds of completely different but compatible algorithms have provided disappearance, and these three kinds of algorithms all are used for by SMBG data estimation HbA
1cWith the danger of hypoglycemia, thereby be used to predict the short-term of hypoglycemia and the long-term danger of long-term danger and hyperglycemia.
Another advantage, the present invention can assess the effectiveness of novel insulin or insulin input (delivery) equipment.The manufacturer of any insulin or insulin dispensing device or research worker can both utilize embodiments of the invention to test the insulin type of its proposition or detection or the success relatively that equipment is thrown in design.
At last, another advantage, the present invention can assess the effectiveness of insulin adjuvant therapy medicaments.
The invention example
I. example No.1
Example No.1 comprises three kinds of algorithms, is used for being estimated simultaneously by daily SMBG data two most important components of diabetes glycemic control, HbA
1cWith hypoglycemia danger.This method directly is adapted to pass through introducing can predict HbA simultaneously
1cImprove the performance of existing domestic BG monitoring equipment with the intelligent data interpretation assembly of hypoglycemia high-risk period.This data analysing method has three parts (algorithm):
● algorithm 1: estimate HbA
1c
● algorithm 2: estimate severe hypoglycemia disease (SH) long-term danger and
● algorithm 3: estimate short-term (in the 24-48 hour) danger of hypoglycemia.
Another important goal of example 1 is with available data check a large amount of hypothesis and viewpoint, and these hypothesis and viewpoint might produce other other algorithm, and they estimate HbA in the open mode that is provided of the conceptive the present invention of being different from
1cWith the danger of calculating hypoglycemia.Target is to find potential better solution, confirms simply that perhaps some viewpoint can not produce better result, and it is mainly used in optimizes and improve analysis to the data of collecting at present in this case study No.2.
Data set(data sets)
For the optimized results that guarantees us can be generalized to level widely, algorithm 1 and 2 is at first optimized with the training data group, checks its precision with incoherent test data set then.For algorithm 3, we have only one group of data that contain parallel SMBG and SH record at present.Being described in detail as follows of patient population investigation (population):
(1)
Training data group 1: 96 patients are suffered from TIDM by diagnosis in this research before 2 years at least.Wherein having 43 patients to be reported in had severe hypoglycemia disease incident in the past year at least 2 times, and 53 patients are reported in does not have this incident in the identical time.38 male and 58 women are wherein arranged.Mean age is 35 ± 8 years old, and the average duration of diabetes is 16 ± 10 years, and the average insulin of every day is 0.58 ± 0.19 unit/kg, average HbA
1cBe 8.6 ± 1.8%.These objects have been collected about 13,000 SMBG readings in 40-45 days time.The frequency of SMBG is about 3 reading/skies.These data collections have been proceeded 6 months, monthly write down moderate and severe hypoglycemia disease incident.This data set (does not have previous HbA as algorithm 1
1c) and the training data group of algorithm 2.
(2)
The training data group: 85 patients are suffered from TIDM by diagnosis in this research before 2 years at least, and all patients all are reported in the SH incident in the past year.44 male and 41 women are wherein arranged.Mean age is 44 ± 10 years old, and the average duration of diabetes is 26 ± 11 years, and the average insulin of every day is 0.6 ± 0.2 unit/kg, average benchmark HbA
1cBe 7.7 ± 1.1%, and average 6 months HbA
1cBe that 7.4 ± 1% (60 objects have 6 months HbA
1c).These objects are at twice HbA
1cAbout 75,500 SMBG readings have been collected in 6 months between the chemical examination.SMBG frequency in the data set 2 is higher, is 4-5 reading/sky.In addition, during 6 months of SMBG, the hold the record date and time of moderate and severe hypoglycemia disease incident and generation thereof of these objects, the result has the SH incident 399 times.This data set (has previous HbA as algorithm 1
1c) with the training data group of all analyses relevant with algorithm 3.
(3)
Test data set: we use the data of N=600 object, and wherein 277 have T1DM, and 323 have T2DM, and they all use insulin for treating diabetes.These data are collected by the pharmacists of cen.am. Santiago Amylin company, comprise 6-8 month SMBG data (about 300,000 readings), have benchmark and 6 months HbA simultaneously
1cResult and some consensus datas.These objects have been participated in the clinical experiment investigation of pramlintide (dosage is the 60-120 microgram) to metabolism control.Object use pramlintide in T1DM and the T2DM group is (table 1) at random.
Table 1: the demographic feature of object in the test data set
| Variable | T1DM meansigma methods (SD) | T1DM meansigma methods (SD) | The P level |
| Age (year) | 38.0(13.4) | 58.1(9.4) | <0.001 |
| Sex: man/woman | 136/41 | 157/166 | Ns |
| Benchmark HbA 1c | 9.74(1.3) | 9.85(1.3) | Ns |
| 6 months HbA 1c | 8.77(1.1) | 8.98(1.3) | 0.04 |
| The diabetes persistent period | 14.6(9.8) | 13.5(7.6) | Ns |
| The beginning age (year) | 23.4(12.8) | 44.6(10.4) | <0.001 |
| #SMBG reading/object/sky | 3.2(1.1) | 2.9(0.9) | <0.005 |
Table 1 has provided the comparison of demographic feature and T1DM and T2DM object.Initial 6 months of research, the average HbA of T1DM and T2DM group
1cAll significantly descend, this may be the use owing to medicine, and this has exceeded the scope of this report book (presentation) (table 1).HbA
1cChange relatively fast and allow the predictive ability of assessment algorithm 1 better.In all data sets, SMBG carries out with the ONE TOUCH II or the ONETOUCH PROFILE meter of Lifescan company.
Algorithm 1: estimate HbA
1c
Example No.1 has provided, but is not limited only to, HbA
1cThe optimization of prediction (algorithm 1), it passes through: (1) SMBG is more near the center, and weight is high more; (2) high BG event time is long more, and weight is high more; (3) usefulness HbA not long ago
1cProofread and correct high BG index; (4) merge other patient's variable, for example age, sex and disease persistent period.
(1) at first, derive the majorized function of the different independent variables of a plurality of uses, best persistent period and best SMBG frequency by two training data groups 1 of collecting in our the research formerly and 2 to T1DM patient;
(2) then, fix all coefficients, algorithm 1 is applied to much bigger test data set, this test data set is included in the data by T1DM that collects in Amylin lumen apothecariorum's the clinical experiment and T2DM object simultaneously under very different conditions;
(3) only use the precision of 1 pair of various majorized function of test data set detailed evaluation algorithm.
The training and testing data component is from allowing our statement, and the estimated accuracy of algorithm 1 can be generalized to any other data of T1DM or T2DM object.And, because Amylin data (test data set) are from just receiving treatment to reduce its HbA
1cObject in collect so their HbA
1cDemonstrated abnormal big variation in 6 months observation cycle, we can state that algorithm 1 not only can be predicted constant relatively HbA
1c, and can predict big and undesired fast-changing HbA
1cIn identical row, algorithm 1 is optimized its HbA for hope
1cThe patient be the most useful, can infer that this patient colony for example estimates HbA continuously probably for the meter of seeking to have advanced
1c, interested.
The summary of results
● the best SMBG data collection cycle is 45 days;
● best SMBG frequency is 3 reading/skies;
● two best HbA have been proposed
1cEstimation function: F1-only uses SMBG data and F2-to use the SMBG data to add at prediction HbA
1cThe HbA that obtained in about before 6 months
1cReading;
● by a plurality of standards of in nextpage (table 2), describing in detail to the middle HbA of test data set (N=573 object)
1cAccuracy of predicting is assessed.We here illustrate, in T1DM, and the overall accuracy (HbA of F1
1cMeasured value 20% within) be 96.5%, and the overall accuracy of F2 is 95.7%.For T2DM, the overall accuracy of F1 is 95.9%, and the overall accuracy of F2 is 98.4%.Therefore, the equal and HbA of the precision of F1 and F2
1cDirect measurement result suitable;
● most important, for HbA
1cChange the patient (N=68) of 2 or more a plurality of units from its datum readings, no matter F1 predicts the precision of this variation be 100% for T1DM or T2DM, and F2 is respectively 71% and 85% for the precision of T1DM and T2DM;
● F1 and F2 were for 6 months HbA
1cThe initial HbA of prediction during all than 0 month
1cEstimate to want much accurate.But use average BG as HbA
1cDirect estimation be inaccurate;
● test a great selection of method of selecting, for example selected the special time point (post-meal readings) of every day to estimate HbA
1c, according to each SMBG reading and HbA
1cInterlude is given the different weight of SMBG reading between determining, to having different average blood sugar/HbA
1cThe object of ratio is assessed respectively etc.But although there are some to obtain than two results that function is better that propose above in these systems of selection, neither one is better on the whole.We can reach a conclusion: will use majorized function F1 and F2 in using the future of algorithm 1.
Detailed results-test data set
Optimal algorithm 1: for each object, selected 45 days subclass of his SMBG reading.The Start Date of this subclass is greatly about 6 months HbA of object
1cChemically examine before 75 days, the Close Date is approximately 30 days before this chemical examination.Because in this data set, HbA
1cThe time of chemical examination is just known roughly, so at the last SMBG reading and the HbA that analyze of once being used to
1cBetween interlude and imprecise.This time cycle is by its persistent period of continuous optimization and its end point (HbA
1cTime before) choose.The best persistent period is 45 days.The best concluding time is HbA
1c1 month before.In other words, 45 days SMBG can shift to an earlier date about 1 month prediction HbA
1cNumerical value.Yet, predict 45-75 days between any one HbA
1cNumerical value is all about the same good---and difference is digital but not clinical significance.Similarly, the difference between the monitoring periods of 45 days monitoring periods and 60 days and little.But, can make that less than 45 days monitoring periods predictive ability descends fast.
The best estimate function is linear, and by under provide:
F1＝0.809098＊BGMM1+0.064540＊LBGI1-0.151673＊RHI1+1.873325
F2＝0.682742＊HBA0+0.054377＊RHI1+1.553277
In these formula, BGMM1 is the average blood sugar of being calculated by 45 days SMBG reader; LBGI1 is the low and high BG index of being calculated by identical reader with RHI1, and HBA0 only is used to estimate 2 benchmark HbA
1cReading.Coefficient value is optimized with the training data group, and in detailed results---provided relevant statistics and drawing in the training data group part.
(1) from HbA
1cThe absolute deviation (AERR) that measured value is estimated;
(2) from HbA
1cThe absolute percent deviation (PERR) that measured value is estimated;
(3) HbA
1cPercentage ratio in the measured value 20% is estimated (HIT20);
(4) HbA
1cPercentage registration (HIT10) in the measured value 10%; With
(5) HbA
1cPercentage registration (MISS 25) around the measured value outside 25% interval.
Table 2A: the precision (N=254 object) of algorithm 1 in T1DM
| F1 | F2 | Average BG | Previous HbA 1c | The P value | |
| AERR | 0.77 | 0.61 | 1.68 | 1.1 | <0.001 |
| PERR(％) | 8.3 | 7.1 | 19.4 | 12.8 | <0.001 |
| HIT20(％) | 96.5 | 95.7 | 61.0 | 81.0 | <0.001 |
| HIT10(10) | 65.4 | 75.5 | 29.9 | 48.2 | <0.001 |
| MISS?25(％) | 2.4 | 1.6 | 28.4 | 9.9 |
Table 2B: the precision (N=319 object) of algorithm 1 in T2DM
| F1 | F2 | Average BG | Previous HbA 1c | The P value | |
| AERR | 0.72 | 0.57 | 1.92 | 0.87 | <0.001 |
| PERR(％) | 7.6 | 6.4 | 20.9 | 11.7 | <0.001 |
| HIT20(％) | 95.9 | 98.4 | 56.4 | 82.8 | <0.001 |
| HIT10(10) | 70.2 | 79.3 | 29.5 | 53.3 | <0.001 |
| MISS?25(％) | 1.2 | 0.6 | 36.7 | 8.2 | <0.001 |
Preceding two row of table 2A and 2B have provided the result of majorized function F1 and F2 respectively.The third line has provided when adopting average BG (mmol/L) as HbA
1cThe precision of estimating during estimation.Fourth line has provided the HbA with 0 o'clock
1cChemical examination was as 6 months HbA
1cEstimate the same precision measurement result of calculating.Obviously, for T1DM and T2DM, F2 is for estimating HbA
1cAll slightly good on the whole than F1.The most important thing is that F1 and F2 are for HbA
1cEstimation all than with its Zao numerical value or on average BG estimate far better.Estimate that for the % that drops on outside the 25% precision interval this is true especially.Carry out F1 and F2 and by previous HbA
1cIt is highly significant (the 4th row) that estimated differences is carried out in chemical examination.
Statistics is noted: emphasis to notice, use traditional recurrence type standard, for example the R that from the ANOVA table, obtains
2Or F and p value, the precision of assessment algorithm 1 is inappropriate.This is because parameter estimation is obtained from another incoherent data set (training data), and just is applied to this test data set.Therefore, (for example in test data set, the summation of residual error is not zero) violated in the statistical assumption of basic (underlying) model, so R
2Or F and p value have lost their statistical significance.
T1DM and T2DM object by checking that SMBG datum readings and 6 months subsequently reading have significant change have carried out further assessment to the precision of algorithm 1 in test data set.Table 3A and 3B have provided HbA
1cAbsolute change be equal to, or greater than the tabulation of the T1DM and the T2DM object of 2 units.In each group of objects, the HbA of 34 objects is arranged
1cHas this variation.Algorithm 1, function F 1, all 100% ground has doped this change in T1DM and T2DM.Owing to contain benchmark HbA in the formula
1c(it partly is withdrawn into HbA with estimated value
1cReference value), so the predictive power of F2 diminishes, in T1DM, be 71%, be 85% in T2DM.Except that two objects, benchmark HbA
1cDrop on 6 months HbA
1cThe outside that is worth 20% interval:
Table 3A:HbA
1c
Change=the T1DM object of 2 units
ID HBA0 HBA6 DHBA F1 F2 HIT?F1 HIT?F2 HIT
HBAO
6504 12.0 7.0 5.00 6.82 9.90 100.00 .00 .00
6613 10.5 6.8 3.70 8.02 9.37 100.00 .00 .00
4003 12.4 8.9 3.50 8.45 10.73 100.00 .00 .00
6204 11.0 7.5 3.50 7.29 9.45 100.00 .00 .00
3709 13.0 9.7 3.30 8.99 11.54 100.00 100.00 .00
4701 12.8 9.5 3.30 9.50 11.61 100.00 .00 .00
3614 11.9 8.7 3.20 8.24 10.30 100.00 100.00 .00
3602 11.5 8.3 3.20 7.93 9.94 100.00 100.00 .00
6008 11.3 8.3 3.00 9.30 10.53 100.00 .00 .00
3723 13.0 10.1 2.90 8.80 11.46 100.00 100.00 .00
7010 12.7 9.8 2.90 8.09 10.89 100.00 100.00 .00
6208 11.5 8.7 2.80 8.42 10.09 100.00 100.00 .00
6202 10.6 7.8 2.80 7.91 9.37 100.00 .00 .00
3924 9.9 7.2 2.70 7.71 8.72 100.00 .00 .00
8211 11.0 8.3 2.70 8.76 10.32 100.00 .00 .00
6012 9.3 6.7 2.60 7.82 8.35 100.00 .00 .00
3913 11.0 8.4 2.60 7.88 9.54 100.00 100.00 .00
6701 11.2 8.6 2.60 8.75 10.07 100.00 100.00 .00
2307 10.6 8.1 2.50 7.95 9.27 100.00 100.00 .00
3516 11.8 9.3 2.50 7.76 10.03 100.00 100.00 .00
5808 9.6 7.2 2.40 7.61 8.52 100.00 100.00 .00
2201 1.8 9.5 2.30 8.90 10.71 100.00 100.00 .00
4010 12.4 10.1 2.30 8.57 11.15 100.00 100.00 .00
6210 11.9 9.6 2.30 8.33 10.40 100.00 100.00 .00
4904 11.3 9.1 2.20 8.63 10.29 100.00 100.00 .00
6709 10.3 8.1 2.20 7.83 9.04 100.00 100.00 .00
6619 9.5 7.3 2.20 7.64 8.57 100.00 100.00 .00
3921 10.9 8.8 2.10 7.20 9.19 100.00 100.00 .00
6603 11.0 8.9 2.10 8.18 9.89 100.00 100.00 .00
7415 10.6 8.5 2.10 7.94 9.27 100.00 100.00 .00
6515 9.8 7.8 2.00 7..13 8.54 100.00 100.00 .00
3611 10.3 8.3 2.00 8.36 9.23 100.00 100.00 .00
3732 13.2 11.2 2.00 9.30 11.99 100.00 100.00 100.00
7409 10.0 8.0 2.00 7.99 9.04 100.00 100.00 .00
Table 3B:HbA
1c
Change=the T1DM object of 2 units
ID HBA0 HBA6 DHBA F1 F2 HIT?F1 HIT?F2 HIT
HBA0
6754 10.8 7.0 3.80 6.90 9.03 100.00 .00 .00
6361 1.3 7.6 3.70 8.51 10.20 100.00 .00 .00
6270 12.0 8.6 3.40 7.85 10.03 100.00 100.00 .00
6264 11.1 7.8 3.30 8.31 9.70 100.00 .00 .00
6355 11.8 8.6 3.20 7.99 9.90 100.00 100.00 .00
3961 10.8 8.0 2.80 9.13 9.73 100.00 .00 .00
6555 11.1 8.3 2.80 8.11 9.55 100.00 100.00 .00
8052 11.7 8.9 2.80 7.68 9.80 100.00 100.00 .00
5356 9.7 7.0 2.70 6.75 8.20 100.00 100.00 .00
3966 10.3 7.7 2.60 8.08 9.07 100.00 100.00 .00
908 9.5 6.9 2.60 7.47 8.23 100.00 100.00 .00
6554 10.7 8.1 2.60 8.16 9.42 100.00 100.00 .00
2353 11.1 8.7 2.40 8.99 9.90 100.00 100.00 .00
4064 11.3 8.9 2.4 0?7.89 9.88 100.00 100.00 .00
6351 10.1 7.7 2.40 7.92 8.63 100.00 100.00 .00
7551 12.2 9.8 2.40 9.17 11.02 100.00 100.00 .00
6358 8.4 6.1 2.30 7.00 7.32 100.00 .00 .00
3965 10.1 7.8 2.30 7.83 8.64 100.00 100.00 .00
914 11.1 8.8 2.30 9.57 10.33 100.00 100.00 .00
1603 10.2 7.9 2.30 8.02 8.88 100.00 100.00 .00
1708 10.8 8.6 2.20 7.62 9.24 100.00 100.00 .00
3761 12.4 10.2 2.20 9.13 10.86 100.00 100.00 .00
3768 11.2 9.0 2.20 8.29 9.74 100.00 100.00 .00
326 10.3 8.2 2.10 7.45 8.78 100.00 100.00 .00
109 9.3 7.2 2.10 7.70 8.18 100.00 100.00 .00
1501 11.9 9.8 2.10 8.52 10.18 100.00 100.00. .00
3964 13.7 11.6 2.10 10.08 12.65 100.00 100.00 100.00
4352 12.2 10.1 2.10 9.51 11.14 100.00 100.00 .00
7858 12.1 10.0 2.10 9.53 11.01 100.00 100.00 .00
4256 10.6 8.6 2.00 8.76 9.69 100.00 100.00 .00
4752 10.1 8.1 2.00 8.51 8.87 100.00 100.00 .00
6556 11.1 9.1 2.00 8.72 9.68 100.00 100.00 .00
6562 7.9 5.9 2.00 7.07 7.04 100.00 100.00 .00
8255 10.9 8.9 2.00 8.90 9.87 100.00 100.00 .00
In table 3A and 3B:
The ID numeral of ID---object;
HBA0---benchmark HbA
1c
HBA6---HbA
1c6 months measured values;
DHBA---HbA
1cAbsolute difference between the value of reference value and 6 months;
F1---by the HbA of function F 1 estimation
1c, only use the SMBG data;
F2---by the HbA of function F 2 estimations
1c, use previous HbA
1cResult of laboratory test;
HitF1=100 is if F1 was at 6 months HbA
1cValue 20% within, otherwise be 0;
HitF2=100 is if F2 was at 6 months HbA
1cValue 20% within, otherwise be 0; With
If Hit HbA0=100 is benchmark HbA
1cAt 6 months HbA
1cReading 20% within, otherwise be 0.
Detailed results---training data group
This part has illustrated the step of optimizing algorithm 1.This optimization comprises two parts: (1) supposition can not obtain previous HbA
1cReading and (2) supposition can be used previous HbA
1cPrediction HbA
1c
We have considered that a plurality of different functions are used to describe SMBG data and HbA
1cRelation.According to precision and the simplicity calculated, if do not use previous HbA
1cReading, optimal function be meansigma methods, the exponential linear function of low and high BG of SMBG reading seemingly, and another is previous HbA
1cWith the exponential linear function of high B.Non-linear relation does not improve the coupling of this model well to be spent, and does not therefore consider to be used for practical application.
Training data group 1-no previous HbA 1c Use the coefficient of linear regression model (LRM) majorized function F1.Provide in the optimum coefficient part in front.Here, we provide the data of the good degree of relevant Model Matching:
Multiple R (Multiple R) .71461
R side .51067
Variance analysis
The DF quadratic sum is mean square
Residual error 90 148.10903 1.64566
F=31.30889 significance F=.0000
The residual analysis of this model demonstrates the normal distribution (seeing Figure 11) that approaches residual error.The SD of residual error is 1.2 (are 0 according to its meansigma methods of definition).Therefore, we can accept, and this model is described these data well.
Training data group 2-previous HbA 1c Again, use the coefficient of linear regression model (LRM) majorized function F3.Provide in the optimum coefficient part in front.Here, we provide the data of the good degree of relevant Model Matching:
Multiple R .86907
R side .75528
Variance analysis
The DF quadratic sum is mean square
F=41.66522 significance F=.0000
The residual analysis of this model demonstrates the normal distribution (seeing Figure 12) near residual error.The SD of residual error is 0.47 (is 0 according to its meansigma methods of definition).Therefore, we can accept, and this model has been described these data well.
In addition, relatively there is not and has previous HbA
1cModel, we can reach a conclusion, if comprise previous HbA in calculating
1c, then final model beguine is according to R
2With all far better according to the residual error error.
Yet, formerly see in the part as us, in uncorrelated data set, previous HbA
1cFor the not contribution of overall accuracy of prediction, in some cases, HbA
1cBe changed significantly even owing to changing the ability that has hindered algorithm fast.Therefore, we can reach a conclusion, even previous HbA
1cMay see better from the viewpoint of statistics, but enough actual utilitys can not be arranged, be used to proofread and correct the reading input of meter in the future.We also do not know HbA
1cChemical examination and SMBG detect interlude between (profile), but still make HbA
1cInput useful.Perhaps, this depends on HbA in this time cycle
1cChange---as what see in our the pro-part, 2 HbA
1cThe variation of unit just makes previous HbA
1cReading to no avail.
SMBG/HbA 1c Ratio
We provide a kind of selectable method now, with the statistical accuracy and the quite high clinical practice of maintenance of improved model coupling.As seen, the meansigma methods of 45 days SMBG readings and HbA
1cRatio be one and have the measuring amount of complete normal distribution (can confirm) by the Kolmogorov-Simov experiment, three group objects have been distinguished on and most important ground, its ratio is＜1.0,1.0-1.2 and 1.2.Preceding two groups each can both be explained about 40% object, can explain about 20% object for the 3rd group.This all is effectively for T1DM and T2DM, and can both observe in training data group and test data set.As if in addition, this ratio is stable especially for the time, therefore may be that (for example, if SMBG mainly carries out when BG is low, then final meansigma methods will be underestimated HbA to the measuring amount that reflects patient SMBG custom
1cThereby corresponding ratio will be less than 1.0).Notice that this is a hypothesis, can not be confirmed that we have carried out some analyses, as if confirm to know everyone ratio to a certain extent at certain time point with obtainable data.As if this be equivalent to know previous HbA
1c, and might be equivalent to the data input, but the application of this ratio and previous HbA
1cUse very different.Except being directly used in this predictor formula, this ratio also is used for the patient is classified, and they are used some in three different predictor formulas.These new formula directly do not comprise HbA
1c, therefore can not comprised HbA
1cThe influence of the inertia that causes (inertia).In addition, average HbA between three groups that limit by this ratio
1cDifference is also little, and uncorrelated with this ratio, so the reason that ratio is different in the different people is certain and HbA
1cUncorrelated.
If we at first are divided into 3 groups and return respectively according to their ratio with object in the training data group, then the no-float degree of regression model significantly increases: (1) in group 1 (ratio＜1.0), we obtain multiple R=0.86, R
2=0.73; (2) in group 2 (ratio is 1.0-1.2), matching degree the best, R=0.97, R
2=0.94 and (3) in group 3 (ratio〉1.2), matching degree is the poorest, R=0.69, R
2=0.47.Because all three kinds of regression models do not comprise previous HbA
1cSo we reach a conclusion, the object for about 80%, the good degree of coupling significantly increases, the object of residue 20%, the good degree of coupling keeps identical, and matching degree can be identified in advance with the object of variation.
Further, the ratio according to object is divided into 3 groups with test data set.The precision of prediction that we obtain and the previous similar accuracy that obtains (table 4A and 4B):
Table 4A: the precision of algorithm 1 in T1DM (N=254 object)
| Ratio＜1.0 | 1.0≤ratio≤1.2 | Ratio〉1.2 | |
| AERR | 0.70 | 0.63 | 0.74 |
| PERR(％) | 7.8 | 7.4 | 7.9 |
| HIT?20(％) | 93.8 | 93.0 | 95.5 |
| HIT?10(％) | 68.8 | 73.4 | 72.7 |
| MISS?25(％) | 3.1 | 2.6 | 0.0 |
Table 4B: the precision of algorithm 1 in T2DM (N=319 object)
| Ratio＜1.0 | 1.0≤ratio≤1.2 | Ratio〉1.2 | |
| AERR | 0.63 | 0.68 | 0.89 |
| PERR(％) | 7.6 | 7.8 | 8.8 |
| HIT?20(％) | 97.4 | 95.0 | 95.3 |
| HIT?10(％) | 67.2 | 65.3 | 57.7 |
| MISS?25(％) | 0.0 | 1.7 | 0.0 |
In brief, know the SMBG/HBA of each object
1cAs if ratio and correspondingly estimating respectively can improve the statistic property of this model and not lose clinical precision.
Tested other hypothesis and thinking
We have tested a large amount of other hypothesis and thinkings, and they are proved at least for promoting and more to analyze the data of being collected by example No.2 be useful in the concentrated area.Concise and to the point presentation of results is as follows:
(1) HbA
1cWith the SMBG reading---12 noon is to 6 pm---the most related (being correlated with) that obtains afternoon, minimum related with empty stomach SMBG reading (4a.m.-8a.m.).Yet, be not only to gather after the meal that the SMBG reading will improve HbA
1cPrediction, on the contrary, if ignore the contribution of whole day all hours (less relatively but important), then prediction can become even worse.If carry out different weightings for the reading of whole day other hour, then might improve HbA
1cPrediction, but this raising is not sufficient to remedy the additional complexity of model;
(2) in T2DM, HbA
1cStronger in T1DM with the relation ratio of average SMBG, even two groups HbA
1cMate mutually.According to directly related property, in whole research, the coefficient among the T1DM is about 0.6, and the coefficient among the T2DM is approximately 0.75;
(3) according to SMBG and HbA
1cInterlude carries out different weighting (for example the approaching more centre of result, weighting is high more) to the SMBG reading and can not obtain HbA between the chemical examination
1cBetter prediction;
(4) comprise demographic variable, for example age, diabetes persistent period, sex etc. can not improve HbA
1cPrediction;
(5) HbA
1cAnd may provide by following formula by the simplest linear relationship between the average SMBG (is that unit is measured with mmol/l): HbA
1c=0.41046*BGMM+4.0775.Although with F1 compare with F2 the statistics on relatively poor, the HbA that this formula provides
1cEstimation in T1DM and T2DM, have about 95% precision (according to HbA
1cThe deviation of chemical examination is less than 20%), if and by calculating low and high BG index is incorporated in it in meter, then may be useful (still, do not calculate low BG index and be the prediction that can not realize hypoglycemia, thus this formula may be only to comprising algorithm 1 but do not comprise that the meter of algorithm 2 and 3 is useful).
The assessment of algorithm 2:SH long-term danger
Example No.1 provides, but be not limited only to, the expansion of algorithm 2, comprise and estimate individual biochemical significantly hypoglycemia (biochemical significant hyoglycemia) (BSH, (BMH is defined as the probability of 39mg/dl＜BG＜=55mg/dl) to be defined as BG reading＜=39mg/dl) or biochemical moderate hypoglycemia.In addition, our Program Assessment is compared with SH in the daytime, and whether the incidence rate of algorithm 2 prediction (midnight is to 7:00am) SH at night is better.
(4) at first, draw a plurality of optimal classification variablees and optimal classification scope, best persistent period and best SMBG frequency from training data group 1;
(5) then, test data set is divided into two parts: initial 45 days and remaining data division.The optimal parameter of algorithm 2 is applied to initial 45 days parts of data, thereby uses the BSH and the MSH of the probability Estimation value prediction data second portion of BSH in the future or MSH;
(6) only algorithm 2 precision are carried out detailed evaluation with test data.
Training separates with test data set and allows our statement, the estimated accuracy of algorithm 2 can be generalized to any other data of T1DM or T2DM patient.And, because the Amylin data are from collecting through the object of intensive treatment, so we can infer that algorithm 2 is tested and confirmed is effective in dangerous constantly variation of hypoglycemia and ever-increasing object.
The summary of results
● estimation BSH in future or required best SMBG data collection cycle of BMH probability are 40-45 days.The optimum frequency of SMBG is 3-4 reading/sky.A large amount of readings can not cause the remarkable increase of algorithm 2 predictive abilities.If the reading of every day is less than 3, then predictive ability descends.Yet this requirement is the average reading with reference to every day in 45 days observation cycles, is not to mean, all needs to carry out 3-4 time reading every day;
● predictor variable the and in the future relation between SH and the MH is strictly non-linear.Therefore, linear method can not be used for optimizing prediction, although can obtain R by direct linear model
2=50% (by contrast, the optimum of DCCT is the SH in future of prediction 8%);
● predict that separately night SH is generally than predicting that SH is more weak in the daytime;
● determined 15 risk ranges of BSH and BMH in the future.The optimal separation of scope just obtains according to low BG index, although the combination of low BG index and other variable can be brought into play similar good effect;
● although the frequency of BSH and BMH is in T1DM and T2DM different (seeing Table 5), and when a given risk range, conditional frequency is between T1DM and T2DM and zero difference.This allows to obtain the unified prediction SH and the method for MH danger;
● be that 15 risk ranges have calculated each empirical probability in the future and compare.All more all have high significance, p ' s＜0.0005.
● these empirical probabilities with two-parameter Weibull distribute (two-parameter Weibulldistribution) be similar to, produced in each risk range the theoretical probability of BSH and BMH in the future.
● the good degree of these proximate couplings is very good---and all definite coefficients are all greater than 85%, and some are up to 98% (seeing Fig. 1-5 and 9-10).
Detailed results---test data set
Determine the individual risk range of SH/MH: the data of 600 objects are used for this analysis altogether.Be the low BG index (LBGI) of each calculation and object by his preceding 45 days SMBG data collection.Then, LBGI is categorized into one of them (scope of variable R CAT is 0-14) in 15 best risk ranges, as in training data group 1, drawing.These risk ranges define by following inequality:
if(LBGI≤0.25)，RCAT＝0
if(0.25<LBGI≤0.5)，RCAT＝1
if(0.50<LBGI≤0.75)，RCAT＝2
if(0.75<LBGI≤1.00)，RCAT＝3
if(1.00<LBGI≤1.25)，RCAT＝4
if(1.25<LBGI≤1.50)，RCAT＝5
if(1.50<LBGI≤1.75)，RCAT＝6
if(1.75<LBGI≤2.00)，RCAT＝7
if(2.00<LBGI≤2.50)，RCAT＝8
if(3.00<LBGI≤3.50)，RCAT＝9
if(3.50<LBGI≤4.00)，RCAT＝10
if(4.00<LBGI≤4.50)，RCAT＝11
if(4.50<LBGI≤5.25)，RCAT＝12
if(5.25<LBGI≤6.50)，RCAT＝13
if(LBGI>6.50)，RCAT＝14
The observed frequency of BSH and BMH: for each object, BSH that after initial 45 day data are collected 1 month, 3 months and 6 months inside countings are indicated by SMBG and the generation of BMH.Table 5A provided and observe 0 in T1DM ,=1 ,=2 ,=frequency of 3 BSH and BMH, table 5B has provided observed identical data in T2DM:
Table 5A: observed BSH and BMH frequency in T1DM
Table 5B: observed BSH and BMH frequency in T2DM
Night BSH and BMHAccount for by about 15% of whole incidents of SMBG indication.In the training data group, night incident and all the dependency between the predictor variable a little less than.We reach a conclusion, and night, the target prediction of incident was invalid.
The empirical probability of BSH and BMH in the future: we have calculated in 15 risk ranges the BSH and the specific experience probability of BMH in future of each.These probability comprise: (1) after the probability of at least BSH or BMH takes place in 1 month, 3 months and 6 months; (2) after the probability of quadratic B SH at least or BMH takes place in 3 months and 6 months; (3) after the probability of at least three BSH or BMH took place in 6 months.Certainly, might calculate any other combined probability as requested.
The most important conclusion that draws from this analysis is, a given risk range, and the probability of BSH and BMH does not have significant difference between T1DM and T2DM in the future.This allows to obtain these probability method among unified empirical and theoretical property prediction T1DM and the T2DM.As a result, T1DM and T2DM patient's data are combined and are used for following analysis.
Fig. 1-5 and 9-10 have provided 6 scatterplot that calculate empirical probability drawing according to 15 risk ranges.The empirical probability of BSH is represented with black triangle, and the empirical probability of BMH is represented with red shape.
Compare with single argument ANOVA empirical probability to all groups in 15 risk ranges, all p levels are all less than 0.0005.Therefore, we observe, the difference highly significant in different risk ranges between BSH and the BMH incident.
The theoretical probability of BSH and BMH in the future: in order to use the probability of direct formula estimation BSH in future and BMH, we are similar to empirical probability with two-parameter WeibulI probability distribution.The Weibull probability function is provided by following formula:
F (x)=1-exp (a, x
b), x〉0; Otherwise F (x)=0
Statistics is noted: parameter a and b be greater than 0, and be called scale parameter and form parameter respectively.In the particular example of b=1, Weibull distributes and becomes exponential type.This distribution is used in the engineering problem of being everlasting, because the distribution of generation technique failure each other and not exclusively irrelevant (if failure is irrelevant fully, then they will form the Piosson processing, and this will be illustrated in exponential, for example b=1) at random.The situation here is very inequality---and we need describe incomplete independence and the distribution of event (inefficacy) that trends towards trooping, and for example our previous institute confirms.
Every group of empirical probability all uses the theoretical formula that provides above to be similar to.With nonlinear least squares method parameter is estimated (estimating with the initial parameter that the Linear Double logarithmic model provides).The good degree of the coupling of every kind of model is determined coefficient (coefficient of determination) (D with it
2) assess.The meaning of this statistics and the R in the linear regression
2Similar, but R
2Can not be applied to nonlinear model.
Provided Model Matching among Fig. 1-6, black line is used for the probability of BSH, and dotted line is used for the probability of BMH.On each figure, we have provided the parameter estimation to corresponding model, so we have provided direct formula, be used to calculate take place 0 after the initial SMBG in 1 month, 3 months and 6 months ,=1 ,=2 ,=frequency of 3 BSH and BMH.In monitoring equipment or software, can comprise some of them formula or its modification indication as SH and MH danger.
The below of each figure has provided D
2Value is as indication of approximation quality.All numerical value is all greater than 85%, and some have reached 98%, and this confirms that this degree of approximation is very good, and confirms can use theoretical probability in research/application in the future, rather than empirical probability.
The theoretical probability that one or many moderate or severe hypoglycemia disease incident take place is provided by formula shown in Figure 1:
P(MH>＝1)＝1-exp(-exp(-1.5839)＊Risk＊＊1.0483)
P(SH>＝1)＝1-exp(-exp(-4.1947)＊Risk＊＊1.7472)
Fig. 1 provided 15 by in each of the danger level of low BG index definition after SMBG estimates moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia in 1 month.Because model is non-linear, so with their definite coefficient D
2Estimate the good degree of coupling, D
2Be R in the linear model
2Analog.Determine that coefficient and square root thereof are as follows:
SH model: D
2=96%, D=98%
MH model: D
2=87%, D=93%
The theoretical probability that one or many moderate or severe hypoglycemia disease incident take place is provided by formula shown in Figure 2:
P(MH>＝1)＝1-exp(-exp(-1.3731)＊Risk＊＊1.1351)
P(SH>＝1)＝1-exp(-exp(-3.2802)＊Risk＊＊1.5050)
Fig. 2 provided 15 by in each of the danger level of low BG index definition after SMBG estimates moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia in 3 months.
Determine that coefficient and square root thereof are as follows:
SH model: D
2=93%, D=97%
MH model: D
2=87%, D=93%
The theoretical probability that one or many moderate or severe hypoglycemia disease incident take place is provided by formula shown in Figure 3:
P(MH>＝1)＝1-exp(-exp(-1.3721)＊Risk＊＊1.3511)
P(SH>＝1)＝1-exp(-exp(-3.0591)＊Risk＊＊1.4549)
Fig. 3 provided 15 by in each of the danger level of low BG index definition after SMBG estimates moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia in 6 months.
Determine that coefficient and square root thereof are as follows:
SH model: D
2=86%, D=93%
MH model: D
2=89%, D=95%
Take place twice or repeatedly the theoretical probability of moderate or severe hypoglycemia disease incident provide by formula shown in Figure 4:
P(MH>＝2)＝1-exp(-exp(-1.6209)＊Risk＊＊1.0515)
P(SH>＝2)＝1-exp(-exp(-4.6862)＊Risk＊＊1.8580)
Fig. 4 has provided at 15 and has taken place 2 times in 3 months after SMBG estimates or repeatedly moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia by in each of the danger level of low BG index definition.
Determine that coefficient and square root thereof are as follows:
SH model: D
2=98%, D=99%
MH model: D
2=90%, D=95%
Take place twice or repeatedly the theoretical probability of moderate or severe hypoglycemia disease incident provide by formula shown in Figure 5:
P(MH>＝2)＝1-exp(-exp(-1.7081)＊Risk＊＊1.19555)
P(SH>＝2)＝1-exp(-exp(-4.5241)＊Risk＊＊1.9402)
Fig. 5 has provided at 15 and has taken place 2 times in 6 months after SMBG estimates or repeatedly moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia by in each of the danger level of low BG index definition.
Determine that coefficient and square root thereof are as follows:
SH model: D
2=98%, D=99%
MH model: D
2=89%, D=95%
Take place three times or repeatedly the theoretical probability of moderate or severe hypoglycemia disease incident provide by formula shown in Figure 9:
P(MH>＝3)＝1-exp(-exp(-2.0222)＊Risk＊＊1.2091)
P(SH>＝3)＝1-exp(-exp(-5.5777)＊Risk＊＊2.2467)
Figure 10 has provided at 15 and has taken place 3 times in 6 months after SMBG estimates or repeatedly moderate (dotted line) and the seriously experience and the theoretical probability of (black line) hypoglycemia by in each of the danger level of low BG index definition.
Determine that coefficient and square root thereof are as follows:
SH model: D
2=97%, D=99%
MH model: D
2=90%, D=95%.
Detailed results---training data group
The training data group comprises SMBG data and the severe hypoglycemia disease record that monthly writes down afterwards.With with determining that the test data set of BSH and BMH is relative by (cutoff) BG numerical value, unconscious, the stupor that the report of the serious symptoms that record monthly is included is defined as being caused by hypoglycemia, can't Heal Thyself or remarkable cognitive impairment.In 6 months after SMBG, the object of study report claims everyone that average 2.24 these incidents are arranged, and wherein 67% object is reported not this incident.From the viewpoint of statistics, this can make the distribution of SH incident significantly biased, and is not suitable for using linear method., can not be used to make up final model although thereby linear regression can be used in the relative distribution prediction SH that estimates various variablees.We have carried out three kinds of following analyses:
(1)
Do not know SH history: ignore any historical knowledge of SH, we are by benchmark HbA
1cWith the SMBG feature, the estimation BG risk that for example average BG, low BG exponential sum change is by in the future SH of regression forecasting (all variablees illustrate in initial open of the present invention).The same with former discovery, HbA
1cWith average BG to the prediction SH without any contribution.Final regression model comprises the variation of low BG exponential sum BG risk, and has following coupling and well spend:
Multiple R .61548
S square of .37882
Variance analysis
F=27.74772 significance F=.0000
------the variable in the equation------
| Variable | B | SE?B | β | T | Significance T |
| LBGI | 4.173259 | .649189 | 2.104085 | 6.428 | .0000 |
| Ratio | -5.749637 | 1.091007 | -1.724931 | -5.270 | .0000 |
| (constant) | -2.032859 | .790491 | -2.572 | .0117 |
(1)
Know previous SH: when we had comprised the number of SH incident the previous year, as reporting in the screening application form, this variable can be explained other 11% SH variation in future:
Multiple R .70328
S square of .49461
Variance analysis
F=29.35999 significance F=.0000
------the variable in the equation------
| Variable | B | SE?B | β | T | Significance T |
| LBGI | .337323 | .704286 | .375299 | 4.541 | .0000 |
| LDR | -4.350779 | 1.036380 | -1.305264 | -4.198 | .0001 |
| RLO | 3.134519 | .631684 | 1.580371 | 4.962 | .0000 |
| (constant) | -2.136619 | .717334 | -2.979 | .0037 |
(3) do not know the number of previous SH, only know that someone had before had or did not have SH, we only can explain that with the SMBG variable 45% SH in future makes a variation;
(4) last, two independent linear models can explain that 55% variation of SH in the daytime and 25% SH at night make a variation.All predictor variablees are also faint with the directly related property of SH at night.Night, incident accounted for 30% of whole SH.
We reach a conclusion, and linear prediction model can directly be explained the SH variation in future of about 40-50%.Yet, according to its residual error, this model not well balanced (this is because the biased distribution of height of SH event number in the diabetics demography).The normal probability paper figure of Figure 13 has provided its statistic evidence, and this figure has demonstrated the remarkable deviation of residual and its predictive value.
Therefore, we have adopted the method for another kind of prediction SH, utilize their SMBG data that object is divided into various risk ranges, and estimate in these scopes the probability of SH subsequently.We have attempted various disaggregated models, make the difference maximum between the risk range and attempt to obtain maximum danger estimation resolution (according to the maximum number of scope).
Obtain 15 risk ranges (Yi Bufen beginning formerly partly provides) and obtained optimal results by only classifying with low BG index.
The optimal separation, this result also has other advantage between scope: (1) does not need previous SH historical knowledge; (2) calculating relatively simply and not needs the tracking time variable, as if for example the pace of change of BG and (3) classification can be applied to T1DM and T2DM patient (without any need for the knowledge of previous SH) comparably.
Algorithm 3: the assessment of hypoglycemia short-term danger
Example No.1 has provided, but is not limited only to, the optimization of algorithm 3:
(1) utilizes benchmark long-term danger (from algorithm 2) and HbA
1c(from algorithm 1);
(2) hazard standard/threshold value of hypoglycemia warning;
(3) SMBG frequency;
(4) hypoglycemia is dangerous to be increased and not to have had SMBG for some time if detect, whether should send then that hypoglycemia is reported to the police and
(5) contribution of demography variable, for example severe hypoglycemia disease history.
Foreword
Different with algorithm 2 with the algorithm 1 with long-run development history, the problem that algorithm 3 is handled all is considered to impossible up to date.In fact, can people still generally believe the BG value (particularly hypoglycemia) that can not predict future according to the numerical value of previously known, and (Bremer T and Gough DA. be from previous numerical prediction blood glucose? data cause. (Is bloodglucose predictable from previous values? A solicitation for data) Diabetes, 1999
48: 445-451).The work that we are previous, it has queried this general view can report being arranged from the original copy that the Lifescan company limited obtains and at length being proposed in the present invention is open.For basis of explaining this query and the principle of illustrating algorithm 3 behinds, the paragraph below we have comprised.
We quantize diabetic character " philosophy ":
According to the hormonal system of being studied, the interaction between hormone is by the biochemical network control of dynamic regulation, and this network has the complicated or simple structure of being made up of main node and pipeline.Diabetes can upset the dynamic (dynamical) network of regulation and control insulin-glucose on various levels.For example, in T1DM, producing naturally of insulin eliminated fully, and in T2DM, insulin is subjected to the obstruction of more powerful insulin resistant in intracellular use.In T1DM (also common in T2DM), the external insulin succedaneum that needs certain form, this makes regulator control system become fragile for defective external factor, these external factor comprise the time limit and the dosage of pill and injection of insulin, the food of being eaten, sports etc.This can cause BG extreme deviation hypoglycemia and hyperglycemia to occur usually.In many but not all case, hypoglycemia excites the endocrine reaction, is referred to as reverse adjusting.Therefore, in art of mathematics, but the fluctuation of BG in a period of time is the measurement result that is subjected to the advanced dynamic system activity of a large amount of inside and outside factor affecting.Yet according to well-known dynamic system theory, when the complexity of regulation and control increased, a pure definitiveness system can develop into demonstration molar behavior at random.Therefore, in short-term (minute), observed BG fluctuation will be near determining on people's level, however in the long run, fluctuation will be near at random, and this comprises extreme transition, for example the SH incident.Therefore, stochastic modeling and statistical inference are suitable in long inner analysis system in period most---the example that algorithm 1 and algorithm 2 adopt, it utilizes our the initial means that propose after the specific observation cycle, for example LBGI and HBGI, the probability of a predicted numerical value scope and an incident.In addition modeling and the prediction of the enough deterministic networks of BG wave energy in a short time, this will realize in can carrying out intelligence insulin delivery in the future equipment of continuous detecting.
(1) some reference information of device collection research object is set up individual basic parameter;
(2) then, device begins to follow the trail of certain stack features of SMBG data;
(3) device has decision rule, determines when to set up to be about to take place the sign of SH, and when reduce this sign when data show that danger weakens;
(4) when having set up sign, we think that object (predicted time) in 24 hours subsequently will receive SH and report to the police.
This dynamic prediction is producing theoretic problem on the Model Parameter Optimization level and on the accuracy evaluation level of the best approach simultaneously.We begin to set forth second problem, because it is most important for the effect of understanding algorithm 3.
(1) makes the prediction maximization that SH will take place in 24 hours;
(2) ratio R in the lasting period of " labelling rising " and " labelling reduction "
UdMinimize.
Although this 2 first is clearly, second may need extra explanation.From the prospect of execution algorithm meter 3, when determining SMBG, meter all determines whether to set up or do not set up the labelling that is about to take place SH at every turn.When labelling was established, it may continue for some time (being accompanied by several SMBG readings subsequently) till making the decision that reduces labelling.Therefore, we will have alternative " sign raises " and " sign reduces " process, and it changes at the SMBG point.Ratio R with reference to upper point (2)
UdBe that a people indicates the average time when reducing divided by sign average time when raising.
Our the previous best result that provides in disclosure of the Invention is to have predicted SH incident in 44% 24 hours, and R
Ud=1:7, high-risk alarm in promptly 1 day and do not have alarm to alternate in 7 days.Because we supposed that alarm cycle was at least 24 hours at that time,, algorithm is no more than 1 time weekly so being optimised for the frequency of setting up sign.If this analysis is to use the data of the object with height ratio SH incident to realize that then this ratio is acceptable.
In the example No.1 of this research, we have to use identical data set to improve algorithm 3, because there are not other available data that comprise SMBG record and SH record simultaneously.We have also used the precision of similar criterion evaluation algorithm 3.Yet we have changed other all things basically.The tracking of data, parameter estimation, all threshold value is all no longer the same with decision rule.These variations are because a new thinking causes, and promptly body has certain " loss " to the deposit (reserve) of reverse adjusting before SH, and this loss can be followed the tracks of with the SMBG data.The definite execution of this thinking has description in " decision rule " part.Because decision rule comprises a successive standard and an artificial termination to a certain extent, so there be a plurality of separating, we have selected one of the best to be used for further research.Yet according to these results' statement, we can determine to select another to separate, thereby are carried out in using the future of algorithm 3.
The summary of results
At first, emphasis to notice that whole result given below surpasses the significance of statistics just.To see in several examples of next part as us, viewed difference is always highly significant (the p value is lower than any possible significant level).The viewpoint of algorithm 3 is generations of prediction SH incident on individual's basis.The result is:
(1) the Zui Xiao benchmark observation cycle be with every day 3-4 reading frequency in the time in about two weeks, obtain 50 SMBG readings.Afterwards each object is categorized in two dangerous groups, the latter uses different decision rules;
(2) we find from six months data that we have, and are enough to carry out the distribution of this group when observing beginning.Therefore, we suppose that approximately per six months meters will use 50 readings to reappraise its set of dispense;
(3) optimal delay of SMBG tracking is the frequency collection 100-150 reading with 3-4 reading every day.In other words, Zui Jia criterion is according to using whole 150 calculating that reading carries out in the meter memorizer.This memory capacity by simulation ONE TOUCH ULTRA realizes.Usually, use in the delay of having only 20 readings that in week, obtains and to obtain good result, but longer delay can produce better prediction;
(4) decision rule is according to new calculation procedure, and low other relevant parameters of BG exponential sum of tracing object are calculated in its use " interim meansigma methods ".We have designed special-purpose software in order to carry out this program and to handle the data that we can access.From the viewpoint of programming, carry out the required coding of this program and have only about 20 row, comprise the calculating of LBGI;
(5) we have investigated a plurality of decision rules (using various parameters).Ignore the SMBG frequency, SH predicts from 43.4% in 24 hours of these rule acquisitions, R
Ud=1:25 to 53.4%, R
Ud=1:7.Therefore, compare with our previous result, the prediction of SH has improved 10% in 24 hours;
(6) as the further optimum solution of studying, SH in 24 hours of the decision rule measurable 50% that we select, and R
Ud=1:10.Below the result with reference to the result of this optimum solution under different condition:
(7) optimum frequency of SMBG is 4 readings every day.If reached this frequency, then the prediction of SH can be increased to 57.2% in 24 hours, and has identical R
Ud=1:10.Other SMBG frequency also has investigation and report;
(8) if we extend to 36 or 48 hours with predetermined period, then the prediction of SH is increased to 57% and 63% respectively, and has same R
Ud=1:10;
(9) utilize reference information can significantly increase the prediction of SH.In fact, increasing by 10% than algorithm 3 previous versions is owing to used the benchmark tracking fully.But this benchmark is followed the tracks of and is modeled now as the self-correction of the meter that does not use any extra input of patient in two periods in week;
(10) individual/demographic information, for example historical or previous HbA of SH
1c, for not contribution of better SH short-term forecast;
(11) it is inappropriate whenever setting up sign when not having SMBG movable for a long time.Have only meter to send to be about to the number of times that the SH alarm takes place to be only access times.This is because the major part of SH prediction is the reproduction (trooping) according to very low BGs.To the estimation of this reproduction be provide in the summary prepared of in June, 2002 ADA meeting (Kovatchev etc. have recurrent hypoglycemia and severe hypoglycemia disease (SH) among the T1DM patient of a large amount of SH history) (seeing appendix).
The detailed description of date processing
Meter is stored together date of SMBG reading and each reading and correct time (hour, divide, second).Therefore, in the training data group, we have the SMBG record of each object certain hour order.During studying, 75,495 SMBG readings (average for each person every day 4.0 ± 1.5) from participant's storage meter, have been downloaded altogether.In the moon note by object, we obtained the date and time of SH incident takes place.Object has been reported 399 SH incidents (everyone is 4.7 ± 6.0 times).There are 68 participants (accounting for 80%) once to go through once or SH incident repeatedly.According to their Demographics, these objects do not have different with people's (object of residue 20%) that those do not live through SH.
The data pretreatment:
We have developed special-purpose software and have been used for the data pretreatment.This comprises: (1) with the storage measurement count of each object according to continuous 6-8 month combined sequence of BG reading together and (2) according to date and time the SH record of every object is complementary with this sequence.The latter's execution is as follows: for each SMBG reading calculates the time (hour/minute) of the last SH incident and the time of experiencing from last SH incident.Therefore possible: 24 hours or 48 hours equal time cycles before or after (1) each SH incident, the time cycle between (2) SMBG reading.Because the essence of SH (stupor, unconscious) does not have SMBG when SH, therefore for the purpose of algorithm 3, the SH incident does not comprise the biochemical significantly hypoglycemia that is used for algorithm 2.Average each SH incident SH with before minimum interval between the nearest SMBG reading be 5.2 ± 4.1 hours; There are 29 SH incidents (accounting for 7%) in 15 minutes before, to have the SMBG reading.To each SH incident, we have counted the number of times of carrying out the SMBG reading before this incident in 24h, 36h, 48h and the 72h.
The calculating of benchmark dangerous values and self-correction:
Be the low BG index of each calculation and object according to his/her SMBG reading.Determine that calculating the required minimum reading of benchmark LBGI is to gather 50 in about 2 weeks.Therefore for each new meter, we need expect the self-correction period in one period initial two week, and meter will scan its owner's totally danger of SH during this period.After the initial period, people are assigned in two dangerous groups: low poor risk (LBGI≤3.5, LM group) or camber danger (LBGI〉3.5, the MH group).Our test data shows that more accurate grouping is unnecessary.This grouping allows to use different decision rules in LM and MH group, thereby compares with its initial hit rate that provides during the present invention is open, and the hit rate of algorithm has improved about 10%.
The use test data do not need correction reference danger again.Therefore we can suppose, if people's not experience variation in treatment was then approximately once proofreaied and correct in per six months again.This result with algorithm 2 is consistent, and it is presented at after the initial observation cycle 6 months, and the long-term forecast of SH is very accurately.
Yet, if having experienced in its glycemic control rapidly, the people changes, may need to proofread and correct again more continually.Again gauged judgement may be automatically, and constantly increases according to the difference between viewed operating risk value (section as follows) and the benchmark LBGI.Yet obtainable data can not allow us to illustrate this problem, because the dangerous not significant change of the hypoglycemia of our observed object.
Calculate the SMBG parameter: after pre-treatment step, we have designed another software and have calculated the SMBG parameter that is used to predict imminent SH.This software comprises:
(1) be that each BG reading calculates low BG dangerous values (RLO), this by following coding realize (BG is that unit is measured with mg/dl here, if unit be mmol/l then coefficient can difference):
scale＝(ln(bg))＊＊1.08405-5.381
risk＝22.765＊scale＊scale
if(bg_1≤112.5)then
RLO＝risk
else
RLO＝0
endif
(2) have the SMBG reading of sequence number n for each, BG (n) calculates runtime value and another statistic of LBGI (n), SBGI (n), and it is the standard deviation of low BG dangerous values.These two parameters are calculated with certain delay (k) after each SMBG reading, for example comprise this reading, BG (n), and (k-1) individual reading of gathering before of BG (n).
(3) a new interim method program (newprovisional means procedure) is adopted in the calculating of LBGI (n) and SBGI (n), and it is encoded based on following recursion:
Initial value n-k (perhaps accurately in maximum (1, n-k), so that explain the meter readings of ordinal number) less than k:
LBGI(n-k)＝rlo(n-k)
rlo2(n-k)＝0
Numerical value for any subsequent iteration j between n-k and the n:
LBGI(j)＝((j-1)/j)＊LBGI(j-1)+(1/j)＊RLO(j)
rlo2(j)＝((j-1)/j)＊rlo2(j-1)+(1/j)＊(RLO(j)-LBGI(j))＊＊2
After this circulation was finished, we had obtained the value of LBGI (n), and calculated
SBGI(n)＝sqrt(rlo2(n))
Because for ONE TOUCH ULTRA meter, maximum n is 150, so in the k=150 scope, seek best delay k at k=10.Though the difference on the performance is not remarkable, best delay is defined as k=150 (seeing for example next part).
Decision rule: when each SMBG reading, whether programmed decision sets up the sign that warning is about to take place SH.If set up sign, then program determines whether to make its reduction.These judgements are according to three threshold parameter: α, β, γ, and its operation is as follows:
For object to low poor risk (LM group):
FLAG＝0
if(LBGI(n)≥α?and?SBGI(n)≥β)FLAG＝1
if(RLO(n)≥(LBGI(n)+γ＊SBGI(n)))FLAG＝1
For the object of the dangerous group of camber, have only the 2nd if narration to work.In other words, if the runtime value of LBGI (n) and its standard deviation S BGI (n) have surpassed certain threshold value, then set up sign (promptly equaling 1), and if the currency of the low dangerous RLO of BG (n) surpass LBGI (n) and add the value of γ standard deviation then indicate also and set up.
A suggestive explanation: the numerical value of LBGI (n) and SBGI (n) has reflected the slower variation of hypoglycemia danger---could significantly change these numerical value through the SMBG of a couple of days.Because LBGI (n) meansigma methods is high more, hypoglycemia is frequent more and extreme in the recent period, so we can reach a conclusion, LBGI (n) and SBGI (n) have reflected the loss (perhaps going short of supply) that reverse adjusting deposit continues in the period of a couple of days.In addition, SBGI (n) is a sign of system stability---bigger SBGI (n) shows that the BG fluctuation of object increases, so control system becomes unstable and is subjected to extremely unusual influence easily.Therefore, the viewpoint of first logical expression reflection is no matter when to defend depleted and control (outside or inner) when becoming instability SH to take place when oppositely regulating.Second logical expression has illustrated the rapid change of low BG danger, no matter when then can send sign when at present low BG dangerous values is moved meansigma methods greater than it suddenly.In fact for the object of the dangerous group of camber, have only second logical expression relative consistent with final " forever exhausting " and " permanent unstable " state of these objects.Because these objects are the low BG value of operation continuously, and their BG is unsettled, so any rapid hypoglycemia incident all may excite SH.Substantially, severe hypoglycemia disease sign or set up after the low unstable BG cycle is perhaps set up acute hypoglycemia incident after, acute hypoglycemia incident substantial deviation (in dangerous space) nearest operating risk meansigma methods (may morning through very high).Still can not be explained with this algorithm without any the SH incident of alarm signal before these.Among the table 5C below, we have provided the sample output of the effect of 3 pairs of a plurality of objects of interpretation algorithms:
Table 5C: the sample output of the effect of 3 pairs of a plurality of objects of interpretation algorithms:
Each provisional capital of this output has provided a SMBG reading, or a SH incident (not having reading).ID is the id number of object, and BG is the BG level in mg/dl, SH=1 when the SH incident takes place.If algorithm 3 decisions are set up flag then FLAG=1; TIME be in hour time of the last SH incident.
The optimization that interim method program postpones: in the publication formerly, we have reported in 48 to 24 hours periods before SH, and average BG level reduces and the BG variance increases.In 24 hours periods before SH, meansigma methods BG level further reduces, and the variance of BG continues to increase, and LBGI sharply increases.In 24 hours periods after SH, the average horizontal normalization of BG, and the BG variance still significantly increases.In 48 hours after SH, the meansigma methods of BG and variance turn back to datum-plane and (see Kovatchev et, detectable blood glucose disturbance (Episodes ofSevere Hypoglycemia in Type 1 Diabetes are Preceded before the severe hypoglycemia disease incident and in 48 hours afterwards in the al.1 type diabetes, and Followed, within 48 Hours by Measurable Disturbances in Blood Glucose).
J of Clinical Endocrinology and Metabolism, 85: 4287-4292,2000).We use the delay according to the interim method program that the meansigma methods optimization algorithm 3 of observed LBGI (n) and SBGI (n) is adopted in 24 hours before SH of these observed values, k now.In brief, be used to calculate LBGI (n) and SBGI (n)) the selected SH of making of delay before in 24 hours these measure and show and compare maximization with remaining research, but system still is in unbalanced period after not comprising SH.Find that optimal delay is k=150.Table 6A and 6B have provided LBGI (n) and SBGI (n) organizes the meansigma methods of two group of objects for a plurality of values of parameter k with for hanging down the danger of poor risk group and camber.Obviously, the difference between each value of k is little, and the numerical value of therefore actual use any k 〉=10 all is suitable.Yet according to present data, we recommend k=150, and all further calculate this delay of use.This recommendation still according to the variance of LBGI (n) and SBGI (n) year length of delay increase reduce, this can be reflected by following bigger t-value:
Table 6A: the different LBGI (n) that postpone before the SH down in 24 hours with all the other times
Table 6B: the different SBGI (n) that postpone before the SH down in 24 hours with all the other times
*Optimum solution
From table 6A and 6B as can be seen, in before SH 24 hours of LBGI and SBGI all highly significant ground increase.Therefore, people attempt to move a kind of direct discriminant or logarithmic model is predicted imminent SH.Unfortunately, the effect of this canonical statistics is not fine, although two kinds of models highly significant all on statistics.Discriminant model (it is good that its effect returns than logarithm) can correctly be predicted 52.6% imminent SH incident.Yet its sign raises (f.lag-up) and indicates the non-constant of ratio---the R that reduces (flag-down)
Ud=1:4.Therefore, this model tends to bigger quantity data point, and this is the tendency that all can predict in any statistical procedure.Therefore, we have to adopt the decision rule that provides above.
The precision of prediction of severe hypoglycemia disease
Threshold parameterα
,β
Withγ
OptimizationBelow we describe the predictive ability of the algorithm 3 of the various combinations of using threshold parameter α, β and γ in detail.Because these parameters and desired result (high predicted SH and ratio R
UdMinimum) relation between is very complicated, so the optimizer that we adopted can not obtain single separating.In addition, as if do not need to obtain single separating yet.It seemingly a commerce but not the judgement of mathematics promptly under the ratio of given " sign raises " and " sign reduces ", obtains acceptable SH prediction percentage rate.Therefore, we do not state that any separating given below is optimal solution.But in order further to explore this theme, we accept to predict SH and R in future of 50%
Ud=1:10 is as benchmark, be used to study except that 24 hours predetermined period and for obtain better dangerous distribute (profile) required every day the SMBG number of readings various requirement.
Table 7 has provided the performance of algorithm 3 under the various combinations of values of α, β and γ, and on behalf of the prediction percentage rate (hit rate) of SH, these combinations be referred to as the ratio R of " bothering index " (annoyance index) with us
UdBetween relation.Each object that is to say, with having illustrated ratio R being warned and not having the Average Total Time (is unit with the sky) that experiences under the alarm condition during table 7 also was included in research
UdThe algorithm of meaning is to the summary result of being warned of object of study experience-no warning period alternation procedure.
Table 7:SH prediction: hit rate, bother exponential sum average time
Best separating is used for further analysis.Suppose that the participant in this research on average experiences the SH incident 4.7 times, can prevent 50% SH, then 19 days the height phase acceptable seemingly of reporting to the police if report to the police so.In addition, the ground (in clusters) that trends towards trooping arrives the high warning phase.Therefore, we can suppose that in practice, can replace long and period relatively placidity with several days high-risk warning phases.Last column in the table 7 has provided R
Ud=1:7 separates, and what provide during itself and the present invention disclose separates quite.Yet current separating has approximately high 10% hit rate: 53.4% to 44% of our previous algorithm.When hit rate is suitable with algorithm that we are previous, bother than less than 1:20, that is to say, got well 3 times.
Figure 14 has provided hit rate and the ratio R of representing with percent
UdBetween level and smooth dependence.Obviously, when the hit rate of algorithm 3 increased, the ratio between " sign raises " and " sign reduces " increased sharply.Therefore, given these data, it is inappropriate that searching can produce the parameter combinations that is higher than 50% hit rate:
Selectable predetermined period: when beginning algorithm 3 is described, we have formulated basic assumption, if set up sign in before the SH incident 24 hours, think then that the SH incident is predicted to have arrived.This hypothesis has produced the hit rate of pro-part report.We are according to scope other predetermined period calculating hit rates from 12 hours to 72 hours now.In whole experiment, parameter alpha, β and γ are fixed as 5.0,7.5 and 1.5 respectively, just, are fixed on the numerical value of their optimum solutions in table 7.Therefore, sign rising ratio is separated with this and is kept identical, R
Ud=1:10, just variation has taken place in hit rate, because we have changed the definition of hitting.Figure 15 has provided the dependence between predetermined period and the corresponding hit rate.
Obviously, along with predetermined period is increased to about 24 hours, hit rate increases sharply, and the increase of hit rate slowly descends then.Therefore, we can reach a conclusion, and 24 hours in advance are best and reasonably forecast the cycle.
Every day the SMBG reading optimal numberAt last, we test, and research forecasts to need what readings every day in order to produce best SH.
The time say in beginning as us, reported 399 SH incidents altogether.In these incidents, 343 SMBG readings (reading before 3 incidents have in addition in 48 hours, other has 4 incidents to have 72 hours reading before) that had before 24 hours.Have in addition and surpass 50 SH incidents (14%) without any the SMBG reading in advance rationally that helps to predict.At least the hit rate of a part before 343 incidents that have a SMBG reading in formerly 24 hours are used for calculating.Its complementary event is ostracised naturally outside calculating.
Further analyze and show that along with the increase of the reading of gathering before the SH incident, hit rate increases fast.Yet,, we consider a SH incident if being strict with the obtained reading of some by force, we can find, satisfy the SH event number decline (table 8) rapidly of this requirement.This is because object is not observed the requirement of research, and may be the good reason that occurs certain alarm in meter in the future, if promptly do not gather the SMBG reading with proper speed, then algorithm 3 will be die on and meter can be closed.
Table 8 has provided the number of the SH incident with the previous SMBG reading of some and algorithm 3 hit rate to these incidents.Best row has comprised the optimal solution of table 7 in the table, and it is as the basis of all calculating of back.All hit rates all provide with 24 hours predetermined period, just, and the sign before the SH in 24 hours.We can reach a conclusion, and along with the increase of object compliance, the precision of algorithm 3 prediction SH also significantly increases.Carry out the SMBG reading every day 5 times, then precision improves 10% from 50% hit rate of benchmark:
Table 8: when providing the previous SMBG reading of some, the performance of algorithm 3
| The number of previous SMBG reading | Satisfy the SH incident (accounting for the % of SH sum) that the 1st row require | Hit rate |
| Has 1 in 24 hours at least | 343(86％) | 49.9％ |
| Have 3 in 24 hours at least | 260(65％) | 54.2％ |
| Have 4 in 24 hours at least | 180(45％) | 57.2％ |
| Have 5 in 24 hours at least | 103(26％) | 64.1％ |
| Have 4 in 36 hours at least | 268(67％) | 52.6％ |
| Have 5 in 36 hours at least | 205(51％) | 54.6％ |
| Have 6 in 36 hours at least | 146(37％) | 60.3％ |
| Have 7 in 36 hours at least | 107(27％) | 60.7％ |
| Have 6 in 48 hours at least | 227(57％) | 53.3％ |
| Have 7 in 48 hours at least | 187(47％) | 54.0％ |
| Have 8 in 48 hours at least | 143(36％) | 55.9％ |
| Have 9 in 48 hours at least | 107(27％) | 59.8％ |
The improvement that tested other are possible
By comprising external parameter, for example the previous year SH number or HbA
1cReference value, the trial that improves algorithm 3 predictive abilities is also unsuccessful.Obviously, the short-term forecast of SH depends primarily on current or recent incident.Yet the restrictive condition of this research is that all participants are at 〉=2 SH of experience the previous year.
At last, we have checked when the danger that detects hypoglycemia increases and had not had SMBG, the then warning that whether should send SH at that time for some time.Doing like this is the SH incident of wanting not have before predicting at least a portion the SMBG reading.But also unsuccessful, what produced mainly is wrong warning.This result has confirmed to observe the importance of SMBG agreement more, and this agreement will have enough frequent SMBG reading.
Appendix: summary
Example No.1 has estimated that hypoglycemia (frequency of hypoglycemia and SH (being defined as the stupor that can't carry out Heal Thyself or unconscious) takes place after the incident of BG＜3.9mmol/l) again.
85 last 1 year experience〉the T1DM patient (41 women) of 2 SH incidents, carry out 3-5 SMBG every day and reach 6-8 month, and write down the SH incident according to date and time.The mean age of object is 44 ± 10 years old, and the persistent period of diabetes is 26 ± 11 years, HbA
1cBe 7.7 ± 1.1%.
All SMBG readings (75,495) combine (n=399 according to date and time with the SH incident of object; The SH incident does not generally have corresponding SMBG reading).For each SMBG reading or SH incident are calculated (＜3.9mmol/l) time of experience afterwards from low BG of nearest last time.Below table 9 provided the percentage ratio of reading among 3 hypoglycemia scope: BG＜1.9mmol/l, 1.9-2.8mmol/l and the 2.8-3.9mmol/l, and 24 hours before, 24-48 hour, 48-72 hour and surpass and had low BG reading in 72 hours the (percentage ratio of the SH incident of BG＜3.9mmol/l).Last string has provided operation test, and it has refused following hypothesis, promptly comprises date random distribution on whole time range of low BG reading (perhaps SH incident).The negative Z value of check demonstrates the date " appearance of trooping " that has and do not have hypoglycemia reading or SH incident.
Table 9: the percentage ratio that before has hypoglycemia/SH of low BG:
We reach a conclusion, and have above hypoglycemia SMBG reading of half and about 2/3 SH incident at least one hypoglycemia reading was arranged in 24 hours before it.In addition, the hypoglycemia incident trends towards occurring with trooping.Therefore, initial hypoglycemia incident may be the alarm signal that soon occurs hypoglycemia once more.
II. example No.2
The present invention adopts daily self-monitoring of blood glucose (SMBG) data, and directly is adapted to pass through introducing and can predicts HbA
1cImprove the performance of domestic SMBG equipment with the intelligent data interpretation logic of remarkable hypoglycemia high-risk period.This method comprises two parts: (1) algorithm 1 is estimated HbA
1cAnd (2) algorithm 2﹠amp; 3 predict the remarkable hypoglycemia of long-term and short-term (in 24 hours) respectively.In this report, we have described proposition, optimization and checking HbA
1cThe step of algorithm for estimating 1 with and the HbA that obtains in the estimating experiment chamber
1cPrecision.
Target:
Our main purpose is to reach to have 95% measurement result to be positioned at laboratory reference value ± 1HbA
1cPrecision in the unit, this is HbA
1cNational glycosylated hemoglobin standardization program (NGSP) criterion (National Glycohemoglobin StandardizationProgram Criterion) of chemical examination precision.
Method:
Object: the SMBG data be by 100 type 1 diabetes objects and 100 type 2 diabetes mellitus objects (T1DM, T2DM) respectively through obtaining in 6 months and 4 months, and in T1DM in the 0th, 3 with carried out HbA in 6 months
1cDetect, and in T2DM in the 0th, 2 with detected in 4 months.
The proposition of algorithm 1 and optimization: the training data group comprises SMBG and the HbA that was collected by T1DM 3 months
1cData and collected 2 months SMBG and HbA by T2DM
1cData.These training datas are used to optimize algorithm 1 and are used to assess a large amount of being used for guarantee more high-precision sample choice criteria.The sample choice criteria is to the requirement of the SMBG sample of collecting by meter, accurately estimates HbA if can satisfy then can guarantee by sample
1cTherefore, meter can scan each SMBG sample, if satisfy the sample choice criteria, then calculates and shows HbA
1cEstimate.At a large amount of by analysis point of contacts (cut point) afterwards, we have selected following standard:
1. test frequency: in order to produce HbA
1cEstimation, meter needs in the past 60 days carry out 2.5 times or more times test average every day, just, carries out that 150 SMBG readings will be arranged altogether in the past two months.Notice that this is the average of every day, and failed call tests every day, this point is important.
2. data randomization: the sample that only had test after the meal or test at night not enough (＜3% sample) in 60 days in the sample will be excluded.Comprise the test of avoiding high concentration in one of every day the most frequently used time, to carry out in addition.In these master copy reports detailed description is arranged.
Result: the perspective checking and the precision of algorithm:
Then, with algorithm, comprise the sample choice criteria, be applied to test data set 1 and independent test data set 2, wherein test data set 1 comprises that T1DM and T2DM object are at last once HbA
1cSMBG and HbA before the test in 2 months
1cData, and independent test data set 2 is made of the data of 60 T1DM objects having participated in previous NIH research.For the purpose of checking, will be by the estimation of algorithm 1 acquisition and with reference to HbA
1cLevel compares.In test data set 1, algorithm has reached the NGSP standard, is positioned at laboratory reference value ± 1HbA
1cPrecision in the unit is 95.1%.In test data set 2, algorithm has reached the NGSP standard, is positioned at laboratory reference value ± 1HbA equally
1cPrecision in the unit is 95.1%.Studies show that of sample choice criteria has 72.5% object can produce so accurate estimation every day, has 94% object can produce 1 so accurate estimation in approximately per 5 days.
Conclusion:Daily SMBG data allow accurately to estimate HbA
1c, and satisfy direct HbA
1cThe NGSP standard of chemical examination precision.
Dui Xiang ﹠amp; Standard choice
We have obtained 100 object and 100 permissions of suffering from the object of type 2 diabetes mellitus (T2DM) of suffering from type 1 diabetes (T1DM).179 objects have been finished the major part of SMBG data collection, wherein suffer from T1DM for 90, suffer from T2DM for 89.The data of these 179 objects are used for check algorithm 2 and 3.Yet check algorithm 1 requires object not only to have the SMBG data, and the HbA that gathered in 60 days before SMBG will be arranged
1cData and SMBG record.3rd month (is dead 2 months for T2DM) of this research, 153 objects (78 suffer from T1DM) have been finished the HbA that satisfies above-mentioned standard
1cData and SMBG record.In addition, we use the N=60 name to suffer from the data detection algorithm 1 of the object of T1DM, and these objects have participated in our previous NIH research (NIH).Table 10 has provided the Demographics of all objects.
Table 10: the Demographics of object
| Variable | T1DM | T2DM | NIH |
| Age (year) | 41.5(11.6) | 50.9(8.1) | 44.3(10.0) |
| Sex: % women | 41％ | 43％ | 46％ |
| The persistent period of diabetes (year) | 20.1(10.1) | 11.7(8.2) | 26.4(10.7) |
| Body mass index | 25.4(4.7) | 34.2(8.1) | 24.3(3.4) |
| Benchmark HbA 1c | 7.5(1.1) | 8.5(2.1) | 7.6(1.0) |
| The 2nd HbA 1c | 7.3(1.2) | 7.9(1.6) | 7.4(0.8) |
| The 3rd HbA 1c | 7.0(0.9) | 7.5(1.1) | - |
| #SMBG reading/object/sky | 5.4(2.3) | 3.5(0.8) | 4.1(1.9) |
| At the HbA second time 1cThe natural law that has the SMBG reading in 2 months before | 56.9(5.4) | 57.3(4.3) | 37.5(14.3) |
Observed meter error
Our investigation shows, HbA
1cThe main cause of not finishing data before chemical examination or other chemical examinations in 60 days is not that object is not obedient to, but meter lost efficacy.If obviously the patient presses " M " button long time, then date (for example in November, 2017) is at random arrived in the time and date meeting " redirect " of ONE TOUCH ULTRA meter.Check the date of each meter once returning us, we find have 60 meters that this incident has taken place in whole research process.The bias effect on time/date 15,280 readings, perhaps nearly all 10% of readings.We preserve these readings separately and allow a student check them.In many cases, but be not in all cases, he can both recover the date sequence of reading.This error and lost the minority meter in the mailing process, the object number that causes having the good data that is used for parser 1 has reduced to 140 from 179.Recovered the data of 12 objects, finally collected the data of 153 objects, wherein suffered from T1DM for 78, suffered from T2DM for 75, they are at HbA
1cThe time series of data does not before have multilated, therefore is suitable for check algorithm 1.
Program
All objects have all been filled in the allowance form of IRB recommendation and have been participated in guidance meeting (orientation meeting), have at the meeting introduced ONE TOUCH ULTRA meter and have allowed them finish the screen questionnaire to them.Instruct after the meeting, all objects are visited the UVA laboratory immediately and have been obtained benchmark HbA by blood sampling
1cThe T1DM object has carried out 6 months, in the 3rd and the chamber HbA that experimentizes June
1cChemical examination; The T2DM object has carried out 4 months, in the 2nd and the chamber HbA that experimentizes April
1cChemical examination.Self-monitoring (SMBG) data are downloaded and are kept at the data base from meter regularly.Per two weeks are carried out parallel record by automatic e-mail/ phone tracking system to remarkable hypoglycemia and hyperglycemia incident.
Data storage and cleaning
The ONE TOUCH ULTRA raw data of T1DM and T2DM object is respectively stored among the InTouch data base.Object and meter error that the software of use customized development is cleared up these raw datas are carried out manual data cleaning (seeing top meter error) in some cases.In the time can not revising, then abandon these data.
In order to guarantee that our optimization result can be generalized to entire population's level, at first optimize algorithm with the training data group, test with test data set then.
The training data groupComprise that the T1DM object is at definite 3rd month HbA
1c60 days SMBG data of Cai Jiing before.This data set is used to optimize the formula of algorithm 1.The T2DM object is being determined their 2nd month HbA
1cThe data of Shou Jiing are used to identify the sample choice criteria before, and this standard is not obvious in the T1DM data.Yet the data of T2DM object are not used in the formula of optimizing algorithm 1.The file that contains these data is PASS01.DAT.
Test data set 1Comprise that the T1DM object is at definite 6th month HbA
1cBefore, the T2DM object is at definite 4th month HbA
1c60 days SMBG data of Cai Jiing before.Below we are called data set 1 with these data.The file that contains these data is PASS02.DAT.
Test data set 2The data that comprise N=60 name T1DM object in the previous NIH research.These data are collected with ONE TOUCH PROFILE meter.Below we are called data set 2 with these data.The file that contains these data is HAT0.XLS.
Variable among PASS01.DAT, PASS02.DAT and the HAT0.XLS is as follows:
ID, the moon, day, the time, year---the ID numeral and the reading duration of self-explanation.
PLASBG---by the BG (N/A among the HAT0.DAT is because used One Touch Profile) of One Touch Ultra record.
RISKLO, RISKHI---expression data transaction result's control variable (seeing below).
BG and BGMM---convert BG to whole blood BG, express (seeing below) with mmol/l then.
(each object) total data, HbA
1c, its estimated value and estimation difference be stored among Excel file PASS01.XLS and the PASS02.XLS.
Variable among PASS01.XLS, PASS02.XLS and the HAT1.XLS is as follows:
ID, (diabetes) type
HBA1---basis reference HbA
1cValue
HBA2---the 3rd month reference HbA
1c(T2DM is 2 months)---this need be predicted;
EST2 and ERR2---HbA
1cAnd the estimation of error;
Control variable (variable that all are used by algorithm 1):
BGMM1---with mmol/l is the average BG (part 2 of face as follows) of unit;
RLO1, RHI1---low and high BG index (part 2 of face as follows);
Hanging down the BG index L06-night---the reading according to midnight-6:59a.m is calculated (if (0.1e.HOUR.1e.6) just);
The number of NC1=SMBG reading in past 60 days;
NDAYS=has the natural law of SMBG reading in 60 days in the past.
N06-interval 0-6:59; The percent of SMBG reading among the 7-12:59;
ECLUDE=0, if 1---ECLUDE=1, then proposed algorithm is got rid of this sample.
File PASS01.DAT and PASS1.XLS can be mated by the ID numeral of object.Similarly, file PASS02.DAT and PASS2.XLS, HAT0.XLS and AHAT1.XLS also can be mated by the ID numeral of object.Raw data and whole second filial generation data file are sent to the LifeScan company limited.
The proposition of algorithm 1
The derivation of equation:
In the example No.1 of this project, provided algorithm 1 most explanation and proposition.Example No.1 does not comprise data collection.On the contrary, we use the data set of being collected by the Amylin pharmacists in clinical trial.Example No.1 has proposed three kinds of possible formula and has been used for by SMBG data estimation HbA
1c: (1) uses average SMBG, the exponential formula of low and high BG; (2) use average SMBG and before with reference to HbA
1cThe formula of reading; (3) only use the simple linear formula (seeing example No.1) of average SMBG.
Also propose another and be used to assess HbA
1cThe objective criterion of precision (in example No.1, we use least square estimation, % sum of errors absolute error to estimate the precision of each formula).This new requirement is transformed into a different optimisation criteria of algorithm 1, i.e. the optimization of formula no longer is to produce minimum error sum of squares (least square estimation), but makes estimated value be fixed on HbA
1cIn the homogeneity range of reference value ± 1.
In order to do like this, we only analyze our initial linear model (formula of example No.1) error with respect to this even coupling with the training data of T1DM object.We find, the high BG index positive correlation (r=0.3) of these errors and object, and we use this relation to revise our initial linear model.We find, preferably use high BG index as the grouping variable, and object samples is divided into the cumulative group of high BG index, and will revise in each group and introduce this linear model.Our idea is to use the low exponential correction of BG to be incorporated in each special group, rather than resemble advise among the example No.1 be incorporated into all samples.This variation is used based on the Different Optimization scheme of NBSP standard and is indicated.
Therefore, according to the training data of T1DM object, we have finished following algorithm 1:
The pretreatment of part 1-data
BG=PLASBG/1.12 (convert blood plasma BG to whole blood BG, it is general).
BGMM=BG/18 (converting BG to mmol/l).
Below each row calculate low and high BG index of each SMBG reading:
COM?SCALE＝(in(G))＊＊1.08405-5.381.
COM?RISK1＝22.765＊SCALE＊SCALE.
COM?RISKLO＝0.
IF(BG≤112.5)RISKLO＝RISK1.
COM?RISKHI＝0.
IF(BG>112.5)RISKHI＝RISK1.
Below each row added up to data of each object:
Average (BGMM) of each object of BGMM1=;
Average (RISKLO) of each object of RLO1=;
Average (RISKHI) of each object of RHI1=;
L06=only is average (RISKLO) that night, reading calculated, if there is not reading then default at night.
N06, N12, N24---be respectively interval 0-6:59; The percent of SMBG reading among 7-12:59 and the 18-23:59, for example if (0≤HOUR≤6); If (7≤HOUR≤12) and if (18≤HOUR≤24).
The sum of NC1=SMBG reading in past 60 days;
NDAYS=has the natural law of SMBG reading in 60 days in the past.
Part 2-estimation routine:
This estimation routine is based on the linear model of we example No.1:
HbA
1c＝0.41046＊BGMM+4.0775.
Analyze this formula error we find that error depends on high BG index.Therefore, we classify to whole objects according to the high BG index of object, in each group linear model are revised then, and are as follows:
A. each object is assigned a group according to his/her high BG index:
if(RHI1≤5.25?or?RHI1≥16) GRP＝0.
if(RHI1>5.25?and?RHI1<7.0)GRP＝1.
if(RHI1≥7.0?or?RHI1<8.5) GRP＝2.
if(RHI1≥8.5?or?RHI1<16) GRP＝3.
B. for each group, we carry out following estimation:
E0＝0.55555＊BGMM1+2.95.
E1＝0.50567＊BGMM1+0.074＊L06+2.69.
E2＝0.55555＊BGMM1-0.074＊L06+2.96
E3＝0.44000＊BGMM1+0.035＊L06+3.65.
EST2＝E0
if(GRP＝1) EST2＝E1.
if(GRP＝2) EST2＝E2.
if(GRP＝3) EST2＝E3.
C. only a few value not in the know (outlier) is revised:
if(missing(L06)) EST2＝E0.
if(RL01≤0.5and?RHI1≤2.0) EST2＝E0-0.25.
if(RL01≤2.5and?RHI1>26) EST2＝E0-1.5＊RLO1.
if((RLO1/RHI1)≤0.25?and?L06>1.3) EST2＝EST2-0.08.
Accuracy standard
For the precision of assessment algorithm 1, we have used a plurality of standard criterions:
1) the NGSP accuracy standard has at least 95% to be positioned at HbA in requiring all to estimate
1cReference value ± 1HbA
1cIn the unit.
2) HbA
1cThe mean absolute deviation of estimated value and measured value;
3) HbA
1cThe average percent deviation of estimated value and measured value.
Emphasis is noted: the NGSP accuracy standard is for checking direct measurement HbA
1cEquipment design.Here, we use this standard to by the SMBG data to HbA
1cEstimation.But the purpose of this estimation is not to replace HbA
1cLaboratory measurement, but help patient and doctor to carry out the daily management of diabetes.Relative with laboratory measurement, this estimate to adopt can obtain with any method and can be in the data of daily acquisition, and do not require special equipment or remove Doctor's office.
In order to confirm HbA
1cOther whether directly measure be consistent with traditional laboratory measurement, we have detected 21 IDDM patients' blood sample and simultaneously with DCA2000 and clinical assay analysis HbA
1cIn these 21 testing results, a big error being arranged, is 2.5 HbA
1cUnit.Table 11 has provided the precision result of the office equipment of this FDA recommendation:
The precision of table 11:DCA2000 in T1DM
| DCA2000 | |
| The NGSP standard---be positioned at ± 1HbA 1cPercent in the unit | 95.2％ |
| Mean absolute error (HbA 1cUnit) | 0.45 |
| The average percent error | 5.7％ |
The sample choice criteria
The derivation of equation:
HbA
1cThe continuous 60 days SMBG of estimated service life.We with these continuous 60 days SMBG data as sample.During his/her SMBG, everyone can produce a large amount of samples.In fact, each new measurement all can produce one with first previous different slightly new samples.Therefore, nature can suppose that meter has some control point, to guarantee to be used to estimate HbA
1cThe quality of SMBG sample data.
Therefore, after the vague generalization algorithmic formula is optimised, just be applied to whole training data group (T1DM and T2DM object data), can produce inaccurate HbA with investigation SMBG sample
1cThe condition of estimating.
This investigation concentrates on the following pattern that causes inaccurate estimation that occurs in SMBG:
1) SMBG is not frequent---and need the middle of the month reading of some so that estimate HbA at two
1cIf this number does not reach, then estimating may be inaccurate;
2) mainly testing after the meal or taking oral drugs and when mainly paying close attention to high BG, the pattern that SMBG is partial to hyperglycemia can occur when object;
3) biased pattern of the time of SMBG is promptly mainly tested in the set time of every day, makes the BG fluctuation of object not have good distribution in a few days.
Investigated after these patterns, we have selected best sample choice criteria according to the point of contact of the most accurate and minimum eliminating.Please refer to appendix A about the detailed description of this programmed logic and the narration of coding purpose.
The final sample choice criteria:
2a) oral medication/detection after the meal: (RLO1/RHI1 〉=0.005).As if in some SMBG samples, hyperglycemia is partial in the distribution of SMBG very much.As if this mainly takes place in the T2DM object, and they only measure BG at night.We suppose that these samples do not comprise the test in the hypoglycemia scope.Our investigation shows that these samples have about 1/3 meeting to HbA
1cProduce too high estimation (2/3 still can produce estimation accurately).We recommend in view of the above, if run into biased sample, and meter display result not then, this calculating is formulated as LBGI and is at least 1/2% of HBGI.
2b) test at night: (NO6 〉=3%).This standard guarantees to have at least a part of night hypoglycemia can access explanation.3% of the whole readings of this standard-required carry out at night (midnight-7:00am).
In other wordsIf, there are at least 5 at night in 150 readings gathering in 2 months, then this sample is an acceptable.Notice that the patient been proposed in usually and tests night, so this standard can promote good management.
2c) prevent highly unusual test pattern: if the reading above 3/4 carries out in any 6 hours intervals of every day, and then this sample can not produce estimated value.For example, be just after breakfast, to carry out if 80% test is arranged in the sample, then do not estimate.This standard is that the LifeScan company limited is desired, attempts " puzzled algorithm " (beat thealgorithm) to prevent people, allows us to guarantee effectively whereby, particularly to the clinician.
Order adopts the precision of choice criteria in the training data group
Following table has illustrated the influence of selected sample choice criteria to precision in the training data group and eliminating number.Note to propose as the precision of algorithm 1 final version of this research (final algorithm) part and the precision (seeing first linear model) that in example No.1, proposes and be included in the simplest linear function among the example No.1.
We provided each model without any the sample choice criteria time and order used sample choice criteria 1---detect frequency, # reading NR 〉=150 and standard 2---data randomization, the time precision, as mentioned above.
As seeing in all tables, the precision of algorithm 1 adopts the sample choice criteria along with order and improves, and has reached 95% of NGSP requirement after having used all standards.The back is one result give prominence in table.
Table 12A: the final sample choice criteria in the training data group---all objects:
Table 12B: final sample choice criteria---the T1DM in the training data group: optimized the coefficient of algorithm 1 in this sample, it has explained the high accuracy when sample is selected.
Table 12C: final sample choice criteria---the T2DM in the training data group: sample choice criteria 2 (data randomization) mainly proposes with this sample, and it has explained 5% precision that increases when this standard of application.
Sample in the training data is got rid of frequency:
The meter HbA that all has an opportunity to estimate when each new reading
1cIf sample does not satisfy choice criteria, then meter can not show HbA
1cEstimation, and meeting:
(c) wait for up to collecting suitable sample, perhaps
(d) if do not collect suitable sample, for example someone has permanent biased metering system, and then meter can send the prompting of revising the SMBG pattern.
Our investigation shows, most object (〉 95%) in 60 days, can obtain at least 10 HbA
1cEstimation (as long as their measuring frequency is enough), and have only object meeting acquisition estimation of 2% owing to biased metering system.Need this their metering system of object correction of 2% of prompting.The final result of this investigation provides hereinafter:
We have calculated (in 60 days) how many days meter can not satisfy choice criteria and can not show HbA to the user owing to sample
1cThe result:
5) for 72.5% of whole objects, meter can be reported HbA every day
1c
6) other 7.5% for whole objects, meter can be reported (in 60 days) HbA of 45-59 days
1c
7) other 10% for whole objects, meter can be reported (in 60 days) HbA of 12-44 days
1c
8) for 9 objects (5.9%), meter can not be reported HbA
1cUnless they change the SMBG mode.
Emphasis is noted: the great majority in these objects can not obtain to estimate that because they do not satisfy test frequency standard 1, promptly their sample is always less than 150 readings.Therefore, all have at least 94% can obtain at least one HbA in approximately per 5 days in the object
1c, and need not to change their metering system (this comprises T1DM and T2DM).
If we required to have 150 readings at least in 60 days, then have only 3 objects can not obtain HbA
1cEstimate:
1) 95.6% in 60 days, obtains at least 10 HbA
1cEstimate;
2) 2.2% can not obtain any estimation.
Therefore, about 98% measure 2.5 times object average every day and can obtain HbA arranged in 60 days
1cEstimate to have〉95% will obtain 1 estimation weekly at least.We reach a conclusion, sample choice criteria 2---the data randomization in one period to showing HbA
1cEstimation effect is very small.Have only about 2% object need be prompted to improve their SMBG mode.
Should be noted that the sample choice criteria can be used in any estimation HbA of raising
1cThe precision of formula.This choice criteria is independent of any special algorithm/formula, and uses before estimating beginning.For example, when being used, the sample choice criteria will improve the precision of algorithm 1 of this research of the conduct part of up-to-date proposition, and the precision of our the initial linear model that proposes in example No.1.
In addition, check that the effect of some other sample choice criteria shows that we can further improve precision, this is supposed to.For example, when one of them original test frequency standard application during in these data, can be confirmed to have more effectiveness.This standard has further instruction in appendix E.
The perspective checking of algorithm 1:
Precision in test data set 1:
With algorithm, comprise final sample choice criteria subsequently, be applied to test data set 1 (T1DM 1 and the last HbA of T2DM object
1c2 months before SMBG) to produce HbA
1cEstimate.Then with these estimated values and HbA
1cReference value compares, thus perspective verification algorithm 1.Table 13 has provided the summary of this checking.About each sample choice criteria being described in more detail of influence of algorithm asked for an interview appendix C.
Table 13: the precision of the perspective application of algorithm:
Precision in the test data set 2:
Another independently NIH data set (the N=60 name suffers from the object of T1DM) confirmed that the result has similar precision, promptly have 95.5% to be in laboratory reference value ± 1HbA
1cWithin the unit (table 14):
Table 14: the precision of algorithm 1 in independent NIH data set:
| Whole objects | |
| NGSP standard---± 1HbA 1cPercentage rate in the unit | 95.5％ |
| Mean absolute error (HbA 1cUnit) | 0.42 |
| The average percent error | 5.9％ |
The comparison of the office equipment precision that the precision of algorithm 1 and FDA recommend:
As what show in the following table 15, the HbA that uses in the precision of algorithm 1 and the Doctor's office
1cThe chemical examination precision is suitable.As explanation in the accuracy standard part, the DCA2000 data are used to confirm HbA
1cOther direct measurement results whether consistent with the laboratory measurement result.We have analyzed 21 T1DM patients' HbA simultaneously with DCA2000 and clinical assay
1cBlood sample.In these 21 times tests, bigger error is arranged, be 2.5HbA
1cUnit:
Precision and the algorithm 1 of table 15:DCA2000 in T1DM compares:
| DCA2000 | Test data set 1 | Test data set 2 | |
| NGSP standard---± 1HbA 1cPercentage rate in the unit | 95.2％ | 95.1％ | 95.5％ |
| Mean absolute error (HbA 1cUnit) | 0.45 | 0.45 | 0.42 |
| The average percent error | 5.7％ | 6.2％ | 5.9％ |
The frequency that sample is got rid of in the test data set:
Discuss in proposing the part of algorithm 1 as us, in each new reading, the meter HbA that all has an opportunity to estimate
1cIf sample does not satisfy choice criteria, meter just can not show HbA so
1c
The frequency that our use test data set 1 and 2 perspective sample estimateses are got rid of.For this reason, our calculating is can have according to meter how many days (in 60 days) show HbA to the people
1c, just people can have how many days have the sample that satisfies the sample choice criteria.Table 16A and 16B have provided the summary of these results in the test data set 1 and 2.We have comprised the data of all objects, and these data are divided into the data of the object of the object (90 SMBG readings were arranged) of average measurement 1.5 times/day and 2.5 times/day in 60 days:
Table 16A: the frequency that sample is got rid of in the test data set 1
| Whole objects (N=148) | The object (N=146) of average measurement 〉=1.5 time/day | The object (N=130) of average measurement 〉=2.5 time/day | |
| Meter can be reported HbA every day 1cThe percentage rate of object | 69.6％ | 72.6％ | 77.7％ |
| Meter can be reported one time HbA in per 3 days 1cThe percentage rate of object | 87.8％ | 91.1％ | 93.1％ |
| Meter can be reported HbA weekly one time 1cThe percentage rate of object | 91.9％ | 95.5％ | 96.9％ |
Table 16B: the frequency that sample is got rid of in the test data set 2
| Whole objects (N=60) | The object (N=55) of average measurement 〉=1.5 time/day | The object (N=30) of average measurement 〉=2.5 time/day | |
| Meter can be reported HbA every day 1cThe percentage rate of object | 51.7％ | 83.6％ | 80.0％ |
| Meter can be reported one time HbA in per 3 days 1cThe percentage rate of object | 95.0％ | 100.0％ | 100.0％ |
| Meter can be reported HbA weekly one time 1cThe percentage rate of object | 96.7％ | 100.0％ | 100.0％ |
Conclusion:
Table 13-16 confirms that meter can produce a HbA who satisfies the 95%NGSP accuracy standard in an average week
1cAccurately estimate, for the people who measures 2.5 average every day then be 96%.
The software logic of appendix-sample choice criteria
Sending the sample choice criteria to object---suggestion is got rid of some SMBG sample, perhaps message by algorithm.This sample choice criteria is programmed as follows:
EXCLUDE＝0
if(NC1>＝150) EXCLUDE＝1
2a)
Oral medication/test after the meal: as if in some SMBG samples, hyperglycemia is partial in the distribution of SMBG very much.As if this mainly takes place in the T2MD object, and they only measure when high BG.We suppose that these samples are not included in the result that the hypoglycemia scope is measured.Our investigation shows that nearly 1/3 understood highland estimation HbA in these objects
1c(2/3 object still produces accurate estimation).In view of the above, biased if our suggestion sample has occurred, meter display result not then, this calculating is formulated as LBGI and is at least 1/2% of HBGI.
if(RLO1-RHI1<0.005)EXCLUDE＝1。
2b)
The test at night: (NO6 〉=3%).This standard guarantees can explain at least the hypoglycemia at night of a part.3% of the whole readings of this standard-required carry out at night (midnight-7:00am).In other words, if having 5 at least at night in 150 readings that obtain two middle of the month, then this sample can be accepted.Note, advise that usually the patient tested at night, so this standard can promote good management.
if(NO6≤3.0) EXCLUDE＝1。
2c)
Guarantee to prevent highly unusual test pattern: if having surpass in 3/4 the reading one day any 6 hours and carry out at interval, then sample can not produce estimated value.For example, if there is 80% test after breakfast, to carry out just in the sample, then can not produce estimated value.To be that LifeScan company requires attempt " puzzled algorithm " so that guarantee people to this standard, allows us to guarantee effectiveness whereby, particularly for the doctor.In our data, there is not sample unusual so to heavens so that trigger this standard (details are seen appendix B---standard 2c).Dependence software is carried out, the frequency that need be calculated as follows from the SMBG data:
The SMBG reading % (breakfast) at M12-6:00am-noon
The SMBG reading % (lunch) of M18-noon-6:00pm
The SMBG reading % (dinner) of M24-6:00pm-12:00
The SMBG reading % (night) of M06-12:00-6:00am
The SMBG reading % of M15-9:00am-3:00pm
The SMBG reading % of M21-3:00pm-9:00pm
The SMBG reading % of M03-9:00pm-3:00am
The SMBG reading % of M09-3:00am-9:00am
Then for above-mentioned any combination (i, j):
if(Mij>75.0)EXCLUDE＝1。
Appendix B-sample choice criteria 2C
This standard is that LifeScan company requires, to guarantee to prevent highly unusual test pattern.The purpose of this standard is to prevent people's " puzzled algorithm ".
Basically, this standard code: if 3/4 (the perhaps number of other expectations) of your reading carried out at interval or in the interval of other expectations at one day any 6 hours, then you can not obtain estimated value.
Therefore, for example, after supper, carry out, then can not obtain estimated value if having above 3/4 test.This will support our general statement more: do not comprise not at random people of test in this calculating.I think, specific calculations and coding to this may look complicated, but crucial is that we can only import " you must test at random " in one day or other are similarly talked about as the panel statement, just can cover our whole exclusion standard (eliminating test frequency).If we need, then we can with concise statement (fine print) import simply precise definition more " in one day any 6 hours readings at interval all cannot surpass whole readings 75% ".This guarantees, follows our standard, can improve the clinical acceptance of algorithm.
Detailed content more:
46 hours intervals are defined as follows:
6:00am-noon (breakfast)
Noon-6:00pm (lunch)
6:00pm-12:00 (supper)
12:00-6:00am (night)
This standard can move twice with different intervals, thereby prevents that people from concentrating on 6 hours and testing but still satisfy improperly this first standard near the point of interface at interval.For example, if people have 40% at 11:50pm, having 40% at 12:10pm, then still is the test of concentrating, though satisfy first pass (pass) at interval, does not satisfy second of interval and closes.
Second group is at interval:
9:00am-3:00pm
3:00pm-9:00pm
9:00pm-3:00am
3:00am-9:00pm
Notice that selectively from the viewpoint of coding, people can move as follows and obtain identical result:
The reading of any 18 hours periods must not be less than 25% of total indicator reading.
You must be in 18 overlapped 3 hours hours periods bootup window:
9:00am-3:00am
12:00 noon-6:00am
3:00pm-9:00am
6:00pm-12:00 noon
9:00pm-3:00pm
12:00 midnight-6:00pm
3:00am-9:00pm
6:00am-12:00 midnight
The sample choice criteria is to the increase effect of arithmetic accuracy in appendix C-test data set 1
As proposing explanation in the part at algorithm, the precision (seeing initial linear model) that following form has related to the algorithm 1 (finally algorithm) that is suggested as this a research part and proposed in example No.1 and be included in the simplest linear function among the example No.1.This form has provided algorithm 1 no specimen in test data set 1 and has got rid of the precision of also using two sample choice criteria successively:
Standard 1---test frequency, # reading NR 〉=150 and
Standard 2---data randomization, as in the sample choice criteria, illustrating:
Table 17A: the precision of algorithm 1---whole objects:
Table 17B: the precision of algorithm 1 in T1DM:
Table 17C: the precision of algorithm 1 in T2DM:
Appendix D-can select the test frequency standard
High-grade test frequency standard helps improving more significantly the precision of algorithm 1.This is because adopt test frequency standard 1 and not only be based on data analysis, but also based on other consideration.If it is too strict to find to require to have in 2 months the standard 1 of 150 readings, then can adopt interchangeable solution.Be exactly primary test frequency standard, its requirement has 35 days, and the SMBG reading frequency of (in 60 days) is 1.8 reading/skies, and the total indicator reading that has 35 days in just 60 days is 63.Table 18 adds that with this primary loose test frequency standard standard 2 (data randomization) confirms this point, and the precision of algorithm 1 surpasses 95%:
Table 18: use the test frequency standard (35 days reading is 1.8 reading/skies) of selecting Precision with the algorithm 1 of data randomization standard:
Emphasis is noted: in addition, but should choice criteria can screen sample, for example, recover again afterwards in 4 weeks, and then can not show HbA if SMBG is interrupted to reach with a large amount of missing datas
1cEstimate.An obvious example of this pattern appears in the test data set 2---his/her HbA
1cEstimated value has the object of maximum error and only collected 159 readings in 60 days 30 day.Therefore, by collected reading fast in several days, object still can satisfy the requirement of collecting 150 readings, but can cause inaccurate HbA
1cEstimate.
The typical definition of example No.2 (but without limits) to this paper
6) severe hypoglycemia disease (SH) be defined as cause can't Heal Thyself stupor, seizure of disease or unconscious hypoglycemia (BG);
7) moderate hypoglycemia (MH) is decided to be the activity of upsetting object but does not hinder the serious neuroglycopenia of Heal Thyself;
8) biochemical severe hypoglycemia disease (BSH) is defined as blood plasma BG reading＜=39mg/dl;
9) to be defined as blood plasma BG reading be 39-55mg/dl to biochemical moderate hypoglycemia (BMH);
10) whole diseases all are called above
Remarkable hypoglycemia
Attached purpose
The data of this example are used for the following algorithm of perspective checking:
Object
We have obtained 100 object and 100 permissions of suffering from the object of 2 shape diabetes (T2DM) of suffering from type 1 diabetes (T1DM).170 objects wherein suffer from T1DM for 90, suffer from T2DM for 89, have finished the major part of SMBG data collection.
Program
All object has all distributed the permission form of an IRB recommendation and has participated in the guidance meeting, has at the meeting introduced ONE TOUCH ULTRA meter and has finished the screen questionnaire survey to them.Instruct after the meeting, all objects have been visited the UVA laboratory immediately and have been taken a blood sample and measured HbA
1cReference value.The T1DM object in subsequently 6 months in the 3rd and the chamber HbA that experimentizes June
1cChemical examination; The T2DM object in subsequently 4 months in the 2nd and the chamber HbA that experimentizes April
1cChemical examination.Self-monitoring (SMBG) data are downloaded and are kept at the data base from meter regularly.Automatic e-mail/ phone tracking system by custom design writes down remarkable hypoglycemia and hyperglycemia incident abreast, and its per 2 weeks get in touch with the full entry person.
Table 19 has provided the summary of SMBG and severe hypoglycemia disease/moderate hypoglycemia [SH/MH] data collection.
Table 19: data collection summary
| Variable | T1MD (N=90 object) | T2DM (N=89 object) |
| The #SH incident | 88 | 24 |
| The #MH incident | 1,660 | 190 |
| The #SMBG reading | 92,737 | 35,306 |
| The #BSH reading | 1,039 | 39 |
| The #BMH reading | 5,179 | 283 |
Because algorithm 1 and 2 keeps not becoming, so we can regard the data collection of whole example No.2 as the perspective test of these algorithms.
The formula of algorithm 2
The execution of algorithm 2 is as follows:
1), as follows he is categorized in 15 risk ranges (RCAT) one according to the low-glycemic of each object according to one month SMBG data:
if(LBGI≤0.25)，RCAT＝0
if(0.25<LBGI≤0.5)，RCAT＝1
if(0.50<LBGI≤0.75)，RCAT＝2
if(0.75<LBGI≤1.00)，RCAT＝3
if(1.00<LBGI≤1.25)，RCAT＝4
if(1.25<LBGI≤1.50)，RCAT＝5
if(1.50<LBGI≤1.75)，RCAT＝6
if(1.75<LBGI≤2.00)，RCAT＝7
if(2.00<LBGI≤2.50)，RCAT＝8
if(2.50<LBGI≤3.00)，RCAT＝9
if(3.00<LBGI≤3.50)，RCAT＝10
if(3.50<LBGI≤4.25)，RCAT＝11
if(4.25<LBGI≤5.00)，RCAT＝12
if(5.00<LBGI≤6.50)，RCAT＝13
if(LBGI>6.50)，RCAT＝14
2) by two-parameter Weibull probability distribution and utilize the following distribution function that provides to calculate in the future the significantly theoretical probability of hypoglycemia:
F (x)=1-exp (a, x
b) for any x 0; Otherwise 0.The parameter of this distribution depends on the prediction persistent period of expectation, and in the report of example No.1 explanation is arranged.If carry out with meter, this step will provide the continuous of remarkable hypoglycemia danger to estimate, for example " next month 50% ".
3) each object is categorized into hypoglycemia in the future significantly minimum, low, in or in the high-risk group: these scopes are defined as follows: minimal risk (LBGI≤1.25); Low dangerous (1.25＜LBGI≤2.5); Poor risk (2.5＜LBGI≤5) and high-risk (LBGI〉5).If carry out with meter, then this step will provide the discrete type of remarkable hypoglycemia danger to estimate for example " being high-risk next month ".
The formula of algorithm 3
At first, for fear of calculate the benchmark dangerous values that provides in algorithm 3 example No.1 report description, we have revised the delegation in the coding.Now, algorithm 3 transfers to use the result of algorithm 2.We introduce the sample results that this variation is used to introduce two objects on the 28th October in 2002.At this moment, obviously the action with simple Excel spreadsheet illustration algorithm 3 is easily, and if avoided the calculating of reference value then be possible.This step does not change the precision of algorithm 3, so keep as the permanent change that makes things convenient for algorithm 3 programmings.Introducing no longer for after on October 28th, 2002 algorithm 3 changes.Here, we provided with example No.1 report in identical algorithm 3 formula, reformed row labelling comes out.
1) by following be encoded to each BG reading calculate low BG dangerous values (RLO) (BG measures with mg/dl here, if unit be mmol/l coefficient could be different):
scale＝(ln(bg))＊＊1.08405-5.381
risk＝22.765＊scale＊scale
if(bg_1≤112.5)then
RLO＝risk
else
RLO＝0
endif
2) for each SMBG reading, we have calculated runtime value LBGI (n) and another statistic SBGI (n), and it is the standard deviation of low BG dangerous values.These two CALCULATION OF PARAMETERS are calculated from each SMBG reading backstepping with specific markers (n), just comprise this reading and this reading (n-1) individual reading before.
3) interim average program (provisionalmeans procedure) is used in the calculating of LBGI (n) and SBGI (n), and it makes according to following recurrence coding:
Initial value n (perhaps accurately in maximum (1, n-k), so that explain the meter readings of ordinal number) less than k:
LBGI(n)＝rlo(n)
Rlo2(n)＝0
Falling to count any subsequent iteration j value between n and 1:
LBGI(j)＝((j-1)/j)＊BLGI(j-1)+(1/j)＊RLO(j)
rlo2(j)＝((j-1)/j)＊rlo2(j-1)+(1/j)＊(RLO(j)-LBGI(j))＊＊2
Finish after this circulation, we have obtained the numerical value of LBGI (n), then calculate
SBGI(n)＝sqrt(rlo2(n))
By this calculating, we have preserved two groups of data: n=150 and n=50 (for example preceding 150 and preceding 50 observed values).
4)
Decision rule: at each SMBG reading, program all judges whether set up the caution sign that is about to take place SH.If sign is set up, then whether programmed decision reduces this sign.Three threshold parameter α are depended in these judgements, and beta, gamma moves as follows:
Object (LM group) for danger in low:
FLAG＝0.
if(LBGI(150)≥2.5?and?LBGI(50)≥(1.5＊LGI(150)and?SBGI(50)≥SBGI(150)) FLAG＝1.
if(RLO≥(LBGI(150)+1.5＊SBGI(150)) FLAG＝1.
In other words, at each SMBG reading, indicate if satisfy in two conditions then set up:
1) be in moderate high-risk SH according to algorithm 2 by preceding 150 object of classification of carrying out of test, and the SD of LBGI and LBGI increases in 50 tests of pro-;
2) or, determine low BG index by second inequality and uprush.
The enlightenment of these narrations is explained and is provided in the report of example No.1.As described above, the statement of first " if " has changed its initial form, so that avoid using benchmark LBGI, thus the output of use algorithm 2.
Illustrate in the report as example No.1 that in case set up sign, then it will keep 24 hours.For the precision of assessment algorithm 3, we use previously presented technology---calculate two measured values:
1) be about in 24 hours to take place the SH/MH incident prediction % and
2) " sign raises " and " sign reduces " continues the ratio R in cycle
Ud(bothering index).
The prediction % of SH incident need be than higher, and ratio R
UdNeed lower.This is that it increases the number of potential " false alarm " conversely because by increasing the percentage rate of prediction SH incident, we can increase the number of " setting up sign " inevitably.Because " false alarm " be not by clearly definition (seeing the report of example No.1), so we use R
UdIndex as algorithm 3 effectiveness.
Our the previous optimal results that provides in the report of example No.1 is, dope 50% 24 hours in the SH/MH incident, R
Ud=1:10 promptly after high-risk was reported to the police one day, did not and then report to the police in 10 days.Here we will keep identical sign rising/sign to reduce ratio, and be respectively the % prediction of SH and MH incident in T1DM and the T2DM calculation and object 24 hours.For this prediction, we do not use BSH and BMH incident, because therefore this be the part of anticipation function by the meter record.
Estimate the significantly danger of hypoglycemia in 1-3 month: the precision of algorithm 2
We have estimated the predictive ability of algorithm 2 as follows:
3) at first, we by one month SMBG data computation LBGI, and as described above each object is categorized into remarkable hypoglycemia minimum, low, in and the high-risk group.
4) then, in 1-3 the middle of the month subsequently, we count the number of SH, BSH, MH and the BMH incident of the perspective record of each object.
Below Figure 16-19 be respectively the number that T1DM and T2DM have provided each object observed SH, BSH, MH and BMH incident in 1 month or 3 months future one month SMBG after.Also comprised statistical simultaneously.
In addition, directly linear regression use LBGI, in the screen questionnaire according to the SH history and the benchmark HbA of SH event number report in past 1 year
1c, predicted (R significantly
2=0.62, f=48, p＜0.0001) after be about to take place the sum (SH+MH+BSH+BMH) of remarkable hypoglycemia incident in 3 months.Predictor variable according to the order of its importance is: 1) LBGI (t=8.2, p＜0.0001), can explain (the R for example of the variation of hypoglycemia 55% in the future significantly separately
2=0.55); 2) SH history (t=3.6, p＜0.0005), soluble other 5% variation, and HbA
1c(t=2.2, p＜0.03), soluble other 2% variation.This has confirmed previous result, and promptly LBGI is the most important prediction index of hypoglycemia in future, and HbA
1cContribution to this prediction is medium.
By the Weibull Model Calculation future remarkable hypoglycemia theoretical probability very consistent with following observed remarkable hypoglycemia incident---for seriously and the moderate incident, determine that coefficient is all above 90%.
Imminent (in 24 hours) are the prediction of hypoglycemia significantly: the precision of algorithm 3
Following table has provided the precision of T1DM and T2DM object SH and MH incident short-term forecast (in 24 hours) respectively.Be used for prediction if can obtain the SMBG reading of some in 24 hours periods, then table 20 and each row of 21 have provided percentage rate of the incident that predicts.For example, first row of each table has provided the percentage rate of the predicted incident that arrives, and no matter whether the SMBG reading was arranged in 24 hours before certain incident.As seen, accuracy of predicting increases along with the increase of number of readings per taken before the incident.Therefore, if someone measures 3 times or more times every day, then meter can be reported to the police and might be helped to avoid to surpass half remarkable hypoglycemia incident.
Emphasis is noted: for the purpose of assessment algorithm 3 precision, we have only used SH that is independent of SMBG and MH incident by the report of e-mail/ telephone system, the date and time of per two week report SH of this system requirements participant and MH.Show as our inquiry agency, sometimes the participant in their report, uses before certain incident on the time and date of a SMBG reading, rather than the precise time of this incident and date, help to help them to recall because inquire about meter.As a result, a large amount of incidents is arranged, moment of the last SMBG reading and this incident interval between the moment approaches zero before it.In order to explain this suspicious time sheet, the 3rd row of each table have provided the only strict precision that is limited to the algorithm 3 of following incident, and promptly leading time of fire alarming was at least 15 minutes.The leading time of fire alarming of assumed average is 11 hours, and we reach a conclusion, in most of the cases, thus the enough Heal Thyselfs fully that early helps of the appearance of warning.
In table 20 and 21, bother index and be set to R
Ud〉=10 so that and the report of example No.1 be complementary.
Table 20: the precision of algorithm 3 in T1DM
Table 20: the precision of algorithm 3 in T2DM
The present invention can be realized with other specific form, and not deviated from its spirit or basic feature.Therefore from any aspect, should think that all aforesaid embodiment is an illustration, but not the invention of this paper explanation is had restriction.Therefore scope of the present invention is specified by appended claims, rather than by aforesaid explanation, therefore all should be included within the scope of this paper with the meaning of claim equivalence and any variation in the scope.
Claims (21)
1. BG data estimation patient HbA that basis is collected in first predetermined lasting time
1cSystem, described system comprises:
Database component is used to preserve the data base who discerns described BG data; With
Processor, it is programmed and is used for:
Use serial mathematical formulae to prepare to be used to estimate HbA
1cData:
The data pretreatment,
Sample by sample choice criteria checking BG data and
If sample effectively then estimate HbA
1c
2. according to the system of claim 1, wherein said first predetermined lasting time is about 60 days.
3. according to the system of claim 1, the scope of wherein said first predetermined lasting time is 45 days to 75 days.
4. according to the system of claim 1, the scope of wherein said first predetermined lasting time is 45 days to 90 days.
5. according to the system of claim 1, wherein each patient's data pretreatment is comprised:
Blood plasma BG is transformed into whole blood BG mg/dl;
To be transformed into mmol/l unit with the BG that mg/dl measures; With
Calculate low-glycemic and hyperglycemic index.
6. according to the system of claim 1, wherein use predetermined mathematical formula that each patient is carried out the data pretreatment as giving a definition:
By BG=PLASBG/1.12 blood plasma BG is transformed into whole blood BG, wherein BG is that unit is measured with mg/dl, and PLASBG is a plasma glucose;
By BGMM=BG/18 will be that the BG that unit is measured is transformed into mmol/l unit with mg/dl; With
The predetermined mathematical formula that utilizes as give a definition calculates low-glycemic and hyperglycemic index:
Scale=[ln (BG)]
1.0845-5.381, wherein BG is that unit is measured with mg/dl,
Risk1=22.765 (Scale)
2, wherein
If RiskLO=Risk1 BG＜112.5, thereby have the danger of LBGI, otherwise RiskLO=0 and
If RiskHI=Risk1 BG is greater than 112.5, thereby there is the danger of HBGI, otherwise RiskHI=0,
Each patient's of BGMM1=average BGMM,
The average RiskLO of each object of RLO1=,
The average RiskHI of each object of RHI1=,
L06=is only to the average RiskLO that night, reading calculated, and is then default if there is not the reading at night,
N06, N12, N24 are the percentage rate of SMBG reading in each interval,
The sum of SMBG reading in NC1=first predetermined lasting time; With
The natural law that has the SMBG reading in NDAYS=first predetermined lasting time.
7. according to the system of claim 6, wherein N06, N12, N24 are respectively the percentage rate of SMBG reading in the following interval, i.e. 0-6:59; 7-12:59 and 18-23:59.
8. according to the system of claim 6, comprise according to giving the group assignment with the high BG index of patient that calculates as the predetermined mathematical formula of giving a definition:
If if group=0 is then specified in RHI1≤5.25 or RHI1 〉=16,
If RHI1〉if 5.25 and RHI1＜7.0, then specify group=1,
If if RHI1 〉=7.0 and RHI1＜8.5, then specify group=2 and
If if group=3 is then specified in RHI1 〉=8.5 and RHI1＜16.
9. system according to Claim 8 comprises that the predetermined mathematical formula that uses as give a definition estimates:
E0＝0.55555＊BGMM1+2.95，
E1＝0.50567＊BGMM1+0.074＊L06+2.69，
E2＝0.55555＊BGMM1-0.074＊L06+2.96，
E3=0.44000*BGMM1+0.035*L06+3.65; With
If Group=1, EST2=E1 then, if perhaps Group=2, EST2=E2 then, if perhaps Group=3, EST2=E3 then, otherwise EST2=E0, wherein EST2 is the estimation of blood sugar level.
10. according to the system of claim 9, comprise that the predetermined mathematical formula that uses as give a definition further revises estimating:
Default if (L06), EST2=E0 then,
If RL01≤0.5 and RHI1≤2.0, EST2=E0-0.25 then,
If RL01≤2.5 and RHI1〉26, then EST2=E0-1.5*RLO1 and
If (RL01/RHI1)≤0.25 and L06 1.3, EST2=EST2-0.08 then.
11. according to the system of claim 10, according to the BG data estimation patient HbA that in first predetermined lasting time, collects
1c, described method comprises:
Use in described as four predetermined mathematical formula of giving a definition at least one to HbA
1cCarry out described estimation:
A) HbA
1c=by according to the EST2 of system's correction of claim 10 or
B) HbA
1c=0.809098*BGMM1+0.064540*RLO1-0.151673*RHI1+1.873325, wherein
BGMM1 is the average BG according to the system of claim 6, and unit is mmol/l
RLO1 is the low BG index according to the system of claim 6
RHI1 is the high BG index according to the system of claim 6; Perhaps
C) HbA
1c=0.682742*HBA0+0.054377*RHI1+1.553277, wherein
HBA0 is adopted in second predetermined period before estimating previous with reference to HbA
1cReading, wherein
RHI1 is the high BG index according to the system of claim 6; Perhaps
d)HbA
1c＝0.41046＊BGMM+4.0775
Wherein BGMM1 is the average BG according to the system of claim 6, and unit is mmol/l.
12. according to the system of claim 11, wherein said second predetermined lasting time is about 3 months.
13. according to the system of claim 11, the scope of wherein said second predetermined lasting time is 2.5 months to 3.5 months.
14. according to the system of claim 11, the scope of wherein said second predetermined lasting time is 2.5 months to 6 months.
15., wherein have only and when the first predetermined lasting time sample satisfies in following four standards at least one, just use HbA according to the system of claim 11
1cThe sample choice criteria checking sample of estimating:
A) test frequency standard, wherein the first predetermined lasting time sample comprises average every day at least 1.5 to 2.5 tests;
B) can select the test frequency standard, as long as the predetermined lasting time sample comprises at least one the 3rd predetermined sample cycle, the average frequency of its reading is about 1.8 reading/skies;
C) as long as data randomization standard-1 is wherein ratio R LO1/RHI1 〉=0.005, then verify and show HbA
1cEstimation,
Wherein
RLO1 is the low BG index according to the system of claim 6
RHI1 is the high BG index according to the system of claim 6; Perhaps
D) as long as data randomization standard-2 is wherein ratio N06 〉=3% checking and show HbA
1cEstimation,
Wherein
N06 be according to the system of claim 6 night reading percentage rate.
16. according to the system of claim 15, wherein said the 3rd predetermined lasting time was at least 35 days.
17. according to the system of claim 15, the scope of wherein said the 3rd predetermined lasting time is 35 days to 40 days.
18. according to the system of claim 15, the scope of wherein said the 3rd predetermined lasting time be 35 days to the same with first predetermined lasting time long.
19. the system according to claim 1 also comprises:
BG obtains mechanism, and the described mechanism that obtains is used for obtaining the BG data from the patient.
20., wherein have only and when the first predetermined lasting time sample satisfies in following four standards at least one, just use HbA according to the system of claim 11
1cThe sample choice criteria checking sample of estimating:
A) test frequency standard, wherein the first predetermined lasting time sample comprises and measuring at least average every day 1.5 times; With
B) data randomization standard-1 wherein has only as ratio R LO1/RHI1 〉=just verify 0.005 the time or show HbA
1cEstimate,
Wherein
RLO1 is the low BG index according to the system of claim 6
RHI1 is the high BG index according to the system of claim 6; Perhaps
C) data randomization standard-2 wherein has only as ratio N06 〉=just verify 3% the time or show HbA
1cEstimate,
Wherein
N06 is the reading percentage rate at night according to the system of claim 6.
21. according to the system of claim 1 or 19, the wherein said HbA that is used for according to collected BG data estimation patient
1cSystem, it not need to realize previous HbA
1cInformation.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US40297602P | 2002-08-13 | 2002-08-13 | |
| US60/402,976 | 2002-08-13 | ||
| US60/478,377 | 2003-06-13 |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101032423A Division CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1688244A CN1688244A (en) | 2005-10-26 |
| CN100466965C true CN100466965C (en) | 2009-03-11 |
Family
ID=35306237
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101032423A Pending CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
| CNB038240092A Expired - Lifetime CN100466965C (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101032423A Pending CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN101079082A (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101785701B (en) * | 2009-01-23 | 2015-10-07 | 理康互联科技(北京)有限公司 | The device of information processing, method, equipment and reagent carrier |
| EP2393412B1 (en) * | 2009-02-04 | 2017-08-30 | Sanofi-Aventis Deutschland GmbH | Medical device and method for providing information for glycemic control |
| US20100330598A1 (en) * | 2009-06-26 | 2010-12-30 | Roche Diagnostics Operations, Inc. | METHOD, SYSTEM, AND COMPUTER PROGRAM PRODUCT FOR PROVIDING BOTH AN ESTIMATED TRUE MEAN BLOOD GLUCOSE VALUE AND ESTIMATED GLYCATED HEMOGLOBIN (HbA1C) VALUE FROM STRUCTURED SPOT MEASUREMENTS OF BLOOD GLUCOSE |
| CN102549587B (en) | 2009-08-10 | 2016-08-03 | 糖尿病工具瑞典股份公司 | For processing the apparatus and method of dextrose equivalent |
| US8532933B2 (en) * | 2010-06-18 | 2013-09-10 | Roche Diagnostics Operations, Inc. | Insulin optimization systems and testing methods with adjusted exit criterion accounting for system noise associated with biomarkers |
| JP5997453B2 (en) * | 2011-04-25 | 2016-09-28 | アークレイ株式会社 | Information processing apparatus and user terminal |
| JP2014524791A (en) * | 2011-06-23 | 2014-09-25 | ユニバーシティ オブ ヴァージニア パテント ファウンデーション | An integrated platform for monitoring and controlling blood glucose levels in diabetics |
| US20140012511A1 (en) * | 2012-07-09 | 2014-01-09 | Dexcom, Inc. | Systems and methods for leveraging smartphone features in continuous glucose monitoring |
| RU2640172C2 (en) * | 2013-02-21 | 2017-12-26 | Университи Оф Вирджиния Патент Фоундэйшион | Tracking changes to averaged values of glycemia in diabetics |
| CN104183104A (en) * | 2014-08-13 | 2014-12-03 | 上海电机学院 | Blood sugar test result communication method and system |
| AU2015335913A1 (en) * | 2014-10-22 | 2017-04-20 | Dexcom, Inc. | User interfaces for continuous glucose monitoring |
| WO2016103390A1 (en) * | 2014-12-25 | 2016-06-30 | 株式会社日立製作所 | Device for analyzing insulin secretion ability, system for analyzing insulin secretion ability provided with same, and method for analyzing insulin secretion ability |
| WO2017007775A2 (en) * | 2015-07-06 | 2017-01-12 | Abbott Diabetes Care Inc. | Systems, devices, and methods for episode detection and evaluation |
| CN113450906A (en) * | 2020-03-26 | 2021-09-28 | 苏州佳世达光电有限公司 | Clinical detection and diagnosis system and clinical detection and diagnosis method |
| CN113505925B (en) * | 2021-07-09 | 2022-07-15 | 重庆邮电大学 | Laboratory hazardous chemical anomaly information early warning method based on ANFIS |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5206144A (en) * | 1985-03-29 | 1993-04-27 | Novo Industri A/S | Determination of glycated (glycosylated) hemoglobin in blood |
| US5882935A (en) * | 1994-10-19 | 1999-03-16 | Fuji Photo Film Co., Ltd. | Analysis element and method for analyzing glycated hemoglobin content ratio |
| US5971922A (en) * | 1998-04-07 | 1999-10-26 | Meidensha Electric Mfg Co Ltd | System and method for predicting blood glucose level |
| US6027692A (en) * | 1995-04-07 | 2000-02-22 | Lxn Corporation | Apparatus for combined assay for current glucose level and intermediate or long-term glycemic control |
-
2003
- 2003-08-08 CN CNA2007101032423A patent/CN101079082A/en active Pending
- 2003-08-08 CN CNB038240092A patent/CN100466965C/en not_active Expired - Lifetime
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5206144A (en) * | 1985-03-29 | 1993-04-27 | Novo Industri A/S | Determination of glycated (glycosylated) hemoglobin in blood |
| US5882935A (en) * | 1994-10-19 | 1999-03-16 | Fuji Photo Film Co., Ltd. | Analysis element and method for analyzing glycated hemoglobin content ratio |
| US6027692A (en) * | 1995-04-07 | 2000-02-22 | Lxn Corporation | Apparatus for combined assay for current glucose level and intermediate or long-term glycemic control |
| US5971922A (en) * | 1998-04-07 | 1999-10-26 | Meidensha Electric Mfg Co Ltd | System and method for predicting blood glucose level |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1688244A (en) | 2005-10-26 |
| CN101079082A (en) | 2007-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1534121B1 (en) | Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management | |
| JP4891511B2 (en) | Method, system and computer program product for assessing diabetes blood glucose control from self-monitoring data | |
| US7824333B2 (en) | Diabetes management methods and systems | |
| CN101278847B (en) | Systems, methods for hyperglycemia and hypoglycemia, glucose variability, and ineffective self-monitoring | |
| CN100466965C (en) | Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management | |
| CN103959291B (en) | Glucose predictions device based on the regularization network with adaptively selected core and regularization parameter | |
| US20060025931A1 (en) | Method and apparatus for real time predictive modeling for chronically ill patients | |
| Musenge et al. | Glycaemic control and associated self-management behaviours in diabetic outpatients: a hospital based observation study in Lusaka, Zambia | |
| US20090171589A1 (en) | Method, System and Computer Program Product for Evaluation of Blood Glucose Variability In Diabetes From Self-Monitoring Data | |
| US20040078232A1 (en) | System and method for predicting acute, nonspecific health events | |
| CA2565918A1 (en) | Method and apparatus for real time predictive modeling for chronically ill patients | |
| US20220230751A1 (en) | Decision support system, and method in relation thereto | |
| CN117672516A (en) | Cardiovascular disease health management system based on user behaviors | |
| Colloff | The promise of technology in the management of the patient with diabetes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1084845 Country of ref document: HK |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1084845 Country of ref document: HK |
|
| CX01 | Expiry of patent term |
Granted publication date: 20090311 |
|
| CX01 | Expiry of patent term |