A Physiological Model Of Glucose-insulin Interaction In .
A physiological model of glucose-insulininteraction in type 1 diabetes mellitusE.D. Lehmann”and T. Deutsch *Diabetic Research Laboratory, Medical Unit (4NW), Department of EndocrinologyandChemical Pathology, United Medical and Dental Schools, St Thomas’ Hospital, LambethPalace Road, London SE1 7EH, UK; and ‘Computer Centre, Semmelweis University ofMedicine, Budapest, Hungary and Department of Systems Science, Centre for Measurement and Information in Medicine, City University, London EClV OHB, UKABSTRACTA clinical model ofglucose-insulin interactionin insulin-dependentdiabetesmellitushas been developedforpatient andmedical staff education. The model attempts to reflect the underlying (pathojphysiologyof insulin action andcarbohydrateabsorptionin quantitativeterms such as insulin sensitivity,volume ofglucose and insulin distributionandmaximal rate ofgastric emptying. The model'spredictions also allow a 24 h simulation ofpatient bloodglucoseprofilesto be generated. A description of the model is provided and its operation illustrated by clinical case studies of insulintreated diabeticpatients. The possible use of the model as a toolfor automated insulin dosage adjustment is explored.Keywords: Computer simulation, physiological model, type 1 diabetes mellitus, insulin dosage adjustmentINTRODUCTIONDiabetesmellitus is a major chronic disease inindustrialized countries. It affects 3% of the population of Europe and approximately100 million peopleworldwide’.While the incidence of the disease iscurrently on the increase in western society, theincidence and severity of the later life complicationswhich accompanyit can be considerablyreducedif the diabetic patient receives effective treatmentleading to good glycaemic control’. In general suchtreatment attempts to achieve normoglycaemiabymaintaining a careful balance between diet, physicalactivity and insulin therapy. However, education ofthe diabetic patient to achieve this balance requires alevel of clinical expertise which, althoughresent inspecialized diabetes units and some genera P racticeswith an interest in diabetes, is not always to l! e foundin other sectors of the health service3. One way ofmaking this clinical ex ertise more widely available isto use information tee nology*.A number of computer-basedapproaches to aid inthe treatment or long-term management of diabeticpatients have been previously reported in the literature. These include knowledge-basedsystems toadvise on patient management in out-patient clinics5,computer algorithms for insulin dosage adjustment”and mathematical models for predicting or simulatingpatient blood glucose levels7.Berger and Rodbard’ have developed a computerRCorrespondence and reprint requests to: E.D. Lehmann, DiabeticResearch Laboratory, Medical Unit (4NW), Department of Endocrinology and Chemical Pathology, United Medical and Dental Schools,St Thomas’ Hospital, Lambeth Palace Road, London SE1 7EH, UKprogram for the simulation of insulin and glucosedynamics following the subcutaneousinjection ofinsulin. Their program incorporatesa pharmacokinetic model of insulin action to calculate the timecourse of plasma and ‘active’ insulin for variouscombinationsof popular insulin preparations. Withthe use of a harmacodynamicglucose model todescribe the B ependenceof glucose dynamics onplasma insulin and glucose levels, their program canalso predict the expected time course of plasmaglucose in response to a change in insulin dose,timing or regimen. However, their system has notbeen designed for individual patient parameterization and simulation.Numerous other stand-alone mathematical modelsof the lucoregulatosystem in insulin-dependent(type 1B diabetes me7 litus exist%15. We have, however, developed a physiological model of glucoseinsulin interactionforatient and medical staffeducation about insulin- x ependent diabetes as partof a more complexdiabetesdata managementsystem 1G20. The integration of the model into thiscomplex system is described elsewhere in this journalin a separate paper by Lehmann and Deutsch.The model developed, in part, draws on Bergerand Rodbard’spharmacokineticmodel of insulinaction’ which has been implemented in conjunctionwith a new model of glucose phannacodynamicsbased on experimental data from the literature*‘. Themodel can provide a suitable framework to characterize individual diabetic patients quantitatively andpredict the blood glucose profile which is expected tobe produced by an adjustment in the diet and/orinsulin dosage regimen. This new model attempts toprovide an anatomicallyexplicit basis for patient0 1992 Butterworth-Heinemann for BES0141-5425/92/030235-08J. Biomed. Eng. 1992, Vol. 14, May235
A m&l of type 1 diabetes: ED. Lehmann and T. Dertschsimulations,with different functions for differentorgans within the bod . The model potentially hasapplication as a tool ror automated insulin dosageadjustment based on home blood glucose monitoringdata. Clinical case studies will be presented and thepossible use of the model as a vehicle for automatedinsulin dosage adjustment will be explored.MODEL DESCRIPTIONThe glycaemic response of an insulin-treateddiabeticpatient goes through transitorybases leading to asteady-state gl caemic profile fo Plowing a change ineither the insu rin regimen or diet. The purpose of ourmodel is to simulate these steady-state glycaemic andplasma insulin responses independently of the initialvalues from which the simulation is started.F@M 7 shows the anatomical basis of the modelwhich assumes a patient completely lacking endogenous insulin secretion. It contains a single glucoseglucose (including001 re resenting extracellularE lood gPucose) into which glucose enters via bothintestinal absorption and hepatic glucose production.Glucose is removed from this space by insulininde endent glucose utilization in red blood cells(RB8s) and the central nervous system (CNS) as wellas by insulin-dependentglucose utilization in the liveranderiphery; the latter taking place mostly inmust Pe and adipose tissue. Hepatic and peripheralhandling of glucose in the model are dealt withseparately. Glucose excretion takes place above thelucose as a function of therenal threshold ofcreatinine clearance g lomerular filtration) rate.By separating the hepatic and peripheral handlingof glucose in the model it is possible to assigndifferentpatient-specificinsulin sensitivityparameters to glucose-insulininteractions in the liver andperiphery.As shown schematicallyin Figure 1,eripheral glucose uptake takes place as a function ofE0th insulin and plasma glucose levels; the formerenhancingglucoseutilizationaccordingto theperipheral insulin sensitivity parameter, S,, which hasa normalizedvalue between 0 and 1. SP multilied by the insulin level gives the effective insulinPeve1 responsible for the control action.As the liver both produces and utilizes glucoseCNS RBCslcarbohvdratemealpdegradationRenal threshold of glucoseSubcutaneousinsulinin’ection;. PeripheryIPeripheralglucoseuptakefuTime(2)Renal threshold ofglucose: 9.ammolll(31 Creafinineclearancerate: 100 mllmin(4) “epaiic insulinsensilivityIShI: 0.5(5) Peripheral insulinFigure 1236Schematic summarizing the anatomical basis of the modelJ. Biomed. Eng. 1992, Vol. 14, May-I
A model of qpe 7 diabetes:E.D. Lehmann and T. DeutschTable 1Net hepaticarterial blood glucosefrom Guyton et al. “.parameter which hasEffectiveplasmainsulin(&. m&d)023468910glucose balance (mmol h-‘) as a function of thelevel, AG, and plasma insulin level, I; calculated& is a patient-specific hepatic insulin sensitivitya normalized value between 0 and 1AGsAG AGaI.lmmoll-’3.3 mm01 I-’4.4 mm01 .7160.0114.666.046.322.64.3- 10.0-25.3-43.3-47.3-49.378.353.3- 1.7-54.3- 76.0-85.0-92.0-97.3-101.0- 104.0- 106.7depending on the blood glucose and insulin levels,we have modelled hepatic glucose handling in termsof the ‘net hepatic glucose balance which is computed asthe sum of gluconeogenesis, glycogen breakdown andglycogen synthesis data derived for different bloodglucose and insulin levels from nomograms given inGuyton et al. 2‘. This represe ntation of hepatic glucosehandling was chosen in order to avoid the use of nonphysiologicallybased mathematicalfunctions todescribe hepatic timction7”‘ ‘2”5. Table 1 shows howthe net hepatic glucose balance varies as a function ofglucose and normalized insulin levels. &, the hepaticinsulin sensitivity parameter, which also has a normalized value between 0 and 1, allows computationof the effective insulin level which controls hepaticglucose handling.The net hepatic glucose balance for any arteriallevel between1.1 mmol 1-l andblood glucose4.4 mmoll-’is computed by interpolation betweenthe values shown on the curves in Figure 1. Capillaryblood glucose values measured using home monitoring blood glucose meters are approximately25%lower than the arterial blood glucose levels which aregiven in Guyton et al. 2’ . Hence, in our model we usethe capillary blood glucose levels which correspondto the arterial blood glucose data given in Table 1.The data shown in Tabb 7 are based on the steadystate plasma insulin level which is normalized withrespect to a basal level, Ibasal. Note that for low bloodglucose values there is an automatic compensatoryincrease in hepatic glucose production(positivebalance) and at high blood glucose levels the netaction of the liver is to take up glucose from the blood(negative balance).Glucose enters the portal circulation via first-orderabsorption from the gut. The rate of gastric emptyingwhich provides the glucose flux into the smallintestine in the model is assumed to be controlled bya complex process maintaining a relatively constantglucose supply to the gut during carbohydrate absorption apart from the ascending and descending phasesof the gastric emptying process.The duration of the period in which glucose entryfrom the stomach into the duodenum is constant andmaximal has been defined as a function of thecarbohydrate content of the meal ingested. Thus thetime course of the systemic appearance of glucose isdescribed by either a trapezoidal or a triangularfunction depending on the quantity of carbohydratein the meal.The function of the kidneys to excrete glucose hasbeen modelled in terms of two patient-specific modelparameters: the renal threshold of glucose and thecreatinine clearance @omen&rfiltration) rate.The model contains separate compartmentsforplasma and ‘active’ insulin. Insulin is removed fromthe former by hepatic degradation while the latter isresponsible for glycaemic control. The activation anddeactivation of insulin are assumed to obey first-orderkinetics. The only insulin input into the model comesfrom the absorptionsite following subcutaneousinjection.MODELEQUATIONSFour differential equations along with twelve auxilliary relations and the experimentaldata fromGuyton et aL2’ constitute the model which is solvedby numerical integration. The change in the lasmainsulin concentration,Z, is given by the fo Plowingequation:dZlabs-b-Zdt Ir:where k is the first-order rate constant of insulinelimination, Z& is the rate of insulin absorption andF is the volume of insulin distribution. The build-upand the deactivation of the ‘active’ insulin pool, la, isassumed to obey first-order kinetics:dZa kreZ-k2.Z,dtwhere kr and kz are first-order rate constants whichserve to describe the delay in insulin action. The rateof insulin absorption is modelled according to Bergerand Rodbard’.s. tS. T&. Dlabs(t) t[T;, (3)t’]’where t is the time elapsed from the injection, T50 isthe time at which 50% of the dose, D, has beenabsorbed and s is a preparation-specificparameterdefining the insulin absorption pattern of the differenttypes of insulin catered for in the model (regular,intermediate, lente and ultralente).A linear dependency of T50 on dose is defined as:T& a.D b(4)where a and b are preparation-specificparametersthe values of which are given in Berger and Rodbard’along with values for s. If the insulin regimen consists of more than one injection and/or components,Zabsbecomes the sum of the individual ZAs contributions resulting from the different multicomponentinjections.The steady-state insulin profile, Zss,corresponding toa given regimen, is computed by using the superposition princi le assuming three days to be enough toreach stea By-state conditions:ZJ t) Z(t) Z(t 24) Z(t 48)(5a)Z, (t) Z,(t) Za(t 24) Za(t 48)(WJ. Biomed. Eng. 1992, Vol. 14, May237
A model of lyPe 1 diabetes: E.D. Lehmann and T. Deutxhi.e. the steady-state response results from the composite effect of injections given for three subsequentdays. It is evident that this summation is not neededfor re lar insulin preparations (e.g. actrapid) but itshoul LY be used for other, longer acting, insulinpreparations whose half time of absorption is higher,especially when larger doses are given.Since the experimentaldata provided by Guytonet aLzl refer to equilibrium conditions, the insulinlevel equilibrated with the steady-state active insulinis considered when computing the net hepatic glucosebalance and peripheralglucose uptake. In otherwords, at any time durin the simulation, we havesteady-state I,,(t) and I,,,, t) values, but use:l&(t) k&,,,(Wk at) - G,“,(t) - G,,,(t)-(7)VGwhere G is the plasma glucose level, Gin is thesystemic a pearance of glucose via glucose absorption from tKe gut, Gout is the overall rate of peripheraland insulin-independentglucose utilization, NHGB isthe net hepatic glucose balance, G,& is the rate ofrenal gh.KOSeexcretion and VG is the volume ofdistribution for glucose.Assuming a classical Michaelis-Mentenrelationship between glucose utilization and the plasmaglucose concentration,with a constant K, such thatinsulin concentration is reflected in different values ofthe maximal rate of the transport process, we canwrites:G(cx! /Z,*, I&) GI)(K,,, Gx.(K, Gx)G)(8)where c is the slope of the peripheral glucose utilization versus insulin level relationship, GI is the insulinindependent glucose utilization and Gx is a referenceglucose level. The NHGB value at any combinationof G and I,* has been derived from the data summarized in Ta8 Ze 1 using &, *l& as the effective insulinlevel. The amount of glucose in the gut, G,,, following the ingestion of a meal containing Ch millimolesof glucose equivalent carbohydrate is defined as:W&t)dtG-empt -kgabs’G,twhere kgabSis the rate constant of glucose absorptionfrom the gut into the systemic circulation and Gemptisthe rate of gastric emptying which is shown as afunction of time in Figure 2a. The duration of theis constantperiod Tmax, for which gastric emptyinand maximal (Vmax,)is a function of tB e carbohydrate content of the meal ingested:Tmaxs, [ Ch -l/2Vmax,,*2(Tascs, where Vmax,238VmaxgeGi,( t) N.GB(G&G, (t)(6)as the insulin level responsible for the hepatic anderi heral control action, where l&(t) is the insulinPeve P in equilibrium with l&t).Assuming a single corn artment for extracellularglucose, the change in g Pucose concentrationwithtime is given by the differential equation:dG- dtTimeTasc9e’ Tdesb9eTime(t)Figure 2Rate of gastric emptying, Gempt, as a function of time, t ; a,for carbohydrate intake 3 10 g and b, for carbohydrate intake 10 gand Tax e and Tdess, are the respective lengths ofthe ascen 8.mg and descending branches of the gastricemptying curve which have default values in themodel of 30 min (0.5 h) (Figure 2a).However,for small quantities of carbohydrate(below approximatelylog) such values cannot beused because there will never be time for the gastricem tying curve to plateau out. In such cases Tascs,an B Tdes, are defined as:Tascs, Tdess, 2 Ch/ Vmax,(11)giving a triangular function as shown in Figure 2b.Equation (11) is only used when the quantiofcarbohydrateingested falls below a critical 7 eve1(Ch,-,iJ which is defined a :Chcrit [(Tax, Tdess,)Vmaxs,]/2Using linear interpolation the rate offor meals containing Ch millimolesgreater than Chcrib can therefore being to the time elapsed from the startfollows:G empt ( m ,/T ge)G empt vmage;t Tax,t;Tax,(12)gastric emp ‘ngof carbohy rrratedefined, accordof the meal, t, as t s Tax,,(134 Tmax,(13b)G empt Vmax,- ( VmaxsJTdes,)(t - Tascs, - Tmaxs,);Tascs, Tmaxs, S t Tmaxs, Tdes&]/Vmaxs,(10)is the maximal rate of gastric emptyingJ. Biomed. Eng. 1992, Vol. 14, MayTascs, Tdes,G empt 0;elsewhere(13 )(13d)
A model of type 1 diabetes: E.D. Lehmann and T. DeutxhTable 2from BergerPatient-independentmodel parameterand Rodbard”and Guyton et al.”valuescalculcatedk, 5.4 h-’Insulin eliminationconstantk, 0.025 h-’Parameter for insulinpharmacodynamicsk, 1.25h-’Parameter for insulinpharmacodynamicsI basal IOmUI-’Referenceinsulin’K, 10mmoll‘kg-’Insulin-independentglucose utilizationper kg body weightGx 5.3mmoll ’k r 1hVmax,,basal level ofMichaelis constant forenzyme-mediatedglucoseuptakeG, 0.54mmolhReferenceutilizationc O.O15mmolhrate‘kg’m*U ‘I’value for glucoseSlope of peripheralutilizationuenur insulin lineglucoseRate constant for glucoseabsorptionfrom the gut 120 mmol h’Maximalemptyingrate of gastricV, 0.142 1kg’Volume of distributionforinsulin per kg body weightVo O.22lkg’Volume of distributionforglucose per kg body weightGlucose input via the gut wall, Gin, can be modelledby:Gin kgabs. G,,(14)Values for these model parameters, which have beenderived from Berger and Rodbard’,and Guyton et 1. are given in Table 2. All parameters except S,and Sh are assumed to be patient independent.The rate of renal glucose excretion,G,,, in themodel is defined as:G,,, GFR(GGrm 0;RTG);elsewhereif G RTG(1W(15b)for blood glucose values (G) above the renalthreshold of glucose (RTG) where GFR is the glomerular filtration (creatinineclearance)rate. Defaultparameter values in the model have been set for RTGand GFR at 9.0 mmol 1-r and 100 ml min-’ respectively. These default values are used for all patientcases except where renal dysfunction is suspected andthe clinical parametersare actually measured. Asshown in Figure 7, the renal excretion of glucose( Gren) is zero for blood glucose values below the renalthreshold of glucose (Equation [ 15b]).It is noted that the insulin and glucose parts of themodel are only linked by equation (8) and whencomputingthe net hepatic glucose balance as afunction of G and I&. This means that the plasmaand ‘active’ insulin profiles as well as the glucoseabsorption profiles for any meal can be computedseparately. This characteristic of the model is utilizedwhen implementing the system for computer simulations.COMPUTER IMPLEMENTATIONAs the only exogenous source of glucose in the modelis carbohydrateintake during meals, the systemicappearance curves of glucose following any meal witha carbohydratecontent betwen 0-60gcan be computed apriori and stored for use as appropriate duringthe simulation. The storage is made for 6 h at 15minintervals.Since in patients completely lacking endogenousinsulin secretion the plasma insulin level followingsubcutaneous injection does not depend on the bloodglucose level, a library of plasma insulin profiles and‘active’ insulin levels can be corn uted apriori for anydose (currently less than 40 units7 and preparation ofinsulin (regular, intermediate,lente and ultralente).This computation assumes that insulin absorption andelimination are not patient specific, apart from thevolume of insulin distribution (VI) which is a linearscaling factor dependenton the patient’s bodylasma andweight. The precomputedsteady-state‘active’ insulin levels are stored for 1 B ay (24 h) at15 min intervals.Simulations are carried out over a 2 day (48 h)period using first-order Euler integrationwith a15 min step size. The second day’s blood glucose andplasma insulinrefiles are assumed to representsteady-state profi Pes as responses to the current insulintherapy and diet plan. These profiles are displayed onthe computer screen as the results of the simulation.A parameter estimation routine has been implemented whereby values for Sh and S, which give thebest ‘fit’ between the observed and predicted data areautomatically determined. The fit is assessed for anycombination of &, and S, in the range of 0 to 1 usinga step size of 0.1 for both parameters. In the presentform of the model these two parameters are used toatient-specific. This is in addition tomake the modelthe patient’s bo B y weight, renal threshold of glucoseand creatinine clearance rate which can be assessedindependentlyin the clinic.In determining the fit hypoglycaemicepisodes areassigned a blood glucose value of 1 .Ommol 1-l. Fit isassessed using modifiedleast-squarescriteriatocalculate the difference between the two data sets atthe observed time points. Parameter values for whichthere is a conflict of trends between the observed andpredicted data in any time period are assigned a verypoor fit by using a ‘penalty’ score for such cases. Forexampleif the observeddata shows a markeddecrease in a given period while there is an increasein the simulated glycaemic profile in the same period,a penalty score is associated with the blood glucoselevel at the end of this period although the absolutedeviation might be minor.An insulin dosage optimization routine has alsobeen implementedwherebydifferentqualitativetherapeutic strategies can be automatically selectedby the system depending on the deficiencies in theblood glucose control observed. Minimization of thetotal amount of daily insulin injected is the overallobjective of the optimization algorithm which tries tofind the simplest change in the insulin regimenrequired to achieve normoglycaemia.As such thedefault strategy is to ‘decrease insulin’ but an alternativestrategy to ‘increase insulin’ has also been provided forJ. Biomed. Eng. 1992, Vol. 14, May239
A model oftypt1dkbetcs: E.D. Lxhmann and T. Deutschcasesof persistenthype caemia.Strategiesto‘decrease regular insulin’ an d ‘&crease logger actinginsulin’ have also been implemented to cater for caseswhen h oglycaemicepisodes occur - the exactstrategy r:c osen being dependent on the timing of the‘hypo’ in relation to the preceding insulin in’ection.The current system runs under DOS on an T/BM PCor compatible. A multitasking version is also availablefor 80386 based machines running WINDOWS 3.0. Thisallows the displaof multiple windows showingFordifferent parts o r the system in operation.example, the data entry screens can be displayed inone window with the results of a simulation in asecond and patient-specific model parameters in athird. The number of windows displayed at any onetime is wholly dependent on the memory capabilitiesof the machine being used.All code for the model and connecteddatarecessing has been implemented in TURBOPASCALPBorland International, v.5.5). The current implementation, running on an IBM PS/2 Model 70 386 at25 MHz with an Intel 80387 numerical co-processor,takes less than 1 s to perform a simulation, less than25s to perform parameter estimation and less than20s to perform insulin dosage optimization. On a25MHzIBM PS/2 Model 95 486 insulin dosageoptimization and parameter estimation both take lessthan 10 s to perform. This speed is, to a great extent,recomputing and storing the plasmaachieved binsulin leve Us ollowing subcutaneous insulin injectionand the systemic appearancecurves for glucosefollowing a ke(h)glucoseLiver:0.60l%n@wy: 0.30Fit: 1.Ommollotm-levelTime: 1O:lS/7-x4.0 mm01 litre0b”36912TimeBlood7151821L.e;24glucoseLNW 0.60Pwiphq: 0.30VP:x.x Mldliie-level15Time:and-1Ch)20CLINICALEXAMPLES240 J. Biomed. Eng. 1992, Vol. 14, MayLiver:0.50Periphery:1.00,Fit: x.x mmol litrelevel0InsulinFigure 3 shows the front end used for accessing themodel. The upper panel dis lays the observedrecorded by ain(measured) blood glucose rea dp70 kg male, insulin-dependent dia%setic patient on athree times daily insulin injection regimen usinghome blood glucose monitoring equipment. Theblood glucose data displayed can either be readingsfrom a single day or the averaged glycaemic profilecomputed from a number of days’ data. The averaging process used to generate such ‘modal’ day bloodglucose profiles has been previouslydescribedelsewhere 16. The lower p anel of the screen representsa composite display of information regarding insulinand carbohydrate intake as well as hypoglycaemicreactions. The distribution of bread equivalent units(10 g carbohydrate) can be seen as can the three timesdaily actrapid and NPH injections that the patientwas prescribed.Figure 3a shows the screen display while parameterregress. Given the insulin andestimation is incarbohydrate in tal! e shown in the lower panel, thesystem performs simulations at 10% increments of thevalues of both &., and Sp. As such the light grey areaon the graph, made up m this case from 95 separatesimulations, constitutes the ‘search space’ for theparameter estimation routine.In the example shown in Figure 3a a penalty score(x.x) has been assigned to the fit for parameter valuesof & and S, of 0.5 and 1.0 respectively, because inthe period between supper and bedtime there was aglucosecarbohydrate13:30intake 0C36912Time15182124(h)Figure 3 Front end display used for accessing the model: a,parameter estimation (‘Wing’) in progress based on clinical data froma 7Okg, male, insulin-dependent diabetic patient; b, results of asimulation after fitting has been performed. Upper panel: observed (0)and predicted blood glucose levels. Lower panel: insulin andcarbohydrate intake with predicted plasma insulin curve; c, ‘teachingmode’ illustrating the effect on the patient’s blood glucose profile ofmissing a morning injectionclear increase in the observed blood glucose level(from 5.3 mmoll-’ to 9.3 mmoll-‘);however, in thecorresponding period the simulator predicted a cleardecreme in the blood glucose level (from 6.5 mmoll-’to 5.0mmoll’for the same parameter values). Suchconflict in the trends means that the predicted curvedoes not match the observed data closely enough.Figure 3b shows the best traditional least-squares fit,obtained by the parameter estimation routine, whichalso closely matches the trends in the observed data.The curve shows the redicted blood glucose profilefor the patient’s car g ohydrate and insulin intakeshown in the loweranel in Figure 3a. The meandeviation between o E served and computed valueswas l.Ommoll-’for hepatic and peripheral insulinsensitivities of 0.6 and 0.3 respectively.
A model of type 7 diabetes: E.D. Lehmann and T. DeutschA clock function has been implemented to permitcloser inspection of the simulated data. This allowstwo cursors to be moved along the blood glucose andplasma insulin profiles, and for the predicted valuesof each variable to be read off the curves at relevanttime points. This function is particularly useful forassessing the maximum and minimum blood glucoselevels at different times during the day.Figure 3c demonstrates the use of the model as aneducational tool where the glycaemic effect of missingLiver 0 2;20t0:E-IBlood155glucosePenphery.0.2F,, 0 8 mm0 IltfeClevel---- Time:aTime9:45(h)Liver:0.20Periphery:0.20,Fit: x.x mmol litreInsulinand carbohydrateinfake plasma insulinleveln NPH/monotardrbohydrateintakea morning injection is simulated for this patient. Thehyperglycaemiawhich would result is predicted toreach a maximum of 16.8mmoll-’at 13:30h.Another clinical example of the use of the model isgiven in Figure 4-a which shows graphicallydatacollected by a 22-year-old,female, insulin-treateddiabetic patient who was receiving twice daily NPHinjections. As the patient was overweight (75 kg) shehad been placed by a dietitian on a restricted diet(60g carbohydrate per day) to lose weight. Figure 4ashows the predicted blood glucose and plasma insulinlevels for this patient’s current treatment regimenafter parameter estimation was performed. The meandeviation between observed and predicted bloodglucose values was 0.8mmolll’for hepatic andperipheral insulin sensitivity parameters of 0.2.Using these parameter values insulin dosage optimization has been performed as shown in progress inFigure4b. For this a target therapeutic range has beendefined (from 4 to lOmmoll ‘)and the system hasdetermined that the guiding strategy to reach thistarget range should be to ‘increase insulin’. Thisstrategy was chosen because no ‘hypos’ occurredduring the day and the overall blood glucose profilewas raised, with an observedmaximumbloodglucosevalue of 10.6 mmolll’at 21:30 h anda predictedmaximumof 11.9mmolll’at 9:45 h(Figure 4 ).The guiding strategy is implementedwithin theoverall strategy of the optimization algorithm which isthe minimization of the total amount of daily insulin.This is achieved, as shown in Figure4c, b increasingthe 18:00 h and 8:00 h insulin by 10 an dy 15 units ofNPH respectively. Having implemented this changethe simulated blood glucose profile now gives apredicted maximum of 9.9 mmol 1-l at 10:00 h with apredicted minimum of 4.2 mmol 1-r overnight; bloodglucose values which would be totally acceptable to 24[h)Liver:0.20Periphery:0.20,Fit: x.x mmol litrelevelTime:9.9mmollit&’0'InsulinEOt -c::40 5;f--;3020100C10.00and carbohydrateintake plasma me15182124(h)Figure 4 a, Results of a simulation for a 75 kg, female, insulindependent diabetic patient after parameter estimation has beenperformed; b, insulin dosage optimization in progress; repeatedsimulations are being performed with an increasing insulin dose untilthe predicted blood glucose profile falls within the therapeutic range;c, insulin dosage optimization complete; the effect of the advice hasbeen simulatedThe model presented here focuses on the adjustmentof insulin and/or diet in the insulin-treated diabeticpatient. In contrast to previously developed heuristicrule-basedexpert systems and
respect to a basal level, Ibasal. Note that for low blood glucose values there is an automatic compensatory increase in hepatic glucose production (positive balance) and at high blood glucose levels the net action of the liver is to take up glucose from the
table of contents introduction to alberta postpartum and newborn clinical pathways 6 acknowledgements 7 postpartum clinical pathway 8 introduction 9 physiological health: vital signs 12 physiological health: pain 15 physiological health: lochia 18 physiological health: perineum 19 physiological health: abdominal incision 20 physiological health: rh factor 21 physiological health: breasts 22
The Dario Blood Glucose Test Strips are for use with the Dario Blood Glucose Meter to quantitatively measure glucose (sugar) in fresh capillary whole blood samples drawn from the fingertip. The Dario Glucose Control Solutions are for use with the Dario Blood Glucose Meter and the Dario Blood Glucose Test Strips to check that the meter and test
The concentration of blood glucose level is maintained constant by the action of two general opposing factors: The rate of glucose enterance to the blood Rate of removal of blood glucose Regulation of blood glucose. Blood Glucose F
A blood glucose meter uses blood glucose test strips to mea-sure the blood glucose level of a diabetic patient. The mea-surement is quick and only requires a small amount of blood. The measured value is recorded in a blood glucose meter which also contains blood glucose data from previous measurements [1-4]. The reaction zone on the electrode of
Sesquicocoate, Methyl Glucose Sesquiisostearate, Methyl Glucose Sesquilaurate, Methyl Glucose Sesquioleate, and Methyl Glucose Sesquistearate, is typically achieved via transesterification of an appropriate fatty acid methyl ester (e.g., methyl laurate to get Methyl Glucose Laurate) with methyl glucoside (releasing methanol as a by-product).
Current AOAC LC Methods 984.22 Purity of lactose Lactose 2000.17 Raw cane sugar Glucose, fructose Glucose, fructose, sucrose, lactose Cane & beet final molassess 996.04 Glucose, fructose, sucrose, maltose 982.14 Presweetened cereals Fructose, glucose, lactose, maltose, sucrose 980.13 Milk chocolate 977.20 Honey Fructose, glucose, sucrose
Glucose High: 600 mg/dL The Livongo meter can measure blood glucose accurately from 20 mg/dL to 600 mg/dL. If your blood glucose is higher than 600 mg/dL, the screen will display this message. Please call your Doctor if you receive this message. Glucose Low: 20 mg/dL Livongo meter can measure blood glucose accurately from 20mg/dL to 600 mg/dL.
ASME BPV CODE, EDITION 2019 Construction Code requirements Section VIII, Div. 1, 2 a 3 ; Section IX ASME BPV Section V, Article 1, T-120(f) ASME BPV Section V, Article 1, Mandatory Appendix III ASME BPV Section V, Article 1, Mandatory Appendix II (for UT-PA, UT-TOFD, RT-DR, RT-CR only ) SNT-TC-1A:2016; ASNT CP-189:2016 ASME B31.1* Section I Section XII ASME BPV Section V, Article 1, Mandatory .
