Standardized Mixed-MealTolerance and ArginineStimulation Tests ProvideReproducible and ComplementaryMeasures of b-Cell Function:Results From the Foundation forthe National Institutes of HealthBiomarkers ConsortiumInvestigative SeriesDiabetes Care 2016;39:1602–1613 | DOI: 10.2337/dc15-0931

OBJECTIVE

Standardized, reproducible, and feasible quantification of b-cell function (BCF) isnecessary for the evaluation of interventions to improve insulin secretion andimportant for comparison across studies. We therefore characterized the re-sponses to, and reproducibility of, standardized methods of in vivo BCF acrossdifferent glucose tolerance states.

RESEARCH DESIGN AND METHODS

Participants classified as having normal glucose tolerance (NGT; n = 23), predia-betes (PDM; n = 17), and type 2 diabetes mellitus (T2DM; n = 22) underwent twostandardized mixed-meal tolerance tests (MMTT) and two standardized argininestimulation tests (AST) in a test-retest paradigm and one frequently sampledintravenous glucose tolerance test (FSIGT).

RESULTS

From theMMTT, insulin secretion in T2DMwas>86% lower comparedwith NGT orPDM (P < 0.001). Insulin sensitivity (Si) decreased from NGT to PDM (∼50%) toT2DM (93% lower [P < 0.001]). In the AST, insulin secretory response to arginine atbasal glucose and during hyperglycemia was lower in T2DM compared with NGTand PDM (>58%; all P < 0.001). FSIGT showed decreases in both insulin secretionand Si across populations (P < 0.001), although Si did not differ significantly be-tween PDM and T2DM populations. Reproducibility was generally good for theMMTT, with intraclass correlation coefficients (ICCs) ranging from ∼0.3 to ∼0.8depending on population and variable. Reproducibility for the AST was very good,with ICC values >0.8 across all variables and populations.

CONCLUSIONS

Standardized MMTT and AST provide reproducible and complementary measuresof BCF with characteristics favorable for longitudinal interventional trials use.

1Lilly Research Laboratories, Eli Lilly and Com-pany, Indianapolis, IN2Division of Endocrinology, Diabetes&Metabolism,Mayo Clinic College of Medicine, Rochester, MN3R-Squared Solutions, Skillman, NJ4Kelly Government Solutions for National Insti-tute of Diabetes and Digestive and Kidney Dis-eases, Rockville, MD5Pfizer, Cambridge, MA, and Groton, CT6Cedars-Sinai Diabetes and Obesity ResearchInstitute, Los Angeles, CA7TakedaDevelopment Center Americas, Deerfield, IL8Department of Information Engineering, Universityof Padova, Padova, Italy9Janssen Research & Development, San Diego, CA10Pacific Northwest Diabetes Research Institute,Seattle, WA11Division of Endocrinology, Departments ofMedicine and Pharmacology, University ofWashington, Seattle, WA12Sanofi, Paris, France13University of Pennsylvania, Philadelphia, PA14Foundation for the National Institutes ofHealth, Bethesda, MD15Joslin Diabetes Center, Boston, MA16ROI Biopharma Consulting, East Lyme, CT

Corresponding author: David A. Fryburg, dfryburg@roibiopharma.com.

Received 5 May 2016 and accepted 15 June2016.

Clinical trial reg. nos. NCT01454973, NCT01663207,and NCT01663207, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc15-0931/-/DC1.

S.S.S. and A.V. share first authorship.

© 2016 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More in-formation is available at http://diabetesjournals.org/site/license.

Sudha S. Shankar,1 Adrian Vella,2

Ralph H. Raymond,3 Myrlene A. Staten,4

Roberto A. Calle,5 Richard N. Bergman,6

Charlie Cao,7 Danny Chen,5

Claudio Cobelli,8 Chiara Dalla Man,8

Mark Deeg,1 Jennifer Q. Dong,5

Douglas S. Lee,5 David Polidori,9

R. Paul Robertson,10,11 Hartmut Ruetten,12

Darko Stefanovski,13 Maria T. Vassileva,14

Gordon C. Weir,15 and David A. Fryburg,16

on behalf of the Foundation for the

National Institutes of Health b-Cell Project

Team

1602 Diabetes Care Volume 39, September 2016

PATH

OPHYS

IOLO

GY/COMPLICATIONS

The pathophysiologic hallmarks of type 2diabetes mellitus (T2DM) are defects ininsulin action and b-cell function (BCF)(1), with the latter manifesting as inade-quate insulin secretion in response to hy-perglycemia (2). Early intervention mayameliorate these defects (3). However,determination of whether a given inter-vention has clinically relevant effects onBCF requires long-term testing in largecohorts. Therefore, the ability to simplyand reproducibly test BCF is critical.Quantification of BCF would enableevaluation of interventions tailored tospecific functional b-cell defects in mul-tiple populations, as well as their poten-tial to alter disease progression.It is readily apparent that assessment of

BCF has been accomplishedusing differentchallenge routes (oral vs. intravenous),stimuli (e.g., arginine, glucagon, glucose,mixed meals), and sampling times, oftenin single-center studies (4–7). Even for thesame method, such as the meal tolerancetest, the composition and caloric load ofthe test meals often differ (5,8,9). Addi-tionally, different approaches have beenused concurrently to assess insulin secre-tion and sensitivity. Some reports usesimple ratios or calculations for insulinsensitivity (Si) and secretion using mea-surements under basal (e.g., homeostasismodel assessment of BCF and quantitativeinsulin sensitivity check index) or postchal-lenge conditions (e.g., Matsuda index andDI0–30/DG0–30). However, due to the com-plex physiology of appearance of the nu-trient stimuli and the sites and timing ofinsulin action and clearance, indices ob-tained using these simple calculationsmay have limited utility (6,7). Some ofthe most widely used methods of measur-ing BCF are described in Table 1, alongwith a summary of the strengths and lim-itations of these methods.Recognizing the need for more con-

sensus methods for the measurementof BCF in humans, the multistakeholderb-Cell Project Team (BCPT) of the Foun-dation for the National Institutes ofHealth (FNIH) Biomarkers Consortium(http://www.biomarkersconsortium.org/) was formed to standardize andcharacterize select BCF tests. The goalof this consortium-driven work is to en-able inclusion of such assessments in fu-ture longitudinal clinical trials examiningresponse to therapeutic interventionsand disease progression. Members ofthe BCPT are listed in the APPENDIX. In

accordance with the principles of theFNIH partnership, project results aremade publically available.

We sought to characterize the repro-ducibility of two methods that could beused in a multicenter clinical trial andthat could quantify insulin secretionacross the glucose tolerance spectrum.An important aspect of quantify insulinsecretion is to do so as a function of pre-vailing insulin action. This is the basis ofthe minimal model as applied to data ob-tained from an intravenous (frequentlysampled intravenous glucose tolerancetest [FSIGT]) or oral challenge (mixed-meal tolerance test [MMTT] or oral glu-cose tolerance test [OGTT]) (10,11).Although the FSIGT is the only in vivotest that potentially replicates the invitro finding of first-phase insulin secre-tion, it is more difficult to implementroutinely in large multicenter interven-tional trials (12,13). For the BCPT, thisspurred interest in alternative BCF tests,with particular emphasis on operationalfeasibility and in the context of a physio-logically relevant challenge (e.g., mealingestion) (14).

Additional considerations forBCF testinginclude attempts to elicit near-maximalstimulation (15) of insulin secretion andde-termination of insulin secretory reserve.Arginine and glucagon in pharmacologicdoses have been used for these pur-poses, with both stimuli yielding similarresponses, although arginine is bettertolerated (16). The response to eithersecretagogue is potentiated by simulta-neous administration of glucose. The argi-nine stimulation test (AST) has been usedin islet autotransplantation studies inwhich the insulin secretory response to ar-ginine correlateswith the number of trans-planted islets in this patient population(17,18). Although these various testshave been used to detect and quan-tify phenotypic differences across healthand disease, their application to evaluatepharmacologic interventions has beendemonstrated in a limited number of stud-ies (19–21). In particular, there is scantinformation on their variability and repro-ducibility characteristics.

Following review and discussion ofeach method, the BCPT selected theMMTT (with minimal model calcula-tions) and AST for further study. The se-lection was based on both scientificand practical reasons. The MMTT offersa simple-to-administer physiologic test

that incorporates the incretin response.When analyzed with the minimal model,the MMTT provides a simultaneous esti-mate of insulin secretion and sensitivity.The AST generates a supraphysiologic insu-lin secretory response and, like the MMTT,is far less technically demanding thanmethodologies such as the hyperglycemicclamp. The BCPT also chose to include theFSIGT as a widely accepted comparator toallow for a reference method, especially asthe FSIGT and MMTT both use minimalmodel approaches for data analysis.

The methodologies for the MMTT,AST, and FSIGT were first standardized.Subsequently, experiments to estimatebetween and within subject variability(and reproducibility) as well as contrastthe means and distributions of BCF pa-rameters across the glucose tolerancespectrum were undertaken. To mini-mize differences in body compositionfor comparisons across glucose toler-ance groups, all subjects were requiredto be obese, including those with nor-mal glucose tolerance (NGT).

RESEARCH DESIGN AND METHODS

SubjectsThree groups of subjects classified bytheir fasting and postchallenge glucosetolerance status (2-h post–75-g OGTT)(and balanced for gender) were stud-ied. NGT subjects had a fasting glu-cose ,100 mg/dL and postchallenge,140 mg/dL; prediabetes mellitus (PDM)subjects had impaired fasting glucose($100 and,126mg/dL) and impaired glu-cose tolerance ($140 and ,200 mg/dL);subjects with T2DM had fasting glucosevalues of 126–270 mg/dL and HbA1c 6.5–10.0%onastabledoseofmetforminmono-therapy (500–2,000 mg/day).

Study DesignAfter obtaining Institutional ReviewBoard approval, local advertisementwas used to recruit for trials, conductedat two study sites (ICON Develop-ment Solutions in San Antonio, TX, andOmaha, NE). After written informedconsent was obtained and the subjectswere screened, all subjects underwenteach procedure on separate days. TheMMTT and ASTwere administered twiceand the FSIGT once, grouped into threeseparate visits during which the subjectresided at the research center from theevening before the first test until com-pletion of all of the tests that were part

care.diabetesjournals.org Shankar and Associates 1603

of that visit. During each of the first twovisits, the overall order of procedureswas fixed, with an MMTT on day 1 andAST on day 2, after which the subjectwas discharged. At the third and sepa-rate visit, an FSIGT was performed. Alltesting was completed within 28 dayswith at least 5 days between visits.

ProceduresIn those subjects with T2DM, metfor-min was held on themorning of the pro-cedure. After a 10-h overnight fast, asingle indwelling intravenous catheterwas placed in the forearm for theMMTT, whereas for the AST and FSIGT,

indwelling catheters were placed in bothupper extremities for infusion and sam-ple acquisition, respectively. The proce-dures are briefly described below withextensive detail in the Study OperationsManual (available at http://www.fnih.org/what-we-do/current-research-programs/biomarkers-consortium-beta-cell-project).

MMTT

Following baseline sampling (230, 215,and 0 min), a test meal (470 kcal, ;66%carbohydrate, 18% fat, and 16% pro-tein) composedof one 8-fluid-ounceBoostnutrition supplement drink (Nestle Health

Science) and one PowerBar (Nestle Nutri-tion) was administered. The meal wasconsumed within 10 min, with the barconsumed first and serial sampling for an-alytes performed at 10, 15, 20, 30, 60, 90,120, 180, and 240 min postmeal.

AST

Following baseline sampling (210, 25,and 0 min), an intravenous injection of5 g of arginine (given as 10% arginine HCl[as R-Gene; Pfizer]) was administeredover 60 s followed by serial samplingat 2, 4, 5, 7, and 10 min. Subsequently,glucose levels were raised by a continu-ous infusion of glucose (20% dextrose at

Table 1—Comparison of current methods for measuring BCF

Test Description Advantages Limitations

Hyperglycemicclamp

A variable IV glucose infusion isadministered to maintain theglucose level at a steady state

Provides measures of insulin secretion(first and second phases) and withmodeling insulin action

No GI incretin effectsRequires continuous adjustment

of IV glucoseFrequent blood sampling and

minute-to-minute adjustmentsof glucose infusion rate atbedside are required

Does not require modeling of data forinsulin secretion

Technically challenging toconduct testing

Widely reported and accepted Expertise limited to select centers

Graded glucoseinfusion

IV glucose is administered atprogressively increasing rates(each rate maintained for;40 min)

Provides measure of insulin secretionover a range of glucose levels

No GI incretin effects

Requires frequent bloodsampling

Provides measure of b-cell glucosesensitivity

Not as widely studied and reported ashyperglycemic clamp, especially in thecontext of therapeutic interventions

Data analyses often require expertise inmodel-based methods

FSIGT Rapid IV injection of glucose isfollowed 20 min later by an IVinjection of insulin

Provides insulin secretion and actionduring rapidly changing glucose levels

No GI incretin effect

Requires very frequent bloodsampling

Provides first-phase insulin releasemeasures

Technically challenging to conduct

Insulin action results correlate well withthose from euglycemic clamp

Expertise to conduct limited to selectcenters

Widely used and reported

Requires computer modeling for theoutcome measures, requiringspecialized expertise

With C-peptide modeling, providessecond-phase insulin release

Requires IV administration of insulin

AST IV arginine is administeredfollowed by combinedglucose/arginine infusions

Measures of insulin secretion known tocorrelated with b-cell mass in islettransplant recipients

Mixed effect on incretin response

Frequent blood samplings over ashort period of time arenecessary

Provides a measure of near-maximalinsulin secretion (insulin secretoryreserve)

Requires IV administration of arginineand glucose

Does not inform on insulin action

Glucagonstimulation test

IV glucagon is given twicesequentially (at baselineand after glucose has beeninfused to achieve elevatedglucose)

Robust insulin secretory response similarto that of arginine but throughdifferent mechanism of action

No oral incretin effectRequires IV administration of glucagonDoes not inform on insulin actionSide effects of nausea and vomiting are

common and potentially confounding

MMTT/OGTT Oral meal or glucose solution isingested

Easy to administer Assumptions must be made for rate ofnutrient absorption into systemiccirculationMMTT physiologically highly

relevant, mimicking oralchallenges routinelyencountered daily

Effect of incretins included

Technically challenging to modeloutcome measure of insulin secretionof sensitivity, requiring software andexpert analysisBlood samples taken at specified

intervals up to 5 hpostchallenge

Provides insulin secretion and actionduring changing glucose levels

Lack of standardized test meal

OGTT standardized and simple as singlesubstrate

MMTT with minimal modeling not aswidely reported as the hyperglycemicand euglycemic clamps

Insulin action and secretion resultscorrelate with those fromhyperglycemic and euglycemic clamps

GI, gastrointestinal; IV, intravenous.

1604 Standardized Tests of b-Cell Function in Humans Diabetes Care Volume 39, September 2016

900 mg/min) over 60 min with repeatsampling at 50, 55, and 60 min, followedby a second dose of 5 g of arginine at60 min with sampling at 62, 64, 65, 67,and 70 min.

FSIGT

Following baseline sampling (230,215,and 0 min), a 300 mg/kg glucose boluswas administered with sampling at 2, 4,8, and 19 min. At 20 min, a single dose(0.03 units/kg) of U100 regular humaninsulin (Humulin R; Eli Lilly and Com-pany, Indianapolis, IN) was adminis-tered intravenously with sampling at22, 30, 40, 50, 70, 100, 180, and 240min.

Analyte Assays

Samples were assayed in the Immuno-chemical Core Laboratory, Mayo Clinic(Rochester, MN). Glucose was mea-sured on the Roche Cobas c311 (RocheDiagnostics, Indianapolis, IN) using a hexo-kinase reagent. C-peptide was measuredby a two-site immunometric assay onthe Roche Cobas e411 (Roche Diagnos-tics). Insulin (plasma) was measuredby a two-site immunometric (sandwich)assay using electrochemiluminescencedetection on the Roche Cobas e411(Roche Diagnostics). All intra-assay co-efficients of variation (CVs) were ,3%and interassay CVs ,6%.

Data and Statistical Analyses

MMTT

Baseline glucose, insulin, and C-peptidewere calculated as the average of230-,215-, and 0-min values. Si was esti-mated using the oral glucose minimalmodel (22). b-Cell responsivity index(Vtot), a measure of insulin secretion,was estimated from the individualsubject plasma glucose and C-peptideconcentrations observed during theexperiment using the oral C-peptideminimal model (23) incorporatingC-peptide kinetics as reported by VanCauter et al. (24). Disposition index(DI) was calculated as the product ofSi and Vtot. For the purposes of thisseries, a standardized approach to mod-eling across glucose tolerance popula-tions was used. MMTT analyses wereperformed using Matlab US R2010B(MathWorks, Natick, MA) with codeprovided by C. Cobelli. (also availablethrough The Epsilon Group, Charlottes-ville, VA). No time points were excludedin the derivation of individual subject-level parameters.

FSIGT

Si was calculated by fitting the glucoseprofiles from the intravenous glucosetolerance tests usingMinmodMillennium(Version 6.02; Minmod Inc., Los Angeles,CA). The acute insulin response to glu-cose (AIRg) was calculated as the areaunder the curve (AUC) of insulin con-centration above the average basalvalue, from 0 to 10min after the glucoseinjection. The DI was calculated as theproduct of the average Si and AIRg fromeach experiment.

AST

At basal glucose, the insulin secretoryresponse to arginine at basal glucose(AIRarg) was derived as the mean ofthe three highest insulin values fromminutes 2, 3, 4, and 5 minus baselineinsulin at basal glucose (average of210,25, and 0 min). At elevated glucose,maximal insulin secretory response toarginine under hyperglycemic condi-tions (AIRargMAX) was derived as themean of the three highest insulin valuesat minutes 62, 63, 64, and 65 minusbaseline insulin at elevated glucose(average of 50, 55, and 60 min). Insulinsecretory reserve, ISR, represents thedifference between insulin secretion atelevatedandatbasal glucose (AIRargMAX2AIRarg).

Data Conventions and Handling

Analyses for AST and MMTT were per-formed on natural log-transformed datafor subjects having matched pairs (bothvisits). Final results for transformedparameters were exponentiated andreported on the original scale. TheGrubbs’test for a single outlier was used to assessextreme values, and, if found to be signif-icant at the one-sided, 0.001 significancelevel, a secondary, sensitivity analysis wasperformed excluding these data forre-estimation of variance componentsand intraclass correlation coefficients(ICCs) and presented as secondary anal-yses. All other analyses, including testsfor glucose tolerance population differ-ences, and correlation analyses are pre-sented with all evaluable subjects includingextreme values.

Variance Component Estimation

Between- and within-subject variancecomponent estimates were derivedusing a mixed-effects model, treatinggender as a fixed effect, subjects groupedby gender as a random effect, and visitsas a repeated effect. Initial modeling

allowed for the potential of separatebetween- and within-subject variancesacross genders. Log-likelihood ratio testsat the two-sided 0.05 significance levelwere used to select the most appropriatemodel. As parameterized, the model ac-counts for any gender differences inmean response with simultaneous esti-mation of between- and within-subjectvariance components. Estimates (95%CI) for log-normally distributed datawere reported as geometric CV andmodelpredicted adjusted geometric means(95% CI). Tukey contrasts adjusting formultiplicity were used for comparison ofmeans across glucose tolerance popula-tions at the two-sided 0.05 significancelevel.

Using the estimated variance com-ponents, the ICC was calculated as(s2between/[s2between + s2within]). TheICC is a measure of the degree to whichrepeated measures within the samesubject resemble each other, a measureof relative repeatability (25). ICC val-ues .0.80 were considered highly re-producible. ICC values .0.50 wereconsidered moderately reproducible,whereas those ,0.50 were consideredweakly reproducible. Overall correla-tions across populations were derived,including correlations within each popu-lation to make a general assessment ofthe consistency of results (concordance)within and across populations. Correla-tions among the different parametersare reported as Spearman rank correla-tions. The MMTT and AST values usedwere the subject averages.

RESULTS

The evaluable subjects included 23 NGT(12men/11women), 17 PDM (6men/11women), and 22 T2DM (11 men/11women), for a total of 62 subjects. Com-plete demographic characteristics are sum-marized in Table 2. All groups, includingthose with NGT, were obese. No seriousprocedure-emergent adverse events werereported. A complete summary of proce-dure-emergent adverse events is reportedin Supplementary Table 1.1 and 1.2.

Measured ParametersdGlucose,Insulin, and C-PeptideConcentrationsSummary plasma profiles of glucose, in-sulin, and C-peptide for the MMTT andthe AST are shown in Fig. 1, reflectingthe mean values of both pairs of tests.

care.diabetesjournals.org Shankar and Associates 1605

Additional figures displaying the excur-sion of these analytes during each of thetwo MMTTs, as well as the two ASTs, canbe found in Supplementary Fig. 1.1–1.6(with additional information found inSupplementary Table 2.1–2.3). As ex-pected, fasting glucose rose progres-sively across NGT, PDM, and T2DM.Fasting insulin and C-peptide werecomparable in NGT and T2DM andhigher in PDM. In the MMTT, follow-ing the meal challenge, glucose rose pro-gressively across populations on bothtest days, achieving peak levels inT2DM . PDM . NGT, whereas insulinand C-peptide responses were highest inPDM, followed by NGT and then T2DM.AUC for glucose, insulin, and C-peptideall differed among the three groups(Supplementary Table 2.2).In the AST, in response to arginine,

increases over baseline in insulin andC-peptide were observed in all groupson both test days during basal andhyperglycemic conditions. The insulinsecretory response to arginine wasblunted in those with T2DM comparedwith PDM and NGT. In the FSIGT, follow-ing the glucose bolus, glucose rose pro-gressively over baseline in all groups,achieving peak levels greater in T2DMthan in PDM or NGT (Supplementary Fig.1.7–1.9 and Supplementary Table 2.4). En-dogenous insulin andC-peptide responseswithin the first 19 min were blunted inT2DM. Following exogenous insulin, re-turn to baseline for glucose occurredwithin 50 min in NGT, but was more de-layed in PDM and T2DM.

Indices of BCF From the MMTT, AST,and FSIGTOverall, insulin secretion, as measuredby either the MMTT or AST, was similar

between the NGT and PDM groups, yetboth differed significantly from theT2DM subjects. Fig. 2 summarizes thederived responses to the MMTT andAST, with the actual values presentedin Table 3. In the MMTT, b-cell res-ponsivity (Vtot, an index of insulinsecretion) in the T2DM group was86 and 87% lower compared with NGTand PDM (P , 0.001), respectively. Al-though Vtot was numerically higher inPDM compared with NGT, the differ-ence was not statistically different. Sidecreased progressively across glucosetolerance populations, with T2DM being78 and 52% lower compared with NGTand PDM (P , 0.001), respectively,whereas PDM was 55% lower comparedwith NGT (P , 0.01). Correspondingly,DI was 51% lower in the PDM groupcompared with NGT (P , 0.001) andwas 93% lower in T2DM comparedwith PDM (P , 0.001).

In 3 of the 44 (;7%) individual subjectvisits in the MMTT in the T2DM popula-tion, the standardized analytical ap-proach yielded near-zero values for Si.As predefined in the analysis plan, the Siand DI data from these subjects werenot included in the primary analyses.An alternate, slightly modified approachto the minimal model as described byBasu et al. (26) (see SupplementaryData) that allowed for inclusion of Sidata from these three subjects did notyield values for the point estimates(geometric means and 95% CI) of Si orDI (with these three subjects, Si = 1.2[0.9–1.6]; DI = 19 [13–38]) that weredifferent from the original analysis(without these three subjects, Si = 1.2[0.9–1.6] and DI = 21 [14–31]). Thiswas true for the variance componentestimates as well.

In the AST, AIRarg, AIRargMAX, and,most notably, the ISR were all lower(P , 0.001) in T2DM compared withNGT and PDM (58 and 63%, 79 and78%, and 85 and 84% lower, respec-tively, for NGT and PDM). No significantdifferences were observed betweenNGT and PDM.

In the FSIGT, AIRg progressively andsignificantly decreased across glu-cose tolerance populations with themean value for T2DM being nearlyzero (Fig. 2). AIRg was statistically sepa-rable among the three glucose toler-ance groups. Likewise, Si decreasedprogressively across glucose tolerancepopulations, although the differencebetween PDM and T2DM did not reachstatistical significance. DI decreasedsignificantly and progressively acrossgroups.

Between- and Within-SubjectVariability and Reproducibility forMeasured Parameters and IndicesFrom MMTT and ASTGlucose, insulin, and C-peptide re-sponses within the MMTT and AST foreach population had good reproduc-ibility (Supplementary Fig. 1.1–1.6 andSupplementary Table 2.1–2.3). TheAUCs (0–4 h) for glucose, insulin, andC-peptide from the MMTT generally dis-played moderate to high reproduc-ibility (as per ICC values), with thesole exception of glucose in the NGT. Re-producibility, as indexed by the ICC,ranged from weak to strong in theMMTT for all model-based parametersin all populations (Table 3). The inclusionof the three additional subjects withT2DM whose Si was numerically noni-dentifiable did not affect that conclusion(not shown). For the AST, reproducibility

Table 2—Summary demographics and anthropometrics for evaluable subjects by glucose tolerance population

Parameter/population NGT PDM T2DM

Number of evaluable (paired)observations [N (men/women)] 23 (12 men/11 women) 17 (6 men/11 women) 22 (11 men/11 women)

Age (years 6 SD) 41.9 6 8.4 44.8 6 9.8 54.7 6 8.1

Weight (kg 6 SD) 85.8 6 14.3 97.4 6 12.9 91.1 6 14.1

BMI (kg/m2 6 SD) 31.5 6 2.8 35.0 6 3.8 32.8 6 3.9

Ethnicity distributionWhite, not Hispanic or Latino 5 3 3White, Hispanic, or Latino 14 10 16Black, not Hispanic 4 3 2Other 1 1

HbA1c 5.7 6 0.38% (39 mmol/mol) 8.28 6 0.79% (67 mmol/mol)

For the PDM population, there were 16 evaluable subjects for the AST.

1606 Standardized Tests of b-Cell Function in Humans Diabetes Care Volume 39, September 2016

was strong across all parameters/popu-lations (all values .0.8).Four extreme values from one of the

two visits for two subjects (Ftot and DI

in NGT and Si and DI in T2DM) for MMTTparameters were flagged during outlieranalyses (P , 0.001). ICCs excludingthese points as secondary, sensitivity

analyses are provided in the legend forTable 3. When these outliers were ex-cluded, reproducibility for the MMTT(per ICC values) rose considerably.

Figure 1—Mean6 SEM glucose, insulin, and C-peptide concentration time course profiles by stimulation test and glucose tolerance group. Squares,NGT; circles, PDM; triangles, T2DM.

care.diabetesjournals.org Shankar and Associates 1607

Figure 2—Indices of BCF. Model predicted geometric means (95% CI) for AST and MMTT and arithmetic means (95% CI) for FSIGT. Note: lower CIsuppressed for graphical purposes for FSIGT AIRg and DI in T2DM. GT, glucose tolerance. *P , 0.05, **P , 0.01, ***P , 0.001.

1608 Standardized Tests of b-Cell Function in Humans Diabetes Care Volume 39, September 2016

Comparative Assessment of Model-Based Indices of BCF Across Tests andMetabolic SpectrumCorrelation analyses were undertakento better understand how these differ-ent measures tracked with one anotherwithin the same subject populations andacross populations. Within the AST, theoverall correlation between AIRarg andAIRargMAX was notably high across(0.923) and within (0.794–0.930) allpopulations (Table 4), indicating con-cordance in results across populations.In addition, the overall correlation be-tween the AST-derived measures (AIRargand AIRargMAX) and Vtot from theMMTT was high (0.858) and statisti-cally significant across and within allpopulations. Overall correlations acrossglucose tolerance between AIRg fromthe FSIGT and Vtot from the MMTT aswell as that between AIRg and AIRargMAXwere high (0.753 and 0.826), whereaswithin-population correlation was less so,especially in T2DM. Si estimated in theMMTT and FSIGT showed a high overallcorrelation (0.695), consistent within thethree populations. The overall correlationbetween DI from the MMTT and FSIGT

was high (0.779), whereas no relationshipfor this parameter was observed withingroups.

CONCLUSIONS

The current series of studies was un-dertaken to characterize the responsesto and reproducibility of a standard-ized MMTT and AST for assessment ofBCF in subjects with NGT, PDM, andT2DM. We report that the MMTT andAST are able to detect differences inBCF across the metabolic spectrum.The reproducibility of the MMTT is, ingeneral, moderate (ranging from weakto strong), depending on parameterand population. The reproducibility ofthe AST is very strong across all popu-lations. Importantly, for both MMTTand AST, the observed variability pre-dicts reasonably sized clinical studiesto detect clinically meaningful changesin insulin secretion. The MMTT- andAST-based measures of BCF are direc-tionally and proportionally concordantand generally concur with indices de-rived from the reference FSIGT. Itshould be noted that despite thesetests having been in use for quite

some time, this is the first report ofwithin- and between-subject variabil-ity for outcome parameters from thestandardized MMTT and AST, espe-cially in subjects across the metabolicspectrum. These data should be ofvalue for the computation of samplesize in interventional studies in relevantpopulations.

The standardized test meal used inthe current series was able to elicit re-sponses in glucose, insulin, and C-peptideduring the MMTT that were similar toprior reports in NGT, PDM, and T2DM(27,28). Our findings for between-groupdifferences for model-based estimates ofBCF (Ftot) in those with and without di-abetes are generally consistent withthose reported by Bock et al. (27) andFerrannini et al. (29), as well as with othermethods such as the graded glucoseinfusion (30), OGTT (29), and hypergly-cemic clamp (31). The significant pro-gressive decrease in Si from NGT toT2DM is concordant with prior reportsusing clamp-based assessments (29).Notably, the DI, a measure of the ap-propriateness of insulin secretion toprevailing levels of insulin action,

Table 3—Variability and reproducibility metrics: geometric means, between- and within-subject geometric CVs, and ICCs forMMTT and AST and arithmetic means and total CVs for FSIGT

Stimulation test parameter Population (N)

Model predictedgeometric mean

(95% CI)

Geometric CV%between subjects

(90% CI)

Geometric CV%within subjects

(90% CI) ICC (90% CI)

MMTT: Ftot (1029min21) NGT (N = 23) 104 (83–132) 27.1 (14.7–51.6) 41.4 (32–54.1) 0.31* (0.30–0.77)

PDM (N = 17) 115 (90–147) 34.8 (24.1–51.1) 19.6 (14.7– 26.2) 0.753 (0.70–0.9)T2DM (N = 22) 15 (12–20) 57.8 (41.8–82.1) 26.1 (20.2–33.8) 0.814 (0.69–0.96)

MMTT: Si (1024min21 [mU/mL]21) NGT (N = 23) 5.5 (4.1–7.3) 48.2 (34.1–69.8) 32.6 (25.4–42.3) 0.674 (0.521–0.874)

PDM (N = 17) 2.5 (1.8–3.3) 44 (28.4–70.5) 35.5 (26.5–48.2) 0.598 (0.469–0.837)T2DM (N = 19) 1.2 (0.9–1.6) 53.5 (26.9–121.5) 83.2 (60.1–120.6) 0.323† (0.034–0.772)

MMTT: DI (10213min22 [mU/mL]21) NGT (N = 23) 585 (405–845) 42.3 (22.7–84.9) 66.1 (50–89.6) 0.312‡ (0.264–0.769)

PDM (N = 17) 289 (197–424) 54.4 (35.3–87.9) 40.3 (29.9–55) 0.633 (0.543–0.852)T2DM (N = 19) 21 (14–31) 67.3 (35.1–152.2) 97.1 (69–145) 0.36§ (0.254–0.754)

AST: AIRarg (mU/mL) NGT (N = 23) 84 (67–106) 38.9 (29.4–52.1) 11.6 (9.1–14.8) 0.914 (0.902–0.961)

PDM (N = 16) 94 (73–120) 38.9 (27.6–55.7) 11.5 (8.6–15.4) 0.915 (0.874–0.972)T2DM (N = 22) 35 (28–45) 53.7 (39.1–75.4) 22.8 (17.7–29.5) 0.833 (0.734–0.966)

AST: AIRargMAX (mU/mL) NGT (N = 23) 391 (301–509) 57.5 (42.8–79.2) 16.6 (13–21.2) 0.914 (0.858–0.967)

PDM (N = 16) 389 (293–517) 31.2 (22.1–44.5) 11.2 (8.3–15) 0.882 (0.815–0.969)T2DM (N = 22) 84 (65–110) 63.1 (46.6–87.9) 13.8 (10.7–17.7) 0.947 (0.931–0.982)

AST: ISR (mU/mL) NGT (N = 23) 306 (226–414) 64.1 (47.2–89.6) 20.1 (15.7–25.7) 0.897 (0.83–0.96)

PDM (N = 16) 293 (212–406) 34.4 (24.3–49.3) 12.4 (9.2–16.6) 0.88 (0.759–0.976)T2DM (N = 22) 47 (35–64) 79.3 (57.3–114.5) 19.6 (15.3–25.3) 0.928 (0.916–0.968)

FSIGT: no within-subject variability estimable, total variance estimated, and analyses on arithmetic scale.

c Mean (95% CI) CV% AIRg ([mU/mL]) 3 min): NGT, 915 (708–1,122), 81%; PDM, 489 (241–737), 77%; and T2DM, 10 (2201 to 222), 413%.

c Mean (95% CI) CV% Si (1024min21 [mU/mL]21): NGT, 1.9 (1.5–2.3), 60%; PDM, 1.2 (0.7–1.6), 37%; and T2DM, 0.8 (0.4–1.2) 101%.

c Mean (95% CI) CV% DI (1024): NGT, 1,339 (1,093–1,585), 64%; PDM, 576 (282–871), 90%; and T2DM, 9 (2242 to 261), 557%.

Exclusion of single-visit extreme value produces good reproducibility results for MMTT across all GT populations: *ICC = 0.59 (0.551–0.801) excluding single-visitextremevalue (P,0.001) fromNGT (N=22);†ICC=0.559 (0.387–0.795)excluding single-visit extremevalue (P,0.001) fromT2DM(N=18);‡0.527 (0.476–0.797)excludingsingle-visit extremevalue (P,0.001) fromT2DM(N=18); §ICC=0.527 (0.476–0.797)excluding single-visit extremevalue (P,0.001) fromT2DM(N=18).

care.diabetesjournals.org Shankar and Associates 1609

decreased across populations, indicat-ing that the methodology was able todetect a progressive decrease in BCF.Together, these observations suggestthat the MMTT used in this seriesrecapitulated prior reports of BCF andinsulin action measured using differenttests.The AST was able to elicit a b-cell re-

sponse at baseline glycemia (AIRarg)and hyperglycemia (AIRmax) in all threepopulations. The responses were similarto those previously reported separatelyin each population (32,33). AIRarg, AIR-max, and ISR were highly reproducibleacross populations. The ability of b-cellsto respond to arginine appears to bepreserved in prediabetes.An approach to better understand the

utility of the reproducibility metrics ofthe MMTT and AST is to compute thesample sizes required to detect prede-termined differences using parametersderived from the current trial. For exam-ple, in the MMTT, to detect a 25% in-crease in Ftot, 18 subjects per groupare required for 80% power at a one-sided, 0.05 significance level. Althoughthe variability observed in Si and DI isgreater than with Ftot, in the currentstudy, the MMTT detected differencesacross groups in Si and DI with modestnumbers of subjects (;20) per group. Itshould be noted that the weaker repro-ducibility observed in some cases wasdriven by a single outlier, as shown bythe sensitivity analyses. For the AST, fivesubjects per group are needed at thesame power and significance level.Thus, both tests exhibit variances that

permit reasonably sized clinical trials inrelevant populations.

The glycemic excursions and insulinsecretory responses in the FSIGT de-tected differences in insulin secretionacross the metabolic spectrum, whichrecapitulated previous experimental re-sults (10,34,35). An interestingobservationin the current series was that althoughVtot in the MMTT did not differ betweenNGT and PDM, AIRg in the FSIGT was de-creased in PDM subjects compared withthe NGT subjects. There are two salientpoints to recognize, however. First, previ-ously published studies have generallymade similar observations regarding theMMTT. Previous work from Bock et al.(27) and Ferrannini et al. (29) reportedthat insulin secretion in subjects with IGTis similar to that of obese subjects withNGT. Distinctions in insulin secretion aredetectable, in general, only when con-trasted to lean subjects with NGT (29).The subjects with NGT in the current serieshad a BMI that was close to those of thePDM group and had high normal fastingplasma glucose and insulin levels; theylikely had overlap of some aspects of BCFwith PDM.

Second, it should be noted that al-though the MMTT and the FSIGT aredifferent tests, they arrived at thesame conclusiondi.e., a progressivedecrease in DI from NGT to PDM toT2DM. Ultimately, the DI provided bythe minimal model yields an inte-grated estimate of insulin secretion inthe context of insulin action. In thecontext of a meal, enteral delivery ofsubstrate to subjects with PDM elicits

an increase in insulin secretion that, onan absolute basis, is similar to that ofobese subjects with NGT. However, in-sulin secretion is inadequate for theprevailing insulin action, as quantifiedby the DI and as expected by inspec-tion of the glucose profiles in eachgroup. The FSIGT, in contrast to theMMTT’s mixed substrate delivery andelicitation of incretin (and other) re-sponses, uses intravenous glucose asthe sole stimulus that results in alesser insulin secretory response inPDM than NGT subjects. Consistentwith prior reports, the FSIGT in the cur-rent series was able to detect differ-ences in acute insulin secretion inresponse to glucose across the meta-bolic spectrum (10,34,35). At the sametime, the integrated parameter repre-sented in DI showed a progressive de-crease fromNGT to PDM to T2DM. Thus,with some differences noted, the MMTTand the FSIGT provided the same overallconclusion.

Given the fairly unique circumstanceof having access to data from multipletests of BCF in the same subjects andacross the glucose tolerance spectrum,it was informative to ask how the mea-sures of BCF related to one another. Sev-eral key findings emerged across testsandwithin and across populations. In gen-eral, correlation across all three glucosetolerance populations was high for everycomparison, and comparisons of parame-ters of BCF obtained with the FSIGT andMMTT showed correlation characteristicsconsistent with previous reports (36).However, in some instances, within each

Table 4—Spearman correlations and significance tests within and across stimulation tests for key parameters

Parameter associationAcross GTpopulations

Within NGTpopulation

Within PDMpopulation

Within T2DMpopulation

MMTT: Ftot (1029min21) vs. AST:AIRargMAX (mU/mL) 0.858*** 0.492* 0.638** 0.799***

AST: AIRarg (mU/mL) vs. AST: AIRargMAX (mU/mL) 0.923*** 0.914*** 0.794*** 0.930***

MMTT: Ftot (1029min21) vs. AST: ISR (mU/mL) 0.853*** 0.482* 0.553* 0.744***

MMTT: Ftot (1029min21) vs. FSIGT: AIRg([mU/mL] 3 min) 0.753*** 0.394# 0.279 (NS) 0.010 (NS)

AST: AIRargMAX (mU/mL) vs. FSIGT: AIRg([mU/mL] 3 min) 0.826*** 0.779*** 0.379 (NS) 0.082 (NS)

MMTT: Si (1024min21 [mU/mL]21) vs.FSIGT: Si (1024min21 [mU/mL]21) 0.695*** 0.827*** 0.650** 0.486*

MMTT: DI (10213min22 [mU/mL]21) vs.FSIGT: DI (1024) 0.779*** 0.179 (NS) 0.085 (NS) 20.284 (NS)

AST: ISR (mU/mL) vs. FSIGT: AIRg ([mU/mL] 3 min) 0.854*** 0.824*** 0.476# 0.156 (NS)

GT, glucose tolerance. *P , 0.05; **P , 0.01; ***P , 0.001; #P , 0.1; NS, P $ 0.1.

1610 Standardized Tests of b-Cell Function in Humans Diabetes Care Volume 39, September 2016

population, the associations were lessprominent. This suggests that some asso-ciations may in part be due to significantdifferences between populations and lessso to a correlation of the tests, per se,within a subject. A notable examplewould be comparisons between Ftot(MMTT) and AIRg (FSIGT), as well as be-tween AIRarg/AIRmax (AST) and AIRg. Al-though these showed reasonable overallcorrelations across glucose tolerancegroups, they were weaker within eachgroup, especially within T2DM. Taken to-gether, the above observations suggestthat although the various indices fromeach of these tests likely quantify differ-ent facets of BCF, there is concordance inthe directionality and proportionalityacross tests (29).Operationally, the MMTT is simple

enough to routinely perform at mostclinical research sites and centers.The test meal used in this series had abalanced macronutrient composition,with solid and liquid components, andis readily available from a manufac-turer, assuring consistency of the testmeal across sites. For similar opera-tional reasons, the AST can be per-formed at most clinical research sites.Some procedural choices weremade tosimplify the testing, including samplingof nonarterialized venous blood (i.e.,no hot hand). We recognize that theuse of the hot hand method may haveimproved reproducibility. Our ap-proach, however, is similar to previousstudies that did not obtain arterializedsamples for assessment of insulin se-cretion and sensitivity (29,37) andavoids potential issues due to the tech-nique (38,39).Analytically, in contrast to the AST,

the MMTT and FSIGT require model-based analyses and occasionally needoperator intervention to reconcile near-zero or negative values, respectively, forsome parameters (35,40,41). In the cur-rent series, we opted for a more stan-dardized analysis strategy that led tononidentifiable Si parameters for MMTTin three subject visits in the T2DM popu-lation. Various methods can mitigatesuch model-related vulnerabilities (42,43),although inclusion of these three sub-jects did not change the conclusions.Furthermore, a novel method for assessingparameter reproducibility applied to mini-mal model data supports the reliability ofthe MMTT-derived indices (44). Although

this series used the Cobelli oral minimalmodel for MMTT (12), it is important torecognize that there are other methodsfor estimating BCF from MMTT or OGTTdata (37).

In summary, the current series of ex-periments shows that the variabilityand reproducibility characteristics ofthe MMTT and AST are sufficient to de-tect the types of changes in measuredand modeled parameters of BCF thatare generally expected in response totherapeutic interventions in relevantpatient populations. Although thechoice of testing methodology will ul-timately be determined by the scien-tific question, the standardized MMTTand AST can provide reliable, repro-ducible, and complementary measuresof BCF, with characteristics favorablefor use in large, longitudinal interven-tional studies. Further studies arerequired to evaluate these tests’ per-formance characteristics in responseto a specific pharmacotherapy or life-style intervention.

Acknowledgments. The authors thank Jes-sica Ratay and Cheryl Melencio for the con-tinuous support in planning, executing, andmanaging these studies and the overall proj-ect; Kathryn Wright for excellent manage-ment of the many samples generated in thisstudy; Clay Dehn and Angelica Guerrero forcontributions to the study operations manualand operationalizing the methodologies;Hilary Blair and the staff of the Mayo ClinicEndocrine Laboratory; and the research anddata management staff at ICON DevelopmentSolutions.Funding. The methodological study describedin this report was designed and implementedunder the auspices of the FNIH BiomarkersConsortium. The Biomarkers Consortium is apublic private partnership that develops andvalidates biological markers, which speed upthe development of therapies for the detection,prevention, diagnosis, and treatment of diseaseand are ultimately aimed at improving patientcare. For this study, the Consortium broughttogether diabetes experts from leading aca-demic institutions, the U.S. Food and DrugAdministration (FDA), the National Institutesof Health (NIH), the nonprofit sector, and thepharmaceutical industry to develop the project.The results of the partnership, discussed in thisstudy, are important because they address acritical unmet medical need and involve keystakeholders in the diabetes treatment field.FNIH has acted as a neutral convener for thepartners and provided the project manage-ment expertise needed for the execution of theoverall project. In the future, this type ofpartnership can be used as a model for estab-lishing common standards for testing in othertherapeutic areas as well. Under the auspices

of the FNIH, this project was jointly funded bythe NIH/National Institute of Diabetes andDigestive and Kidney Diseases (NIDDK) and theFDA and the following participating pharma-ceutical companies: Amylin (now AstraZeneca),Janssen, Eli Lilly and Company, Merck, Novartis,Novo Nordisk, Pfizer, Sanofi, and Takeda. Anin-kind donation of P800 tubes was made byBecton Dickinson (D. Craft). Additional supportwas also received from the American DiabetesAssociation and JDRF.Duality of Interest. S.S.S. is an employee andshareholder of Eli Lilly and Company. R.H.R. is ashareholder of Bristol-Myers Squibb. R.A.C.,D.C., J.Q.D., and D.S.L. are employees andshareholders of Pfizer. R.N.B. is responsiblefor the MinMod program, which is also mar-keted. C.Ca. is an employee and shareholder ofTakeda Pharmaceuticals. M.D. is an employeeand shareholder of Eli Lilly and Company and amember of the American Diabetes Associationgrant review panel. D.P. is an employee andshareholder of Johnson & Johnson. H.R. is anemployee and shareholder of Sanofi. D.A.F. is ashareholder of Pfizer. No other potential con-flicts of interest relevant to this article werereported.Author Contributions. S.S.S., A.V., and R.H.R.designed the experiment, researched data,wrote, edited, and reviewed the manuscript,and coordinated overall integration of themanuscript. M.A.S., R.A.C., R.N.B., C.Ca.,D.C., C.Co., M.D., D.S.L., D.P., R.P.R., H.R., D.S.,M.T.V., and G.C.W. designed the experiment,researched data, and wrote, edited, and re-viewed the manuscript. C.D.M. and J.Q.D. re-searched data and reviewed the manuscript.D.A.F. designed the experiment, researcheddata, oversaw study execution, edited andreviewed the manuscript, and coordinatedoverall integration of the manuscript. D.A.F. isthe guarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.Prior Presentation. Parts of this study werepresented in abstract form at the 73rd ScientificSessions of the American Diabetes Association,Chicago, IL, 21–25 June 2013, and at the 29thAnnual Meeting of the European Associationfor the Study of Diabetes, Barcelona, Spain,23–27 September 2013.

Appendix

Current Members of the b-Cell ProjectTeam. Richard Bergman, PhD, Cedars-Sinai Di-abetes & Obesity Research Institute; RobertoCalle, MD, Pfizer; Charlie Cao, PhD, Takeda De-velopment Center Americas; Danny Chen, PhD,Pfizer; Claudio Cobelli, PhD, University of Pa-dova; Mark Farmen, PhD, Eli Lilly and Company;David Fryburg, MD, FNIH and Roi BiopharmaConsulting; Atalanta Gosh, PhD, Janssen; IlanIrony, MD, Center for Drug Evaluation and Re-search/FDA; David Kelley, MD, Merck; DouglasLee, PhD, Pfizer; Frank Martin, PhD, JDRF;Henriette Mersebach, MD, Novo Nordisk; LoriMixson, PhD, Merck; Stephanie Moran, MD,Takeda Development Center Americas; DavidPolidori, PhD, Janssen; Jessica Ratay, MS, FNIH;Ralph Raymond, MS, FNIH and R-Squared Solu-tions; R. Paul Robertson, MD, Pacific Northwest

care.diabetesjournals.org Shankar and Associates 1611

Diabetes Research Institute and University ofWashington; Hartmut Ruetten, MD, PhD, Sanofi;Peter Savage, MD, NIH/National Institute of Dia-betes and Digestive and Kidney Diseases; SudhaShankar, MD, Eli Lilly and Company; MyrleneStaten, MD, Kelly Government Solutions on con-tract to NIH/NIDDK; Darko Stefanovski, PhD,University of Pennsylvania; Maria Vassileva,PhD, FNIH; Adrian Vella,MD,Mayo Clinic; GordonWeir, MD, Joslin Diabetes Center; and MarjorieZakaria, MD, Novartis.Previous Members of the b-Cell ProjectTeam Who Contributed to This Work.Richard Chen, PhD, formerly of NIDDK/NIH; MarkDeeg, MD, Eli Lilly and Company; Ying Ding, PhD,formerly of Eli Lilly and Company; ChristianDjurhuus, MD, PhD, Novo Nordisk; Cong Han,PhD, Takeda; David Maggs, MD, formerly ofAmylin; Mads Rasmussen, MD, PhD, Novo Nor-disk; Thomas Strack, MD, formerly of Takeda;Krystyna Tatarkiewicz, PhD, formerly of Amylin;and Adrianne Wong, formerly of JDRF.

References1. Sathananthan A, Dalla Man C, ZinsmeisterAR, et al. A concerted decline in insulin secretionand action occurs across the spectrum of fastingand postchallenge glucose concentrations. ClinEndocrinol (Oxf) 2012;76:212–2192. Bergman RN. Lilly lecture 1989. Towardphysiological understanding of glucose toler-ance. Minimal-model approach. Diabetes1989;38:1512–15273. Knowler WC, Barrett-Connor E, Fowler SE,et al.; Diabetes Prevention Program ResearchGroup. Reduction in the incidence of type 2 di-abetes with lifestyle intervention or metformin.N Engl J Med 2002;346:393–4034. Cersosimo E, Solis-Herrera C, TrautmannME,Malloy J, Triplitt CL. Assessment of pancreaticb-cell function: review of methods and clinicalapplications. Curr Diabetes Rev 2014;10:2–425. Marena S,MontegrossoG, DeMichieli F, PisuE, Pagano G. Comparison of the metabolic ef-fects of mixed meal and standard oral glucosetolerance test on glucose, insulin and C-peptideresponse in healthy, impaired glucose toler-ance, mild and severe non-insulin-dependentdiabetic subjects. Acta Diabetol 1992;29:29–336. Muniyappa R, Lee S, Chen H, Quon MJ. Cur-rent approaches for assessing insulin sensitivityand resistance in vivo: advantages, limitations,and appropriate usage. Am J Physiol EndocrinolMetab 2008;294:E15–E267. Pisprasert V, Ingram KH, Lopez-Davila MF,Munoz AJ, Garvey WT. Limitations in the useof indices using glucose and insulin levels topredict insulin sensitivity: impact of race andgender and superiority of the indices derivedfrom oral glucose tolerance test in AfricanAmericans. Diabetes Care 2013;36:845–8538. Selimoglu H, Duran C, Kiyici S, et al. Compar-ison of composite whole body insulin sensitivityindex derived from mixed meal test and oralglucose tolerance test in insulin resistant obesesubjects. Endocrine 2009;36:299–3049. Wolever TM, Chiasson JL, Csima A, et al. Var-iation of postprandial plasma glucose, palatabil-ity, and symptoms associated with a standardizedmixed testmeal versus 75 g oral glucose. DiabetesCare 1998;21:336–340

10. Bergman RN, Phillips LS, Cobelli C. Physio-logic evaluation of factors controlling glucosetolerance in man: measurement of insulin sen-sitivity and beta-cell glucose sensitivity from theresponse to intravenous glucose. J Clin Invest1981;68:1456–146711. Cobelli C, Dalla Man C, Toffolo G, Basu R,Vella A, Rizza R. The oral minimal modelmethod. Diabetes 2014;63:1203–121312. Nesher R, Cerasi E. Modeling phasic insulinrelease: immediate and time-dependent effectsof glucose. Diabetes 2002;51(Suppl. 1):S53–S5913. Caumo A, Luzi L. First-phase insulin secre-tion: does it exist in real life? Considerations onshape and function. Am J Physiol EndocrinolMetab 2004;287:E371–E38514. Xiang AH, Watanabe RM, Buchanan TA.HOMA and Matsuda indices of insulin sensitiv-ity: poor correlation with minimal model-basedestimates of insulin sensitivity in longitudinalsettings. Diabetologia 2014;57:334–33815. Toschi E, Camastra S, Sironi AM, et al. Effectof acute hyperglycemia on insulin secretion inhumans. Diabetes 2002;51(Suppl. 1):S130–S13316. Robertson RP, Raymond RH, Lee DS, et al.;Beta Cell Project Team of the Foundation for theNIH Biomarkers Consortium. Arginine is pre-ferred to glucagon for stimulation testing ofb-cell function. Am J Physiol Endocrinol Metab2014;307:E720–E72717. Robertson RP, Bogachus LD, Oseid E, et al.Assessment of b-cell mass and a- and b-cellsurvival and function by arginine stimulationin human autologous islet recipients. Diabetes2015;64:565–57218. Teuscher AU, Kendall DM, Smets YF, Leone JP,Sutherland DE, Robertson RP. Successfulislet autotransplantation in humans: functional in-sulin secretory reserve as an estimate of survivingislet cell mass. Diabetes 1998;47:324–33019. Derosa G, Carbone A, Franzetti I, et al. Ef-fects of a combination of sitagliptin plus met-formin vs metformin monotherapy on glycemiccontrol, b-cell function and insulin resistance intype 2 diabetic patients. Diabetes Res Clin Pract2012;98:51–6020. Derosa G, Franzetti IG, Querci F, et al. Exena-tide plus metformin compared with metforminalone on b-cell function in patients with Type 2diabetes. Diabet Med 2012;29:1515–152321. Derosa G, Ragonesi PD, Carbone A, et al.Vildagliptin added to metformin on b-cell func-tion after a euglycemic hyperinsulinemic andhyperglycemic clamp in type 2 diabetes pa-tients. Diabetes Technol Ther 2012;14:475–48422. Dalla Man C, Caumo A, Cobelli C. The oralglucose minimal model: estimation of insulinsensitivity from a meal test. IEEE Trans BiomedEng 2002;49:419–42923. Breda E, Cavaghan MK, Toffolo G, PolonskyKS, Cobelli C. Oral glucose tolerance test mini-mal model indexes of beta-cell function and in-sulin sensitivity. Diabetes 2001;50:150–15824. Van Cauter E, Mestrez F, Sturis J, PolonskyKS. Estimation of insulin secretion rates fromC-peptide levels. Comparison of individual andstandard kinetic parameters for C-peptide clear-ance. Diabetes 1992;41:368–37725. Koch GG, Amara IA, Davis GW, Gillings DB. Areview of some statistical methods for covari-ance analysis of categorical data. Biometrics1982;38:563–595

26. Basu A, Dalla Man C, Basu R, Toffolo G,Cobelli C, Rizza RA. Effects of type 2 diabeteson insulin secretion, insulin action, glucose ef-fectiveness, and postprandial glucose metabo-lism. Diabetes Care 2009;32:866–87227. Bock G, Dalla Man C, Campioni M, et al.Pathogenesis of pre-diabetes: mechanisms offasting and postprandial hyperglycemia in peo-ple with impaired fasting glucose and/or im-paired glucose tolerance. Diabetes 2006;55:3536–354928. Smushkin G, Sathananthan M, Piccinini F,et al. The effect of a bile acid sequestrant onglucose metabolism in subjects with type 2 di-abetes. Diabetes 2013;62:1094–110129. Ferrannini E, Gastaldelli A, Miyazaki Y,Matsuda M, Mari A, DeFronzo RA. beta-Cellfunction in subjects spanning the range fromnormal glucose tolerance to overt diabetes: anew analysis. J Clin Endocrinol Metab 2005;90:493–50030. Kjems LL, Holst JJ, Vølund A, Madsbad S.The influence of GLP-1 on glucose-stimulatedinsulin secretion: effects on beta-cell sensitivityin type 2 and nondiabetic subjects. Diabetes2003;52:380–38631. Sjaarda LG, Bacha F, Lee S, Tfayli H,Andreatta E, Arslanian S. Oral disposition indexin obese youth from normal to prediabetes todiabetes: relationship to clamp disposition in-dex. J Pediatr 2012;161:51–5732. Ward WK, Bolgiano DC, McKnight B, HalterJB, Porte D Jr. Diminished B cell secretory capac-ity in patients with noninsulin-dependentdiabetes mellitus. J Clin Invest 1984;74:1318–132833. van der Zijl NJ, Goossens GH, Moors CC,et al. Ectopic fat storage in the pancreas, liver,and abdominal fat depots: impact on b-cellfunction in individuals with impaired glucosemetabolism. J Clin Endocrinol Metab 2011;96:459–46734. Buchanan TA, Xiang AH, Peters RK, et al.Preservation of pancreatic beta-cell functionand prevention of type 2 diabetes by pharma-cological treatment of insulin resistance inhigh-risk hispanic women. Diabetes 2002;51:2796–280335. Mayer-Davis EJ, Monaco JH, Hoen HM,et al.; The Insulin Resistance AtherosclerosisStudy (IRAS). Dietary fat and insulin sensitivityin a triethnic population: the role of obesity. AmJ Clin Nutr 1997;65:79–8736. Cobelli C, Toffolo GM, Dalla Man C, et al.Assessment of beta-cell function in humans, si-multaneously with insulin sensitivity and he-patic extraction, from intravenous and oralglucose tests. Am J Physiol Endocrinol Metab2007;293:E1–E1537. Utzschneider KM, Prigeon RL, Tong J, et al.Within-subject variability of measures of betacell function derived from a 2 h OGTT: implica-tions for research studies. Diabetologia 2007;50:2516–252538. Nauck MA, Blietz RW, Qualmann C. Com-parison of hyperinsulinaemic clamp experi-ments using venous, ‘arterialized’ venous orcapillary euglycaemia. Clin Physiol 1996;16:589–60239. Petrie JR, Ueda S, Morris AD, Elliott HL,Connell JM. Potential confounding effect of

1612 Standardized Tests of b-Cell Function in Humans Diabetes Care Volume 39, September 2016

hand-warming on the measurement of insulinsensitivity. Clin Sci (Lond) 1996;91:65–7140. Watanabe RM, Valle T, Hauser ER, et al.Familiality of quantitative metabolic traits inFinnish families with non-insulin-dependent di-abetes mellitus. Finland-United States Investi-gation of NIDDM Genetics (FUSION) Studyinvestigators. Hum Hered 1999;49:159–16841. QuonMJ, Cochran C, Taylor SI, Eastman RC.Non-insulin-mediated glucose disappearance in

subjects with IDDM. Discordance between ex-perimental results and minimal model analysis.Diabetes 1994;43:890–89642. Pillonetto G, Sparacino G, Cobelli C. Numer-ical non-identifiability regions of the minimalmodel of glucose kinetics: superiority ofBayesian estimation. Math Biosci 2003;184:53–6743. Pillonetto G, Sparacino G, Magni P, BellazziR, Cobelli C. Minimal model S(I)=0 problem in

NIDDM subjects: nonzero Bayesian estimateswith credible confidence intervals. Am JPhysiol Endocrinol Metab 2002;282:E564–E57344. Dalla Man C, Pillonetto G, Riz M, Cobelli C.An index of parameter reproducibility account-ing for estimation uncertainty: theory and casestudy on b-cell responsivity and insulin sensitiv-ity. Am J Physiol Endocrinol Metab 2015;308:E971–E977

care.diabetesjournals.org Shankar and Associates 1613