Diabetes Care Volume 41, August 2018 1717

Impact of and The RISE Consortium* Versus Metformin Alone on b-Cell Function in Youth With Impaired Tolerance or Recently Diagnosed Type 2 Diabetes Care 2018;41:1717–1725 | https://doi.org/10.2337/dc18-0787

OBJECTIVE Pediatric type 2 diabetes prevalence is increasing, with b-cell dysfunction key in its pathogenesis. The RISE Pediatric Medication Study compared two approachesd glargine followed by metformin and metformin alonedin preserving or improving b-cell function in youth with impaired glucose tolerance (IGT) or recently diagnosed type 2 diabetes during and after therapy withdrawal.

RESEARCH DESIGN AND METHODS

Ninety-one pubertal, overweight/obese 10–19-year-old youth with IGT (60%) PATHOPHYSIOLOGY/COMPLICATIONS or type 2 diabetes of <6 months duration (40%) were randomized to either 3 months of insulin glargine with a target glucose of 4.4–5.0 mmol/L followed by 9 months of metformin or to 12 months of metformin alone. b-Cell function (insulin sensitivity paired with b-cell responses) was assessed by hyperglycemic clamp at baseline, 12 months (on treatment), and 15 months (3 months off treatment).

RESULTS No significant differences were observed between treatment groups at baseline, RISE Coordinating Center, Rockville, MD 12 months, or 15 months in b-cell function, BMI percentile, HbA1c, fasting glucose, Corresponding author: Sharon L. Edelstein, rise@ or oral glucose tolerance test 2-h glucose results. In both treatment groups, clamp- bsc.gwu.edu. measured b-cell function was significantly lower at 12 and 15 months versus Received 10 April 2018 and accepted 13 May 2018. baseline. HbA1c fell transiently at 6 months within both groups. BMI was higher in the glargine followed by metformin versus metformin alone group between Clinical trial reg. no. NCT01779375, clinicaltrials 3 and 9 months. Only 5% of participants discontinued the interventions, and .gov. both treatments were well tolerated. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/suppl/ CONCLUSIONS doi:10.2337/dc18-0787/-/DC1. *A complete list of the RISE Consortium Inves- In youth with IGT or recently diagnosed type 2 diabetes, neither 3 months of glargine tigators can be found in the Supplementary Data followed by 9 months of metformin nor 12 months of metformin alone halted online. the progressive deterioration of b-cell function. Alternate approaches to preserve © 2018 by the American Diabetes Association. b-cell function in youth are needed. Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More infor- The SEARCH for Diabetes in Youth (SEARCH) epidemiological study has highlighted a mation is available at http://www.diabetesjournals continued increase in the incidence of type 2 diabetes in adolescents aged 10–19 .org/content/license. years, particularly among ethnic minority groups. The Treatment Options for See accompanying articles, pp. 1560, Type 2 Diabetes in Adolescents and Youth (TODAY) study demonstrated that in 1696, and 1707. 1718 b-Cell Function in Youth With Dysglycemia Diabetes Care Volume 41, August 2018

50% of youth with type 2 diabetes, Participants was withdrawn, and follow-up continued initial assigned treatment failed, re- A summary of study visits is shown in while off study medication for an additional quiring the initiation of insulin ther- Supplementary Fig. 1, and a summary of 3 months to determine the 15-month apy after an average follow-up of 3.9 participants enrolled and their rates of primary outcome of treatment durability. years, which is in keeping with faster dis- completion of study visits is shown in a Anthropometrics, HbA1c,andsafetyeval- ease progression than reported in adults CONSORT (Consolidated Standards of Re- uations were performed every 3 months, (1–3). In TODAY, lower b-cell function porting Trials) diagram (Supplementary and hyperglycemic clamps and OGTTs atbaselineaswellasagreaterdeclinein Fig. 2). Enrollment occurred between were repeated at 12 and 15 months b-cell function over time predicted worse July 2013 and April 2016. Youth aged (8–10) (Supplementary Fig. 1). Labora- glycemic control and treatment failure 10–19 years with a BMI $85th percen- tory assessments were performed in a (4). Thus, interventions to preserve or tile for age and sex but ,50 kg/m2,IGT central laboratory at the University of improve b-cell function in youth are or recently diagnosed (,6 months dura- Washington as summarized below and critically needed. tion) type 2 diabetes, and Tanner stage described in greater detail elsewhere (9). Metformin and insulin each has been $II (females, Tanner stage $II breast Medication adverse effects, symptoms showntoimproveb-cell function in development; and males, testicular vol- of and , and adults with impaired glucose tolerance ume .3mL) were eligible. Screening in- adherence (returned medication bottle (IGT) or recent-onset type 2 diabetes. The cluded a history and physical examination, pill count, returned insulin pen residual Diabetes Prevention Program (DPP) dem- laboratory tests, and a 75-g oral glucose volume) were assessed at all study visits. A onstrated that metformin improves b-cell tolerance test (OGTT). Eligibility criteria detailed algorithm was used for address- function and reduces diabetes progres- included the simultaneous presence of ing hypoglycemia, hyperglycemia, and sion by 31% over 3 years in adults with all the following: negative GAD and IA2 acute metabolic decompensation at each IGT (5,6). Weng et al. (7) showed in adults autoantibodies, fasting glucose $5 mmol/L, visit (Supplementary Study Methods). If $ with newly diagnosed type 2 diabetes OGTT 2-h glucose 7.8 mmol/L, and protocol-specified HbA1c safety thresh- that ,2 weeks of intensive insulin ther- HbA1c #8.0% (64 mmol/mol) if drug olds were exceeded at any visit, outcome apy improves and maintains b-cell func- na¨ıve. In youth with type 2 diabetes al- assessments were performed whenever tion, resulting in prolonged remission ready taking metformin, eligibility was possible. The participant was then with- from requiring . modified to HbA1c #7.5% (58 mmol/mol) drawn from the study and referred for Whether such interventions have similar if on metformin for ,3 months and additional diabetes treatment as appropri- benefits in youth is unclear. The Re- #7.0% (53 mmol/mol) if on metformin ate. Safety was monitored by an indepen- storing Insulin Secretion (RISE) Pediatric for 3–6 months. dent data and safety monitoring board. Medication Study was designed as a Final eligibility required $80% com- proof-of-principle trial to test in youth pliance with 3 weeks of placebo medi- Procedures and Calculations across a continuum of b-cell dysfunction cation (or study-provided metformin if A two-step hyperglycemic clamp (11.1 from IGT to mild, recently diagnosed already taking metformin) and atten- mmol/L, .25 mmol/L) paired with the type 2 diabetes the hypothesis that initial dance at scheduled run-in visits. Eligi- insulin secretagogue arginine at the short-term treatment with insulin glargine ble participants underwent a baseline second step was performed at baseline, for 3 months followed by metformin for hyperglycemic clamp and OGTT followed 12 months, and 15 months as previously 9 months would preserve or improve b-cell by random 1:1 treatment assignment by described (8–10). If the participant was function compared with metformin alone, study site, stratified by glycemic status. on metformin, the last dose of metformin with a sustained effect after withdrawal was taken the evening before the hy- of therapy. Interventions perglycemic clamp. The hyperglycemic The glargine followed by metformin clamp was used to quantify simulta- RESEARCH DESIGN AND METHODS groupreceived3monthsofinsulinglar- neouslyinsulinsensitivity(M/I)andthree Study Protocol gine titrated at least twice weekly by a b-cell response measures: 1)steady- The RISE Pediatric Medication Study preset algorithm (Supplementary Study state (second-phase) C-peptide, 2)acute was a four-center, randomized, open- Methods) to achieve a fasting glucose C-peptide response to arginine at maximal label clinical trial funded by the National of 4.4–5.0 mmol/L on the basis of daily glycemic potentiation (ACPRmax), and 3) Institute of Diabetes and Digestive and self-monitored blood glucose (SMBG) acute (first-phase) C-peptide response Kidney Diseases. The rationale and meth- (Freestyle Lite, Abbott Laboratories, Lake to glucose (ACPRg). M/I was calculated ods have been described in detail else- Bluff, IL), which was followed immediately as the mean glucose infusion rate (M) at where (8–10); additional specifics are by 9 months of metformin (titrated to 100, 110, and 120 min of the clamp provided in the Supplementary Data. 1,000 mg twice daily). Youth already divided by the corresponding mean steady- Each center’s institutional review board taking metformin and randomized to state plasma insulin concentration (I) approved the protocol. Written parental glargine therapy discontinued active (11–13). Mean steady-state (second- consent and child assent were obtained metformin at the time of initiation of phase) C-peptide and insulin concentra- consistent with the Declaration of Helsinki glargine treatment. The metformin group tions were calculated at 100, 110, and and each center’s institutional review received metformin alone (titrated to 120 min (13). ACPRmax was calculated as board guidelines. The protocol can be 1,000 mg twice daily) for 12 months. the mean incremental response above found athttps://rise.bsc.gwu.edu/web/ Participants took their assigned inter- concentrations before the arginine bolus rise/collaborators. vention for 12 months, after which it (14) and ACPRg as the mean incremental care.diabetesjournals.org The RISE Consortium 1719

response above baseline for the first 10 individuals taking medication and those baseline demographic and metabolic min after glucose bolus (15). taking placebo as well as expected corre- characteristics of participants by treat- The 15-month (3 months after discon- lations between baseline and follow-up ment arm. The metformin alone group tinuation of metformin) steady-state C- measures. Therefore, sample size was was slightly younger but not different peptide and ACPRmax, each paired with based on a minimum effect size equal to in Tanner stage. All other baseline phys- 15-month insulin sensitivity, represent the the difference between the two groups ical, metabolic, b-cell response, and in- coprimary outcomes. These were paired divided by the SD in favor of the glargine sulin sensitivity measures were similar to express the magnitude of the b-cell re- followed by metformin group at a power between treatment groups. Of the 91 sponse as a function of the prevailing insulin of 80%. Because we targeted improve- participants, 21 with type 2 diabetes sensitivity (15). Secondary outcomes from ments in b-cell function, a one-sided were receiving metformin at random- the hyperglycemic clamp included the 15- significance level of 0.05 to detect a ben- ization or were previously exposed to it, month ACPRg paired with the 15-month efit of glargine followed by metformin with a similar distribution of metformin M/I and the 12-month C-peptide response was used for each comparison, with a exposure and IGT versus type 2 diabetes measures (steady-state C-peptide, ACPRmax, conservative correlation of 0.5 between status between treatment groups. ACPRg) each paired with M/I at the same baseline and follow-up measures. Thirty- time point. nine participants per arm (78 total) at the Adherence to Interventions A 3-h 75-g OGTT also was performed end of the washout provided $80% power During weeks 11 and 12, the mean 6 SD on a separate day and used to determine to detect a minimum effect size of 0.57 glargine dose in the glargine followed fasting and 2-h glucose at baseline, 12 in favor of glargine followed by metfor- by metformin group was 0.7 6 0.4 units/ months, and 15 months. If the participant min for either primary outcome measure- kg/day, and fasting blood glucose had was on metformin, the last dose of ment using baseline-adjusted ANCOVA. decreased from a mean of 6.0 6 0.8 metformin was taken the evening before Ninety participants were sought, allowing mmol/LatthebaselineOGTTtoafasting the OGTT. Additional HbA1c testing was for an ;10% loss to follow-up. SMBG mean of 5.2 6 0.7 mmol/L, with done at baseline and 3, 6, 9, 12, and 15 The primary analysis was the compar- 50% achieving the 4.4–5.0 mmol/L goal months for robust phenotyping of par- ison of b-cell responses paired with M/I (Fig. 1). As assessed by residual volume, ticipants over time. between treatment groups at 15 months, 68% of participants were .80% compliant adjusted for baseline b-cell response and with the insulin. Metformin adherence Assays M/I. Joint models for b-cell response and by pill count was 88 6 18% over the Glucose was measured using the glu- M/I were fit simultaneously using seem- entire study, not differing by treatment cose hexokinase method on a Roche ingly unrelated regression techniques group. c501 autoanalyzer. C-peptide and insulin (16–18), which provide a two degrees were measured by a two-site immunoen- 2 of freedom x test of the treatment arm Treatment Effects on Temporal zymometricassayperformedontheTosoh difference in the joint values of b-cell Changes in b-Cell Function 2000 autoanalyzer (Tosoh Bioscience, Inc., response and M/I between treatment Figure 2 illustrates glucose and C-peptide South San Francisco, CA). Interassay co- groups. The same methods were used to concentrations during the hyperglyce- efficients of variation on quality control compare treatment groups at 12 months. mic clamps performed at baseline, 12 samples with low, medium, medium-high, To evaluate changes within each treat- months, and 15 months. Of note, iden- and high concentrations were #2.0% for ment arm over time, Hotelling T2 distri- tical target glucose concentrations of glucose, #4.3% for C-peptide, and #3.5% bution was used to test changes in b-cell 11.1 and .25 mmol/L were achieved for insulin. responses and M/I simultaneously (19). and not different between treatment Statistics Additional secondary analyses compared groups at baseline, 12 months, and 15 Data were accumulated centrally, and means across treatment arms at specific months. Supplementary Table 1 shows analyses were performed according to a time points by using ANCOVA adjusted b-cell responses and M/I by treatment prespecified analysis plan. For this proof- for the baseline value of the measure and group. of-principle trial, we chose two mea- by repeated-measures ANCOVA across the No significant differences were found sures of b-cell function as coprimary active 12-month intervention period ad- between treatment groups at 12 or 15 outcomes: clamp-derived steady-state justed for the baseline value of the measure. months in either primary measure of C-peptide and ACPRmax at 15 months Two participants had an HbA1c eleva- b-cell function (e.g., response [steady- (both evaluated jointly with M/I). These tion requiring study withdrawal before state C-peptide, ACPRmax]pairedwith two measures evaluate different compo- their 15-month primary outcome visit. The M/I) or in the secondary measure of nents of b-cell function and were selected clamp-derived outcome data for these b-cell function (ACPRg paired with M/I). because there were no data in youth to participants were imputed using the mid- Figure 3 shows a vector plot of changes determinewhether one or the other was point between zero and the lowest value from baseline to 12 and 15 months for likely to change after the interventions. in their respective treatment arm. each of the two coprimary clamp-derived No preliminary data were available in b-cell responses (steady-state C-peptide, youth on the expected differences in RESULTS ACPRmax) paired with M/I, and Supple- either primary outcome after a period Demographic, Physical, and Metabolic mentary Fig. 3 shows the vector plot for of intervention withdrawal (washout). Characteristics the secondary b-cell response (ACPRg Several unpublished analyses provided We screened 236 youth and randomized paired with M/I). Explanations of the a range of possible differences between 91 (Supplementary Fig. 2). Table 1 lists vector plots appear in the respective 1720 b-Cell Function in Youth With Dysglycemia Diabetes Care Volume 41, August 2018

, Table 1—Baselinedemographic and physicalcharacteristics, insulin sensitivity, P 0.001), and ACPRg with M/I (glargine and b-cell responses from the hyperglycemic clamp and OGTT by treatment group followed by metformin P , 0.001, met- Glargine followed by Metformin alone formin alone P = 0.014). metformin (n = 44) (n = 47) Approximately 60% of participants had Female, n (%) 27 (61.4) 38 (80.9) IGT at baseline; by study design, none had received metformin before randomiza- Age (years) 14.9 6 2.0 13.9 6 2.1* tion. In this IGT subgroup, we observed Tanner stage V, n (%) 32 (72.7) 25 (53.2) patterns similar to those seen in the full Race/ethnicity, n (%) fi White 13 (29.5) 12 (25.5) cohort, with no evidence of a bene cial Black 14 (31.8) 9 (19.1) effect on b-cell function of glargine fol- Hispanic 14 (31.8) 20 (42.6) lowed by metformin versus metfor- Other 3 (6.8) 6 (12.8) min alone, and a worsening of b-cell Weight (kg) 102.0 6 25.7 97.7 6 23.3 function through 12 and 15 months in BMI (kg/m2) 36.5 6 6.4 36.9 6 6.4 both treatment groups (Supplemen- BMI percentile 98.4 6 2.5 98.8 6 1.3 tary Fig. 4). BMI z score 2.3 6 0.4 2.4 6 0.3 Waist circumference (cm) 109.8 6 15.8 109.6 6 12.4 Temporal Patterns of BMI and

HbA1c Glycemic Measures Over 15 Months % 5.7 6 0.6 5.7 6 0.6 Several transient changes were seen mmol/mol 39.2 6 6.6 38.6 6 6.3 in body size and glycemic parameters. IGT, n (%) 26 (59) 28 (60) BMI did not differ between treatment Type 2 diabetes, n (%) 18 (41) 19 (40) groups at 12 or 15 months. However, Metformin use at baseline, n (%) 10 (22.7) 11 (23.4) BMI declined from baseline within the Systolic blood pressure (mmHg) 120.7 6 7.8 119.5 6 8.7 metformin alone group over the first Diastolic blood pressure (mmHg) 67.6 6 7.7 70.1 6 7.9 9 months, resulting in a significantly lower Hypertension, n (%) 9 (20.5) 15 (31.9) BMI across the 12-month intervention Triglycerides (mmol/L) 1.11 (0.39, 3.14) 1.17 (0.45, 3.06) period in the metformin alone versus HDL cholesterol (mmol/L) 1.0 6 0.3 1.0 6 0.2 glargine followed by metformin group fi , LDL cholesterol (mmol/L) 2.3 6 0.8 2.1 6 0.6 (P = 0.008) and speci cally at 3 (P 0.001), 6(P , 0.02), and 9 (P , 0.04) months. Hyperglycemic clamp variables fi Fasting glucose (mmol/L) 6.04 6 0.85 6.11 6 1.10 HbA1c did not signi cantly differ be- Fasting C-peptide (nmol/L) 1.63 6 0.55 1.82 6 0.58 tween treatment groups across the Fasting insulin (pmol/L) 211.1 (54.8, 813.5) 247.4 (74.1, 825.4) 12-month intervention period or at any Steady-state C-peptide (nmol/L) 5.22 (2.44, 11.15) 5.09 (2.32, 11.19) individual time point (Fig. 4). However, ACPRmax (nmol/L) 7.29 (3.29, 16.16) 8.12 (3.42, 19.26) compared with baseline, HbA1c de- ACPRg (nmol/L) 1.06 (0.10, 11.66) 1.14 (0.14, 9.02) clined significantly within the glargine Glucose infusion rate (M) (mmol/kg/min) 0.025 6 0.013 0.023 6 0.010 Steady-state insulin (I) (pmol/L) 1,391 (281, 6,874) 1,371 (286, 6,572) followed by metformin group at 3 months M/I (3 10e25 mmol/kg/min per pmol/L) 1.60 (0.35, 7.38) 1.54 (0.35, 6.76) while on glargine treatment (P , 0.03) OGTT variables and within both groups at 6 months Fasting glucose (mmol/L) 6.0 6 0.8 6.1 6 1.1 (glargine followed by metformin P = 2-h glucose (mmol/L) 10.2 6 2.5 10.2 6 2.8 0.001, metformin alone P = 0.01) before Data are mean 6 SD or geometric mean (95% CI) for nonnormally distributed variables. *P , rebounding to baseline levels in both 0.03; all other P values were not significant. groups at months 9 and 12, despite continued active metformin treatment. By 15 months, HbA1c had increased sig- nificantly from baseline in the glargine figure legends. Two participants in the through 12 months within each treatment followed by metformin group (P = 0.03), metformin alone arm met HbA1c criteria group for steady-state C-peptide with ending at ;6% (42 mmol/mol) in both for study withdrawal at 9 months; thus M/I (glargine followed by metformin P = treatment groups. they had a 12-month visit but did not 0.019, metformin alone P = 0.025) and No differences were found between complete the 15-month visit. The inclu- ACPRmax with M/I (glargine followed by treatment groups in fasting or OGTT 2-h sion of imputed values for their b-cell metformin P , 0.001, metformin alone glucose concentrations at 12 or 15 months. responses and M/I at 15 months did not P = 0.001). b-Cell function remained Within the glargine followed by metformin alter the results. significantly worse at 15 months com- group, fasting glucose was not different at Neither treatment improved nor main- pared with baseline within both treat- 12 months compared with baseline but tained b-cell function at the end of ments for steady-state C-peptide with rose higher than baseline by 15 months or after active treatment (Fig. 3 and M/I (glargine followed by metformin (off treatment) (P = 0.02). Within the Supplementary Fig. 3). Rather, b-cell P , 0.001, metformin alone P = 0.031), glargine followed by metformin group, function declined significantly from ACPRmax with M/I (glargine followed by 2-h glucose was significantly lower than baseline while on active treatment metformin P = 0.001, metformin alone baseline at 12 months (P = 0.02) but care.diabetesjournals.org The RISE Consortium 1721

CONCLUSIONS In youth along the glycemic continuum of IGT to mild, recently diagnosed type 2 diabetes, we tested whether initial short- term treatment with insulin glargine for 3 months followed by metformin for 9 months would preserve or improve b-cell function compared with metformin alone for 12 months and a sustained effect after withdrawal of therapy. No significant dif- ferences were found between groups in b-cell function at 12 or 15 months, and all b-cell measures at 12 months (on treat- ment) and 15 months (off treatment) were worse than at baseline. Thus, the only two agents approved for treating type 2 di- abetes in youth both failed to preserve or improve b-cell function during or after withdrawal of treatment in youth with IGT or recently diagnosed type 2 diabetes. Trials of insulin therapy in adults with IGT or early type 2 diabetes support the concept that early use of insulin may have a beneficial long-term effect on dysglycemia. In the Outcome Reduction With Initial Glargine Intervention (ORIGIN Trial), adults with high cardiovascular dis- ease risk and were treated with insulin glargine to a similar target glucose (#5.3 mmol/L) as we achieved (21). A median of 3.3 months after in- sulin discontinuation, a similar duration of durability as we evaluated, those with prediabetes randomized to glargine treat- Figure 1—Insulin doses and corresponding fasting glucose values over time while on glargine ment had a 20% reduction in the odds of treatment. A: Mean fasting morning SMBG values every 2 weeks over 12 weeks. B: Mean glargine developing diabetes on the basis of OGTT dose corresponding to each SMBG value. C: Corresponding percentage of participants who criteria compared with standard care (21). achieved the goal fasting SMBG of 4.4–5.0 mmol/L every 2 weeks. Data are mean (95% CI). In adults with recently diagnosed type 2 diabetes, Weng et al. (7) found that up to 2 weeks of intensive insulin therapy no longer different from baseline at no participants required rescue therapy 15 months. Within the metformin alone with insulin. as either multiple daily injections or con- group, no significant changes were found Over the 15 months, three participants tinuous subcutaneous infusion, an even in fasting or 2-h glucose at 12 months while in the glargine followed by metformin shorter treatment duration than we on active treatment or at 15 months. group with IGT at baseline and none in tested, resulted in remission and no the metformin alone group with IGT at need for diabetes medications 1 year Safety Outcomes baseline developed diabetes as defined later in 45% and 51% of participants, respectively. Furthermore, after the Seven serious adverse events occurred by OGTT criteria and HbA1c $6.5% (one hospitalization each for suicidal (48 mmol/mol) (20). In each treatment 2-week insulin treatment, the b-cell re- ideation/anxiety, newly diagnosed Ewing group, three participants met hypergly- sponse increased, and this improvement sarcoma, otitis externa, appendicitis/ cemic criteria for study withdrawal. The was sustained in the participants who re- appendectomy, and pneumonia and last outcome visits for these partici- mained in remission while off medications. two hospitalizations for tonsillectomy/ pants were three at 15 months in the Although the study by Weng et al. sug- adenoidectomy), all deemed unrelated glargine followed by metformin group gested benefitfrom,2 weeks of insulin to the study interventions. Other non- and two at 12 and one at 15 months in treatment, we saw no effect from a serious adverse events are summarized in the metformin alone group, after which 3-month duration of treatment, which Supplementary Table 2. No participants the participants were withdrawn from is longer than that used by Weng et al. yet had severe hypoglycemia or acute meta- the study and returned to their primary brief enough to realistically consider in bolic decompensation as defined in Sup- diabetes provider for additional diabe- youth. Moreover, we chose a target plementary Study Methods. Accordingly, tes treatment. glucose lower than that used in the 1722 b-Cell Function in Youth With Dysglycemia Diabetes Care Volume 41, August 2018

Figure 2—Glucose and C-peptide values during the hyperglycemic clamp at baseline, after 12 months of treatment, and 3 months after discontinuing the intervention (15 months). The glargine followed by metformin group (n = 44 [green]) and the metformin alone group (n = 47 [brown]) are shown. The steady-state glucose targets were 11.1 mmol/L between 90 and 120 min and .25 mmol/L at 150 min. Data are mean 6 SEM.

ORIGIN Trial to maximize the effect of the weight gain, with an uncertain net bal- interpretation comes from the appar- insulin treatment. To further increase ance of benefit and harm. Moreover, the ently more rapid progression of type the likelihood of insulin having an effect, injections and glucose monitoring re- 2 diabetes observed in youth in the we added the oral glucose-lowering quired for longer-term insulin treatment TODAYstudythaninasimilarstudy agent metformin shown in the DPP to might be met with resistance by youth, in adults treated with the same be effective in preventing the develop- particularly those who are not yet di- glucose-lowering medications (2,3). Fur- ment of diabetes in adults with IGT (6). In agnosedwithdiabetes. thermore, the current observation of contrast to the adult studies, we found The RISE Consortium (9) reported a lack of a beneficial effect of metfor- no durable effect of insulin glargine tar- that youth with IGT or recently diag- min on b-cellfunctioninyouthafter geted to achieve a plasma glucose of nosed, drug-na¨ıve type 2 diabetes are 12 months of therapy contrasts the 4.4–5.0 mmol/L followed by metformin profoundly more insulin resistant, have improvement in b-cell function ob- on measures of b-cell function or glycemia greater insulin responses for any de- served in adults with IGT (5) or type 2 in youth. Furthermore, the lack of a ben- gree of insulin sensitivity, and have diabetes (22). The RISE Adult Medica- eficial effect of metformin alone on b-cell lower hepatic insulin clearance than tion Study, which includes these same function, even when youth were still on adults. This finding, combined with the two intervention arms as part of a four- the medication at 12 months, contrasts poor response of youth to the RISE arm study, should provide additional with observations after 1 year of therapy interventions, suggests that although insight into these differences (8). in adults with IGT (5) or type 2 diabetes b-cell dysfunction is the foundation The current study has several strengths, (22). Although a longer duration of in- for dysglycemia in both age-groups, including the randomized design and sulin treatment may have had more of greater b-cell demand may place youth a robust, multicenter approach to an effect on b-cell function, longer-term at a higher risk for more rapid b-cell quantification of insulin sensitivity and treatment also may have led to additional damage. Additional support for this b-cell responses to both glucose and care.diabetesjournals.org The RISE Consortium 1723

the inclusion of youth with type 2 diabetes who could not remain untreated. Although adeclineinb-cell function might have been even greater in a placebo group, the marked decline in b-cell function evident during and after stopping the interventions suggests that the medications largely were masking disease progression. In addition, despite the high reported insulin doses, 50% of the participants in the insulin glargine followed by metformin arm did not fully meet the fasting goal of 4.4– 5.0 mmol/L, regardless of IGT or type 2 diabetes status. Of note, adults with pre- diabetes or type 2 diabetes in the ORIGIN Trial were treated with insulin glargine to a target fasting glucose of #5.3 mmol/L, a slightly less-aggressive target than we chose, yet we still observed no significant beneficial effect on b-cell function despite reaching a mean fasting glucose of 5.2 mmol/L (21). Moreover, after 1 year in the ORIGIN Trial, only 50% of the adult par- ticipants achieved the targeted glucose, similar to the proportion of youth in the current study, yet positive effects still were seen on diabetes development in these adults (21). Finally, the sample size was too small to obtain reliable estimates of the effects of the interventions in the subgroups of IGT versus type 2 diabetes. However, in a sensitivity analysis of the Figure 3—Relationship of the two coprimary outcomes: hyperglycemic clamp-derived b-cell subgroup with IGT who had not received responses (steady-state C-peptide and ACPRmax) paired with M/I. Changes are shown from metformin before randomization and com- baseline to 12 and 15 months for the clamp-derived b-cell responses of steady-state C-peptide prised the majority of participants in each b (A)andACPRmax (B) paired with M/I. The black line depicts the joint relationship between -cell arm, results were similar to the full cohort response and M/I at baseline for the full cohort, with the mean value at baseline for the full fi cohort indicated by the black box with a 0. The dotted lines to boxes at months 12 and 15 show (Supplementary Fig. 4). Speci cally, in the the trajectory of values from baseline to 12 months of intervention and then to 3 months after IGT group, treatment with glargine fol- discontinuation of the intervention (15months) for glargine followed bymetformin(green) and lowed by metformin did not have a ben- metformin alone (brown). Values above the black line represent improved b-cell function, and eficial effect on b-cell function outcomes values below the line represent poorer b-cell function. The ellipses depict the 95% confidence fi compared with metformin alone, and bands around the points at months 12 and 15. No signi cant differences were observed at any b time point between treatment arms; however, significant within-group declines were seen -cell function declined within both treat- from baseline while on active treatment through 12 months for steady-state C-peptide with ment groups. M/I (glargine followed by metformin P = 0.019, metformin alone P = 0.025) and ACPRmax with In conclusion, neither 3 months of in- M/I (glargine followed by metformin P , 0.001, metformin alone P = 0.001) and through sulin glargine followed by 9 months of , 15 months for steady-state C-peptide with M/I (glargine followed by metformin P 0.001, metformin nor 12 months of metformin metformin alone P =0.031)andACPRmax with M/I (glargine followed by metformin P = 0.001, metformin alone P , 0.001). alone improved or preserved b-cell func- tion when administered to youth with IGT or recently diagnosed type 2 diabetes. b- nonglucose secretagogue arginine per- multicenter study, we reached matching Cell function deteriorated with both treat- formed to provide thorough, longitu- levels of hyperglycemia during each of ments, the only treatments approved for dinal phenotyping of participants. By three repeated hyperglycemic clamp use in type 2 diabetes in youth, highlight- quantifying both insulin sensitivity and tests, showing a high degree of both ing an urgent need for alternate approaches b-cell responses simultaneously, we were reproducibility and reliability for the out- to preserve b-cell function in youth and able to account for the well-recognized come measures across time points and prevent the progression of dysglycemia. relationship between insulin sensitivity across study sites. Study retention and and b-cell responsivity (15) and thus metformin adherence also were excellent Appendix gain mechanistic insight into the effect despite adolescents presenting a challeng- Writing Group: Kristen J. Nadeau (chair), of the interventions on b-cell function ing study population. Limitations include Tamara S. Hannon, Sharon L. Edelstein, over time. Of key importance for this the absence of a placebo arm because of Silva A. Arslanian, Sonia Caprio, Ellen W. 1724 b-Cell Function in Youth With Dysglycemia Diabetes Care Volume 41, August 2018

Figure 4—BMI and glycemia over time. BMI (A), HbA1c (B), fasting glucose (C), and OGTT 2-h glucose over 15 months (D). In the glargine followed by metformin group (n = 44 [green]), glargine insulin was given from baseline to 3 months followed by metformin from months 3 through 12. In the metformin alone group (n = 47 [brown]), metformin was given from baseline through 12 months. Treatment was discontinued in both groups at 12 months. Data are mean 6 SEM. *P , 0.05 for the difference between treatment groups at the specified time point; §P , 0.05 for the difference within treatment group between the specified time point and baseline.

Leschek, Philip S. Zeitler, Thomas A. S.L.E., S.A.A., S.C., E.W.L., T.A.B., D.A.E., K.J.M., in the evolution of type 2 diabetes in the Dia- Buchanan, David A. Ehrmann, Kieren and S.E.K. (chair) comprise the steering commit- betes Prevention Program: Effects of lifestyle J. Mather, and Steven E. Kahn. tee (principal investigator at each site, the data intervention and metformin. Diabetes 2005; coordinating center, and the National Institute 54:2404–2414 of Diabetes and Digestive and Kidney Diseases 6. Diabetes Prevention Program Research Group. project scientist) and designed and imple- Reduction in the incidence of type 2 diabetes Acknowledgments. The RISE Consortium thanks mented the study, contributed to the discus- with lifestyle intervention or metformin. N Engl the RISE Data and Safety Monitoring Board and sion, and edited the manuscript. K.J.N., T.S.H., J Med 2002;346:393–403 Barbara Linder, the National Institute of Diabetes S.L.E., S.A.A., S.C., E.W.L., P.S.Z., T.A.B., D.A.E., K.J.M., 7. Weng J, Li Y, Xu W, et al. Effect of intensive and Digestive and Kidney Diseases Program Offi- and S.E.K. researched data. S.L.E. performed insulin therapy on b-cell function and glycaemic cial for RISE (Rockville, MD), for support and all data analyses. K.J.N., S.L.E., and S.E.K. are the control in patients with newly diagnosed type 2 guidance. The Consortium also thanks the par- guarantors of this work and, as such, had full diabetes: a multicentre randomised parallel- ticipants who, by volunteering, are furthering the access to all the data in the study and take group trial. Lancet 2008;371:1753–1760 ability to reduce the burden of diabetes. responsibility for the integrity of the data and the 8. The RISE Consortium. Restoring Insulin Secre- Funding and Duality of Interest. RISE is sup- accuracy of the data analysis. tion (RISE): Design of studies of b-cell preservation ported by grants from the National Institutes Prior Presentation. Parts of this study were in prediabetes and early type 2 diabetes across of Health (U01-DK-094406, U01-DK-094430, U01- presented in oral form at the 78th Scientific the life span. Diabetes Care 2014;37:780–788 DK-094431, U01-DK-094438, U01-DK-094467, Sessions of the American Diabetes Association, 9. The RISE Consortium. Metabolic contrasts P30-DK-017047, P30-DK-020595, P30-DK-045735, Orlando, FL, 22–26 June 2018. between youth and adults with impaired glucose P30-DK-097512, UL1-TR-000430, UL1-TR-001082, tolerance or recently diagnosed type 2 diabetes: UL1-TR-001108, UL1-TR-001855, UL1-TR-001857, References I. Observations using the hyperglycemic clamp. UL1-TR-001858, UL1-TR-001863), the Depart- 1. Mayer-Davis EJ, Lawrence JM, Dabelea D, Diabetes Care 2018;41:1696–1706 ment of Veterans Affairs, and Kaiser Permanente et al.; SEARCH for Diabetes in Youth Study. Incidence 10. Hannon TS, Kahn SE, Utzschneider KM, et al.; Southern California. Additional financial and ma- trends of type 1 and type 2 diabetes among youths, The RISE Consortium. Review of methods for terial support from the American Diabetes Asso- 2002-2012. N Engl J Med 2017;376:1419–1429 measuring b-cell function: Design considerations ciation, Allergan, Apollo Endosurgery, Abbott 2. TODAY Study Group. A clinical trial to main- from the Restoring Insulin Secretion (RISE) Con- Laboratories, and Novo Nordisk was received. tain glycemic control in youth with type 2 di- sortium. Diabetes Obes Metab 2018;20:14–24 S.A.A. and S.E.K. serve as paid consultants on abetes. N Engl J Med 2012;366:2247–2256 11. Andres R, Swerdloff R, Posefsky T, Coleman advisory boards for Novo Nordisk. S.A.A. is 3. Kahn SE, Haffner SM, Heise MA, et al.; ADOPT DL. Manual feedback technique for the control a participant in a Novo Nordisk–sponsored Study Group. Glycemic durability of rosiglitazone, of blood glucose concentration. In Automation clinical trial. T.A.B. has received research support metformin, or glyburide monotherapy. N Engl J in Analytical Chemistry. Skeegs LT Jr., Ed. New from Allergan and Apollo Endosurgery. K.J.M. Med 2006;355:2427–2443 York, Mediad, 1966, p. 486–491 holds an investigator-initiated research grant 4. TODAY Study Group. Effects of metformin, 12. DeFronzo RA, Tobin JD, Andres R. Glucose from Novo Nordisk. No other potential conflicts metformin plus rosiglitazone, and metformin plus clamp technique: a method for quantifying in- of interest relevant to this article were reported. lifestyle on insulin sensitivity and b-cell function sulin secretion and resistance. Am J Physiol 1979; Author Contributions. Members of the RISE in TODAY. Diabetes Care 2013;36:1749–1757 237:E214–E223 Consortium recruited participants and collected 5. Diabetes Prevention Program Research 13. Elahi D. In praise of the hyperglycemic clamp. study data. K.J.N. wrote the first draft. K.J.N., Group. Role of insulin secretion and sensitivity A method for assessment of b-cell sensitivity and care.diabetesjournals.org The RISE Consortium 1725

insulin resistance. Diabetes Care 1996;19:278– 16. Zellner A. An efficient method of estimating 20. American Diabetes Association. 2 Classifica- 286 seemingly unrelated regression equations and tion and diagnosis of diabetes: Standards of 14. Ward WK, Bolgiano DC, McKnight B, Halter tests for aggregation bias. J Am Stat Assoc 1962; Medical Care in Diabetesd2018. Diabetes Care JB, Porte D Jr. Diminished B cell secretory 57:348–368 2018;41(Suppl. 1):S13–S27 capacity in patients with noninsulin-dependent 17. Srivastava VK, Giles DEA. Seemingly Un- 21. The ORIGIN Trial Investigators. Basal insu- diabetes mellitus. J Clin Invest 1984;74:1318– related Regression Equations Models: Estimation lin and cardiovascular and other outcomes in 1328 and Inference. New York, Marcel Dekker, 1987 dysglycemia. N Engl J Med 2012;367:319–328 15. Kahn SE, Prigeon RL, McCulloch DK, et al. 18. Henningsen A, Hamann J. Systemfit: a pack- 22. Kahn SE, Lachin JM, Zinman B, et al.; ADOPT Quantification of the relationship between in- age for estimating systems of simultaneous equa- Study Group. Effects of rosiglitazone, glyburide, sulin sensitivity and b-cell function in human tions in R. J Stat Softw 2007;23:1–40 and metformin on b-cell function and insu- subjects. Evidence for a hyperbolic function. Dia- 19. Hotelling H. The generalization of student’s lin sensitivity in ADOPT. Diabetes 2011;60:1552– betes 1993;42:1663–1672 ratio. Ann Math Stat 1931;2:360–378 1560