Abstract
Objective
We examined predictors of early and late loss of glycemic control in individuals with youth-onset type 2 diabetes, as well as predictors of short-term deterioration in youth from the Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study.
Methods
Demographic, physical, and biochemical measures at baseline and 48 months, and change over time, were examined in 584 participants separated into those with loss of glycemic control (sustained HbA1c ≥ 8%) before 48 months or at 48 months or later, and those who remained in control until the end of the study (median 6.8 years). Univariate and multivariate models, and receiver operating characteristic curve analyses were performed.
Results
Approximately 45% of youth remained in control at 48 months; of these, 30% subsequently lost glycemic control prior to the end of follow-up. Predictors of early loss of glycemic control included baseline HbA1c, C-peptide index, oral disposition index, proinsulin, and proinsulin to insulin ratio. Predictors of late loss included baseline measures of insulin secretion and change in HbA1c and insulin processing at 48 months. A baseline HbA1c cutoff of ≥ 6.2% was optimally predictive of loss of glycemic control at any time, while an absolute rise in HbA1c > 0.5% related to loss of glycemic control within 3 to 6 months.
Conclusion
This analysis demonstrates that youth with type 2 diabetes at risk for loss of glycemic control, including impending rapid deterioration, can be identified using available clinical measures, allowing for closer monitoring of at-risk youth, and facilitating the design of research on better therapeutic options.
Keywords: glycemic control, type 2 diabetes, youth, risk factors
The Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study was the first and largest randomized clinical trial to examine approaches to management of youth-onset type 2 diabetes. Among the primary findings of TODAY (2004-2011) were that metformin monotherapy was inadequate for maintenance of glycemic control (glycated hemoglobin [HbA1c] < 8.0%) in almost half of participants after a median time of approximately 11 months (1). This degree of failure of metformin monotherapy in youth with type 2 diabetes was higher than has been reported from studies of adults with type 2 diabetes (2, 3). The addition of rosiglitazone significantly but modestly reduced the rate of failure. These findings suggest that type 2 diabetes among adolescents can be more rapidly progressive than in adults (4-7). In 2011, TODAY participants enrolled in a follow-up observational study (TODAY2) to monitor long-term glycemic and nonglycemic outcomes.
However, the finding that 50% of the randomized participants did maintain durable glycemic control suggests that type 2 diabetes among adolescents is a heterogeneous disorder with half of patients able to maintain glycemic control on initial oral therapy with metformin. We recently reported that accumulation of complications and comorbidities is also rapid in these youth and related to glycemia, among other determinants (8). Taken together, these findings underscore the substantial clinical importance of identifying high-risk individuals early in the course of the disease so that appropriate treatment can be planned and implemented. We previously reported that easily available demographic, clinical, and metabolic measures can help predict the likelihood of loss of glycemic control by 48 months of treatment (9). In the current analysis, we extend this examination to include up to an additional 5 years of follow-up, depending on the date of original TODAY enrollment. Because of the initial rolling enrollment in TODAY (2004-2009), this analysis includes follow-up ranging from 5 to 9 years from randomization until 2014. In addition, we explore the subsequent course of those who had durable glycemic control 48 months after randomization to evaluate: 1) the percentage of participants who lost glycemic control after 48 months of treatment; 2) the factors at baseline and at 48 months that predict loss of glycemic control after 48 months; and 3) changes in HbA1c that predict subsequent loss of glycemic control.
Methods
TODAY Design and Primary Findings
The TODAY clinical trial (2004-2011) rationale, design, and methods have been reported in detail (10) and are described here briefly, with the timeline illustrated in Fig. 1. At study entry, the participants (N = 699) at 15 centers across the United States were 10 to < 18 years old with < 2 years duration of type 2 diabetes, based on the American Diabetes Association 2002 criteria (11), and had body mass index (BMI) ≥ 85th percentile, fasting C-peptide > 0.6 mmol/L, and negative pancreatic antibodies (12). After screening to determine eligibility, participants completed a 2- to 6-month prerandomization run-in period during which they demonstrated mastery of standard diabetes education, were weaned from all diabetes medications except metformin, demonstrated tolerance of metformin 500 to 1000 mg twice daily, maintained HbA1c < 8% (<64 mmol/mol) monthly for at least 2 months on metformin alone, and demonstrated adherence to study medication and visit attendance (13) (Fig. 2). Participants who successfully completed run-in were randomized to metformin monotherapy, metformin plus rosiglitazone, or metformin plus an intensive lifestyle intervention and followed for an average of 3.9 (range, 2 to 6) years. Study baseline data were collected prior to start of randomized treatment assignment (10). Participants attended clinic visits for purposes of medical management and data collection every 2 months in the first year and quarterly thereafter. The primary objective was to compare the 3 treatment arms on time to treatment failure, defined as either HbA1c ≥ 8% over a 6-month period or inability to wean from temporary insulin therapy within 3 months following acute metabolic decompensation. Forty-six percent reached the primary outcome. Metformin plus rosiglitazone (38.6%) was superior to metformin alone (51.7%; P = 0.006) and metformin plus lifestyle intervention (46.6%) was intermediate but not significantly different from metformin alone or metformin plus rosiglitazone (1). In 2011, 572 of the original TODAY clinical trial participants (82%) enrolled in a predominately observational follow-up study (TODAY2). For the next 3 years, from 2011 to 2014, participants no longer received randomized treatment, but continued to receive diabetes-related care every 3 months from the TODAY study team and were treated with metformin and/or insulin as needed to maintain glycemic control.
Figure 1.
Study timeline during 2004-2014*. *Due to the staggered entry in the cohort, participants’ length of follow-up varied across study periods. For example, a participant followed for 48 months may have his/her entire follow-up during TODAY or split across TODAY and TODAY2 (ie, 2 years of follow-up during TODAY and 2 years of follow-up during TODAY2).
Figure 2.
Flow of participants according to groups of glycemic failure within 48 months of randomization during 2004-2014, across TODAY and TODAY2. *Total of 410 participants (n = 320 + 90) reached glycemic failure during 2004-2014. **Durable responders, still in study and at risk of glycemic failure at month 48 since randomization.
Demographic and baseline characteristics in the analysis were: age, sex, self-report of race/ethnicity (classified as non-Hispanic Black [NHB], Hispanic [H], non-Hispanic White [NHW], or other), household annual income, highest education level of parents or guardians; first-degree family and maternal history of diabetes, and maternal diabetes during pregnancy. Also included were Tanner stage (by physician examination); months since diagnosis of type 2 diabetes; body composition by dual-energy x-ray absorptiometry (DXA) scan; and presence of depressive symptoms based on the Children’s Depression Inventory (14) for participants ≤ 15 years of age, or the Beck Depression Inventory (15), for participants ≥ 16 years (by self-report). Measurements of height, weight, and calculated BMI [weight (kg)/height2 (m2)] were obtained at every study visit. Waist circumference was measured at the iliac crest using a fiberglass Gulick II tape measure.
Measurements of a fasting HbA1c, insulin, C-peptide, and proinsulin were performed centrally at the TODAY Central Biochemistry Laboratory (Northwest Lipid Metabolism and Diabetes Research Laboratories, University of Washington, Seattle WA) according to standardized procedures. HbA1c concentration was measured at every visit with a dedicated high-performance liquid chromatography method (TOSOH Biosciences Inc., South San Francisco, CA). Proinsulin was measured by radioimmunoassay (Millipore Inc, Burlington MA). The assay sensitivity is 2 pM and the interassay coefficients of variation for high and low proinsulin level samples are 7.5% and 10%, respectively. A 2-hour oral glucose tolerance test (OGTT) was performed at randomization, at 6 and 24 months, and annually thereafter through 2014. Surrogate estimates of insulin sensitivity and β-cell function were derived from the OGTT as follows: insulin sensitivity = 1/fasting insulin (1/IF); C-peptide index = calculated as the ratio of the incremental C-peptide and glucose responses over the first 30 minutes of the OGTT (∆C30/∆G30); and C-peptide-based oral disposition index (coDI) = a measure of insulin secretion relative to insulin sensitivity calculated as the product of insulin sensitivity and C-peptide index (1/IF x ∆ C30/∆G30) (16). Insulin-based oDI was presented in the previous report examining baseline predictors of loss of glycemic control (9); however, since some participants received insulin during the study, which could have potentially resulted in circulating insulin antibodies interfering with the insulin assay, coDI was chosen in the present study as a more reliable reflection of insulin secretion. For purposes of this analysis, loss of glycemic control was defined as HbA1c ≥ 8% at 2 or more consecutive visits.
The protocol was approved by a Data and Safety Monitoring Board convened by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the National Institutes of Health and by the Institutional Review Boards of each participating institution. All participants provided written informed consent and minor children confirmed assent according to local guidelines.
Statistical Analysis
The association between the predictor variables (baseline/randomization, 0- to 6-month change, 48 months, and 0- to 48-month change) and glycemic failure groups were assessed using univariable and multivariable logistic regression models. For purposes of this study, predictors of glycemic failure evaluated within the first 6 months of the study were considered early predictors, and those at 48 months or after, late predictors. The R-square (R2) was used to provide an overall fit of the regression and the Wald chi-square statistic to rank predictors. A stepwise backward selection procedure (using P value > 0.01 as criteria for removal) identified predictive factors that were significant at the 5% level. An analysis of collinearity was completed for the multivariable analyses. Receiver operating characteristic (ROC) curve analyses were performed to identify optimal cut points for baseline HbA1c that were predictive of glycemic failure (17). The standard logistic regression model and the trapezoidal rule method were used to compute the total area under the curve (AUC) and its associated 95% CI in the overall analysis sample and in subgroups of sex and/or race/ethnicity. The Youden index method (18) was used to select the optimal threshold point from the ROC curve and the positive likelihood ratio (PLR) was used to determine the likelihood of glycemic failure given the identified cutoff. Mean HbA1c by glycemic failure group was plotted relative to time of glycemic failure in the participants who lost control during the 2004-2014 period and relative to the study midpoint for participants who maintained control during that same time period. Change in HbA1c (within 3, 6, 9, or 12 months) relative to the time of glycemic failure or study midpoint was also evaluated and ROC curve analyses performed to identify optimal cut points for change in HbA1c that were predictive of glycemic failure. All analyses were performed using SAS software (version 9.4; SAS Institute, Cary, NC). A 2-sided P value ≤ 0.05 was considered to be statistically significant.
Results
The assignment of participants into the analysis groups is shown in Fig. 2. The analysis sample included all participants in the TODAY study with data collected during 2004-2014 and with a minimum of 48 months of follow-up. Twenty-two individuals who were subsequently identified as having maturity-onset diabetes of the young (MODY) mutations after randomization and 12 individuals who were treated with insulin despite not meeting study criteria for insulin initiation were excluded from the analyses, and 5 individuals who underwent bariatric surgical procedures during the analysis period were censored beginning at the time of surgery. A small number of participants (n = 81) who were still in glycemic control during their follow-up but did not have 48 months of data were not included in the analyses.
Data from n = 584 TODAY participants were used in this study, with a median follow-up of 6.8 years (including 2-6 years of follow-up in the TODAY randomized trial because of rolling enrollment and up to 3 years of follow-up in TODAY2). Participants were grouped as follows: 1) those who reached glycemic failure (ie, HbA1c ≥ 8% at 2 or more consecutive visits) within the first 48 months of the study (n = 320); and 2) those who did not reach glycemic failure within the first 48 months of the study (n = 264). The latter group was further divided into those who had reached glycemic failure after 48 months (n = 90) and those who remained in control until the end of the analysis period (n = 174). The 90 participants who subsequently lost glycemic control after 48 months in the study had a median time of failure of approximately 5.5 years since randomization (range, 4-9 years), while the n = 174 participants who remained in control had a median time of follow-up of 6.7 years (range, 4-9 years).
Table 1 compares baseline factors and change over the first 6 months in TODAY between the 410 participants who lost glycemic control in TODAY and TODAY2 and the 174 participants who had not lost control by the end of the analysis period. Participants with durable glycemic control had shorter diabetes duration at randomization, less frequent family and maternal history of diabetes, significantly lower HbA1c concentration and higher indices of β-cell function, including coDI. In univariate analysis, significant baseline predictors of loss of glycemic control included duration of diabetes, family and/or maternal history of diabetes, exposure to maternal diabetes in utero, HbA1c concentration, proinsulin concentration, proinsulin/insulin ratio, C-peptide index, and coDI (Table 1). Similar results were obtained when insulin-based indices of beta-cell function were evaluated instead of C-peptide based ones (data not shown). However, treatment group, depressive symptoms, pubertal Tanner stage, household level of income and education, and DXA determined fat and lean mass at the time of randomization were not significantly different between those who lost glycemic control and those who did not (data not shown).
Table 1.
Baseline (at randomization) and early change (baseline to 6-month change since randomization) predictors of loss of glycemic control during 2004-2014
| Baseline predictors | Baseline to 6 months change predictors | |||||||
|---|---|---|---|---|---|---|---|---|
| No failure (n = 174) | Glycemic failure (n = 410) | R2 (%) | P value | No failure (n = 174) | Glycemic failure (n = 322a) | R2 (%) | P value | |
| Female (%) | 67.2 | 64.1 | <1 | 0.47 | -- | -- | -- | -- |
| Age (years) | 14.0 ± 1.9 | 13.9 ± 2.1 | <1 | 0.88 | -- | -- | -- | -- |
| Race/ethnicity (%) | ||||||||
| Non-Hispanic Black | 31.6 | 35.6 | <1 | 0.35 | -- | -- | -- | -- |
| Hispanic | 39.1 | 40.2 | -- | -- | ||||
| Non-Hispanic White | 19.0 | 17.8 | -- | -- | ||||
| Other | 10.3 | 6.3 | -- | -- | ||||
| Duration of diabetes (months) | 6.8 ± 5.2 | 8.3 ± 6.2 | 1.5 | 0.002 | -- | -- | -- | -- |
| Pubertal Tanner stage 4-5 (%) | 93.1% | 89.5% | <1 | 0.17 | -- | -- | -- | -- |
| Family history of diabetes (%) | 46.7% | 64.4% | 2.6 | < 0.0001 | -- | -- | -- | -- |
| Maternal history of diabetes (%) | 41.8% | 62.5% | 3.6 | < 0.0001 | -- | -- | -- | -- |
| Maternal diabetes during pregnancy (%) | 24.4% | 36.5% | 1.4 | 0.006 | -- | -- | -- | -- |
| Weight (kg) | 97.6 ± 24.8 | 95.7 ± 24.8 | <1 | 0.28 | 1.31 ± 5.74 | 2.45 ± 4.39 | 1.2 | 0.09 |
| BMI (kg/m2) | 35.2 ± 7.9 | 35.1 ± 7.4 | <1 | 0.94 | 0.17 ± 1.90 | 0.49 ± 1.65 | <1 | 0.15 |
| Waist circumference (cm) | 109.2 ± 15.4 | 109.4 ± 16.9 | <1 | 0.95 | -0.45 ± 4.98 | 1.05 ± 5.75 | 1.6 | 0.01 |
| HbA1c (%) | 5.7 ± 0.5 | 6.2 ± 0.8 | 11.5 | < 0.0001 | -0.04 ± 0.50 | 0.22 ± 0.90 | 2.6 | 0.0003 |
| C-peptide (ng/mL) | 3.8 ± 1.5 | 3.9 ± 1.6 | <1 | 0.40 | -0.03 ± 1.15 | 0.08 ± 1.18 | <1 | 0.17 |
| Proinsulin (pM) | 26.4 ± 19.1 | 41.7 ± 41.4 | 5.3 | < 0.0001 | 4.29 ± 28.26 | 7.28 ± 42.76 | <1 | 0.06 |
| Proinsulin to insulin ratio (pM/µU) | 0.9 ± 0.7 | 1.6 ± 1.7 | 6.7 | < 0.0001 | 0.27 ± 1.38 | 0.09 ± 1.76 | <1 | 0.45 |
| Insulin sensitivity (mL/µU) | 0.047 ± 0.035 | 0.051 ± 0.050 | <1 | 0.91 | 0.009 ± 0.038 | -0.001 ± 0.054 | 1.0 | 0.05 |
| C-peptide Index (ng/mL per mg/dL) | 0.128 ± 0.200 | 0.063 ± 0.062 | 14.5 | < 0.0001 | -0.024 ± 0.206 | -0.008 ± 0.088 | <1 | 0.75 |
| C-peptide oDI (1/IF x ∆ C30/∆G30) | 0.006 ± 0.012 | 0.003 ± 0.004 | 10.1 | < 0.0001 | -0.001 ± 0.011 | 0.001 ± 0.005 | <1 | 0.16 |
Data are mean ± SD or percent.
aExcluding n = 88 participants from the 410 who lost glycemic control during the first 6 months of the study. P values comparing glycemic failure groups were obtained from logistic regression models. The R-square (R2) is indicative of overall fit of the regression, with higher values meaning higher predictive ability. Results for the early predictors (baseline to 6 months change since randomization) were unaffected after adjustment for randomized treatment group.
Table 2 shows the ROC analysis of baseline (randomization) HbA1c as a predictor of loss of glycemic control at any time in TODAY and TODAY2. Similar to our previous findings (9) examining glycemic control up to 48 months, the analysis of outcomes obtained up to 9 years of follow-up identified an overall baseline HbA1c cutoff of 6.2%, which had predictive value in distinguishing participants who did and did not have durable glycemic control on metformin monotherapy (PLR of 2.94 and AUC of 0.71). As in our previous report, the ROC analysis showed a sex difference in HbA1c operator characteristics, with an HbA1c cutoff of 6.4% in females (PLR 7.37) and 5.6% in males (PLR 1.74); the lower PLR in males was a result of the TODAY cohort predominately being female (65%). This sex difference in optimal cutoff was true across race/ethnicity.
Table 2.
Baseline HbA1c cutoffs, AUC and its 95% Wald CI, and PLR overall and by sex and racial/ethnic subgroups (“other” race/ethnicity not presented)*
| Groups | N | HbA1c cutoff | AUC | PLR |
|---|---|---|---|---|
| Overall | 584 | 6.2 | 0.71 (0.66-0.75) | 2.94 |
| Female | 380 | 6.4 | 0.71 (0.66-0.76) | 7.37 |
| Male | 204 | 5.6 | 0.71 (0.63-0.78) | 1.74 |
| Non-Hispanic Black | 201 | 6.4 | 0.71 (0.64-0.79) | 4.27 |
| Hispanic | 233 | 6.2 | 0.68 (0.61-0.75) | 2.93 |
| Non-Hispanic White | 106 | 5.6 | 0.78 (0.67-0.87) | 2.30 |
| Female Non-Hispanic Black | 141 | 6.4 | 0.75 (0.67-0.83) | 5.78 |
| Female Hispanic | 140 | 6.3 | 0.66 (0.57-0.75) | 5.47 |
| Female Non-Hispanic White | 66 | 5.4 | 0.82 (0.68-0.95) | 2.60 |
| Male Non-Hispanic Black | 60 | 6.0 | 0.62 (0.44-0.81) | 1.91 |
| Male Hispanic | 93 | 5.5 | 0.70 (0.59-0.82) | 1.92 |
| Male Non-Hispanic White | 40 | 5.6 | 0.72 (0.55-0.88) | 2.00 |
Abbreviations: AUC, area under the curve; HbA1c, glycated hemoglobin A1c; PLR, positive likelihood ratio.
*Based on overall sample size of n = 584 (n = 174 without loss of glycemic control during 2004-2014 plus n = 410 with loss of glycemic control during 2004-2014; Figure 1). AUC is a measure of diagnostic accuracy ranging from 0 to 1, where 0.5 is equivalent to a coin toss; AUC values 0.6-0.7 represent poor ability to predict failure, values 0.7-0.8 represent fair ability, and 0.8-0.9 represent good ability. The TODAY cohort is predominantly female and minority, which affects the overall estimates (eg, the male baseline HbA1c cutoff identified via the Youden index 0.8% below the female one). The AUC 95% CI includes 0.50 only for the male non-Hispanic Black subgroup (n = 60) indicating that the cutoff is equivalent to simply guessing. A PLR (sensitivity/1-specificity) > 1 indicates the test result (HbA1c) is associated with presence of the disease, and the larger the PLR, the greater the likelihood of disease.
As seen in Table 1, HbA1c was the only variable where both the baseline and baseline to 6-month change were significantly different between those participants who lost glycemic control and those who did not. Furthermore, if both baseline and the change from baseline to 6 months in HbA1c were included in the same model, both remained significant (all P < 0.0001), meaning that the change in HbA1c over the first 6 months added to the prediction of the durability of glycemic control. As indicated by a higher Wald chi-square test value (41.7 for baseline HbA1c vs 17.8 for the baseline to 6-month change; Table 3), the baseline HbA1c value was the stronger predictor.
Table 3.
Multivariable model for early loss of glycemic control with baseline HbA1c and baseline to 6 months change in HbA1c
| Multivariable model | OR | 95% CI | Wald chi-square | P value |
|---|---|---|---|---|
| HbA1c (per 1 %) | ||||
| Baseline | 3.15 | 2.22-4.46 | 41.7 | < 0.0001 |
| Baseline to 6 months change | 2.31 | 1.56-3.40 | 17.8 | < 0.0001 |
Odds ratios (95% CI) from multivariable logistic regression model. Covariates are listed in order of significance as indicated by the Wald chi-square value.
To evaluate the effect of baseline predictors other than HbA1c on glycemic failure, a multivariable model including all significant (P value < 0.05) baseline predictors identified in the univariate analysis (Table 1) was examined. A backward stepwise predictor selection was applied, which gradually eliminated covariates using P value > 0.1 as the criterion for removal. Four significant predictors were left in the model after backward elimination and, as shown in Table 4, the strongest baseline predictor of glycemic failure in the final multivariable model was insulin secretion (C-peptide index; Wald statistic = 21.5), followed by HbA1c (Wald statistic = 11.0), the proinsulin to insulin ratio (Wald statistic = 8.5), a marker of dysfunction of the insulin secretory pathway, and maternal history of diabetes (Wald statistic = 3.7).
Table 4.
Multivariable model for loss of glycemic control during 2004-2014 using baseline (randomization) predictors identified via backward stepwise elimination
| Multivariable model with baseline risk factors | OR | 95% CI | Wald chi-square | P value |
|---|---|---|---|---|
| C-peptide index (ng/mL per mg/dL) | 0.48 | 0.35-0.66 | 21.5 | < 0.0001 |
| HbA1c (%) | 1.89 | 1.30-2.75 | 11.0 | 0.0009 |
| Proinsulin to insulin ratio (pM/µU) | 1.52 | 1.15-2.02 | 8.5 | 0.003 |
| Maternal history of diabetes | 1.51 | 1.00-2.31 | 3.7 | 0.05 |
Odds ratios (95% CI) from multivariable logistic regression model. Variables entered in the model: duration of diabetes, maternal history of diabetes, maternal diabetes during pregnancy, HbA1c, proinsulin to insulin ratio, C-peptide index, and C-peptide oDI. The significant covariates (P < 0.05) remaining after the backward stepwise elimination are shown in the table. Covariates are listed in order of significance as indicated by the Wald chi-square value. There were no multicollinearity issues identified (all VIF values < 10).
Next, we explored the glycemic trajectories of individuals who maintained glycemic control for the first 48 months of participation in TODAY to determine if there was subsequent loss of control and, if so, examined the predictors of loss of subsequent control. As seen in Fig. 2, of the 264 participants who continued to have good glycemic control at 48 months, 90 (34%) had subsequent loss of glycemic control, while the remainder (174; 66%) maintained glycemic control through the end of the analysis period in 2014 (5 to 9 years of follow-up). Separate univariable models (Table 5) compares these 2 groups by predictors at baseline, at 48 months, and change in predictors between baseline and 48 months. At baseline, the predictors of late (after 48 months) decompensation were again markers of beta-cell function (C-peptide index, coDI) and the proinsulin/insulin ratio, but not proinsulin concentration itself. However, participants who reached glycemic failure within the first 48 months (n = 320) had higher HbA1c, proinsulin, proinsulin to insulin ratio and lower beta-cell function (C-peptide index and coDI) at baseline when compared to those participants who reached glycemic failure after 48 months (n = 90) (Table 6).
Table 5.
Predictors (at baseline, at month 48, and during the first 48 months in study) of late loss of glycemic control among the n = 264 durable responders (n = 174 with no glycemic failure vs n = 90 with subsequent failure post month 48), from univariable models
| Baseline predictors | 48-month predictors | Baseline to 48 month changes predictors | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No failure (n = 174) | Glycemic failure (n = 90) | R2 (%) | P value | No failure (n = 174) | Glycemic failure (n = 90) | R 2 (%) | P value | No failure (n = 174) | Glycemic failure (n = 90) | R 2 (%) | P valuee | |
| Weight (kg) | 97.6 ± 24.8 | 92.3 ± 21.0 | 1.2 | 0.08 | 108.8 ± 27.8 | 104.8 ± 22.8 | <1 | 0.28 | 10.5 ± 13.1 | 14.7 ± 11.5 | 2.4 | 0.02 |
| BMI (kg/m2) | 35.2 ± 7.9 | 34.1 ± 6.8 | <1 | 0.25 | 37.9 ± 9.3 | 37.6 ± 7.6 | <1 | 0.77 | 2.7 ± 4.4 | 3.8 ± 3.9 | 1.3 | 0.10 |
| Waist circumference (cm) | 109.2 ± 15.4 | 108.2 ± 16.6 | <1 | 0.62 | 115.3 ± 17.4 | 111.7 ± 17.1 | <1 | 0.32 | 5.9 ± 10.7 | 8.0 ± 7.1 | <1 | 0.32 |
| HbA1c (%) | 5.7 ± 0.5 | 5.7 ± 0.6 | <1 | 0.40 | 5.7 ± 0.7 | 6.6 ± 1.1 | 18.9 | < 0.0001 | 0.10 ± 0.61 | 0.89 ± 1.14 | 17.1 | < 0.0001 |
| C-peptide (ng/mL) | 3.8 ± 1.5 | 3.8 ± 1.4 | <1 | 0.72 | 3.7 ± 1.6 | 3.8 ± 1.7 | <1 | 0.71 | -0.13 ± 1.36 | 0.03 ± 1.49 | <1 | 0.41 |
| Proinsulin (pM) | 26.4 ± 19.1 | 29.6 ± 24.4 | <1 | 0.13 | 28.4 ± 24.1 | 45.9 ± 43.1 | 6.8 | 0.0001 | 2.18 ± 24.11 | 18.26 ± 45.46 | 5.1 | 0.002 |
| Proinsulin to insulin ratio (pM/µU) | 0.9 ± 0.7 | 1.4 ± 1.5 | 2.5 | 0.01 | 1.0 ± 1.0 | 1.3 ± 0.8 | 7.3 | < 0.0001 | 0.04 ± 1.13 | -0.08 ± 1.54 | <1 | 0.53 |
| Insulin sensitivity (mL/µU) | 0.047 ± 0.035 | 0.054 ± 0.053 | <1 | 0.34 | 0.043 ± 0.028 | 0.039 ± 0.023 | <1 | 0.38 | -0.003 ± 0.033 | -0.016 ± 0.049 | 2.5 | 0.02 |
| C-peptide index (ng/mL per mg/dL) | 0.128 ± 0.200 | 0.088 ± 0.075 | 4.0 | 0.002 | 0.098 ± 0.135 | 0.036 ± 0.039 | 24.3 | < 0.0001 | -0.019 ± 0.113 | -0.042 ± 0.045 | 1.5 | 0.12 |
| C-peptide oDI (1/IF x ∆ C30/∆G30) | 0.006 ± 0.012 | 0.004 ± 0.004 | 1.7 | 0.04 | 0.004 ± 0.005 | 0.001 ± 0.002 | 22.5 | < 0.0001 | -0.001 ± 0.006 | -0.002 ± 0.004 | <1 | 0.24 |
Data are mean ± SD. P values comparing glycemic failure groups are obtained from logistic regression models. The R-square (R2) is indicative of overall fit of the regression, with higher values meaning higher predictive ability. Results were unaffected after adjustment for randomized treatment group.
Table 6.
Baseline predictors of glycemic failure comparing the n = 320 TODAY participants who failed within the first 48 months in study and the n = 90 with subsequent failure post month 48
| Baseline predictors | Early failure (n = 320) | Late failure (n = 90) | P value |
|---|---|---|---|
| Female (%) | 64.1 | 64.4 | 0.95 |
| Age (years) | 14.0 ± 2.1 | 13.6 ± 2.1 | 0.07 |
| Race-ethnicity (%) | |||
| Non-Hispanic Black | 37.5 | 28.9 | 0.42 |
| Hispanic | 38.4 | 46.7 | |
| Non-Hispanic White | 17.5 | 18.9 | |
| Other | 6.6 | 5.6 | |
| Duration of diabetes (months) | 8.4 ± 6.1 | 8.0 ± 6.3 | 0.16 |
| Pubertal Tanner stage 4-5 (%) | 90.9% | 84.4% | 0.08 |
| Family history of diabetes (%) | 66.6% | 56.8% | 0.09 |
| Maternal history of diabetes (%) | 63.8% | 58.1% | 0.34 |
| Maternal diabetes during pregnancy (%) | 36.7% | 35.7% | 0.87 |
| Depressive symptoms (%) | 16.1% | 17.2% | 0.79 |
| Weight (kg) | 96.6 ± 25.8 | 92.3 ± 21.0 | 0.26 |
| BMI (kg/m2) | 35.3 ± 7.6 | 34.1 ± 6.8 | 0.17 |
| Waist circumference (cm) | 109.8 ± 17.0 | 108.2 ± 16.6 | 0.34 |
| HbA1c (%) | 6.3 ± 0.8 | 5.7 ± 0.6 | < 0.0001 |
| C-peptide (ng/mL) | 3.9 ± 1.6 | 3.8 ± 1.4 | 0.40 |
| Proinsulin (pM) | 45.1 ± 44.5 | 29.6 ± 24.4 | 0.0001 |
| Proinsulin to insulin ratio (pM/µU) | 1.7 ± 1.7 | 1.4 ± 1.5 | 0.0009 |
| Proinsulin to C-peptide ratio (pM/ng/mL) | 10.9 ± 7.2 | 7.7 ± 4.4 | < 0.0001 |
| Insulin sensitivity (1/IF, mL/µU) | 0.050 ± 0.051 | 0.054 ± 0.053 | 0.30 |
| C-peptide Index (∆C30/∆G30, ng/mL per mg/dL) | 0.056 ± 0.056 | 0.088 ± 0.075 | < 0.0001 |
| C-peptide oDI (1/IF x ∆ C30/∆G30) | 0.002 ± 0.003 | 0.004 ± 0.004 | < 0.0001 |
| Glucose total 2h AUC (mg/dL) | 420.4 ± 99.4 | 361.9 ± 70.3 | < 0.0001 |
| C-peptide total 2h AUC (ng/mL) | 16.2 ± 6.6 | 20.1 ± 8.1 | < 0.0001 |
| C-peptide total 2h AUC/ Glucose total 2h AUC (ng/mL per mg/dL) | 0.042 ± 0.022 | 0.058 ± 0.027 | < 0.0001 |
Data are mean ± SD. P values comparing glycemic failure groups are obtained from logistic regression models.
Abbreviations: AUC, area under the curve; BMI, body mass index; HbA1c, glycated hemoglobin A1c.
HbA1c at 48 months was a strong predictor of subsequent failure, as was seen for baseline HbA1c and early failure. Markers of decreased beta-cell function (C-peptide index and coDI) at 48 months and the proinsulin/insulin ratio remained predictive. Interestingly, proinsulin at 48 months, but not baseline proinsulin, predicted subsequent failure, analogous to the prediction of earlier failure by baseline proinsulin. Among measures of change between baseline and 48 months, increased weight, increased HbA1c, increased proinsulin, and decreased insulin sensitivity were significant predictors, with change in HbA1c and proinsulin the strongest among these. Results were unaffected when models were adjusted for randomized treatment group.
Because HbA1c is easier to obtain in the clinical setting than other measures of change, we further explored the characteristics of HbA1c change over time as a predictor of subsequent loss of glycemic control. Fig. 3A compares the HbA1c trajectory for the 410 participants who lost glycemic control during TODAY and TODAY2 relative to their time of glycemic failure vs the trajectory for the 174 participants who remained in control. Participants who never lost glycemic control had no appreciable change in HbA1c over time, while those who lost control had a gradual rise in HbA1c for many months but experienced accelerated increase in HbA1c approximately 3 months prior to loss of control. This figure also shows that among those who ultimately lost glycemic control, the gradual rise in HbA1c was generally within the HbA1c range (< 7%) that would meet American Diabetes Association treatment targets. Table 7 presents the results of ROC analysis to identify the change in HbA1c within 3, 6, 9, or 12 months relative to time of glycemic failure or study midpoint that was a predictor of loss of glycemic control. As shown, an increase in HbA1c of 0.6% within 3 months or an increase in HbA1c of 0.8% within 6 months were very strong predictors of loss of glycemic control. In addition, Fig. 3B shows that among participants who lost glycemic control at any point during the study, 84.8% experienced a rise in HbA1c of > 0.5% within 3 months prior to meeting the criteria of loss of glycemic control and 94.0% experienced such a rise within 6 months, irrespective of starting HbA1c value.
Figure 3.
Mean HbA1c trajectory within 2 years of glycemic failure or study midpoint by loss of glycemic control status (Panel A) and percent of participants who lost glycemic control with a change in HbA1c prior to time of failure > 0.5% (Panel B)*. *Mean HbA1c (± standard error bars) is plotted relative to time of glycemic failure for the n = 410 participants who lost glycemic control during 2004-2014 and is plotted relative to the study midpoint for the n = 174 participants who did not lose glycemic control during 2004-2014. Panel B: percent of participants who lost glycemic control with a change in HbA1c prior to time of glycemic failure > 0.5%; 84.8% of the participants with loss of control had an increase in HbA1c > 0.5% within 3 months of their failure start date; 9.2% of participants an increase in HbA1c > 0.5% within 6 months of their failure start date; 4.2% of participants had an increase in HbA1c > 0.5% within 9 months of their failure start date; and 1.8% of participants had an increase in HbA1c > 12 months of their failure start date. The 4 categories in the bar graph above are mutually exclusive.
Table 7.
Change in HbA1c from time of glycemic failure/study midpoint cutoffs, area under the curve (AUC) and its 95% Wald CI, within 3, 6, 9, and 12 months of glycemic failure/study midpoint*
| Groups | N | Change in HbA1c cutoff | AUC |
|---|---|---|---|
| Within 3 months | 493 | +0.6 | 0.97 (0.95-0.98) |
| Within 6 months | 470 | +0.8 | 0.91 (0.89-0.94) |
| Within 9 months | 431 | +0.6 | 0.92 (0.89-0.95) |
| Within 12 months | 409 | +0.7 | 0.96 (0.94-0.98) |
*Not all the n = 584 participants included in the analysis (n = 174 without loss of glycemic control during 2004-2014 plus n = 410 with loss of glycemic control during 2004-2014; Figure 1) had an HbA1c collected within 3, 6, 9, or 12 months relative to their time of failure or study midpoint (eg, n = 493 within 3 months, n = 470 within 6 months). For participants who lost glycemic control (ie, reached glycemic failure), change in HbA1c is defined as the difference between HbA1c values collected at a given time point prior to failure and the time point when failure was reached (eg, within 3 months of glycemic failure); for participants who did not lose glycemic control during 2004-2014, change in HbA1c is defined as the difference between HbA1c values collected at a given time prior to the participant’s midpoint in the study and the participant’s study midpoint. AUC is a measure of diagnostic accuracy ranging from 0 to 1, with AUC values > 0.9 indicating very good ability to predict glycemic failure.
Discussion
We have previously demonstrated that the population of individuals with youth-onset type 2 diabetes is heterogeneous (1, 9), with 2 large subsets characterized as those at risk of losing glycemic control rapidly and those who will maintain glycemic control for a more prolonged period. In our previous analysis, we examined baseline anthropometric and metabolic characteristics that distinguish between groups that did or did not lose glycemic control within 48 months after randomization into the TODAY study and reported that HbA1c after a few months of metformin monotherapy during run-in was a predictor of risk for loss of glycemic control (9). We now extend this analysis using a larger set of data assessed for 5 to 9 years of follow-up to describe predictors of loss of glycemic control beyond 48 months of treatment, as well as examining longitudinal predictors of loss of glycemic control not included in the previous analysis. This analysis extends the previous analysis by confirming that individuals at high risk for loss of glycemic control—now out as long as 9 years—can be identified based on easily obtainable clinical characteristics soon after diagnosis; that a third of youth with good glycemic control 4 years after diagnosis will have subsequent loss of control that can be predicted based on both baseline and longitudinal clinical characteristics; and that an absolute rise in HbA1c of 0.5%, even within the normal range, is strongly associated with rapid loss of glycemic control.
As in the previous analysis to 48 months, we report that the predictors of later loss of glycemic control include measures of insulin secretion (C-peptide index) and glycemia after a few months of metformin monotherapy (baseline HbA1c), confirming our previous report (9). In addition, in this analysis, we extend the findings with 2 metrics not included in the previous analysis, proinsulin/insulin (a measure of insulin processing) and maternal history of diabetes and demonstrate that these are also predictive of whether an individual will have durable glycemic control in multivariate analysis, although with lesser impact than either insulin secretion or HbA1c. Furthermore, we examined the impact of changes in metrics over the first 6 months after randomization to determine if the response during the initial period of treatment conferred added ability to predict glycemic control. This analysis demonstrated that, while both baseline HbA1c and change in HbA1c over the first 6 months of treatment remained significant when included in the same multivariable model, baseline HbA1c was the stronger predictor. Since we have previously demonstrated that insulin secretion is the strongest determinant of HbA1c in youth with type 2 diabetes (9), these data suggest that HbA1c after a few months on metformin is a useful indicator of insulin secretion, which is subsequently reflected in glycemic outcomes evaluated after up to 9 years of treatment.
We were next interested in the subsequent outcomes of youth who had demonstrated durable glycemic control at 48 months to determine if there was subsequent loss of control and whether this loss of control could be predicted. We now report that approximately 30% of youth who remained in control at 48 months subsequently lost glycemic control. Late loss of glycemic control (after 48 months) after initial durable control within 48 months since randomization was best predicted by measures of insulin secretion (C-peptide index and coDI) at baseline, though not baseline HbA1c. Importantly, metrics at 48 months, including HbA1c, reduced insulin secretion, and impaired insulin processing (proinsulin/insulin ratio), as well as rise in HbA1c and proinsulin between baseline and 48 months, were stronger predictors than any baseline measure. Of particular interest is that, while elevated proinsulin at baseline was a predictor of glycemic failure early in the course of diabetes, it was not a predictor of late loss of glycemic control (ie, after 48 months since randomization) whereas proinsulin concentration at 48 months was a predictor of subsequent late loss of control. These results suggest that proinsulin, as a measure of dysfunctional insulin processing, may be a more short-term predictor of glycemia than other measures of insulin secretory capacity, perhaps suggesting that dysfunctional insulin processing is an early step in loss of beta-cell function that is followed by falling insulin secretory capacity (19-22).
The observation that change in HbA1c over the first 6 months of therapy contributed to the ability to predict subsequent glycemic outcomes prompted a broader exploration of the implications of change in HbA1c. We demonstrate that an HbA1c that rises more than 0.5% over any interval is predictive of subsequent loss of glycemic control. While it may seem obvious that a rising HbA1c is predictive of a subsequent higher HbA1c, it is critical to note that the predicted loss of control occurred within 3 to 6 months and was irrespective of the starting HbA1c, meaning that a rising HbA1c, even within the non-diabetes range—ie, within current treatment targets—is a cause for concern. Stated differently, a rise in HbA1c of more than 0.5% was not simply associated with continued gradual increase in HbA1c over time, but rather, predicts rapid decompensation even if the overall glycemic control appears initially acceptable. If confirmed, these findings would support earlier intensification of antihyperglycemic therapy in youth with type 2 diabetes who are experiencing HbA1c changes even within currently accepted target ranges.
Limitations
These analyses have several important strengths. First, the TODAY cohort is a large multi-ethnic cohort that closely reflects the demographics of youth-onset type 2 diabetes in the United States. Second, the participants in TODAY have been extensively characterized, allowing analysis of a broad set of demographic, anthropometric, and biochemical measures. Third, the cohort was followed longitudinally with robust outcome measures allowing rigorous assessment of factors associated with differing degree of glycemic control. However, despite the relatively large size of the TODAY cohort, the sample size is too small for more than exploratory analysis of sex by-racial/ethnic subgroups. Also, the fact that most TODAY participants were obese at study entry and remained obese throughout the study with a relatively narrow range of distribution of weight and BMI may have limited our ability to study the effect of weight in relation to glycemic control. Furthermore, newer diabetes medications, such as glucagon-like peptide-1 (GLP-1) receptor agonists, dual GLP-1/glucose-dependent insulinotropic polypeptide (GIP) agents, and sodium-glucose transporter-transporter 2 inhibitors (SGLT2i) were not available when TODAY was designed, and we are unable to determine the impact of these agents on glycemic trajectories in the TODAY cohort. Finally, approximately 40% of participants had received treatments other than metformin monotherapy prior to run-in. However, the distribution of treatments at the time of screening reflects treatment practices in the US (23) and, therefore, we believe that the analysis of the predictive implications of HbA1c after stabilization on metformin monotherapy in this cohort remains clinically meaningful. We did not undertake further analysis of the potential impact of these prior treatments because of limited information on the variables leading to the prior treatment.
Conclusion
In summary, the analyses reported here add to our knowledge of the heterogeneity of the population of youth with type 2 diabetes. In particular, these analyses demonstrate 3 potentially clinically important findings. First, those individuals at high risk for loss of glycemic control—as far out as 9 years from diagnosis—can be identified based on easily obtainable clinical characteristics after a short course of metformin following diagnosis. Second, approximately a third of youth with good glycemic control 4 years after diagnosis will have subsequent loss of control that can be predicted based on both baseline and ongoing clinical characteristics. Together, these findings indicate that clinicians can identify youth who warrant closer monitoring, even if initial control is acceptable. In addition, these analyses demonstrate that an absolute rise in HbA1c of 0.5%, even within the normal range, is strongly associated with rapid loss of glycemic control within as little as 3 to 6 months, providing an opportunity for clinicians to proactively intensify therapy and avoid metabolic decompensation. Finally, these results, by identifying those individuals who will need aggressive intervention, can facilitate research on the design and use of better therapeutic approaches to this high-risk population.
Acknowledgments
We thank all members of the TODAY Study Group not included among the authors: a complete list of individuals can be found at: https://www.nejm.org/doi/suppl/10.1056/NEJMoa2100165/suppl_file/nejmoa2100165_appendix.pdf
Glossary
Abbreviations
- AUC
area under the curve
- BMI
body mass index
- coDI
C-peptide-based oral disposition index
- DXA
dual-energy x-ray absorptiometry
- HbA1c
glycated hemoglobin A1c
- OGTT
oral glucose tolerance test
- PLR
positive likelihood ratio
- ROC
receiver operating characteristics curve
- TODAY
Treatment Options for type 2 Diabetes in Adolescents and Youth
Contributor Information
Philip Zeitler, University of Colorado Anschutz Medical Campus, Aurora, CO 80045, USA.
Laure El Ghormli, The Biostatistics Center, George Washington University, Rockville, MD 20852, USA.
Silva Arslanian, University of Pittsburgh, UPMC Children’s Hospital of Pittsburgh, Pittsburgh, PA 15213, USA.
Sonia Caprio, Yale University, New Haven, CT 06520, USA.
Elvira Isganaitis, Joslin Diabetes Center, Boston, MA 02215, USA.
Megan K Kelsey, University of Colorado Anschutz Medical Campus, Aurora, CO 80045, USA.
Ruth S Weinstock, State University of New York Upstate Medical University, Syracuse, NY 13210, USA.
Neil H White, Washington University in St. Louis School of Medicine, St. Louis, MO 63110, USA.
Kimberly Drews, The Biostatistics Center, George Washington University, Rockville, MD 20852, USA.
Funding Support
Funding Support. This work was completed with funding from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and the NIH Office of the Director (OD) through grants U01-DK61212, U01-DK61230, U01-DK61239, U01-DK61242, and U01-DK61254; from the National Center for Research Resources General Clinical Research Centers Program grant numbers M01-RR00036 (Washington University School of Medicine), M01-RR00043-45 (Children’s Hospital Los Angeles), M01-RR00069 (University of Colorado Denver), M01-RR00084 (Children’s Hospital of Pittsburgh), M01-RR01066 (Massachusetts General Hospital), M01-RR00125 (Yale University), and M01-RR14467 (University of Oklahoma Health Sciences Center); and from the NCRR Clinical and Translational Science Awards grant numbers UL1-TR000003 (Children’s Hospital of Philadelphia), UL1-TR001863 (Yale University), UL1-TR001857 (Children’s Hospital of Pittsburgh), UL1-TR002548 (Case Western Reserve University), UL1-TR002345 (Washington University in St Louis), UL1-TR002541 (Massachusetts General Hospital), and UL1-TR002535 (University of Colorado Denver). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Author Contributions
P.Z. designed and wrote the manuscript. S.A., S.C., E.I., M.K.K., R.S.W., N.H.W., and K.D. contributed to interpretation of data, reviewed, and edited the manuscript. L.E. conducted the statistical analyses, prepared the tables and figures, and wrote portions of the manuscript. All the included authors contributed significantly to accomplish this study. P.Z. and L.E. had full access to all data in the study and take responsibility for the integrity and accuracy of the data analysis.
Industry Contributions
The TODAY Study Group thanks the following companies for donations in support of the study’s efforts: Becton, Dickinson and Company; Bristol-Myers Squibb; Eli Lilly and Company; GlaxoSmithKline; LifeScan, Inc.; Pfizer; Sanofi Aventis. We also gratefully acknowledge the participation and guidance of the American Indian partners associated with the clinical center located at the University of Oklahoma Health Sciences Center, including members of the Absentee Shawnee Tribe, Cherokee Nation, Chickasaw Nation, Choctaw Nation of Oklahoma, and Oklahoma City Area Indian Health Service; the opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the respective Tribes and the Indian Health Service.
Clinical Trial Information
ClinicalTrials.gov registration no. NCT00081328.
Disclosures
None of the authors reported a conflict of interest.
Data Availability
Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in References. Data collected for the TODAY study are available to the public through the NIDDK Repository (https://repository.niddk.nih.gov/studies/today/).
References
- 1. TODAY Study Group. A clinical trial to maintain glycemic control in youth with type 2 diabetes. NEJM. 2012;366:2247-2256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Turner RC, Cull CA, Frighi V, Holman RR; for the UK Prospective Diabetes Study Group. Glycemic control with diet, sulfonylurea, metformin, or insulin in patients with type 2 diabetes mellitus: progressive requirement for multiple therapies. JAMA. 1999;281(21):2005-2012. [DOI] [PubMed] [Google Scholar]
- 3. Brown JB, Conner C, Nichols GA. Secondary failure of metformin monotherapy in clinical practice. Diabetes Care. 2010;33(3):501-506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Constantino MI, Molyneaux L, Limacher-Gisler F, et al. Long-term complications and mortality in young-onset diabetes: type 2 diabetes is more hazardous and lethal than type 1 diabetes. Diabetes Care. 2013;36(12):3863-3869. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Katz LEL, Magge SN, Hernandez ML, Murphy KM, McKnight HM, Lipman T. Glycemic control in youth with type 2 diabetes declines as early as two years after diagnosis. J Pediatrics. 2011;158(1):106-111. [DOI] [PubMed] [Google Scholar]
- 6. UKPDS Group. Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes. Lancet. 1998;352(9139):854-865. [PubMed] [Google Scholar]
- 7. Ostgren CJ, Lindblad U, Ranstam J, Melander A, Rastam L. Glycaemic control, disease duration and β-cell function in patients with type 2 diabetes in a Swedish community: Skaraborg hypertension and diabetes project. Diabet Med. 2002;19(2):125-129. [DOI] [PubMed] [Google Scholar]
- 8. TODAY Study Group, Bjornstad P, et al. Long-term complications in youth-onset type 2 diabetes. N Engl J Med. 2021;385(5):416-426. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Zeitler P, Hirst K, Copeland KC, et al. TODAY Study Group. HbA1c After a Short Period of Monotherapy With Metformin Identifies Durable Glycemic Control Among Adolescents With Type 2 Diabetes. Diabetes Care. 2015;38(12):2285-2292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. TODAY Study Group. Treatment options for type 2 diabetes in adolescents and youth: a study of the comparative efficacy of metformin alone or in combination with rosiglitazone or lifestyle intervention in adolescents with type 2 diabetes. Pediatr Diabetes. 2007;8(2):74-87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. American Diabetes Association. Diagnosis and classification of diabetes. Diabetes Care. 2005;28(Suppl 1):S37-S42. [DOI] [PubMed] [Google Scholar]
- 12. Klingensmith GJ, Pyle L, Arslanian S, et al. TODAY Study Group. The presence of GAD and IA-2 antibodies in youth with a type 2 diabetes phenotype results from the TODAY study. Diabetes Care. 2010;33(9):1970-1975. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Laffel L, Chang N, Grey M, et al. TODAY Study Group. Metformin monotherapy in youth with recent onset type 2 diabetes: experience from the prerandomization run-in phase of the TODAY study. Pediatr Diabetes. 2012;13(5):369-375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Kovacs M. Children’s Depression Inventory. New York: Multi-Health Systems; 1992. [Google Scholar]
- 15. Beck AT, Steer RA, Brown GK. Manual for the Beck Depression Inventory-II. San Antonio, TX: Psychological Corporation; 1996. [Google Scholar]
- 16. TODAY Study Group. Effects of metformin, metformin plus rosiglitazone, and metformin plus lifestyle on insulin sensitivity and β-cell function in TODAY. Diabetes Care. 2013;36(6):1749-1757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kumar R, Indrayan A. Receiver operating characteristic (ROC) curve for medical researchers. Indian Pediatr. 2011;48(4):277-287. [DOI] [PubMed] [Google Scholar]
- 18. Youden WJ. Index for rating diagnostic tests. Cancer. 1950;3(1):32-35. [DOI] [PubMed] [Google Scholar]
- 19. Riahi Y, Israeli T, Cerasi E, Leibowitz G. Effects of proinsulin misfolding on β-cell dynamics, differentiation and function in diabetes. Diabetes Obes Metab. 2018;20(Suppl 2):95-103. [DOI] [PubMed] [Google Scholar]
- 20. Sun J, Cui J, He Q, Chen Z, Arvan P, Liu M. Proinsulin misfolding and endoplasmic reticulum stress during the development and progression of diabetes. Mol Aspects Med. 2015;42:105-118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Egan AM, Laurenti MC, Hurtado Andrade MD, et al. Limitations of the fasting proinsulin to insulin ratio as a measure of β-cell health in people with and without impaired glucose tolerance. Eur J Clin Invest. 2021;51(6):e13469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Tura A, Pacini G, Kautzky-Willer A, Ludvik B, Prager R, Thomaseth K. Basal and dynamic proinsulin-insulin relationship to assess beta-cell function during OGTT in metabolic disorders. Am J Physiol Endocrinol Metab. 2003;285(1): E155-E162. [DOI] [PubMed] [Google Scholar]
- 23. Amutha A, Praveen PA, Hockett CW, et al. Treatment regimens and glycosylated hemoglobin levels in youth with Type 1 and Type 2 diabetes: Data from SEARCH (United States) and YDR (India) registries. Pediatr Diabetes. 2021;22(1):31-39. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in References. Data collected for the TODAY study are available to the public through the NIDDK Repository (https://repository.niddk.nih.gov/studies/today/).
