Novel Approach for Assessing Outcomes of Type 1 Diabetes Prevention Trials Over a Fixed Time Interval

Abstract

We evaluated whether a binary metabolic end point for change (Δ) from baseline to 1-year postrandomization could be useful in type 1 diabetes (T1D) prevention trials. Using 2-h oral glucose tolerance testing data from the stage 1 participants in the recent abatacept prevention trial and similar participants in the observational TrialNet Pathway to Prevention (PTP) study, we assessed Δmetabolic measures, plotted glucose and C-peptide response curves, and categorized vectors for Δ from baseline to 1 year as metabolic treatment failure versus success. Analyses were validated using the teplizumab prevention study. PTP participants with Δglucose >0 and ΔC-peptide <0 from baseline to 1 year were at substantially higher risk for stage 3 T1D than those with Δglucose <0 and ΔC-peptide >0 (P < 0.0001). Based on this, we compared placebo versus treatment groups in both trials for failure (Δglucose >0 with ΔC-peptide <0) versus success (Δglucose <0 with ΔC-peptide >0) after 1 year. Using this end point, a favorable metabolic impact of abatacept was found after 12 months of treatment. An analytic approach using a binary metabolic end point of failure versus success at a fixed time interval appears to detect treatment effects at least as well as standard primary end points with shorter follow-up.

Article highlights: Challenges in time to event type 1 diabetes (T1D) prevention trial design can yield negative results even for treatments that may actually improve disease pathology. We evaluated whether a binary metabolic end point for 12-month change from baseline to 1 year postrandomization could be useful in T1D prevention trials. This approach detected treatment effects at least as well as standard primary end points with shorter follow-up. Fixed interval metabolic end points should be used in combination with traditional T1D end points to better understand treatment effects of preventive agents.

Figure 1

Dynamic metabolic measures reflect metabolic improvement in the abatacept-treated group compared with decline in the placebo group over the 12 months of active abatacept treatment. A and B: Change in Index60 and change in AUC C-peptide/AUC glucose (AUC ratio). *P < 0.05 with and without adjustment for age, BMI, and baseline DPTRS. C and D: Glucose C-peptide response curves showing plots of mean 30-, 60-, 90-, and 120-min values from OGTT at study entry (baseline, shown in solid line) and at the 12-month time point (dashed line) for each treatment arm. Vectors of change from the curve centroid, or centermost point, are shown in green. n = 104 for placebo, and n = 97 for abatacept arm.

Figure 2

Characterization of GCRC vector directionality is linked to differences in T1D progression. A: Vectors of change in GCRCs can be categorized into four quadrants based on increasing vs. decreasing C-peptide and glucose values. Vector 2 reflects metabolic worsening, and vector 4 reflects metabolic improvement, while vectors 1 and 3 reflect more intermediate changes. B: Survival curve for progression to stage 3 T1D in Pathway to Prevention participants with stage 1 disease who meet inclusion criteria for the abatacept prevention study. Cox regression confirmed that a 12-month GCRC vector 2 is associated with increased progression to stage 3 disease compared with other patterns. Those with vector 4 displayed the lowest risk of progression, while progression rates in those with vector 1 or 3 were more intermediate. Median (interquartile range) follow-up after 12-month vector was 2.9 (1.1, 5.5) years. Hazard ratios (HR) were calculated with and without adjustment for baseline DPTRS, age, and BMI.

Figure 3

Individual GCRC vectors for participants in the abatacept prevention study. Individual 0- to 12-month vectors for each participant are shown for the placebo and abatacept treatment arms as labeled. n = 104 for placebo, and n = 97 for abatacept arm. Vectors reflective of the vector 2 category (metabolic decline and treatment failure) are shown in blue, while vectors reflective of the vector 4 category (metabolic improvement and treatment response) are shown in red. Gray arrows represent vectors in intermediate categories (vectors 1 and 3).

Figure 4

Six-month (6M) WQE and ODE values for abatacept prevention study. Aggregate GCRC directional change was quantified at 6 months between the abatacept and placebo groups using ODE and WQE measurements (lower values are indicative of less progression toward stage 3 T1D). Comparisons were made with and without adjustment for age, BMI, and baseline AUC C-peptide and AUC glucose (P < 0.05). WQE values were lower in the abatacept group at 6 months; however, the difference was not significant (P = 0.33 with adjustment and P = 0.28 without adjustment). ODE values were significantly lower in the abatacept group without and with adjustment (P = 0.025 with adjustment and P = 0.026 without adjustment). n = 104 for the placebo arm, and n = 97 for the abatacept arm. Ten participants did not receive OGTTs at the 6-month time point because of dropout or noncompliance (four abatacept and six placebo; one individual from the placebo group developed stage 3 diabetes).

Figure 5

Survival curve for progression to stage 3 T1D in Pathway to Prevention participants with stage 2 disease who meet inclusion criteria for the teplizumab prevention study. Cox regression confirmed that a 12-month GCRC vector 2 is associated with increased progression to stage 3 disease compared with other patterns. Those with vector 4 displayed the lowest risk of progression, while progression rates in those with vector 1 or 3 were more intermediate. Hazard ratios (HR) were calculated with and without adjustment for baseline DPTRS, age, and BMI. Median (interquartile range) follow-up after 12-month vector was 2.3 (1.0, 4.3) years.

Figure 6

Individual GCRC vectors for participants in the teplizumab prevention study. Individual 0- to 12-month vectors for each participant are shown for the placebo and teplizumab treatment arms. Vectors reflective of the vector 2 category (metabolic decline and treatment failure) are shown in blue, while vectors reflective of the vector 4 category (metabolic improvement and treatment response) are shown in red. Gray arrows represent vectors in intermediate categories (vectors 1 and 3).