Lower Baseline Glycemia Reduces Apparent Oral Agent Glucose-Lowering Efficacy

RESEARCH DESIGN AND METHODS

Using previously described criteria (8), a total of 61 studies (some including more than one active therapy group) were identified involving the five major oral agent classes (sulfonylureas, meglitinides, metformin, thiazolidinediones, α-glucosidase inhibitors), published before 2004. Fifteen of these studies included oral agent combinations. Routinely in pharmacologic studies, a period far less than the required 3 months to ensure stability of the patient’s initial A1C will have elapsed between withdrawal of prior treatment and trial initiation. We therefore felt that use of A1C at the beginning of the therapy period underestimated the extent to which it would be anticipated to increase, were the patient allowed to equilibrate in an untreated state. To address this, we analyzed the decreases from baseline in fasting plasma glucose (FPG) as well as in A1C. The number of subjects in each study was used for a weighted regression analysis with SAS statistical software (SAS Institut, Cary, NC).

RESULTS

Overall, mean (±SD) baseline FPG was 11.2 ± 1.8 mmol/l and baseline A1C 8.6 ± 1.0%. There were no statistical differences between groups when sorted by therapeutic class. A significant correlation was observed between mean baseline FPG and mean decrease in FPG (R² = 0.34, F = 50.19, P < 0.0001; Fig. 1) and between mean baseline A1C and mean decrease in A1C (R² = 0.18, F = 21.20, P < 0.0001; Fig. 2). Notably, the correlation between the change in FPG and baseline FPG is stronger than that between change in A1C and baseline A1C. This likely reflects the more immediate change in FPG during washout and the variable washout periods used in these studies. That is, the “baseline” A1C during a short washout phase may not reliably reflect the baseline glycemic status of the patient.

When active therapy groups were stratified by baseline FPG quartile, the mean decrease in FPG was −0.3 mmol/l for groups with initial glucose <10.2 mmol/l, −1.1 mmol/l for groups with initial glucose 10.2–11.2 mmol/l, −2.4 mmol/l for initial glucose 11.2–12.7 mmol/l, and −2.5 mmol/l for initial glucose >12.7 mmol/l (P for linear trend = 0.04). Additionally, groups with subjects in the lowest quartile of mean baseline FPG were nearly 10 times less likely to show a change in A1C of at least −1.0% compared with groups in the highest quartile (8 vs. 73%, P = 0.002). Significantly greater reductions in FPG and A1C were observed in those groups with higher baseline A1C (Table 1). Indeed, when the baseline A1C was <8.0%, the reduction from active therapy is only 0.1–0.2% greater than in the control group.

CONCLUSIONS

Irrespective of drug class, the baseline glycemic status of patients who have been recruited into clinical trials strongly influences the FPG and A1C reduction following pharmacologic intervention. Our therapeutic assumptions as to the intensiveness of treatment required to attain near-normal glycemic levels may be flawed by undue extrapolation from the effects of these agents on individuals with markedly higher glucose levels. These findings should be taken into account by clinical investigators and study sponsors as the effectiveness of newer antihyperglycemic agents are tested, especially since the mean baseline A1C of clinical trial participants has fallen significantly over the past decade. Moreover, these findings should also be considered by approval agencies in proposing expectations for efficacy of new antihyperglycemic agents being tested in individuals with modest but still important degrees of hyperglycemia.