Baseline glycated haemoglobin (HbA(1c)) concentrations vary between clinical trials of glucose-lowering agents and this may affect interpretation of clinical efficacy. The objective of this study is to quantify the relationship between baseline HbA(1c) and reduction of HbA(1c) in clinical trials. PubMed literature searches from 1991 to 2007. Randomized controlled studies with placebo-controlled or comparator arms [> or = 9 patients in the intent-to-treat (ITT) population] ranging in duration from 23 to 52 weeks, in which baseline and change in glycated haemoglobin (HbA(1c)) were reported. The relationship between baseline HbA(1c) and change in HbA(1c) was analysed by a weighted least-squared regression model accounting for ITT population and variance of HbA(1c) change. Fourteen per cent of independently abstracted studies met the selection criteria. Meta-analysis from 59 clinical trials (8479 patients) produced weighted R(2) of 0.35 (P < 0.0001) for the association of baseline HbA(1c) and absolute change in HbA(1c). Subanalysis of eight metformin clinical trials demonstrated a stronger association [weighted R(2) of 0.67 (P = 0.0130)]. Exclusion of metformin clinical trials from the overall meta-analysis (n = 51) yielded a weighted R(2) of 0.31 (P < 0.0001). Subanalyses of clinical trials of glucose-lowering therapies predominantly targeting fasting (n = 37) or postprandial (n = 22) blood glucose produced weighted R(2) values of 0.27 (P < 0.001) and 0.42 (P < 0.005), respectively. These data demonstrate a positive relationship between baseline HbA(1c) and the magnitude of HbA(1c) change across 10 categories of glucose-lowering therapies, irrespective of class or mode of action. These observations should be considered when assessing clinical efficacy of diabetes therapies derived from clinical trials.
