Assessing the potential for precision medicine of glucose-lowering drugs in type 2 diabetes mellitus: a meta-regression analysis considering body weight variability from 120 randomized trials

Abstract Background Thus far, randomized controlled trials (RCTs) of glucose-lowering drugs reporting on treatment effects on body weight (BW) as cardiovascular risk factor in type 2 diabetes mellitus have not included precision medicine approaches. However, taking into account treatment response heterogeneity, their potential may be analyzed from published data of RCTs using meta-regression analyses. Both treatment response variation and predictors are required conditions for precision medicine to work. Purpose To assess the potential benefits of precision medicine by quantifying the variation of body weight as patient response to treatment, and to identify predictors which could explain this variation. Methods We used four recent systematic reviews on RCTs comparing pharmacological treatment of type 2 diabetes (included but not limited to sodium-glucose cotransporter-2 inhibitors, glucagon-like peptide-1 receptor agonists and thiazolidinediones) to placebo. RCTs reporting on BW at baseline and after follow-up to allow for calculation of its logarithmic standard deviation (Log(SD)) in treatment and placebo groups were included. Meta-regression models accounting for treatment means, differing sample sizes, and correlation across treatment from the same trial were used. A separate meta-regression model to evaluate interactions of treatment and individual predictor variables (age, male sex, pre-randomization BW, duration of disease, duration of treatment, year of study and drug class) was applied. Results 120 randomized trials with a total of 43,663 participants were included. The median BW after treatment was 85.2 kg in the placebo groups and 86.2 kg in the therapy groups. A slightly greater treatment response heterogeneity was shown in the therapy groups with a median Log(SD)=2.83 compared to 2.79 from the placebo. After full adjustment in the meta-regression model the difference in BW Log(SD)s was -0.020 (95% CI: -0.063; 0.022) (Figure 1). Scatter plots adjusted for different sample sizes did not show any interaction (i.e. slope divergence) between predictors and the respective treatment (therapy or placebo). Figure 2 demonstrates this using pre-randomization BW as an example. Conclusions We found no major treatment response heterogeneity in RCTs of glucose-lowering drugs reporting on body weight in type 2 diabetes. In addition, none of the considered predictors were able to explain a potential variation in treatment response. The precision medicine appro