Figure 3.

Association between metformin and all-cause mortality in selected subgroups adjusted for multiplicity. Metformin was only associated with lower all-cause mortality in patients whose HbA1c ≥7% [HR (95% CI) = 0.339 (0.117–0.977); P = 0.045]. The widths of the confidence intervals were adjusted for multiplicity, using the stepwise regression, with a threshold of 0.1. Adjusted covariables included age, gender, body mass index, cigarette smoking, systemic blood pressure, diastolic blood pressure, New York Heart Association class, left ventricular ejection fraction, duration of diabetes and heart failure, whether living with hypertension, atrial fibrillation, coronary heart disease, or cerebral infarction, glycated hemoglobin, estimated glomerular filtration rate, hemoglobin, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglyceride, left atrial, sulfonylureas, glinides, glucosidase inhibitors, insulin.