BACKGROUND: Resistance training (RT) is effective for glycemic control in type 2 diabetes mellitus (T2DM) patients. However, the characteristics of an RT program that will maximize its effect and those of patients that will especially benefit from RT are unknown. OBJECTIVE: The objectives of this systematic review were to identify via a comprehensive meta-analysis the characteristics of an RT program for patients with T2DM that might increase the patients' improvement in glycemic control and the characteristics of patients that will benefit from RT. DATA SOURCES: Electronic-based literature searches of MEDLINE and EMBASE entries from 1 January 1966 to 25 August 2014 were conducted to identify clinical trials examining the effect of RT on glycemic control among patients with T2DM. Study keywords were text words and thesaurus terms related to RT and T2DM. STUDY SELECTION: Studies were included if they (1) were clinical trials consisting of two groups with and without RT exercise intervention; (2) had an intervention period of at least 5 weeks; (3) clarified that all patients had T2DM; and (4) reported or made it possible to estimate the effect size [i.e., change in glycosylated hemoglobin (HbA1c) in the RT group minus that in the control group] and its corresponding standard error. STUDY APPRAISAL AND SYNTHESIS METHODS: The effect size in each study was pooled with a random-effects model. Analyses were stratified by several key characteristics of the patients and RT exercise programs; meta-regression analysis was then used to detect a difference in the effect size among strata within each factor. Linear regression analyses were added by entering each of the following profiles: patients' baseline characteristics [mean baseline age, body mass index (BMI), and HbA1c levels] and exercise characteristics (total sets per week, total sets per bout of exercise, frequency, and intensity). RESULTS: There were 23 eligible studies comprising 954 patients with T2DM. The pooled effect size (95% confidence interval) was -0.34% (-0.53 to -0.16). A program with multiple sets (≥21 vs. <21) per one RT bout was associated with a large effect size (P = 0.03); however, the linear correlation between the number of sets and effect size was not significant (P = 0.56). A larger effect size was observed in studies with participants with diabetes of a relatively short duration (<6 vs. ≥6 years; P = 0.04) or a high baseline HbA1c [≥7.5% (58 mmol/mol) vs. <7.5 %; P = 0.01] while a smaller effect size was observed in studies with a particularly high mean baseline BMI value (≥32 vs. <32 kg/m(2); P = 0.03). Linear regression analyses predicted that each increment of 1% in the baseline HbA1c would enlarge the effect size by 0.036%, while each increment of 1 kg/m2 in the baseline BMI decreased it by 0.070% in the range between 22.3 and 38.8 kg/m2. CONCLUSION: In terms of glycemic control, RT could be recommended in the early stage of T2DM, especially for patients with relatively poor glycemic control. More benefit would be elicited in less obese patients within a limited range of the BMI. A substantial amount of exercise might be required to stimulate post-exercise glucose uptake, although the dose-dependency was not specifically clarified.