OBJECTIVES: Despite the needs for comparative effects of active treatments in many areas, direct head-to-head comparison trials were not always available. Indirect comparison (Ic) pooling randomized controlled trials (RCTs) of different drugs vs. a common comparator (e.g. placebo) has been used. Overcome the threat to valid comparison due to risk factors difference between trial populations, Ic methods utilized common comparator have been used. This study elucidated the impact of varied risk factors of trials to Ic analysis. METHODS: Simulations were carried out to illustrate the bias and performance of Ic methods in existence of subject level and/or trial level risk factors in a series of replicated trials with binary or continuous endpoints. RESULTS: “Non-collapsibility” resulted in biased association estimate using ratio (e.g. odds ratio) was due to pooling of subjects with varied risk factors. It resulted in biased Ic even though all the risk factors have been balanced out in each RCT. This bias may lead to wrong comparative effect direction in comparing drugs of interest. It could not be corrected by available Ic methods using aggregated data; unless individual level data with all risk factors for all the trials were available. However, when the common comparator had the effect in between the two drugs of interest, Ic methods would reveal correct effect direction regardless non-collapsibility. Effects using mean difference of continuous variables were not influenced by non-collapsibility, thus Ic might provide valid comparative effect estimate. The bias due to trial level background risk which influence all subjects of one trial homogeneously, but varied across trials, could be adjusted by Ic methods, but not by naïve comparison which ignored common comparator. CONCLUSIONS: Understand the way risk factors within and/or between trials, and types of endpoint variable used in trials influenced Ic should improve the accuracies in conducting and interpreting Ic.