Constraint-induced movement therapy (CIMT) is a neurorehabilitation technique designed to improve upper extremity motor functions after stroke. This review aimed to investigate evidence of the effect of CIMT on upper extremity in stroke patients and to identify optimal methods to apply CIMT. Four databases (MEDLINE, EMBASE, CINHAL, and PEDro) and reference lists of relevant articles and reviews were searched. Randomized clinical trials that studied the effect of CIMT on upper extremity outcomes in stroke patients compared with other rehabilitative techniques, usual care, or no intervention were included. Methodological quality was assessed using the PEDro score. The following data were extracted for each trial: patients' characteristics, sample size, eligibility criteria, protocols of CIMT and control groups, outcome measurements, and the PEDro score. A total of 38 trials were identified according to the inclusion criteria. The trials included were heterogeneous in CIMT protocols, time since stroke, and duration and frequency of treatment. The pooled meta-analysis of 36 trials found a heterogeneous significant effect of CIMT on upper extremity. There was no significant effect of CIMT at different durations of follow-up. The majority of included articles did not fulfill powered sample size and quality criteria. The effect of CIMT changed in terms of sample size and quality features of the articles included. These meta-analysis findings indicate that evidence for the superiority of CIMT in comparison with other rehabilitative interventions is weak. Information on the optimal dose of CIMT and optimal time to start CIMT is still limited.