BACKGROUND: Cardiac resynchronization therapy (CRT) is effective in reducing clinical events in patients with heart failure and prolonged QRS interval. Studies using surrogate measures and subgroup analysis of large trials suggest that only patients with severely prolonged QRS benefit from CRT. Our objective was to determine whether the effect of CRT on adverse clinical events (eg, death, hospitalizations) is different in patients with moderately (ie, 120 to 149 milliseconds) [corrected] vs severely (ie, ≥150 milliseconds) prolonged QRS duration. METHODS: Searches of MEDLINE, SCOPUS, and Cochrane databases were conducted for randomized controlled CRT trials. Trials reporting clinical events according to different QRS ranges were identified. Five randomized trials fulfilling the inclusion criteria (total patients, n = 5813) were included in the meta-analysis. RESULTS: In patients with severely prolonged QRS, there was a reduction in composite clinical events with CRT (risk ratio, 0.60; 95% confidence interval [CI], 0.53-0.67) (P < .001). In contrast, there was no benefit of CRT in patients with moderately prolonged QRS (RR, 0.95; 95% CI, 0.82-1.10) (P = .49), resulting in a significantly different impact of CRT in the 2 QRS groups (P < .001). There was a significant relationship between baseline QRS duration and risk ratio (P < .001) with benefit of CRT appearing at a QRS of approximately 150 milliseconds and above. The differential response of the 2 QRS groups was evident for all New York Heart Association classes. CONCLUSIONS: Cardiac resynchronization therapy was effective in reducing adverse clinical events in patients with heart failure and a baseline QRS interval of 150 milliseconds or greater, but CRT did not reduce events in patients with a QRS of less than 150 milliseconds. These findings have implications for the selection of patients for CRT.