Nine RCTs (n=1,701) were included in the review.
Validity assessment: Two studies did not use adequate allocation concealment and two studies did not have clear criteria for handling withdrawals.
Three-year mortality: The pooled odds ratio showed a statistically significant benefit of adjuvant radiotherapy compared to surgery alone (odds ratio 0.67, 95% CI 0.55 to 0.82, p=0.0001, number needed to treat was 14, no significant heterogeneity; nine RCTs). Sensitivity analyses did not change the result.
Five-year mortality: The pooled odds ratio again showed a statistically significant benefit of adjuvant radiotherapy over surgery alone (odds ratio 0.54, 95% CI 0.43 to 0.68, p<0.00001, number needed to treat was eight, no significant heterogeneity; seven RCTs). Sensitivity analyses did not change the result.
Subgroup analyses: Subgroup analyses revealed that preoperative radiotherapy was beneficial for both three-year mortality (odds ratio 0.57, 95% CI: 0.43 to 0.76, p=0.0001; four RCTs) and five-year mortality (odds ratio 0.62, 95% CI: 0.46 to 0.84, p=0.002, number needed to treat was 10; four RCTs). Post-operative chemoradiotherapy showed a benefit in five-year mortality (odds ratio 0.45, 95% CI: 0.32 to 0.64, p<0.00001, number needed to treat was six; three RCTS).
Subgroup analyses also indicated that the pooled odds ratio for three-year mortality was significant in RCTs given a dose equivalent of 40Gy or greater, but not in those given a dose equivalent of less than 40Gy. Subgroup analyses by radiation machine all indicated statistically significant benefit.
The meta-regression for three-year mortality showed that the result was affected only by study sample size.
Adverse effects: The authors stated that compliance with preoperative radiotherapy was generally satisfactory, but that compliance with postoperative chemoradiotherapy was not. Analysis indicated no increased risk of postoperative mortality or anastomotic leakage with preoperative radiotherapy. Postoperative chemoradiotherapy, however, was associated with greatly increased risks of toxic effects (odds ratio 4.61, 95% CI: 2.89 to 7.36, p<0.00001; five RCTs).
Publication bias: Both tests for publication bias found no statistical evidence of bias, although the authors stated that the risk was substantial.