Eleven RCTs (n=794) were included in the review.
Study quality.
Four of the 11 included studies used an adequate allocation sequence generator; four adequately concealed allocation sequences; blinding was adequate in only one; and follow-up was adequate in all but two. Six (n=451) of the 11 included trials were used in the meta-analyses.
Chemoradiotherapy followed by chemotherapy versus best supportive care.
One small low-quality RCT reported increased survival (HR 0.28, 95% CI: 0.13, 0.60) with combined therapy. Chemoradiotherapy achieved an overall response rate of 31% with a 6.1-month median time to progression. Corresponding data for the best supportive care group were not provided.
Radiotherapy versus chemoradiotherapy. Two RCTs (n=168) reported a significantly longer overall survival with chemoradiotherapy (HR 0.69, 95% CI: 0.51, 0.94), with no difference in disease-free survival or response rates. Haematological toxicity occurred less often with radiotherapy alone.
Chemotherapy versus chemoradiotherapy, followed by chemotherapy (combined modality therapy).
Two of 4 RCTs (n=283) showed comparable overall survival for both groups (HR 0.79, 95% CI: 0.32, 1.95); there was evidence of significant statistical heterogeneity (p=0.01; I-squared 83.4%). Two studies failed to provide sufficient data to be included in the analysis of survival. One study found no significant difference in time to progression between the two groups.
5FU-based chemoradiotherapy followed by chemotherapy versus another agent-based chemoradiotherapy, followed by chemotherapy.
Three trials (n=256) were included, but were too heterogeneous to pool. These compared 5FU-based chemoradiotherapy with gemcitabine alone, with gemcitabine and cisplatin, and with adriamycin, respectively. 5FU and adriamycin were reported to have similar survival rates (HR 0.97, 95% CI: 0.73, 1.29) at the expense of a significantly higher rate of adverse events with adriamycin (p<0.05). Gemcitabine was reported to have improved overall survival, time to progression and response rate in comparison with 5FU; toxicity was similar between the two groups. Gemcitabine and cisplatin chemoradiation had comparable 9-month survival rates to 5FU.
Publication bias.
Funnel plots, which were not shown, suggested publication bias for all comparisons assessed.