We carryied out a meta-analysis of studies on exposure to polychlorinated biphenyls (PCBs) and risk of non-Hodgkin lymphoma (NHL). Through a systematic search of the literature, we identified relative risks (RRs) for PCB exposure and NHL risk in 30 populations (10 occupational exposure, seven high environmental exposure, 13 without special exposure). We performed random effects meta-analyses for exposure to all PCBs, specific PCB congeners and risk of all NHL and NHL subtypes. The meta-RR for studies of occupational exposure, high environmental exposure, and no special exposure were 0.94 [95% confidence interval (CI): 0.84-1.03], 1.05 (95% CI: 0.94-1.16), and 1.03 (95% CI: 0.72-1.34), respectively, and the cumulative meta-RR was 0.96 (95% CI: 0.85-1.07). No positive associations were found for exposure to specific congeners, nor for NHL subtypes. The meta-RR for an increase of 100 ppb serum or fat PCB level was 1.02 (95% CI: 1.00-1.04). There was weak indication of publication bias. Our meta-analysis found no association between PCB exposure and NHL risk, in particular in studies of occupational exposures. We detected a weak dose-response relation; the possibility of residual confounding and other sources of bias cannot be ruled out. PCBs are not likely to cause NHL in humans.