Results from epidemiological studies exploring the association between childhood lymphoma and maternal smoking during pregnancy have been contradictory. This meta‐analysis included all published cohort (n = 2) and case–control (n = 10) articles; among the latter, the data of the Greek Nationwide Registry for Childhood Hematological Malignancies study were updated to include all recently available cases (‐2008). Odds ratios (ORs), relative risks and hazard ratios were appropriately pooled in three separate analyses concerning non‐Hodgkin lymphoma (NHL, n = 1,072 cases), Hodgkin lymphoma (HL, n = 538 cases) and any lymphoma (n = 1,591 cases), according to data availability in the included studies. An additional metaregression analysis was conducted to explore dose–response relationships. A statistically significant association between maternal smoking (any vs. no) during pregnancy and risk for childhood NHL was observed (OR = 1.22, 95% confidence interval, CI: 1.03–1.45, fixed effects model), whereas the risk for childhood HL was not statistically significant (OR = 0.90, 95% CI: 0.66–1.21, fixed effects model). The analysis on any lymphoma did not reach statistical significance (OR = 1.10, 95% CI = 0.96–1.27, fixed effects model), possibly because of the case‐mix of NHL to HL. No dose–response association was revealed in the metaregression analysis. In conclusion, this meta‐analysis points to a modest increase in the risk for childhood NHL, but not HL, among children born by mothers smoking during pregnancy. Further investigation of dose–response phenomena in the NHL association, however, warrants accumulation of additional data.