INTRODUCTION: There is an increasing interest in understanding whether air pollutants modify the quantitative relationships between temperature and health outcomes. The results of available studies were, however, inconsistent. This study aims to sum up the current evidence and provide a comprehensive understanding of this topic. METHODS: We conducted an electronic search in PubMed (MEDLINE), EMBASE, Web of Science Core Collection, and ProQuest Dissertations and Theses. The modified Navigation Guide was applied to evaluate the quality and strength of evidence. We calculated pooled temperature-related mortality at low and high pollutant levels respectively, using the random-effects model. RESULTS: We identified 22 eligible studies, eleven of which were included in the meta-analysis. Significant effect modification was observed on heat effects for all-cause and non-accidental mortality by particulate matter with an aerodynamic diameter of <10 μm (PM(10)) and ozone (O(3)) (p < 0.05). The excess risks (ERs) for all-cause and non-accidental mortality were 5.4% (4.4%, 6.4%) and 6.3% (4.8%, 7.8%) at the low PM(10) level, 8.8% (7.5%, 10.1%) and 11.4% (8.7%, 14.2%) at the high PM(10) level, respectively. As for O(3), the ERs for all-cause and non-accidental mortality were 5.1% (3.9%, 6.3%) and 3.6% (0.1%, 7.2%) at the low O(3) level, 7.6% (6.3%, 9.0%) and 12.5% (4.7%, 20.9%) at the high O(3) level, respectively. Surprisingly, the heat effects on cardiovascular mortality were found to be lower at high carbon monoxide (CO) levels [ERs = 5.4% (3.9%, 6.9%)] than that at low levels [ERs = 9.4% (7.0%, 11.9%)]. The heterogeneity varied, but the results of sensitivity analyses were generally robust. Significant effect modification by air pollutants was not observed for heatwave or cold effects. CONCLUSIONS: PM(10) and O(3) modify the heat-related all-cause and non-accidental mortality, indicating that policymakers should consider air pollutants when establishing heat-health warning systems. Future studies with comparable designs and settings are needed.