OBJECTIVE: To examine cardiovascular disease (CVD) mortality burden attributable to ambient temperature; to estimate effect modification of this burden by gender, age and education level. METHODS: We obtained daily data on temperature and CVD mortality from 15 Chinese megacities during 2007-2013, including 1,936,116 CVD deaths. A quasi-Poisson regression combined with a distributed lag non-linear model was used to estimate the temperature-mortality association for each city. Then, a multivariate meta-analysis was used to derive the overall effect estimates of temperature at the national level. Attributable fraction of deaths were calculated for cold and heat (ie, temperature below and above minimum-mortality temperatures, MMTs), respectively. The MMT was defined as the specific temperature associated to the lowest mortality risk. RESULTS: The MMT varied from the 70th percentile to the 99th percentile of temperature in 15 cities, centring at 78 at the national level. In total, 17.1% (95% empirical CI 14.4% to 19.1%) of CVD mortality (330,352 deaths) was attributable to ambient temperature, with substantial differences among cities, from 10.1% in Shanghai to 23.7% in Guangzhou. Most of the attributable deaths were due to cold, with a fraction of 15.8% (13.1% to 17.9%) corresponding to 305,902 deaths, compared with 1.3% (1.0% to 1.6%) and 24,450 deaths for heat. CONCLUSIONS: This study emphasises how cold weather is responsible for most part of the temperature-related CVD death burden. Our results may have important implications for the development of policies to reduce CVD mortality from extreme temperatures.