Pre‐ and post‐diagnosis body mass index and heart failure mortality: a dose–response meta‐analysis of observational studies reveals greater risk of being underweight than being overweight

Summary Objective To investigate the dose–response association between pre‐ and post‐diagnosis body mass index (BMI) and heart failure (HF) mortality. Methods Eligible observational studies were searched in databases, up to November 2017. We used random‐effects generalized least squares spline models for trend estimation to derive pooled BMI unit–HF mortality relationship. Results Sixteen cohort studies (six pre‐diagnosis and 10 post‐diagnosis BMI) were included, comprising a total of 258,379 subjects with 13,201 deaths due to HF. A nonlinear U‐shaped association was found between pre‐diagnosis BMI and the risk of HF mortality, with a greater risk from being at the lowest extreme, rather than being at the top category. The combined hazard ratio of HF mortality among the highest compared to the lowest category of pre‐diagnosis BMI was 1.24 (0.65–2.37, I2 = 90.7%). No significant nonlinear association was found between post‐diagnosis BMI and HF mortality as well as when comparing the highest to the lowest category of BMI. Conclusions This meta‐analysis showed those with both high and low pre‐diagnosis BMI had higher risk for HF mortality, with a greater risk from being too underweight, rather than being obese. No significant association was found between post‐diagnosis BMI and the risk of HF mortality. Further detailed investigations are needed to accurately examine the potential mechanistic links between BMI and health outcomes.