Diagnostic performance of body mass index to identify obesity as defined by body adiposity: a systematic review and meta-analysis

Top Cited Papers
Publisher Website
Google Scholar
Abstract
Objective: We performed a systematic review and meta-analysis of studies that assessed the performance of body mass index (BMI) to detect body adiposity. Design: Data sources were MEDLINE, EMBASE, Cochrane, Database of Systematic Reviews, Cochrane CENTRAL, Web of Science, and SCOPUS. To be included, studies must have assessed the performance of BMI to measure body adiposity, provided standard values of diagnostic performance, and used a body composition technique as the reference standard for body fat percent (BF%) measurement. We obtained pooled summary statistics for sensitivity, specificity, positive and negative likelihood ratios (LRs), and diagnostic odds ratio (DOR). The inconsistency statistic (I2) assessed potential heterogeneity. Results: The search strategy yielded 3341 potentially relevant abstracts, and 25 articles met our predefined inclusion criteria. These studies evaluated 32 different samples totaling 31 968 patients. Commonly used BMI cutoffs to diagnose obesity showed a pooled sensitivity to detect high adiposity of 0.50 (95% confidence interval (CI): 0.43–0.57) and a pooled specificity of 0.90 (CI: 0.86–0.94). Positive LR was 5.88 (CI: 4.24–8.15), I2=97.8%; the negative LR was 0.43 (CI: 0.37–0.50), I2=98.5%; and the DOR was 17.91 (CI: 12.56–25.53), I2=91.7%. Analysis of studies that used BMI cutoffs ⩾30 had a pooled sensitivity of 0.42 (CI: 0.31–0.43) and a pooled specificity of 0.97 (CI: 0.96–0.97). Cutoff values and regional origin of the studies can only partially explain the heterogeneity seen in pooled DOR estimates. Conclusion: Commonly used BMI cutoff values to diagnose obesity have high specificity, but low sensitivity to identify adiposity, as they fail to identify half of the people with excess BF%.