Sensitivity and Specificity of 50% Excess Weight Loss (50%EWL) and Twelve Other Bariatric Criteria for Weight Loss Success

Criteria for bariatric weight loss success are numerous. Most of them are arbitrary. None of them is evidence-based. Our objective was to determine their sensitivity and specificity. Thirteen common bariatric weight loss criteria were compared to a benchmark reflecting the gold standard in bariatric surgery. We used an elaborate baseline BMI-independent weight loss percentile chart, based on retrospective data after laparoscopic Roux-en-Y gastric bypass (LRYGB), performed between 2007 and 2017. Percentile curves p31.6 (patients' expectation), p25 (interquartile range), p15.9 (1 standard deviation (SD) below median), and p10.9 (surgeons' goal) were used as possible cutoff for success to determine true or false positive and negative results beyond 1 year. We operated 4497 primary LRYGB patients, with mean follow-up 22 (+/- 1 SD 19; range 0-109) months, 3031 patients with last result >= 1 year, 518 >= 5 years. For all four cutoff percentile curves for success, specificities were low (2-72%) for criteria < 35 body mass index (BMI), >= 25percentage excess BMI loss (%EBMIL), >= 50%EBMIL, >= 15 percentage total weight loss (%TWL), >= 20%TWL, >= 25 percentage excess weight loss (%EWL), and high (83-96%) for < 30 BMI. No criterion had > 80% specificity and sensitivity for a cutoff above p15.9. For p15.9, they were both > 80% for criteria >= 10 BMI reduction and >= 50%EWL, both > 90% for >= 25%TWL and >= 35 percentage alterable weight loss (%AWL). All criteria had high sensitivities for all cutoff percentile curves (87-100%), except < 30 BMI (65-78%). For the first time, common bariatric criteria for weight loss success were systematically validated. Most criteria recognized success very well (high sensitivities), but >= 15%TWL, >= 20%TWL, < 35BMI, >= 25%EWL, >= 25%EBMIL, and >= 50%EBMIL left too many poor responders unnoticed (low specificities). Bariatric weight loss success is best assessed by comparing results to percentile curve 1 SD below median (p15.9) in a bariatric baseline BMI-independent weight loss percentile chart. Criteria >= 35%AWL and >= 25%TWL came close to that curve, both with > 90% sensitivity and specificity. Among others, criterion >= 50%EBMIL did not.