Table 2.
Comparative Effectiveness of RYGB, SG, and AGB for TWL Among Adults at 1, 3, and 5 Years*
| Comparison | Time Since Bariatric Procedure | |||||
|---|---|---|---|---|---|---|
| 1 Year | 3 Years | 5 Years | ||||
| Patients, n | TWL (95% CI), % | Patients, n | TWL (95% CI), % | Patients, n | TWL (95% CI), % | |
| SG vs. RYGB | ||||||
| SG | 14 929 | −25.2 (−25.4 to −25.1) | 5304 | −21.0 (−21.3 to −20.7) | 1088 | −18.8 (−19.6 to −18.0) |
| RYGB | 19 029 | −31.2 (−31.3 to −31.1) | 9225 | −29.0 (−29.2 to −28.8) | 3676 | −25.5 (−25.9 to −25.1) |
| Difference | − | 5.9 (5.8 to 6.1) | − | 8.0 (7.6 to 8.4) | − | 6.7 (5.8 to 7.7) |
| P value | <0.001 | <0.001 | <0.001 | |||
| AGB vs. RYGB | ||||||
| AGB | 1681 | −13.7 (−14.0 to −13.3) | 943 | −12.7 (−13.5 to −12.0) | 337 | −11.7 (−13.1 to −10.2) |
| RYGB | 18 684 | −31.4 (−31.5 to −31.3) | 9152 | −29.1 (−29.3 to −28.9) | 3733 | −25.6 (−26.0 to −25.2) |
| Difference | − | 17.7 (17.3 to 18.1) | − | 16.4 (15.6 to 17.2) | − | 13.9 (12.4 to 15.4) |
| P value | <0.001 | <0.001 | <0.001 | |||
| AGB vs. SG | ||||||
| AGB | 1681 | −13.1 (−13.5 to −12.7) | 933 | −12.0 (−12.8 to −11.2) | 306 | −11.4 (−13.2 to −9.6) |
| SG | 14 664 | −25.1 (−25.3 to −25.0) | 5270 | −20.9 (−21.2 to −20.6) | 1088 | −18.7 (−19.5 to −17.8) |
| Difference | − | 12.0 (11.6 to 12.5) | − | 8.9 (8.0 to 9.8) | − | 7.3 (5.2 to 9.3) |
| P value | <0.001 | <0.001 | <0.001 | |||
AGB = adjustable gastric banding; RYGB = Roux-en-Y gastric bypass; SG = sleeve gastrectomy; TWL = total weight loss.
TWL was calculated as follows: (weight in kilograms at 1, 3, and 5 y – weight in kilograms at baseline)/weight in kilograms at baseline × 100. A propensity score model was constructed for each pairwise analysis and time point. Age at index procedure, sex, race/ethnicity, year of index procedure, baseline body mass index, number of days from baseline weight to bariatric surgery, and baseline Charlson–Elixhauser comorbidity index score were forced into all propensity score models. Site, smoking status, inpatient hospitalizations in the year before surgery, baseline blood pressure, and comorbidities at baseline were included subject to the variable selection process. Further, to account for differing effects of confounders on propensity scores by site, interactions between site and all confounders were made available for selection. In addition to adjustment for deciles of the predicted propensity score, we included main effects for baseline weight, sex, age, and all other baseline covariates listed here in the outcome model. For each pairwise comparison, we restricted the analysis to sites that included ≥1 patient who had each procedure at each time point. See the Statistical Appendix section of the Supplement (available at Annals.org) for more details.