Study Selection
Of the 4395 identified studies, 28 were included and analyzed (see Fig. 1). The characteristics of studies are described in Tables 1 and S3. As for country income level, 24 [25,26,27,28,29,30,31,32,33,34,35, 37,38,39,40,41,42,43, 45,46,47,48,49,50] (85.7%) and 4 [36, 44, 51, 52] (14.3%) studies were conducted in HICs and UMICs, respectively. From these, 20 (71.4%) were conducted in mixed obesity group (i.e., with/without diabetes), whereas 8 [45,46,47,48,49,50,51,52] (28.6%) studies were conducted in obesity with diabetes group. All studies, except for one [52], were conducted in obese patients whose BMIs were 35 kg/m2 and higher. The three most common interventions were AGB (N = 15, 53.6%), RYGB (N = 20, 71.4%), and SG (N = 11, 39.3%).
Table 1 Characteristics of the included studies As for methodology, all studies were CUAs. The model-based techniques were Markov model [26, 28, 29, 33,34,35,36,37,38,39,40,41,42,43,44, 47,48,49,50,51,52] (N = 21; 75.0%), discrete-event simulation [30] (N = 1; 3.6%), decision tree [25] (N = 1; 3.6%), decision tree and lifetime natural history model [32] (N = 1; 3.6%), and a de novo economic model [31] (N = 1; 3.6%). Two studies [45, 46] (7.1%) did not specify the type of model, while one study [27] (3.6%) did not state the technique. Most studies (N = 25; 89.3%) adopted a payer perspective. One study [29] adopted societal perspective and two studies [27, 49] did not clearly mention the perspective. Lifetime, 10-year, and less than 10-year horizons were respectively used in 18 [25,26,27, 29, 30, 32,33,34,35, 37,38,39,40,41,42,43, 48, 49], 10 [28, 30, 31, 34, 35, 39, 41,42,43,44], and 5 studies [27, 30, 45, 46, 51].
INBs of Bariatric Surgery in HICs
The estimated INBs from HICs were presented separately by types of patients (see Tables S4 and S5). For mixed obesity group, bariatric surgery was found to be cost-effective in all except for three studies, which adopted 2-year [27], 5-year [30], and 10-year [31] time horizons. For obesity with diabetes group, all studies revealed that bariatric surgery was cost-effective.
Pooled INBs of Bariatric Surgery among Mixed Obesity Group in HICs
The INBs of overall bariatric surgery versus usual care among mixed obesity group under payer perspective were highly varied across 11 studies [26, 33,34,35, 37,38,39,40,41,42,43] (I2 = 83.9%) with a pooled INB (95% CI) over lifetime horizon of $101,897.96 ($79,390.93, $124,404.99) (see Fig. 2). Subgroup analysis indicated that AGB, SG, and mixed types of BS, but not RYGB, were significantly cost-effective as compared with usual care with pooled INBs of $51,143.29 (95% CI $15,735.29, $86,551.29; I2 = 37.1%), $127,578.98 (95% CI $62,139.61, $193,018.36; I2 = 0%), $143,438.56 (95% CI $91,320.26, $195,556.88; I2 = 89.5%), and $110,928.33 (95% CI −$8,677.49, $230,534.14; I2 = 85.2.1%), respectively. A funnel plot of overall bariatric surgery effect indicated asymmetry of the funnel which contrasted to Egger’s test (coefficient = 2.11, SE = 0.47, p = 0.001), Fig. S1. A contour-enhanced funnel (Fig. S2) showed that asymmetry might be caused by heterogeneity.
The INBs of bariatric surgery among mixed obesity group over a 10-year time horizon were highly varied across seven studies [28, 34, 35, 39, 41,42,43] (I2 = 75.6%), with a pooled INB (95% CI) of $53,063.69 ($42,647.96, $63,479.43) (see Fig. 3). A funnel plot (Fig. S3) and Egger’s test (coefficient = 11.13, SE = 6.81, p = 0.163) indicated no evidence of publication bias.
Pooled INBs of Bariatric Surgery among Obesity with Diabetes in HICs
The INBs of bariatric surgery among obesity with diabetes group were pooled from seven studies [33, 37, 42, 43, 47, 48, 50] over life-time horizon (see Fig. 4). The pooled INB was $80,826.28 (95% CI $32,500.75, $129,151.81; I2 = 85.0%), indicating that bariatric surgery was cost-effective relative to usual care. Asymmetry was observed in funnel plot and Egger’s test (coefficient = 2.51, SE = 0.92, p = 0.041) (Fig. S4). A contour-enhanced funnel plot (Fig. S5) showed that asymmetry might be caused by heterogeneity.
INBs of Bariatric Surgery in UMICs
Of the four studies from UMICs [36, 44, 51, 52], three studies [36, 44, 51] with five comparisons were conducted in mixed obesity group, while four comparisons [36, 44, 51, 52] were conducted in obesity with diabetes group [36, 44, 51, 52] (see Table S6). Time horizon varied across studies, i.e., 5 years [51], 10 years [44], 20 years [36], and 50 years [52]. All studies adopted a payer perspective [36, 44, 51, 52]. For mixed group obesity group, the evidence was inconclusive. On the other hand, all studies conducted among obesity with diabetes group revealed that bariatric surgery was cost-effective with the INBs ranging from $4,015.28 to $40,867.97 (see Table S6). Since studies had varying time horizons, the INBs were not pooled.
Risk of Bias
Risk of bias was assessed using ECOBIAS checklist (see Table S7). All studies had no bias in terms of comparator, outcomes measurement, and discounting. Four studies [27, 31, 45, 46] had unclear risk of wrong model bias. Three studies [45, 46, 51] adopted short-term time-horizon of 5 years, four studies [28, 31, 36, 44] used mid-term period of 10 and 20 years, and 18 studies [25,26,27, 29, 30, 32,33,34,35, 37,38,39,40,41,42,43, 48, 49] adopted lifetime horizon. Partial biases related to treatment effect, intermittent data collection, and quality of life weight occurred in almost all studies. Data on long-term treatment effect of bariatric surgery and utility per unit of BMI change were limited. Moreover, half of the studies [25,26,27,28,29, 31, 37, 38, 40, 44,45,46, 48, 49, 51, 52] were subjected to bias related to internal consistency.