The literature search identified eight population-based studies for inclusion [17,18,19,20,21,22,23,24]. Figure 1 demonstrates the PRISMA flowchart for the literature search. In total, there were 635,642 patients included the analysis. There were 114,020 patients who received bariatric surgery (gastric bypass: 69,740; sleeve gastrectomy: 7519; gastric band: 13,609; gastroplasty: 8632; alternative procedure or undetermined: 14,520). There were 521,622 control patients. The majority of studies utilized patients diagnosed as obese to form the control group [17,18,19, 21, 22, 24]. One study matched patients according to age and gender [20], and one study compared results to the background national population [23]. Table 1 provides the patient demographic details for all studies. Quality assessment of studies was undertaken with the Newcastle Ottawa Scale (Supplementary Table 1) [25]. All studies scored four stars for patient selection (rated out of four stars). Two studies scored one star for comparability as controls did not necessarily have a diagnosis of obesity [20, 23]. All other studies scored the maximum score for comparability (two stars). All studies scored three stars for outcome (rated out of three stars).
Table 1 Patient demographics for the studies included in the pooled analysis Overall Cancer Incidence
Four papers reported overall cancer incidence in patients who had undergone bariatric surgery [17, 19, 20, 24]. Bariatric surgery was associated with a significant reduction in overall cancer incidence compared to control patients (pooled odds ratio (POR) = 0.72; 95% confidence interval (CI) 0.59 to 0.87; p = 0.0007). There was some evidence of statistical heterogeneity (Cochran Q = 11.33; p = 0.01; I2 = 73.5%) but no evidence of bias (Egger = −0.06; p = 0.98) (Fig. 2).
Obesity-Related Cancer
Three studies reported rate of obesity-related cancer defined as cancer of the breast, prostate, colorectum, endometrium, ovary, kidney, esophagus (adenocarcinoma only), liver, pancreas, gallbladder, non-Hodgkin lymphoma, leukemia, and thyroid [17, 23, 24]. Myeloma [17, 24], gastric cardia [24], and meningioma [24] were also included in the definition of obesity-related cancer in some studies (Table 2). An additional study described results for “hormone-related cancers” (including breast, endometrium, and prostate), and these were included as obesity-related cancer for the purposes of analysis [21].
Table 2 Outcomes for overall cancer incidence and obesity-related cancer Bariatric surgery was associated with a significant reduction in incidence of obesity-related cancer (POR = 0.55; 95% CI 0.31 to 0.96; p = 0.04). There was some evidence of significant statistical heterogeneity (Cochran Q = 124.34; p = < 0.0001; I2 = 97.6%) but no evidence of bias (Egger = −15.4; p = 0.14) (Fig. 3).
Three studies analyzed rates of obesity-related cancer stratified by gender [21, 23, 24]. When analyzed by gender, it was identified that in male patients bariatric surgery was not associated with a significant reduction in the rate of obesity-related cancer (POR = 0.76; 95% CI 0.87 to 1.32; p = 0.46). However, in female patients who had undergone bariatric surgery, there was a trend towards a reduced rate of obesity-related cancer (POR = 0.50; 95% CI 0.24 to 1.04; p = 0.065) (Supplementary fig. 1). There was evidence of statistical heterogeneity (Cochran Q 1.09.03; p < 0.0001; I2 = 98.2%) and too few strata for an assessment of bias.
Esophageal Cancer
Three studies reported the incidence of esophageal cancer following bariatric surgery [17, 21, 22]. There was no significant effect of bariatric surgery upon the incidence of esophageal cancer (POR = 0.79; 95% CI 0.43 to 1.44; p = 0.43). There was no evidence of statistical heterogeneity (Cochran Q = 0.74; p = 0.69; I2 = 0%). There were too few strata to facilitate an analysis of bias.
Colorectal Cancer
Four studies reported the incidence of colorectal cancer [17, 18, 21, 23]. Bariatric surgery was not associated with a statistically significant change in the incidence of colorectal cancer (POR = 1.39; 95% CI 0.96 to 2.02; p = 0.08). There was some evidence of statistical heterogeneity (Cochran Q = 10.99; p = 0.01; I2 = 72.7%) but no evidence of bias (Egger = − 1.43; p = 0.71) (Supplementary fig. 2).
Breast Cancer
Three studies reported the rate of breast cancer in patients who have undergone bariatric surgery [17, 21, 23]. Bariatric surgery was associated with a significant reduction in the incidence breast cancer compared to controls (POR = 0.50; 95% CI 0.25 to 0.99; p = 0.045) (Fig. 4). There was some evidence of statistical heterogeneity (Cochran Q = 39.4; p < 0.0001; I2 = 94.9%). There were too few strata to facilitate an analysis of bias.
Endometrial Cancer
Three studies reported the rate of endometrial cancer in bariatric surgery patients [17, 21, 23]. There was no significant difference in the rate of endometrial cancer between patients who had undergone bariatric surgery and controls (POR = 0.47; 95% CI 0.08 to 2.65; p = 0.39). There was some evidence of statistical heterogeneity (Cochran Q = 94.8; p = < 0.0001; I2 = 97.9%). There were too few strata to facilitate an analysis of bias.
Prostate Cancer
Three studies reported the rate of prostate cancer in these patients [17, 21, 23]. Bariatric surgery was not associated with any significant differences in the incidence of prostate cancer (POR = 0.82; 95% CI 0.39 to 1.73; p = 0.61). There was evidence of statistical heterogeneity (Cochran Q = 8.86; p = 0.01; I2 = 77.4%). There were too few strata to facilitate an analysis of bias.