Healthcare utilisation across five years among adolescents following metabolic-bariatric surgery; a prospective observational study.

Data from the Adolescent Bariatrics: Assessing Health Benefits & Risks (Teen-LABS) study (Version 4) are available for request at the NIDDK Central Repository (NIDDK-CR) website, Resources for Research (R4R): https://repository.niddk.nih.gov/studies/teen-labs/ .

The Teen Longitudinal Assessment of Bariatric Surgery (Teen-LABS) study ( NCT00474318 ) was a prospective, multicentre, observational study, examining consecutive adolescents (≤19 years, Tanner stage ≥ IV) undergoing MBS for severe obesity (BMI ≥35 kg/m 2 ). Enrolment across 5 centres occurred between February 28, 2007 and December 30, 2011. The sample size was determined by formal power and sample-size work calculations by the Data Coordinating Centre. The study was approved by review the institutional review boards of each of the participating institutions and consent was obtained for all participants prior to enrolment. Healthcare utilisation events included any return to a secondary care facility for any reason other than a planned follow-up visit within the study framework. Data were collected prospectively using protocol mandated self-report and research team standardised forms, completed by interview for consistency across sites, at each regularly scheduled study visit. Research visits occurred at baseline, 6 months after surgery, and annually to 5 years after surgery. Definitions and cutoffs used to determine disease states have been described in detail previously. 3 Independent external adjudication was conducted for all healthcare utilisation events requiring readmission, reoperation, and any deaths up to 2 years postoperatively. After 2 years, adjudication was performed only for reoperations and deaths. For adjudications, de-identified medical records and select study data were provided to the Teen-LABS Adjudication Committee, which consisted of 9 individuals who were not members of the study group, including 5 surgeons, 3 physicians with paediatric or obesity expertise, and one National Institutes of Health representative. Adjudication Committee members independently reviewed records related to each healthcare utilisation event alongside study data related to the participant, before classifying the healthcare utilisation as related or unrelated to the MBS. Relatedness of non-adjudicated events was determined by Teen-LABS investigators, with resolution to consensus where necessary using study records and case records. Adjudication committee outcomes were recorded for analytic purposes. Categorical measures are presented using frequencies and percentages and compared using chi-square or Fisher's exact tests. Continuous variables were summarised using means with standard deviations and compared using t-tests. Negative binomial modelling, with the logarithm of person-years as an offset parameter, was used to calculate 5-year healthcare utilisation event rates. Incidence rates and 95% confidence intervals were calculated using observed event frequencies by surgical procedure and event category. Rates were expressed per 500-person years (i.e., 100 patients followed for 5 years). Rate ratios (RR) and 95% confidence intervals were calculated to compare healthcare utilisation event rates (related, unrelated, admission events) between surgical procedures. The following variables were considered for inclusion in the final model that generated the RR value: age at surgery, sex, race, ethnicity, baseline BMI, baseline sagittal abdominal diameter, baseline body fat percentage, glycated haemoglobin (HbA1c), high-sensitivity c-reactive protein (hs-CRP), baseline type 2 diabetes, baseline dyslipidaemia, baseline hypertension, clinical centre, presence of any 30-day events, caregiver bariatric surgery, caregiver education, and household income. All two-way interactions with surgical procedure were evaluated. For multivariable models, a stepwise variable selection was performed, retaining each variable from univariate models with a p-value less than 0·40. Variables were then eliminated from the multivariable model if the p-value was greater than 0·05. No adjustments were made for multiple comparisons. Specific power and sample size calculations were not performed for these analyses; therefore, the results should be considered exploratory in nature. Over the 5-year analysis period, 96% of the original cohort was maintained, with 87% of all study visits completed. No statistical adjustments or imputation were applied to account for missing data. All statistical analyses were conducted using SAS v9·4; all reported p-values were two-sided and considered statistically significant when less than 0·05. The funding source had no role in the design or conduct of the study.

The mean age at surgery was 17·0 ± 2 years and 76% of the cohort was female. In total, 161/228 (71%) participants underwent Roux-en-Y gastric bypass (RYGB) and 67/228 (29%) underwent vertical sleeve gastrectomy. Mean preoperative BMI was 53 kg/m 2 with a slightly higher BMI (54 kg/m 2 ) in the RYGB group compared to the sleeve gastrectomy group (50 kg/m 2 , p = 0·04). The ethnicity distribution was similar in both groups, reported as white by a majority of participants (164/228, 72%), black for 50/228 (22%), Hispanic for 16/228 (7%), Asian for 1/228 (<1%), American Indian or Alaskan native for 1/228 (<1%), and more than one ethnic background for 12/228 participants (5%). Details of participants with and without healthcare utilisation events are illustrated in Table 1 and patient flow through the study is illustrated in Supplementary Figure S1 . Table 1 Baseline characteristics. Participants with 1 or more healthcare utilisation events Participants with 0 healthcare utilisation events p-value n = 156 n = 72 Age at surgery, years (SD) 17·0 (1·57) 17·0 (1·56) 0·86 Female (n) 80·1% (125) 63·9% (46) 0·009 White (n) 73·1% (114) 69·4% (50) 0·57 Non-Hispanic (n) 93·6% (146) 91·7% (66) 0·60 RYGB (n) 72·4% (113) 66·7% (48) 0·37 BMI, kg/m 2 (SD) 52·3 (8·64) 53·2 (10·84) 0·55 Body fat percentage (SD) 53·4 (5·51) 52·6 (8·37) 0·46 Sagittal abdominal diameter (SD) 31·6 (3·97) 32·0 (4·32) 0·58 T2D (n) 9·8% (15) 19·4% (14) 0·044 Elevated BP (n) 41·2% (63) 46·5% (33) 0·46 Dyslipidaemia (n) 77·3% (119) 73·2% (52) 0·51 Figures are given as mean values with standard deviation (SD), unless presented as percentages with number (n). RYGB, Roux-en-Y gastric bypass; BMI, body mass index; T2D, type 2 diabetes; BP, blood pressure. Baseline characteristics. Figures are given as mean values with standard deviation (SD), unless presented as percentages with number (n). RYGB, Roux-en-Y gastric bypass; BMI, body mass index; T2D, type 2 diabetes; BP, blood pressure. Over the 5-year period, there were 365 healthcare utilisation events among 156 participants [177·0 per 500 person years (95% CI 152·4–205·7)]; thus 68% of all participants experienced one or more healthcare utilisation event(s). The proportion of events adjudicated externally was 37% (134/365). Thirty-eight individuals experienced a complication within 30 days after MBS, of whom 30 (79%; 25 RYGB and 5 sleeve gastrectomy) experienced a total of 99 long-term healthcare utilisation events. Among the 99 events, 45 (45%) were related to the original operation, including 14 of 18 abdominal procedures, and 15 of 18 endoscopic procedures. Of the 190 individuals who did not experience a complication within 30 days, 266 long-term healthcare utilisation events occurred among 126 participants (66%). A total of 141 related healthcare utilisation events among 81 (36%) participants were adjudicated as related to the prior MBS operation. These events occurred at a rate of 67·3 per 500 person years follow up (95% CI 53·4–84·9). Rates were comparable between RYGB [71·5 per 500 person years (95% CI 54·6–93·6)] and sleeve gastrectomy [56·9 per 500 person years (95% CI 36·2–89·3), p = 0·36] ( Table 2 ). Table 2 Related healthcare utilisation events. Gastric Bypass (n = 161) Sleeve Gastrectomy (n = 67) Total (n = 228) Participants, n (%) Events, n Rate a (95% CI) Participants, n (%) Events, n Rate a (95% CI) p-value Participants, n (%) Events, n Rate a (95% CI) Total 63 (39·1) 107 71·5 (54·6, 93·6) 18 (26·9) 34 56·9 (36·2, 89·3) 0·36 81 (35·5) 141 67·3 (53·4, 84·9) HPB 20 (12·4) 20 13·4 (8·7, 20·8) 8 (11·9) 9 15·1 (7·9, 29·1) 0·77 28 (12·3) 29 13·9 (9·7, 20·0) Intestinal 6 (3·7) 8 5·3 (2·2, 12·8) 1 (1·5) 1 1·7 (0·2, 14·9) 0·21 7 (3·1) 9 4·3 (1·9, 9·7) Skin/Wound 15 (9·3) 21 13·9 (7·8, 24·9) 6 (9·0) 11 18·1 (7·7, 42·6) 0·47 21 (9·2) 32 15·1 (9·3, 24·5) Oesophagogastric 13 (8·1) 21 14·1 (7·5, 26·7) 2 (3·0) 2 3·3 (0·7, 16·0) 0·019 15 (6·6) 23 11·1 (6·1, 20·6) General GI Symptoms 19 (11·8) 24 16·1 (10·0, 26·0) 8 (11·9) 11 18·7 (9·1, 38·4) 0·44 27 (11·8) 35 16·9 (11·3, 25·1) Renal 3 (1·9) 3 2·0 (0·6, 6·2) 0 0 0 0·16 3 (1·3) 3 1·4 (0·5, 4·5) Nutritional 8 (5·0) 8 5·4 (2·7, 10·7) 0 0 0 0·16 8 (3·5) 8 3·8 (1·9, 7·7) Psychosocial 2 (1·2) 2 1·3 (0·3, 5·4) 0 0 0 0·25 2 (0·9) 2 1·0 (0·2, 3·8) a Rate per 500 person-years. Related healthcare utilisation events. Rate per 500 person-years. The most frequent healthcare utilisation events related to the prior MBS operation were gastrointestinal tract symptoms, such as nausea and abdominal pain (35 events), skin or wound problems, such as skin lesions and superficial abscesses (32 events), hepatobiliary or pancreatic events, such as symptomatic cholelithiasis and pancreatitis (29 events), and oesophagogastric problems, such as strictures and ulceration (23 events). Additional related events included intestinal problems, such as bleeding and obstruction (9 events), and nutritional events, such as iron deficiency anaemia (8 events; Table 2 and Supplementary Table S1 ). Of the 141 related events, resolution of the presenting problem was documented in 43% (n = 61). Comparing related events between procedures, oesophagogastric events, but none of the other event categories, occurred with greater frequency after RYGB compared to sleeve gastrectomy (14·1 vs. 3·4 per 500 person-years, p = 0·019). Fifty-nine intra-abdominal procedures were performed on 44 individuals (19·3%) between 31 days and 5 years after MBS, including 44 procedures among 32 participants who underwent RYGB (20%), and 15 among 12 participants who underwent sleeve gastrectomy (18%; Table 3 ). Notably, the incidence of gallbladder disease was comparable between surgical groups. An additional 25 participants, who underwent surgical procedures for excess skin or body contouring during the 5-year period, were not included in this category. Twenty-six gastrointestinal endoscopic procedures were performed on 15 individuals (7%) across 5 years; 20 among 10/228 RYGB participants (6%), and 6 among 5/67 sleeve gastrectomy participants (7%) ( Table 3 ). Table 3 Abdominal operations and gastrointestinal endoscopy more than 30 days following surgery. Events (%) Indication RYGB VSG Total Intra-abdominal procedures Symptomatic gallstones 18 (11%) 7 (10%) 25 (11%) Intra-abdominal adhesions 7 (4%) 4 (6%) 11 (5%) Anastomotic/staple line leak 1 (<1%) 1 (1%) 2 (<1%) Post-operative bleeding 2 (1%) 0 (0%) 2 (<1%) Internal hernia 4 (2%) 0 (0%) 4 (1%) Abdominal wall hernia 0 (0%) 1 (1%) 1 (2%) Colocutaneous fistula 2 (1%) 1 (0%) 3 (1%) Retained foreign body 1 (1%) 0 (0%) 1 (<1%) Gastrointestinal stricture 10 (6%) 1 (1%) 11 (5%) Nutritional access (gastrostomy formation) 1 (<1%) 0 (0%) 1 (<1%) Urinary bladder access (appendicovesicostomy) 0 (0%) 0 (0%) 0 (0%) Splenic injury 0 (0%) 0 (0%) 0 (0%) Appendicitis 0 (0%) 0 (0%) 0 (0%) Pregnancy (Caesarean section) 0 (0%) 0 (0%) 0 (0%) Ovarian pathology 0 (0%) 0 (0%) 0 (0%) Endometriosis 0 (0%) 0 (0%) 0 (0%) Endoscopy Gastroesophageal reflux/gastritis/esophagitis 18 (11%) 3 (1%) 21 (9%) Hiatal hernia 0 (0%) 0 (0%) 0 (0%) Gastrointestinal leak 1 (1%) 1 (1%) 2 (1%) Biliary pancreatitis 0 (0%) 1 (1%) 1 (<1%) Remnant stomach access 1 (1%) 0 (0%) 1 (<1%) Vomiting 0 (0%) 1 (1%) 1 (<1%) Haemorrhoids 0 (0%) 0 (0%) 0 (0%) RYGB, Roux-en-Y gastric bypass; VSG, vertical sleeve gastrectomy. Abdominal operations and gastrointestinal endoscopy more than 30 days following surgery. RYGB, Roux-en-Y gastric bypass; VSG, vertical sleeve gastrectomy. Adjusted analyses identified comparable healthcare utilisation event rates between sleeve gastrectomy and RYGB participants (Risk Ratio (RR): 0·92; 95% CI 0·54, 1·55; p = 0·75). Participants with perioperative healthcare utilisation events (within 30 days of index MBS operation) had a significantly higher rate of long-term (31 days–5 years) related healthcare utilisation events compared to those without perioperative events (RR: 2·26; 95% CI 1·31, 3·90; p = 0·004). A total of 224 healthcare utilisation events deemed unrelated to the original MBS operation occurred among 112/228 patients (49%) at a rate of 107·4 per 500 person-years (95% CI 94·2–122·5; Table 4 ). Unrelated events occurred with similar frequency following RYGB (106·0 [95% CI 90·7–123·9]) and sleeve gastrectomy (110·9 [87·2–141·2]). The most common unrelated healthcare utilisation events were related to musculoskeletal, obstetric, dermatological, and psychosocial issues ( Table 4 and Supplementary Table S2 ). Table 4 Unrelated healthcare utilisation events. Roux-en-Y gastric Bypass (N = 161) Sleeve gastrectomy (N = 67) TOTAL (N = 228) Patients, n (%) Events, n Rate a (95% CI) Patients, n (%) Events, n Rate a (95% CI) Patients, n (%) Events, n Rate a (95% CI) TOTAL 79 (49·1) 158 106·0 (90·7, 123·9) 33 (49·3) 66 110·9 (87·2, 141·2) 112 (49·1) 224 107·4 (94·2, 122·5) Skin/Wound 12 (7·5) 15 10·1 (6·1, 16·7) 5 (7·5) 13 21·9 (12·7, 37·6) 17 (7·5) 28 13·4 (9·3, 19·4) Oesophagogastric 2 (1·2) 2 1·3 (0·3, 5·4) 0 0 0 2 (0·9) 2 1·0 (0·2, 3·8) General GI Symptoms 2 (1·2) 2 1·3 (0·3, 5·4) 0 0 0 2 (0·9) 2 1·0 (0·2, 3·8) Renal 1 (0·6) 1 0·7 (0·1, 4·8) 0 0 0 1 (0·4) 1 0·5 (0·1, 3·4) Musculoskeletal 22 (13·7) 32 21·5 (15·2, 30·4) 6 (9·0) 10 16·8 (9·0, 31·2) 28 (12·3) 42 20·1 (14·9, 27·3) HPB 2 (1·2) 3 2·0 (0·6, 6·2) 3 (4·5) 3 5·0 (1·6, 15·6) 5 (2·2) 6 2·9 (1·3, 6·4) Urological 5 (3·1) 11 7·4 (4·1, 13·3) 2 (3·0) 3 5·0 (1·6, 15·6) 7 (3·1) 14 6·7 (4·0, 11·3) Respiratory 5 (3·1) 5 3·4 (1·4, 8·1) 1 (1·5) 1 1·7 (0·2, 11·9) 6 (2·6) 6 2·9 (1·3, 6·4) Neurological 9 (5·6) 10 6·7 (3·6, 12·5) 1 (1·5) 1 1·7 (0·2, 11·9) 10 (4·4) 11 5·3 (2·9, 9·5) Endocrine 2 (1·2) 7 4·7 (2·2, 9·9) 0 0 0 2 (0·9) 7 3·4 (1·6, 7·0) Cardiological 1 (0·6) 1 0·7 (0·1, 4·8) 1 (1·5) 1 1·7 (0·2, 11·9) 2 (0·9) 2 1·0 (0·2, 3·8) Dental 10 (6·2) 10 6·7 (3·6, 12·5) 8 (11·9) 9 15·1 (7·9, 29·1) 18 (7·9) 19 9·1 (5·8, 14·3) Otorhinolaryngological 2 (1·2) 3 2·0 (0·6, 6·2) 1 (1·5) 1 1·7 (0·2, 11·9) 3 (1·3) 4 1·9 (0·7, 5·1) Ophthalmological 0 0 0 1 (1·5) 1 1·7 (0·2, 11·9) 1 (0·4) 1 0·5 (0·1, 3·4) Gastrointestinal 4 (2·5) 4 2·7 (1·0, 7·2) 3 (4·5) 5 8·4 (3·5, 20·2) 7 (3·1) 9 4·3 (2·2, 8·3) Obstetric 14 (11·1) 19 16·2 (10·3, 25·3) 4 (8·9) 6 15·5 (7·0, 34·5) 18 (10·5) 25 16·0 (10·8, 23·7) Gynaecological 9 (7·1) 9 7·7 (4·0, 14·7) 0 0 0 9 (5·3) 9 5·8 (3·0, 11·1) Breast 1 (0·8) 2 1·7 (0·4, 6·8) 0 0 0 1 (0·6) 2 1·3 (0·3, 5·1) Haematological 0 0 0 2 (3·0) 2 3·4 (0·8, 13·4) 2 (0·9) 2 1·0 (0·2, 3·8) Psychosocial 10 (6·2) 17 11·4 (7·1, 18·4) 5 (7·5) 9 15·1 (7·9, 29·1) 15 (6·6) 26 12·5 (8·5, 18·3) Miscellaneous 4 (2·5) 4 2·7 (1·0, 7·2) 0 0 0 4 (1·8) 4 1·9 (0·7, 5·1) a Rate per 500 person-years. HPB, hepatopancreatobiliary. Unrelated healthcare utilisation events. Rate per 500 person-years. HPB, hepatopancreatobiliary. As with events related to surgery, adjusted analyses showed that participants with perioperative healthcare utilisation events had a significantly higher rate of long-term (31 days–5 years) unrelated healthcare utilisation events compared to those without (RR: 1·75; 95% CI 1·13, 2·71; p = 0·014). A higher rate of unrelated events was also observed among adolescents with a caregiver that had also undergone bariatric surgery, (RR 2·10; 95% CI 1·41–3·11; p < 0·001). An interaction between surgical procedure and type 2 diabetes was observed (p = 0·002). Among participants without type 2 diabetes at baseline, the rate of unrelated healthcare utilisation events was significantly higher after sleeve gastrectomy than RYGB (170·3 vs. 103·9, p = 0·026). Conversely, among patients with baseline type 2 diabetes, fewer unrelated healthcare utilisation events occurred after sleeve gastrectomy than RYGB (56·0 vs. 220·8 per 500 person-years, p = 0·10), albeit without reaching statistical significance ( Table 5 ). Table 5 Final multivariable modelling results. Risk Ratio 95% CI p-value Related healthcare utilisation events—final model  30 day event (Yes) 2·26 1·31, 3·90 0·004  Sleeve gastrectomy vs. RYGB 0·92 0·54, 1·55 0·75 Unrelated healthcare utilisation events – final model  30 day event (Yes) 1·75 1·13, 2·71 0·014  Bariatric caregiver (Yes) 2·10 1·41, 3·11 <0·001  Surgery x T2D 0·019  T2D = Yes  Sleeve gastrectomy vs. RYGB 0·17 0·04, 0·80 0·10  T2D = No  Sleeve gastrectomy vs. RYGB 1·31 0·93, 1·85 0·026 Healthcare utilisation events with admission – final model  30 day event (Yes) 2·16 1·22, 3·80 0·007  Sleeve gastrectomy vs. RYGB 1·06 0·62, 1·81 0·84  Body Mass Index 0·97 0·94, 0·99 0·024  Hypertension (Yes) 1·64 0·99, 2·74 0·05 RYGB, Roux-en-Y gastric bypass; T2D, type 2 diabetes. Models adjusted for: age at surgery, sex, race, ethnicity, baseline BMI, baseline sagittal abdominal diameter, baseline body fat percentage, glycated haemoglobin (HbA1c), high-sensitivity c-reactive protein (hs-CRP), baseline type 2 diabetes, baseline dyslipidaemia, baseline hypertension, clinical centre, presence of any 30-day events, caregiver bariatric surgery, caregiver education, and household income. Final multivariable modelling results. RYGB, Roux-en-Y gastric bypass; T2D, type 2 diabetes. Models adjusted for: age at surgery, sex, race, ethnicity, baseline BMI, baseline sagittal abdominal diameter, baseline body fat percentage, glycated haemoglobin (HbA1c), high-sensitivity c-reactive protein (hs-CRP), baseline type 2 diabetes, baseline dyslipidaemia, baseline hypertension, clinical centre, presence of any 30-day events, caregiver bariatric surgery, caregiver education, and household income. A total of 188 healthcare utilisation events requiring admission occurred among 88/228 (39%) participants at a rate of 90·2 per 500 person years follow up (95% CI 78·2–104·0). Rates were comparable between RYGB [87·9 per 500 person years (95% CI 74·1–104·3)] and sleeve gastrectomy [95·8 per 500 person years (95% CI 73·9–124·2)] ( Table 6 ). Table 6 Healthcare utilisation events with admission. Roux-en-Y gastric Bypass (N = 161) Sleeve gastrectomy (N = 67) TOTAL (N = 228) Patients, n (%) Events, n Rate a (95% CI) Patients, n (%) Events, n Rate a (95% CI) Patients, n (%) Events, n Rate a (95% CI) TOTAL 63 (39·1) 131 87·9 (74·1, 104·3) 25 (37·3) 57 95·8 (73·9, 124·2) 88 (38·6) 188 90·2 (78·2, 104·0) Skin/Wound 18 (11·2) 26 17·4 (11·9, 25·6) 9 (13·4) 21 35·3 (23·0, 54·1) 27 (11·8) 47 22·5 (16·9, 30·0) Oesophagogastric 6 (3·7) 11 7·4 (4·1, 13·3) 1 (1·5) 1 1·7 (0·2, 11·9) 7 (3·1) 12 5·8 (3·3, 10·1) General GI Symptoms 11 (6·8) 13 8·7 (5·1, 15·0) 5 (7·5) 7 11·8 (5·6, 24·7) 16 (7·0) 20 9·6 (6·2, 14·9) Renal 4 (2·5) 4 2·7 (1·0, 7·2) 0 0 0 4 (1·8) 4 1·9 (0·7, 5·1) Musculoskeletal 3 (1·9) 5 3·4 (1·4, 8·1) 1 (1·5) 3 5·0 (1·6, 15·6) 4 (1·8) 8 3·8 (1·9, 7·7) HPB 2 (1·2) 4 2·7 (1·0, 7·2) 3 (4·5) 3 5·0 (1·6, 15·6) 5 (2·2) 7 3·4 (1·6, 7·0) Urological 1 (0·6) 4 2·7 (1·0, 7·2) 2 (3·0) 3 5·0 (1·6, 15·6) 3 (1·3) 7 3·4 (1·6, 7·0) Respiratory 3 (1·9) 3 2·0 (0·6, 6·2) 1 (1·5) 1 1·7 (0·2, 11·9) 4 (1·8) 4 1·9 (0·7, 5·1) Neurological 4 (2·5) 4 2·7 (1·0, 7·2) 0 0 0 4 (1·8) 4 1·9 (0·7, 5·1) Endocrine 2 (1·2) 6 4·0 (1·8, 9·0) 0 0 0 2 (0·9) 6 2·9 (1·3, 6·4) Otorhinolaryngological 2 (1·2) 2 1·3 (0·3, 5·4) 0 0 0 2 (0·9) 2 1·0 (0·2, 3·8) Gastrointestinal 2 (1·2) 2 1·3 (0·3, 5·4) 2 (3·0) 2 3·4 (0·8, 13·4) 4 (1·8) 4 1·9 (0·7, 5·1) Obstetric 13 (10·3) 17 14·5 (9·0, 23·2) 3 (6·7) 4 10·3 (3·9, 27·5) 16 (9·4) 21 13·4 (8·8, 20·6) Gynaecological 2 (1·6) 2 1·7 (0·4, 6·8) 0 0 0 2 (1·2) 2 1·3 (0·3, 5·1) Haematological 0 0 0 2 (3·0) 2 3·4 (0·8, 13·4) 2 (0·9) 2 1·0 (0·2, 3·8) Psychosocial 10 (6·2) 17 11·4 (7·1, 18·4) 5 (7·5) 9 15·1 (7·9, 29·1) 15 (6·6) 26 12·5 (8·5, 18·3) Miscellaneous 3 (1·9) 3 2·0 (0·6, 6·2) 0 0 0 3 (1·3) 3 1·4 (0·5, 4·5) Intestinal 6 (3·7) 8 5·4 (2·7, 10·7) 1 (1·5) 1 1·7 (0·2, 11·9) 7 (3·1) 9 4·3 (2·2, 8·3) a Rate per 500 person-years. HPB, hepatopancreatobiliary. Healthcare utilisation events with admission. Rate per 500 person-years. HPB, hepatopancreatobiliary. The most frequent healthcare utilisation events involving admission were categorised as skin or wound problems, such as skin lesions and superficial abscesses (47 events), psychosocial (26 events), obstetric (21 events), and general gastrointestinal symptoms (20 events). Adjusted analyses did not detect a difference in the rate of admission events by surgical procedure (RR: 1·06; 95% CI 0·62, 1·81; p = 0·84). Participants with a perioperative event had a significantly higher rate of admission events (RR: 2·16; 95% CI 1·22, 3·80; p = 0·007). Admission events were more common with higher baseline BMI (RR: 0·97; 95% CI 0·94, 0·99; p = 0·024), and among those with baseline hypertension (RR: 1·64; 95% CI 0·99, 2·74; p = 0·05). Of the 228 participants, three deaths occurred at 3·2, 3·8, and 3·9 years after surgery [1·4 per 500 person years (95% CI 0·5–4·5)]. Each death was adjudicated in a similar fashion to other healthcare utilisation events. Two deaths resulted from acute combined drug toxicity, occurring long after and unrelated to their surgical treatment, at 3·8 and 3·9 years after RYGB, respectively. The third occurred 3·25 years after RYGB, caused by complications of poorly controlled type 1 diabetes. No death was attributable directly to surgery or post-operative care.

This study examines in detail and quantifies the real-world healthcare utilisation events and deaths experienced by adolescents from 30 days to 5 years following MBS for severe obesity. The main findings are that, although many participants experienced an healthcare utilisation event, most were unrelated to the bariatric procedure. The study was not powered to detect procedural differences, and indeed did not detect any statistical difference in these outcomes between RYGB and sleeve gastrectomy. Despite this overall statistical power limitation, healthcare utilisation events appeared to be more common in individuals who experienced an initial 30-day complication. Healthcare utilisation events also were more common in individuals with a caregiver who had also undergone MBS. Among those without type 2 diabetes, a higher risk of long term healthcare utilisation events was observed after sleeve gastrectomy than RYGB. Approximately half of the healthcare utilisation events involved an acute care hospital admission, more commonly in those with higher BMI and those with hypertension at baseline. During the 5-year follow-up period, 3 deaths occurred, all in individuals who had undergone RYGB and 2 of 3 suspected to be a result of drug and alcohol use. Each death occurred at least 3 years following surgery and was not attributed surgery, but can be linked to the high-risk population that receives adolescent bariatric surgery. While it is of concern that all participants had undergone RYGB, that group did have the largest sample size and sleeve gastrectomy procedures did not begin to be offered for several years after study enrolment began, thus selection bias likely influenced the distribution. The aetiology of substance abuse and its relation to bariatric surgery warrants further investigation. The adult LABS cohort observed 4·9 death/1000 person-year, 12 which was higher than the rate in this study, but comparable to the general US population. In general, understanding the healthcare utilisation experience of patients who undergo MBS is challenging, owing to variability in clinical follow-up, limited provider/patient contact time, multiplicity of care settings with medical record systems that sometimes are not linked, and failure to prioritise accurate collection and reporting of such data, even in prospective research studies. This leads to a high likelihood of substantial under-estimation of the healthcare utilisation and complications/events experienced by bariatric patients in the published literature. Detailed, protocol driven long-term data, such as those presented herein, have been absent thus far from even high-quality studies in adults and adolescents alike. Use of healthcare resources increases with rising BMI, 13 , 14 , 15 , 16 and obesity is causally linked to increased healthcare utilisation in childhood. 15 It was striking to observe that events unrelated to surgery occurred more frequently than related events in this surgical population, a novel and important observation for healthcare planning and expectations in this group. In the non-surgical population in the US, healthcare utilisation appears far greater among adolescents with obesity (BMI ≥95th centile), compared with overweight, normal weight or underweight (177·5 vs. 35·4 events per 100 person-years). 17 A >40% greater risk of hospitalisation (adjusted rate ratio [RR] 1·42, 95% CI 0·99, 2·04), and mental health visits (RR 1·48, 95% CI 1·14, 1·92) across 3 years, >25% more visits to the emergency department (RR 1·27, 95% CI 1·19, 1·35), and more visits to the outpatient clinic setting (RR 1·07, 95% CI 1·04, 1·11) have also been observed. 14 However, the validity of comparison between our data and this existing literature is limited since there are no available healthcare utilisation data for a cohort of adolescents with severe obesity (in whom the likelihood of experiencing health problems is greater in comparison to a group with less severe obesity). The same phenomenon is appreciated in adults with obesity, in whom end organ effects of obesity have typically had longer to establish disease states requiring healthcare attention or intervention, 18 , 19 although the associated excess costs reduce with weight loss. 20 Thus, our finding of healthcare utilisation events in more than half of the participants in this study is not surprising, given that these individuals have typically experienced severe obesity and its complications for many years prior to surgery, have also undergone a major gastrointestinal surgical intervention, and are, as demonstrated by their seeking surgery, capable of accessing healthcare. The most striking difference between procedure groups was the occurrence of oesophagogastric events. These predominantly comprised two event types; firstly ulceration, and secondly stricturing phenomena, such as anastomotic stricture. This category of event occurred almost exclusively in participants who underwent RYGB, at a rate of 14·1 per 500 person-years, compared with 3·4 per 500 person-years after sleeve gastrectomy. It should be noted that this rate may be overly pessimistic, as 21 events occurred in only 13 of the 161 RYGB participants (8·1%), compared with 2 events in 2 of the 67 sleeve gastrectomy participants (3·0%). This skew, related to recurrent events within the same participant, illustrates one limitation of using the rate per 500 person-years to compare the incidence of events. It was also notable that the incidence of gastro-oesophageal reflux and its complications, along with endoscopic examination, were unexpectedly low following sleeve gastrectomy in this study. This is particularly important as the future lifespan for a complication resultant from gastro-oesophageal reflux is many decades in this group. This may partly reflect under-reporting related to the timing of the study, between 2007 and 2011, when the expectation and perceived importance of this issue was less prominent. It is also likely that some cases were included in the “general GI symptoms” category, which included nausea and abdominal discomfort, rather than the “oesophagogastric” category. Despite this, we would caution scrutiny for this complication at a potentially greater incidence than identified herein. Both we, 2 , 3 , 4 and others, 5 , 6 have demonstrated notable benefits to physical health following MBS, including sustained improvements in glycaemic control, blood pressure, lipid status, inflammation, liver function and kidney function. However, despite some improvement in mental health from baseline to 2 years, 21 particularly among the physical wellbeing domains assessed within mental health tools, 10 alleviation of mental health problems should not be expected into the long term after MBS in this age group and serious mental health issues may persist or arise after MBS. 2 , 10 In addition, nutritional deficiencies have been reported to be highly prevalent in adolescents with severe obesity and this risk may be accentuated following MBS. 3 , 5 The higher rates of healthcare utilisation in the present study are similar to the Adolescent Morbid Obesity Surgery (AMOS) study, which reported Swedish register data comparing 161 adolescents who underwent RYGB or non-surgical treatment, describing significantly more inpatient hospital days (mean difference [MD] 13·0, 95% CI 7·4, 18·6), more outpatient visits (MD 4·9, 95% CI 1·3, 8·4), and similar prescription costs (MD $611, 95% CI −3252, 2030) after MBS, across 5-year follow-up. 5 These figures were uncontrolled for the index admission. The most rigorous examination of healthcare utilisation after MBS was the Swedish Obese Subjects study, a Swedish register-based cohort study examining 4047 adults across 20 years of follow-up after a range of MBS procedures. 22 This study demonstrated more inpatient and non-primary outpatient care during the first 6 years, but not thereafter. They also demonstrated lower drug costs for surgically managed patients during years 7–20. Another study examining adult outcomes reported increased emergency department visits in the first 2 years after RYGB, but lower prescription and overall costs by 3 years after surgery. 23 Again among adults, admissions have been shown to more than double from 8·4% to 20·2% between the year prior to, and the year after RYGB, before declining across the subsequent 2 years. 24 Healthcare utilisation events were more common among the 38 adolescents that we previously reported to have experienced a 30-day complication. 11 In many cases this is understandable, such as episodes of abdominal pain, nausea, or intestinal obstruction related to intra-abdominal adhesions, or complex abdominal wounds following post-operative wound infection. The interaction of unrelated healthcare utilisation events with type 2 diabetes was interesting. Where participants did not have type 2 diabetes at baseline, unrelated healthcare utilisation events were more frequent following sleeve gastrectomy than RYGB. We did not identify a satisfactory explanation for this finding and believe that further investigation is needed. We acknowledge that the small number of participants reaching some of these subgroups presents the possibility of type II statistical error. This may have contributed to the statistical non-significance of the higher frequency of unrelated events observed after RYGB in participants without diabetes at baseline. The adolescent MBS field typically follows a similar path to the adult MBS field as its own evidence emerges. There was a rapid shift toward greater use of sleeve gastrectomy in adults over the last 10–15 years, subsequently mirrored in adolescents. This has softened more recently in adults, with a reversion to greater use of RYGB. While we observed an excess in gastrointestinal complications in the RYGB group, more recent, high-quality, large randomised trials in adults have directly compared the two procedures, confirming similar safety profiles, 25 , 26 alongside favourable weight, health, and quality of life-related end-points at 3 years after RYGB. 26 The RYGB procedure has matured between our study and these randomised studies. Technical procedural advances, such as the use of linear rather than circular staplers, and additional paediatric surgical bariatric experience, may have positively influenced the incidence of stricture and ulceration after RYGB. The proportion of participants admitted to hospital across 5 years was high, at 39%, but independent of the procedure type. This is likely a reflection of the type of healthcare utilisation event leading to admission in the majority of cases, with most being related to issues independent of the procedure type, such as superficial skin infection, psychosocial events and childbirth. It should also be borne in mind that the threshold for admission of any patient who has undergone MBS, particularly the young, may be lower in general than that of the wider population, recognising the important potential complications of the procedures. Intra-abdominal operations related to the MBS procedure were also common, performed in 18·4% of participants, explaining the need for a number of the admissions. MBS leads to rapid and substantial liberation of cholesterol from fat stores, promoting gallstone formation, and cholecystectomy was performed in 25 of our participants across 5 years. Research emerging after the initiation of this study has identified an opportunity to reduce this; administration of ursodeoxycholic acid post-operatively has been shown to reduce the incidence of gallstones by 80%, and the cholecystectomy rate by more than 75%. 27 Healthcare utilisation events related to nutrition were less common than anticipated, with just 8 events recorded. MBS impacts upon macro- and micronutrition, and rates of nutritional deficiency after MBS in adolescents have been as high as 72%. 5 However, these are mostly subclinical and supplementation practice has evolved, largely through the adult experience, 28 reducing the incidence of nutritional deficiency or insufficiency. It is reassuring, therefore, that the rate of healthcare utilisation in relation to nutrition was low, particularly in a group in whom non-compliance with supplementation has been reported as being between 44 and 61%. 29 The institution of regular study visits may also have encouraged health-seeking behaviour, promoting routine contact with healthcare services. The similar rate of healthcare utilisation events observed between the sexes was unanticipated. In the wider adult population healthcare utilisation is greater among females, 30 largely owing to the disproportionate effect of reproductive biology in females. 30 Adolescence usually precedes the additional healthcare associated with female reproduction, albeit not menarche and early fertility, yet female children are more likely to have a usual source of healthcare. 31 It may be that the study's structured healthcare pathway softened the typical barrier to healthcare seeking sufficiently to close the gap between sexes. Healthcare utilisation appeared independent of ethnicity and measured socioeconomic factors, including caregiver education status and household income. Ethnicity and lower income have been shown to influence healthcare utilisation, with children of a ‘non-Hispanic, white’ background more likely than children from a ‘Hispanic’ or ‘black’ background, to have a usual care provider and to attend that provider, rather than the emergency department. 31 Similarly, a lower income is associated with the absence of a usual source of care, poor continuity of care, and failure to visit healthcare providers. 31 , 32 Again, the study pathway, along with modest reimbursement payments for travel and inconvenience, may have overridden the underlying causes for such disparities. Baseline comorbidities, weight and BMI were not significantly associated with healthcare utilisation events, perhaps related to the young, and therefore pre-morbid, age of participants, along with the health improvements and regular contact within the study. Strengths of this study include its unique design, with a detailed inquisition into all healthcare utilisation events after MBS, designed a priori . Prospective data collection was uniform, using standardised data proformas, completed by the healthcare team, minimising recall bias. With data reaching 5 years after surgery for all participants, these results represent long-term follow-up. This analysis represents the most in-depth interrogation for, and analysis of, healthcare utilisation events after MBS, including complications. Several limitations are also acknowledged, the principal of which is the absence of a non-surgical comparison group. The observational nature may lead to heterogeneity between groups according to the procedure performed, introducing possible selection bias. Not all events were independently assessed for relatedness and binary coding of relatedness may be overly simplistic for a minority of cases. Any events occurring outside of participating centres may not have been captured. However, participants were specifically asked about any such event at each follow-up visit within the study and additional forms were completed for any missing events. Events were classified by researchers according to the bodily systems involved, although this was prior to, and independent of, subsequent analyses. Generalisability within the US is likely to be good owing to the multicentre design and spread of participant ethnicity. Beyond the US, this may not apply as strongly. Finally, some events happened concurrently or as a consequence of one another. These are coded separately wherever possible, which may have led to overestimation of healthcare utilisation. Healthcare utilisation is high across 5 years after MBS in adolescents, both related and unrelated to the procedure. Closer surveillance is indicated in individuals who experience a 30-day complication, as their healthcare utilisation requirements may be greater. In patients with type 2 diabetes, a greater potential burden of healthcare utilisation events unrelated to the procedure should be expected following RYGB than after sleeve gastrectomy. In participants without diabetes, more unrelated events may be expected after sleeve gastrectomy. Together, these observations provide novel insight into the realistic healthcare utilisation requirements across 5 years after adolescent MBS.

Conception and design: All authors contributed to the conception and design of the study, as well as developing the methodology and study protocols. Acquisition and curation of data and formal analysis: All authors contrubted to data acquisition. Andrew J. Beamish, Todd Jenkins, Justin R. Ryder, and Thomas H. Inge had full access to the data, and led data curation and formal analysis. Original draft and critical review: Andrew J. Beamish, Todd Jenkins, Justin R. Ryder, and Thomas H. Inge wrote the original draft and all authors critically reviewed and edited the manuscript for intellectual content. Funding: Thomas H. Inge led the funding acquisition. Todd Jenkins, Justin R. Ryder, Robert M. Dorman, Marc Michalsky, Carrol Harmon, and Michael Helmrath also contributed to securing funding for the study. Project Administration: Andrew J. Beamish, Todd Jenkins, Justin R. Ryder, Anita Courcoulas, Robert M. Dorman, Marc Michalsky, Carrol Harmon, Michael Helmrath, Thomas H. Inge. All authors were involved in the administrative and logistical coordination of the project. Study supervision: Andrew J. Beamish, Todd Jenkins, Justin R. Ryder, Anita Courcoulas, Robert M. Dorman, Marc Michalsky, Carrol Harmon, Michael Helmrath, Thomas H. Inge. All authors contributed to study supervision. Manuscript submission: Thomas H. inge was responsible for the decision to submit the manuscript, and all authors agreed with this decision.

Metabolic and bariatric surgery (MBS) is a highly effective therapy for obesity and is recommended for use in adolescents with severe obesity. 1 , 2 , 3 , 4 , 5 Its safety and effectiveness have been demonstrated in youth in several high-quality studies, including a recent randomized trial 2 , 3 , 4 , 5 , 6 and uptake of MBS among adolescents is increasing. 7 While the expanding evidence base has been highly encouraging, the high-quality studies in the field have also prospectively examined adverse outcomes associated with this treatment, documenting early and longer-term side effects, complications and reoperations in order to more fully understand and address the challenges of this therapy in this younger population. 2 , 3 , 4 , 5 , 8 , 9 , 10 However, when assessing outcomes beyond 30 days, whether examining adults or adolescents, most studies focus on a narrow set of outcomes, predominantly incorporating weight and comorbidity status outcomes, with far less information on complications and, more generically, longer-term healthcare utilisation events. The real-world healthcare journey of adolescents who undergo MBS is, therefore, unclear beyond short-term surgical complications. We previously reported on perioperative adverse medical and surgical outcomes up to 30 days post procedure, 11 and other select adverse outcomes over a 3 or 5 year timeline. 2 , 3 The aim of this present study was to conduct an in-depth analysis of a broad spectrum of healthcare utilisation after MBS in youth over 5 years, and to assess the relationship of healthcare utilisation with baseline characteristics. Further, this study aimed to examine the relatedness of healthcare utilisation events with MBS and analyse whether experiencing an adverse event within the first 30 days was associated with increased healthcare needs across 5-year follow-up. We hypothesised that postoperative healthcare utilisation would be high in this population and that baseline characteristics and post-operative (30-day) complications would be predictive of healthcare utilisation across 5 years.

Dr. Beamish reports receiving funding from the Royal College of Surgeons of England in the form of a clinical research fellowship and reimbursed advisory board work for Johnson and Johnson, unrelated to the present work. Dr. Inge has received consulting fees from Standard Bariatrics, Teleflex, and Mediflex, Eli Lilly, and royalties from Wolters Kluwer (UpToDate). Dr. Ryder receives a donation of drug and placebo for a clinical trial from Boehringer Ingelheim, has received grants from Eli Lilly, Recordati, ADA, and AHA, and serves on the advisory board for Calorify. Dr. Michalsky reports consulting fees from Ascensus Surgical, Inc., and education honoraria from Intuitive Surgical Inc. Drs. Jenkins, Courcoulas, Dorman, Harmon and Helmrath report no conflicts.