Machine learning-based prediction of 5-year weight trajectories after bariatric surgery in adolescents: a retrospective cohort study from France and Sweden

doi:10.21203/rs.3.rs-5014399/v1

Download PDF

Article

Machine learning-based prediction of 5-year weight trajectories after bariatric surgery in adolescents: a retrospective cohort study from France and Sweden

https://doi.org/10.21203/rs.3.rs-5014399/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Metabolic-bariatric surgery is an efficient therapy in selected adolescents with severe obesity. However, predicting the postoperative weight loss is challenging. A machine learning calculator predicting 5-year weight loss trajectory has been developed in adults, based on seven simple preoperative features. The aim of the present study was to adapt and test this model in adolescents.

Methods

Retrospective cohort study in patients aged 12 to 20 years undergoing Roux-en-Y gastric bypass (RYGB), sleeve gastrectomy (SG), or adjustable gastric band (AGB) in France and Sweden between 2001 and 2022. Primary outcome was the accuracy of 5-year BMI prediction, expressed as the median absolute deviation (MAD) between predicted and observed BMI. The model, initially developed in adults, was trained in a subset of adolescents consisting of 80 % of randomly selected participants, and secondly tested in the remaining 20%.

Results

We enrolled 2255 patients (1705 female [75.6%], 12 to 20 years [median 19]). The median (IQR) 5-year total weight loss was 30.2% (23.9–38.6) for RYGB, 23.4% (13.7–32.8) for SG, and 13.4% (0.0–30.1) for AGB. The adapted model predicted the observed 5-year BMI with a MAD of 3.7 kg/m² (95% CI [3.3–3.9]). The accuracy of the model was maximal for bypass (3.2 kg/m² [3.0–3.7]), good for SG (3.9 kg/m² [3.1–5.0]), and lower for AGB (7.3 kg/m² [5.5–8.4]), and accuracy decreased with time and in adolescents under 19 years. Age, height, weight, and type of intervention influenced 5-year weight loss. Coexisting type 2 diabetes influenced weight loss until 2 years after surgery but not after 5 years.

Conclusion

The model had an acceptable accuracy to predict 5-year postoperative weight loss trajectory when adapted in adolescents. Accuracy decreased over time of follow up and was influenced by type of intervention and age.

Health sciences/Endocrinology

Health sciences/Endocrinology/Endocrine system and metabolic diseases/Obesity

The prevalence of severe obesity is still raising globally in children and adolescents (1,2), and is a major health concern. Children living with obesity have a 70% risk to remain having obesity in adulthood (3), which expose them to precocious onset of comorbidities and major psychological and social burden (4).

In young people living with severe obesity, multidisciplinary lifestyle interventions often fail to result in long term significant weight loss (5,6). Nevertheless, they remain essential for patients’ empowerment and for maintaining results after more intensive treatments (7–9).

Despite recent promising results of pharmacotherapies in adolescents (10), bariatric surgery remains, to date, the most efficient treatment for sustained weight loss and improvement of comorbidities (7,9,11,12). In some countries, bariatric surgery is now a validated therapeutic option for selected adolescents with severe obesity when classical approaches have failed (13–15).

Adolescents need thorough preparation and careful assessment before bariatric surgery. Despite high-quality training programs, even in experienced teams, there is a risk of acute and chronic complications and variable results (16).

Information about the expected weight loss is particularly important in adolescents. Unrealistic expectations can indeed result in postoperative disappointment and may reinforce mental health problems after bariatric surgery (17). There is therefore a crucial need to identify factors that can predict the individual response.

A user-friendly tool applying machine learning for bariatric weight trajectory prediction has been developed for adults and has been internationally validated (18). The algorithm, based on decision trees, is able to predict 5-year postoperative weight loss trajectories, based only on seven preoperative features: weight, height, type of intervention, age at intervention, smoking history, type 2 diabetes (T2D) status, and diabetes duration (19).

The aims of the present study were (1) to train and validate an adapted prediction model from adults to adolescents, (2) to assess if patients’ characteristics and surgical technique: Adjustable gastric band (AGB); Vertical sleeve Gastrectomy (VSG) and Roux-en-y gastric bypass (RYGB) influence accuracy of weight loss prediction in youth in comparison to in adults.

Study design and participants

In this international retrospective observational study, we used data from two cohorts of adolescents living with obesity and submitted for the first time to bariatric surgery (RYGB, SG and AGB), one from France gathering patients from 5 academic centers, and one from Sweden based on the national registry. Participants were aged 12 to 20 years, and were followed up to 5 years after surgery. If patients had a subsequent bariatric intervention within 5 years following the first one, data were censored at the time of reoperation. Data points for patients lost to follow-up were kept in the analysis but censored after the last visit.

In France, data was collected in the following University Hospitals Lille, Angers, and Paris (Bicêtre, Robert-Debré, Necker and Armand-Trousseau Hospitals) for the period from March 2001 to December 2022. Data from all patients under 18 years came from electronic or paper medical records. In Lille, we also enrolled participants from the ABOS cohort (NCT01129297) aged between 18 and 20 years. In order to reduce the number of missing data, some patients were also contacted by telephone and e-mail, particularly for long term follow-up. In Sweden, virtually all patients undergoing metabolic, bariatric surgery are registered in the nationwide research and quality registry, SOReg (the Scandinavian Obesity Surgery Registry). The registry is continuously validated and has so far been shown to have very high validity of data (20). Anonymous data from the SOReg were included in the current study.

The study protocol was approved by the Ethics Review Committee (n°2023-070) of University of Lille. Non-opposition forms were sent out to all French participants by post. In Sweden, all patients are informed of the registry and that research will be conducted based on information from the registry (“opt-out”). Given the anonymous extraction of data, no additional ethics approval in Sweden was necessary.

Data analyzed for each patient at baseline included the seven clinical items which have been shown to be the most relevant for predicting weight loss in adults: height, weight and age at surgery, smoking status, diagnosis and duration of T2D. Smoking status was defined as 'smoking' or 'nonsmoking' at surgery. We then analyzed postoperative height and weight measured at 12, 24, 36 and 60 months. Total weight loss in percent was calculated as:

TWL = (visit weight – preoperative weight)/preoperative weight × 100

The preoperative weight is those, assessed before preoperative weight reduction.

Study outcomes

The primary outcome was the accuracy of 5-year BMI prediction, expressed as the median absolute deviation (MAD), between predicted and observed BMI expressed in kg/m². Secondary outcomes were weight loss prediction accuracy at earlier postoperative visits (at months 12, 24), expressed by MAD and normalized MAD as percentage of BMI.

Adaptation of the adult prediction model to teenagers’ cohort

For the purpose of this study, the French and Swedish cohorts were first merged into a single dataset, which was further divided into two subsets as follows. Among each of the three intervention subgroups (AGB, SG, RYGB), 80% of patients were randomly selected to form the training subset, and the remaining 20% were assigned to the testing subset. This assignment criterion ensures that both the training and testing subsets reflect the repartition of interventions observed in the merged cohort.

Following the methodology developed for adults, we then trained a class of machine learning algorithms called decision trees on the training subset to, first, learn meaningful subgroups of patients that share statistical similarities in their baseline characteristics, and second, to fit a TWL prediction model for each subgroup. As detailed in the initial report (19), decision trees are non-linear predictors that are able to learn customized predictions depending on the type of bariatric interventions, as well as other clinical variables.

To specialize the tree-based prediction model to teenagers, we used the seven attributes that were previously selected by the LASSO (least absolute shrinkage and selection operator) analysis performed in the adult study (19), namely type of intervention, preoperative weight, height, age at intervention, smoking status, diabetes status and duration, and we subsequently applied the regression trees (CART) algorithm on the training subset to learn adolescent-specific subgroups and predicted TWL. We further compared the predicted weight losses to the observed outcomes of patients in the testing subset.

Statistics

The baseline variables used to predict the weight loss trajectories were extracted from the retrospective French and Swedish patients’ medical records using the least absolute shrinkage and selection operator (LASSO) algorithm (21). The prediction model consisted of decision trees, trained at each outcome date (month 12, 24 and 60 after bariatric surgery) to predict percent of total weight loss (TWL) from the baseline variables, using the classification and regression trees (CART) algorithm (22). Predicted TWL were then converted into predicted weight and BMI.

The accuracy of the predicted outcomes with respect to the true outcomes was reported in terms of MAD. 95% confidence intervals (CI) for MAD were estimated by bootstrap (BCa method, n=10,000 replications). The predictive model for adolescents was developed by comparison with the adult predictive model, itself validated by comparison with other predictive models in the literature. Comparison is made with MAD.

We also used Bland-Altman plots to evaluate model calibration and 1-sample t-tests to assess the presence of systematic bias. Weight and BMI median (Interquartile Range [IQR]) 5-year trajectories of participants submitted to each operation were illustrated as a function of time using a nonlinear smoothing.

Subgroup analyses were performed by splitting the study cohort by intervention, pregnancy status, age (younger vs. older than 19 years old) and AGB removal.

The analysis was performed using the R software version 4.3.1, the library rpart for the CART implementation. We used the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD AI) guidelines to report the prediction model’s development and validation.

The overall study enrolled participants from the two cohorts, from France (n=329) and Sweden (n=1926), representing a total of 2255 patients.

Dropout rates and patients excluded for analysis are presented in a flow chart according to each follow-up time (Figure 1). At 5 years, we had 59% and 38% of expected visits in French and Swedish cohorts respectively. Main reasons for excluding patients were: reoperation by another bariatric surgical technique; AGB removal; patients who had less than 5 years follow up; pregnancy within the 5 years period and death.

The baseline characteristics of participants of each cohort and the proportion of each type of operation performed are shown in Table 1.

During the study period, AGB was the most frequent intervention in France (50%), while in Sweden the majority of patients underwent RYGB (87%). Eighteen Swedish patients (0.9%) had another surgical technique and were excluded. At baseline, individual weights and BMIs ranged from 72 to 235 kg and from 25.6 to 79.5 kg.m^-² respectively. The median age was 19 years which corresponds to the WHO definition of the end of the adolescence (10 to 19 years) (23). Thus, the cohort included 1 514 adolescents and, additionally 741 young adults aged 19 to 20 years.

The overall trajectories of the median (IQR) BMI and TWL observed during the 5 years after each operation are shown in Figure 2. Those data are shown first, for each country, and second, for merged cohort.

At 5 years, the median TWL was 30.2% (IQR 23.9–38.6) for RYGB, 23.4% (IQR 13.7–32.8) for the SG and 13.4% (IQR 0.0–30.1) for the AGB. Although Swedish patients were overall older and with a lower BMI, the general shapes of weight loss trajectories for each operation were similar among cohorts. The nadir weight loss was reached between 1 and 2 years, followed by limited weight regain: median 14.5% (IQR 0.0–31.0) of maximal weight loss across operations. Weight regain was significantly (p<0.001) greater after sleeve gastrectomy 27.1% (IQR 0.0–49.3), compared to gastric bypass 13.9% (IQR 0.0–27.3) (Figure 2).

Using decision trees, we found that the only variable which had a significant impact on weight loss at five years, among the seven predictive features previously selected in adults (19), was the surgical technique. For earlier visits (up to 2 years), the presence of diabetes impacted weight loss negatively for patients who underwent SG and RYGB. There was no other meaningful predictive feature for weight loss after AGB at any postoperative time. Smoking status had no impact.

The model’s prediction performances were evaluated at 1, 2, and 5 years in the testing subset of the training cohort. Overall, the mean difference between predicted and observed BMI at 5 years was 3.7 kg/m² (95% CI 3.3–3.9), resulting in a 12 percent normalized difference. The predictive performances of the model at 1 year, 2 years, and 5 years in the testing cohort and for each intervention are shown in Figure 3. At 5 postoperative years, predictive performances were significantly better after RYGB (MAD 3.2) than after AGB (MAD 7.3, p<0.001), and moderate after sleeve gastrectomy (MAD 3.9, p=0.06) (Figure 3).

Finally, we tested the influence of age at surgery on the accuracy of weight loss prediction. We took as a cut-off the median age of the cohort (19 years). As presented in Figure 4, the accuracy of weight loss prediction was lower in adolescents (higher MAD under 19 years), compared to young adults.

Finally, the machine learning model developed in the present study for adolescent was integrated into the existing software (bwtp. Univ-lille.fr). The resulting user-friendly calculator displays the predicted weight trajectory at any given time during 5 years after surgery, alongside prediction intervals corresponding to IQR of prediction errors. By default, individual predicted trajectories are expressed in kilograms. According to the user’s choice, results can also be displayed in kg.m^-², percentage of TWL, or percentage of Excess Weight Loss. To improve readability, predicted trajectories over time for each of these metrics are displayed as smooth curves.

Two illustrative examples of BMI and TWL trajectories predicted for individual patients from the current version of the bariatric weigh trajectory prediction calculator are shown in Figure 5A and 5B.

In this study, we adapted a machine learning based model, predicting 5-year weight loss trajectory after bariatric surgery in adolescents. This model, previously developed and tested in adults (19), was retrained and validated in a large international collaborative cohort of younger patients. The resulting calculator was able to predict 5-year BMI in adolescent with a precision of 3.7 kg.m^-2 (12 %), as compared to the observed values, based on collection of five simple preoperative variables, namely age, weight, height, type of intervention and type 2 diabetes status.

Compared to adults, the calculator has less, but still acceptable accuracy for predicting 5-year BMI in youth (MAD 2.8 kg.m^-2 in adults vs MAD 3.7 kg.m^-2in adolescents) (19). Surgical technique had a major impact on the accuracy of the weight loss prediction, which was good for RYGB, moderate for SG, but lower for patients who had an AGB, since in the latter group there weas a greater variability of postoperative results. Discrepancies in postoperative results according to surgical technique have been reported by other authors in adolescents, and the choice of the optimal surgical technique is still a subject of debate (15,24–37).

Of note, the accuracy of prediction increased with age, and decreased with the duration of follow-up. Overall a younger age was associated with less precise weight loss prediction, illustrated by a greater MAD for BMI in the youngest half of the cohort (patients under 19). As previously reported by Inge et al for RYGB, the range of weight trajectories over time was wider in teenagers, compared to adults (38). In adult cohorts, younger age has been described as a favorable factor for weight loss (39), although weight regain is more common in younger patients, resulting in a greater dispersion of results with time (40). The explanation of this finding remains elusive, since precise data is lacking (41). One can however speculate that major changes in hormonal status, behavior and lifestyle, during and at the end of adolescence, as well as genetic background might explain these discrepancies in long term results.

Additionally, the presence of type 2 diabetes had a negative impact on weight loss in first postoperative years. Type 2 diabetes was however, rare (3.4%) in adolescents and its duration was short. Our results are consistent with findings of Borges-Canha et al, who reported that insulin resistance and beta cell dysfunction impact negatively post-surgical weight loss, even before type 2 diabetes onset (42). In adults, type 2 diabetes is known to be a major limiting factor of weight loss after bariatric surgery, particularly when the disease is ancient and poorly controlled (19,43,44).

Interestingly, postoperative weight loss trajectories converged at 5 years between patients with and without diabetes at baseline (figure 6B). This result illustrates that in adolescents, the negative impact of type 2 diabetes on weight loss disappeared with time, in line with other reports showing sustainable diabetes remission after bariatric surgery in that age group (15,32,33,38).

Our study had several limitations. First this retrospective study gathered data from only two European countries with a majority of Caucasian patients. We had no extensive prospective clinical biological and genetical data for testing predictive features, neither had we any social, developmental and psychological aspects, which are known to be present with significative differences between adolescents with severe obesity compared to normally weighted peers (45). However, the present study population was one of the largest reported cohorts of bariatric surgery in adolescents (26,32,41). The sex ratio and the quality of follow up (lost-of-view rates) were similar to those previously reported (26,37,38,46–48).

Second, we tested here the impact of seven relevant clinical features previously identified in adults. It is unclear if these are the most relevant for youth, but in the lack of comprehensive data for adolescents made us unable to explore other factors. Some authors have identified parameters associated with the magnitude and sustainability of postoperative weight loss in young patients, beyond baseline BMI, such as the duration of preoperative training, intentional weight loss before surgery (8,49,50), other authors didn’t find any predictive factors (51). Finally, we did not consider the genetic background of our cohort. It is well known that some genetic variants have negative impact on long term weight loss results after bariatric surgery (52–55).

In summary, we have adapted and internationally validated a machine-learning-based tool, previously developed for adults and easy-to-use for clinicians, and demonstrate a good accuracy to predict weight loss after bariatric surgery for adolescents. Based on the present work, we also found that surgical technique and, to a lesser extent, age and type 2 diabetes had an influence on the weight loss and on the accuracy of its prediction.

Author contributions

Iva Gueorguieva and François Pattou designed the protocol and wrote the manuscript.

Celine Lespagnol collected and analyzed French data.

Patrick Saux and Philippe Preux analyzed data, adapted calculator to adolescents and provided statistical analysis.

Torsten Olbers provided Swedish data, updated reference list and provided feedback on the report.

Kajsa Jarvholm updated reference list and provided feedback on the report.

Erik Stenberg contributed to extracting of the Swedish data and provided feedback on the report.

Estelle Aubry, Robert Caiazzo, Hélène Verkindt, Naima Oukhoiuya, Jessica Amsellem-Jager, Regis Coutant, Myriam Dabbas, Naziha Khen-Dunlop,Gianpaolo De Filippo, Beatrice Dubernand Catherine Terrazcontributed to collecting and extracting French data.

Regis Coutant, Myriam Dabbas, Gianpaolo De Filippo and Beatrice Dubern provided feedback on the report.

Author conflict of interest

Authors declare not having any competing financial interests in relation with the present work.

Hales CM, Fryar CD, Carroll MD, Freedman DS, Ogden CL. Trends in Obesity and Severe Obesity Prevalence in US Youth and Adults by Sex and Age, 2007-2008 to 2015-2016. JAMA. 24 avr 2018;319(16):1723‑5.
Phelps NH, Singleton RK, Zhou B, Heap RA, Mishra A, Bennett JE, et al. Worldwide trends in underweight and obesity from 1990 to 2022: a pooled analysis of 3663 population-representative studies with 222 million children, adolescents, and adults. The Lancet. mars 2024;403(10431):1027‑50.
Clarke WR, Lauer RM. Does childhood obesity track into adulthood? Critical Reviews in Food Science and Nutrition. 1 janv 1993;33(4‑5):423‑30.
Rankin J, Matthews L, Cobley S, Han A, Sanders R, Wiltshire HD, et al. Psychological consequences of childhood obesity: psychiatric comorbidity and prevention. Adolescent Health, Medicine and Therapeutics. 14 nov 2016;7:125‑46.
Ells LJ, Rees K, Brown T, Mead E, Al-Khudairy L, Azevedo L, et al. Interventions for treating children and adolescents with overweight and obesity: an overview of Cochrane reviews. Int J Obes. nov 2018;42(11):1823‑33.
Al‐Khudairy L, Loveman E, Colquitt JL, Mead E, Johnson RE, Fraser H, et al. Diet, physical activity and behavioural interventions for the treatment of overweight or obese adolescents aged 12 to 17 years - Al‐Khudairy, L - 2017 | Cochrane Library. [cité 27 août 2024]; Disponible sur: https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD012691/full
Järvholm K, Janson A, Peltonen M, Neovius M, Gronowitz E, Engström M, et al. Metabolic and bariatric surgery versus intensive non-surgical treatment for adolescents with severe obesity (AMOS2): a multicentre, randomised, controlled trial in Sweden. The Lancet Child & Adolescent Health. 1 avr 2023;7(4):249‑60.
Fennig U, Snir A, Halifa-Kurzman I, Sela A, Hadas A, Fennig S. Pre-surgical Weight Loss Predicts Post-surgical Weight Loss Trajectories in Adolescents Enrolled in a Bariatric Program. OBES SURG. 1 avr 2019;29(4):1154‑63.
Inge TH, Jenkins TM, Xanthakos SA, Dixon JB, Daniels SR, Zeller MH, et al. Long-term outcomes of bariatric surgery in adolescents with severe obesity (FABS-5+): a prospective follow-up analysis. The Lancet Diabetes & Endocrinology. 1 mars 2017;5(3):165‑73.
Weghuber D, Barrett T, Barrientos-Pérez M, Gies I, Hesse D, Jeppesen OK, et al. Once-Weekly Semaglutide in Adolescents with Obesity. New England Journal of Medicine. 14 déc 2022;387(24):2245‑57.
Olbers T, Beamish AJ, Gronowitz E, Flodmark CE, Dahlgren J, Bruze G, et al. Laparoscopic Roux-en-Y gastric bypass in adolescents with severe obesity (AMOS): a prospective, 5-year, Swedish nationwide study. The Lancet Diabetes & Endocrinology. 1 mars 2017;5(3):174‑83.
Beamish AJ, Ryan Harper E, Järvholm K, Janson A, Olbers T. Long-term Outcomes Following Adolescent Metabolic and Bariatric Surgery. The Journal of Clinical Endocrinology & Metabolism. 1 sept 2023;108(9):2184‑92.
Haute Autorité de Santé [Internet]. [cité 27 août 2024]. Définition des critères de réalisation des interventions de chirurgie bariatrique chez les moins de 18 ans. Disponible sur: https://www.has-sante.fr/jcms/c_2010309/fr/definition-des-criteres-de-realisation-des-interventions-de-chirurgie-bariatrique-chez-les-moins-de-18-ans
Nationella riktlinjer för vård vid obesitas: prioriteringsstöd till beslutsfattare och chefer 2023. Stockholm: Socialstyrelsen; 2023. 60 p.
Janson A, Järvholm K, Sjögren L, Dahlgren J, Beamish AJ, Gronowitz E, et al. Metabolic and Bariatric Surgery in Adolescents: For Whom, When, and How? Hormone Research in Paediatrics. 9 mars 2022;96(6):609‑19.
McClelland PH, Kabata K, Gorecki W, Jano A, Zenilman ME, Gorecki P. Long-term weight loss after bariatric procedures for morbidly obese adolescents and youth: a single-institution analysis with up to 19-year follow-up. Surg Endosc. 1 mars 2023;37(3):2224‑38.
Järvholm K, Olbers T, Peltonen M, Marcus C, Flodmark CE, Gronowitz E, et al. Depression, anxiety, and suicidal ideation in young adults 5 years after undergoing bariatric surgery as adolescents. Eat Weight Disord. 1 mai 2021;26(4):1211‑21.
Bariatric Weight Trajectory Prediction [Internet]. [cité 27 août 2024]. Disponible sur: https://bariatric-weight-trajectory-prediction.univ-lille.fr/
Saux P, Bauvin P, Raverdy V, Teigny J, Verkindt H, Soumphonphakdy T, et al. Development and validation of an interpretable machine learning-based calculator for predicting 5-year weight trajectories after bariatric surgery: a multinational retrospective cohort SOPHIA study. The Lancet Digital Health. 1 oct 2023;5(10):e692‑702.
Sundbom M, Näslund I, Näslund E, Ottosson J. High acquisition rate and internal validity in the Scandinavian Obesity Surgery Registry. Surgery for Obesity and Related Diseases. 1 mars 2021;17(3):606‑14.
Lim M, Hastie T. Learning Interactions via Hierarchical Group-Lasso Regularization. Journal of Computational and Graphical Statistics. 3 juill 2015;24(3):627‑54.
Breiman L, Friedman J, Olshen RA, Stone CJ. Classification and Regression Trees. New York: Chapman and Hall/CRC; 2017. 368 p.
Adolescent health [Internet]. [cité 27 août 2024]. Disponible sur: https://www.who.int/health-topics/adolescent-health
Zitsman JL, DiGiorgi MF, Zhang AZ, Kopchinski JS, Sysko R, Devlin MJ, et al. Adolescent Gastric Banding: a 5-Year Longitudinal Study. OBES SURG. 1 mars 2020;30(3):828‑36.
Silberhumer GR, Miller K, Pump A, Kriwanek S, Widhalm K, Gyoeri G, et al. Long-term results after laparoscopic adjustable gastric banding in adolescent patients: follow-up of the Austrian experience. Surg Endosc. 1 sept 2011;25(9):2993‑9.
Black JA, White B, Viner RM, Simmons RK. Bariatric surgery for obese children and adolescents: a systematic review and meta-analysis. Obesity Reviews. 2013;14(8):634‑44.
Khen-Dunlop N, Dabbas M, De Filippo G, Jais JP, Hervieux E, Télion C, et al. Primordial Influence of Post-operative Compliance on Weight Loss After Adolescent Laparoscopic Adjustable Gastric Banding. OBES SURG. 1 janv 2016;26(1):98‑104.
Hervieux E, Baud G, Dabbas M, Pigeyre M, Caiazzo R, Verhaeghe R, et al. Comparative results of gastric banding in adolescents and young adults. Journal of Pediatric Surgery. 1 juill 2016;51(7):1122‑5.
Schmitt F, Riquin E, Beaumesnil M, Dinomais M, Topart P, Weil D, et al. Laparoscopic adjustable gastric banding in adolescents: Results at two years including psychosocial aspects. Journal of Pediatric Surgery. 1 mars 2016;51(3):403‑8.
Dewberry LC, Jalivand A, Gupta R, Jenkins TM, Beamish A, Inge TH, et al. Weight Loss and Health Status 5 Years After Adjustable Gastric Banding in Adolescents. OBES SURG. 1 juin 2020;30(6):2388‑94.
Shoar S, Mahmoudzadeh H, Naderan M, Bagheri-Hariri S, Wong C, Parizi AS, et al. Long-Term Outcome of Bariatric Surgery in Morbidly Obese Adolescents: a Systematic Review and Meta-Analysis of 950 Patients with a Minimum of 3 years Follow-Up. OBES SURG. 1 déc 2017;27(12):3110‑7.
Wu Z, Gao Z, Qiao Y, Chen F, Guan B, Wu L, et al. Long-Term Results of Bariatric Surgery in Adolescents with at Least 5 Years of Follow-up: a Systematic Review and Meta-Analysis. OBES SURG. 1 juin 2023;33(6):1730‑45.
Vilallonga R, Himpens J, van de Vrande S. Long-Term (7 Years) Follow-Up of Roux-en-Y Gastric Bypass on Obese Adolescent Patients (<18 Years). Obesity Facts. 2 avr 2016;9(2):91‑100.
de la Cruz-Muñoz N, Xie L, Quiroz HJ, Kutlu OC, Atem F, Lipshultz SE, et al. Long-Term Outcomes after Adolescent Bariatric Surgery. Journal of the American College of Surgeons. oct 2022;235(4):592.
Al Sabah S, Haddad EA, Shuaibi S, Qadhi I, Al-Saidan L, Khayat A. Long-term Outcomes of Sleeve Gastrectomy in Adolescent Patients: The Effect of Weight Loss in Younger Years to Outcomes in Adulthood. BMC Surg. 28 avr 2023;23(1):103.
Alqahtani AR, Elahmedi M, Abdurabu HY, Alqahtani S. Ten-Year Outcomes of Children and Adolescents Who Underwent Sleeve Gastrectomy: Weight Loss, Comorbidity Resolution, Adverse Events, and Growth Velocity. Journal of the American College of Surgeons. 1 déc 2021;233(6):657‑64.
Ogle SB, Dewberry LC, Jenkins TM, Inge TH, Kelsey M, Bruzoni M, et al. Outcomes of Bariatric Surgery in Older Versus Younger Adolescents. Pediatrics. 1 mars 2021;147(3):e2020024182.
Inge TH, Courcoulas AP, Jenkins TM, Michalsky MP, Brandt ML, Xanthakos SA, et al. Five-Year Outcomes of Gastric Bypass in Adolescents as Compared with Adults. New England Journal of Medicine. 30 mai 2019;380(22):2136‑45.
Parri A, Benaiges D, Schröder H, Izquierdo-Pulido M, Ramón J, Villatoro M, et al. Preoperative Predictors of Weight Loss at 4 Years Following Bariatric Surgery. Nutrition in Clinical Practice. 2015;30(3):420‑4.
Shantavasinkul PC, Omotosho P, Corsino L, Portenier D, Torquati A. Predictors of weight regain in patients who underwent Roux-en-Y gastric bypass surgery. Surgery for Obesity and Related Diseases. 1 nov 2016;12(9):1640‑5.
Trooboff SW, Stucke RS, Riblet NB, Kulkarni AS, Anand R, Casey A, et al. Psychosocial Outcomes Following Adolescent Metabolic and Bariatric Surgery: a Systematic Review and Meta-Analysis. OBES SURG. 1 nov 2019;29(11):3653‑64.
Borges-Canha M, Neves JS, Mendonça F, Silva MM, Costa C, M. Cabral P, et al. Beta Cell Function as a Baseline Predictor of Weight Loss After Bariatric Surgery. Front Endocrinol [Internet]. 12 août 2021 [cité 27 août 2024];12. Disponible sur: https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2021.714173/full
Courcoulas AP, Christian NJ, O’Rourke RW, Dakin G, Patchen Dellinger E, Flum DR, et al. Preoperative factors and 3-year weight change in the Longitudinal Assessment of Bariatric Surgery (LABS) consortium. Surgery for Obesity and Related Diseases. 1 sept 2015;11(5):1109‑18.
Raverdy V, Cohen RV, Caiazzo R, Verkindt H, Petry TBZ, Marciniak C, et al. Data-driven subgroups of type 2 diabetes, metabolic response, and renal risk profile after bariatric surgery: a retrospective cohort study. The Lancet Diabetes & Endocrinology. 1 mars 2022;10(3):167‑76.
Kelly AS, Barlow SE, Rao G, Inge TH, Hayman LL, Steinberger J, et al. Severe Obesity in Children and Adolescents: Identification, Associated Health Risks, and Treatment Approaches. Circulation. 8 oct 2013;128(15):1689‑712.
Zitsman JL, DiGiorgi MF, Zhang AZ, Kopchinski JS, Sysko R, Devlin MJ, et al. Adolescent Gastric Banding: a 5-Year Longitudinal Study. OBES SURG. 1 mars 2020;30(3):828‑36.
Furbetta N, Gragnani F, Cervelli R, Guidi F, Furbetta F. Teenagers with obesity: Long-term results of laparoscopic adjustable gastric banding. Journal of Pediatric Surgery. 1 avr 2020;55(4):732‑6.
Dreber H, Thorell A, Torgerson J, Reynisdottir S, Hemmingsson E. Weight loss, adverse events, and loss to follow-up after gastric bypass in young versus older adults: A Scandinavian Obesity Surgery Registry study. Surgery for Obesity and Related Diseases. 1 sept 2018;14(9):1319‑26.
Livhits M, Mercado C, Yermilov I, Parikh JA, Dutson E, Mehran A, et al. Preoperative Predictors of Weight Loss Following Bariatric Surgery: Systematic Review. OBES SURG. 1 janv 2012;22(1):70‑89.
Fennig U, Brik S, Sela A, Shachar-Lavie I, Tahar T, Halifa-Kurzman I, et al. Weight and Waiting for Adolescent’s Bariatric Surgery: Changes in Weight During Waiting Periods for Bariatric Surgery in Adolescents. OBES SURG. 1 août 2020;30(8):2920‑6.
Ryder JR, Gross AC, Fox CK, Kaizer AM, Rudser KD, Jenkins TM, et al. Factors associated with long-term weight-loss maintenance following bariatric surgery in adolescents with severe obesity. Int J Obes. janv 2018;42(1):102‑7.
de Toro-Martín J, Guénard F, Tchernof A, Pérusse L, Marceau S, Vohl MC. Polygenic risk score for predicting weight loss after bariatric surgery. JCI Insight. 3(17):e122011.
Antoine D, Guéant-Rodriguez RM, Chèvre JC, Hergalant S, Sharma T, Li Z, et al. Low-frequency Coding Variants Associated With Body Mass Index Affect the Success of Bariatric Surgery. The Journal of Clinical Endocrinology & Metabolism. 1 mars 2022;107(3):e1074‑84.
Gantz MG, Driscoll DJ, Miller JL, Duis JB, Butler MG, Gourash L, et al. Critical review of bariatric surgical outcomes in patients with Prader-Willi syndrome and other hyperphagic disorders. Obesity. 2022;30(5):973‑81.
Poitou C, Puder L, Dubern B, Krabusch P, Genser L, Wiegand S, et al. Long-term outcomes of bariatric surgery in patients with bi-allelic mutations in the POMC, LEPR, and MC4R genes. Surgery for Obesity and Related Diseases. 1 août 2021;17(8):1449‑56.

Table 1: Baseline characteristics (M0)

Legend: AGB, adjustable gastric band; BMI, body mass index; F, female; IQR, interquartile range; NA, not available; others: another surgical technique; RYGB, Roux-en-y gastric bypass; SG, sleeve gastrectomy; T2D, type 2 diabetes.

There is NO conflict of interest to disclose

Download PDF

Reviewers invited by journal
08 Sep, 2024
Submission checks completed at journal
04 Sep, 2024
First submitted to journal
03 Sep, 2024
Unknown event
02 Sep, 2024
Editor assigned by journal
01 Sep, 2024

You are reading this latest preprint version

Machine learning-based prediction of 5-year weight trajectories after bariatric surgery in adolescents: a retrospective cohort study from France and Sweden

Status:

Version 1

Abstract

Figures

INTRODUCTION

METHODS

RESULTS

DISCUSSION

Declarations

References

Table

Additional Declarations

Status:

Version 1