Abstract P4-11-34: An integrated clinical, behavioral and biological model to predict the risk of weight gain among breast cancer survivors (BCS)

Abstract

Abstract Background: Weight management is an integral part of survivorship care. Excess weight in BCS is associated with worse clinical outcomes and quality of life. Early identification of BCS at risk of gaining substantial weight could lead to prompt and tailored interventions. We aimed at developing a predictive model of weight gain that integrates clinical, behavioral and biological data. Methods: We included patients with stage I-III BC from the CANTO cohort (NCT01993498). CANTO collects longitudinal data, including objective anthropometric measures, at diagnosis (dx), 1 (T1), 2 (T2) and 4 (T3) years after dx. In addition, profiling of blood samples obtained at dx was performed for two sub-cohorts with HR+/HER2- BC for quantification of: (1) inflammatory and metabolic biomarkers (IL6, TNFα, IL1RA, CRP, IL2, IL1β, IFNγ, IL10, IL1A, IL4, IL8, ADPN, LEPT, INS, RETN) and (2) detectable proteins using hyper reaction mass spectrometry (Biognosys). Our outcome of interest was weight gain (increase ≥ 5%) compared to dx. First, multivariable logistic regression with bootstrapped Augmented Backwards Elimination (ABE) retained associations between weight gain and clinico-behavioral covariates. To assess contribution of biologic data, ABE retained associations between weight gain and biomarkers, correcting for significant covariates. Models were validated using internal cross-validation and overoptimism-correction. For proteomics, proteins relative intensity was calculated, and a bootstrapped differential protein expression analysis identified proteins associated with weight gain that were then included in logistic regression. Models performance was assessed in terms of Area Under the Curve (AUC). Results: In the overall cohort (N=9541) mean age was 56.8 (SD 11.4), mean BMI was 25.9 Kg/m2 (SD 5.4), 48.9% of pts were overweight or obese, and 52.9% received chemotherapy (CT). Overall, 16.9% (T1), 23.4% (T2), and 27.2% (T3) BCS gained weight (absolute mean change (95% CI): 6.1 kg (5.9-6.2), 6.7 kg (6.5-6.9) and 7.2 kg (6.9-7.3) at T1, T2, T3, respectively). In clinico-behavioral models, younger age, current smoking, lower income and education, receipt of CT and radiotherapy were associated with increased risk of weight gain (Table). Among 1261 BCS with biomarkers data, higher levels of IL1α (OR for 1-unit log increase [95%CI] 0.11 [0.02 - 0.65]) and of ADPN (1.36 [1.01 - 1.85]) were associated with lower and higher risk of weight gain at T2 and T3, respectively. Performance of models integrating these biomarkers was similar to clinico-behavioral models. Among 462 BCS with proteomic profiling, preliminary data showed that higher relative abundance of IgG Fc Binding Protein (OR 0.44, p<.05) and Tubulin-1 (OR 0.73, p<.05) was associated with lower risk of weight gain at T1. AUC of model integrating clinical and proteomics data was 0.74 (0.58-0.90). Conclusions: Over one-in-four BCS in the CANTO cohort experienced meaningful weight gain 4 years after dx. This large, multidimensional study confirms the role of clinico-behavioral risk factors for weight gain. However, the predictive ability of clinico-behavioral models seems suboptimal. The exploitation of additional data dimensions, including serum proteins and proteomic data, may help improve predictive ability and inform underlying biological processes implicated in weight gain after BC. Further studies will aim at improving model stability, particularly for proteomics analyses. Table. Models of weight gain in the overall cohort.T1 (N= 8397)T2 (N= 7663)T3 (N= 5802)Clinical predictors OR* (95% CI)OR* (95% CI)OR* (95% CI)Age, 1-year increase0.96 (0.94 - 0.97)0.96 (0.95 - 0.97)0.96 (0.95 - 0.97)BMI, 1-unit increaseNRNS0.97 (0.94 - 0.99)Education, primary vs collegeNS1.57 (1.04 - 2.39)NREducation, high school vs college1.38 (1.04 - 1.83)1.54 (1.21 - 1.98)NRIncome, ≥ 1500 and <3000 vs >3000NRNR1.29 (1.00 - 1.66)Smoke, current vs never1.70 (1.24 - 2.33)NR1.53 (1.12 - 2.08)Chemotherapy, yes vs no1.40 (1.07 - 1.82)1.31 (1.01 - 1.69)NRRadiotherapy, yes vs no2.10 (1.10 - 3.99)NR1.83 (1.08 - 3.12)AUC (95% CI) - clinical models0.65 (0.63 - 0.68)0.64 (0.61 - 0.67)0.65 (0.63 - 0.68)AUC (95% CI) - clinical + inflammatory and metabolic biomarkers models, [N]0.65 (0.60 - 0.70), [1179]0.66 (0.62 - 0.70), [948]0.67 (0.63 - 0.71), [1017]AUC (95% CI) - clinical and proteomics models, [N]0.74 (0.58 - 0.90), [462]0.65 (0.50 - 0.81), [462]NEOR= Odds Ratio, CI= Confidence Interval, NR= Not Retained, NS= Not significant, NE= Not evaluated *Adjusted by age, menopause, smoke, socioeconomic, psychological, tumor and treatments **Significant covariates from previous models were forced and ABE selected significant variables among all circulating biomarkers. Citation Format: Davide Soldato, Antonio Di Meglio, Caroline Pradon, Antonin Della Noce, Daniele Presti, Julie Havas, Florine Dubuisson, Barbara Pistilli, Valerie Camara-Clayette, Fabrice André, Alexandra Jacquet, Sibille Everhard, Sandrine Boyault, Paul-Henry Cournede, Stefan Michiels, Ines Vaz-Luis, Stergios Christodoulidis. An integrated clinical, behavioral and biological model to predict the risk of weight gain among breast cancer survivors (BCS) [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P4-11-34.