Adaptation of a modified Elixhauser comorbidity index to isolate the impact of body mass index on survival outcomes in patients with cancer in a large integrated healthcare system.

124 Background: Variable impacts of body mass index (BMI) on overall survival (OS) and cancer-specific survival (CSS) have been described in patients with cancer, with existing data highlighting both detrimental effects and malignancy-specific benefits. Real-world data derived from integrated healthcare systems are underexplored in this context, and tight control of confounding using validated tools such as the Elixhauser Comorbidity Index (ECI) is rarely employed in pre-existing studies. Herein, we offer a highly refined description of the impact of BMI on cancer-specific outcomes in a large patient cohort using a targeted ECI model. Methods: 153,270 patient records were abstracted from an integrated healthcare system from 2010-2018, including descriptors of age, race/ethnicity, income, disease stage, and treatment regimens. Standardized derivation of BMI within one year prior to diagnosis was performed and categorized by World Health Organization definitions. A custom ECI was designed to exclude ICD codes for obesity to effectively isolate the impact of BMI on downstream analyses. Kaplan-Meier survival estimates of OS and CSS were constructed, and univariate and multivariate Cox proportional hazard (PH) modeling was performed to determine hazard ratios (HR). Results: When normalized to healthy BMI in multivariate modeling, we found a favorable dose-dependent relationship of overweight and obese BMI on OS and CSS independent of age, race/ethnicity, income, modified ECI score, cancer stage, or therapeutic regimen (Table). This advantage to OS and CSS was not as strongly reflected in severely obese patients, though still significant compared to healthy BMI counterparts. Being underweight was a negative prognostic indicator for OS and CSS in all patients compared to a healthy BMI. Conclusions: We demonstrate a beneficial impact of increased BMI on OS and CSS in a large cohort of patients treated in an integrated healthcare model independent of tightly controlled confounding. Specifically, we highlight the utility of isolating obesity using a targeted ECI model for comorbidity adjustment, which may provide a more robust measure of the influence of BMI on survival. Our findings raise questions regarding attention to adequate control of confounding in prior studies, as well as the potential benefit of removing variation in healthcare delivery in understanding these outcomes. Adjusted HR of OS and CSS stratified by BMI category (*95% Confidence Interval). BMI Category Adjusted OS Adjusted CSS Normal Weight(18.5 – 24.9) 1 (reference) 1 (reference) Underweight(< 18.5) 1.35 (1.25 – 1.45*) 1.28 (1.15 – 1.41) Overweight(25 – 29.9) 0.83 (0.81 – 0.85) 0.84 (0.81 – 0.87) Obese(30 – 39.9) 0.81 (0.79 – 0.83) 0.81 (0.78 – 0.84) Severe Obesity(≥ 40) 0.86 (0.82 – 0.90) 0.82 (0.77 – 0.87) Unknown 1.33 (1.24 – 1.43) 1.19 (1.07 – 1.31)