Cancer
Beyond Body Surface Area: Microsimulation of Muscle-informed Chemotherapy Dosing to Improve Treatment Completion for Colon Cancer Anlan Cao* Anlan Cao Cao Cao Cao Cao Cao Kaiser Permanente Northern California
Chemotherapy is primarily dosed by body surface area (BSA) despite wide variability in drug distribution and clearance, contributing to frequent toxicities that lead to dose reductions/delays and compromise efficacy. Since low muscle mass is a predictor of toxicities, muscle-informed dosing may improve tolerability. We used microsimulation to evaluate whether muscle-informed dosing improves treatment completion.
We analyzed 94 patients with stage II-III colon cancer treated with FOLFOX at Kaiser Permanente. Body composition was quantified from CT scans. Chemotherapy delivery was obtained from the electronic health record (EHR), and toxicities were collected each cycle. We developed a cycle-by-cycle microsimulation to model: (1) delivered doses of 5-FU and oxaliplatin, (2) dose modifications, delays, and discontinuation, and (3) toxicities. Delivered doses influenced toxicities, and toxicities affected the subsequent dose changes. Predictors were selected using cross-validated grouped LASSO; discrimination and calibration were evaluated with cross-validation and recalibrated out-of-sample if needed. We simulated BSA- and muscle-informed dosing, respectively. The primary outcome was relative dose intensity (RDI; 0-100%, higher=better completion).
Sub-models showed moderate-to-good performance (AUC 0.68-0.92 with acceptable calibration). Under BSA-dosing, the mean RDI was 77.0% (IQR: 72.7-82.0), similar to observed RDI (78.7%, IQR: 72.0-92.0, P=0.34), indicating acceptable simulation performance. Under muscle-informed dosing (38.5 mg/cm2 for 5-FU and 1.2 mg/cm2 for oxaliplatin), projected RDI was 82.0% (IQR: 68.2-94.8), representing a 5% increase compared to the BSA-dosing (95% CI: 2.1-7.8, P<0.001).
Muscle-informed dosing may improve chemotherapy completion. Next, we will perform a target trial emulation in an EHR cohort (N~2,000) to estimate survival benefits. Our findings could transform personalized cancer treatment by informing future trials and dosing guidelines.