Allostatic load connects tumor genomics, disease trajectory, and pre-diagnosis wearable activity

Background: Allostatic load (AL), a composite index of physiological stress from routine labs and vitals, reflects multisystem strain and predicts survival in cancer. Building on prior validation of cancer-modulated AL (cmAL), we examined how cmAL varies across disease states, relates to tumor genomics, and aligns with wearable-measured activity before diagnosis. Methods: We analyzed 12,689 adults treated at MSK with pre-treatment labs and vitals across NSCLC, colorectal, prostate, ovarian, breast, pancreatic, endometrial, and bladder cancers. cmAL was computed from ten cardiovascular, metabolic, renal, and immune biomarkers and compared across disease phases using rank-sum tests with multiple-testing correction. Per-cancer-type genomic models used multivariable logistic regression relating cmAL to recurrent oncogenic alterations while adjusting for clinical covariates and comorbidity burden. Wearable activity was evaluated in 77 MSK patients (Apple HealthKit) and 1,867 All of Us participants (Fitbit) using median daily step counts across cmAL levels, Spearman correlation, and continuous trend tests. Pre-diagnosis steps were summarized using a 90-day window before first cancer diagnosis. Results: cmAL decreased during remission (median change -0.42, p<0.001) and increased with progression (median change +0.61, p<0.001), with the highest values observed within six months of death (p<1×10⁻⁴⁰). Higher cmAL correlated with cardiometabolic comorbidity (diabetes OR=0.74; renal OR=0.54). In NSCLC, higher cmAL was less common in patients with EGFR mutations and more common in those with KRAS mutations. Across both wearable datasets, higher cmAL was linked to lower pre-diagnosis activity. In the MSK cohort, median daily steps declined from 4,914 at (cmAL=1) to 1,092 at (cmAL=5) (Spearman ρ=-0.388, p=0.0005). In All of Us, median steps declined from 6,276 (cmAL=1) to 4,091 (cmAL=5) (ρ=-0.248, p<0.0001), demonstrating reproducibility in a population cohort. Conclusions: cmAL increases with disease progression and quantifies cumulative physiologic strain across body systems. Elevated cmAL is associated with specific genomic patterns in NSCLC, greater comorbidity burden, and lower pre-diagnosis step counts in both institutional and population cohorts. These findings support cmAL as a scalable and potentially modifiable biomarker integrating molecular, physiologic, and behavioral domains in real-world cancer survivorship.