Abstract / Summary
Background Cancer incidence and mortality are rising in Africa, yet the distribution and determinants of cancer in high HIV-prevalence regions remain limited. We characterized cancer epidemiology, treatment, and survival among adults in western Kenya, where HIV prevalence ranges between 10.3% and 11.7%. Methods This retrospective cohort study included adults (18 years and older) with cancer treated at Jaramogi Oginga Odinga Teaching and Referral Hospital (JOOTRH) in Kisumu, Kenya, between Jan 1, 2014, and Dec 31, 2024, followed through June 30, 2025. Data were abstracted from paper records into REDCap. Negative binomial models estimated annual trends, Kaplan-Meier curves evaluated survival, and Cox proportional hazards models evaluated mortality. Findings Among 3978 patients (64% female; median age 53.9 years, IQR 42.0-65.1), annual case counts rose from 29 to 708, averaging a 38% annual increase (95% CI 29-48). Cervical (24%), esophageal (13.6%), breast (12%), prostate (8%), and colorectal (4%) cancers accounted for 62% of cases. 1305 (33%) of cancer patients were living with HIV and 802 (20%) had unknown HIV status. Among patients whose cancer was assigned a stage, 72% presented at stage III-IV. 1973 (50%) received cancer-directed treatment, and 51% had an unscheduled clinic visit lapse of at least 180 days. Cancer-directed treatment was the strongest protective factor for mortality (adjusted hazard ratio [aHR] 0.27, 95% CI 0.22-0.32); HIV-positive status (aHR 1.43, 1.19-1.71), HIV-unknown status (aHR 1.48, 1.20-1.81), and poor or missing performance status were independently associated with higher mortality, robust across sensitivity analyses. Interpretation Cancer mortality was strongly associated with treatment and HIV status at diagnosis. Closing the treatment-access gap and adopting routine opt-out HIV testing at the first oncology visit are essential for integrated cancer care in high-HIV-prevalence settings. Funding Research reported in this publication was supported by the American Cancer Society; the Washington University in St. Louis Institute of Clinical and Translational Sciences through a grant from the National Center for Advancing Translational Sciences (NCATs) of the National Institutes of Health; Just-in-Time funding, and the National Cancer Institute Center Support Grant P30CA091842.