Background
Virus-associated cancers cause significant mortality in Africa. Understanding risk factors for mortality is crucial for effective prevention and treatment. We analyzed demographic and clinical characteristics of people with virus-associated cancers in Kisumu County, western Kenya – a region with a high HIV prevalence of 14.5% – and evaluate factors associated with mortality.
Methods
We conducted a retrospective cohort study of adult patients who presented for cancer care at Jaramogi Oginga Odinga Teaching and Referral Hospital (JOOTRH) in western Kenya between January 2014 and December 2023. We included cases where viral infection is either causally associated with cancer (cervical, anal, Kaposi sarcoma) or is the strongest risk factor for cancer (penile, vaginal, vulvar). Data were extracted from paper-based patient records. We used Kaplan-Meier analysis and competing risk regression to assess survival and mortality risk factors, with loss to follow-up (LTFU, defined as >6 months late for the last appointment) and transfer-out considered as competing events.
Results
Among 3,437 patients, 1,073 (31%) met eligibility criteria. Median age at diagnosis was 47 years (inter-quartile range 39-59), with 933 (87%) female, and 553 (51.5%) people with HIV. Cervical cancer was the most prevalent (n=823, 76.7%), followed by Kaposi sarcoma (n=112, 10.4%), penile (n=57, 5.3%), vulva (n=48, 4.5%), anal (n=28, 2.6%), and vaginal (n=5, 0.5%). One year after enrolling in cancer care, 538 (62%) were LTFU. Median overall survival was 6% (95% CI: 5%-7%). In multivariable analyses, age <20 years and advanced stage at diagnosis (stage III and IV) were associated with increased risk of death. HIV status was not significantly associated with mortality (subhazard ratio 1.29, 95% CI: 0.83-2.01; p=0.3).
Conclusions
Virus-associated cancers are highly prevalent in western Kenya, with high LTFU and low survival. Older age and early-stage diagnosis (but not HIV status) were associated with lower mortality. Urgent strategies are needed to encourage early cancer detection and referral to treatment, and improve retention in care. Future research should consider testing for viral infections to attribute carcinogenicity for cancer types excluded in this analysis, such as Epstein-Barr virus (nasopharyngeal carcinoma; gastric cancer; Burkitt, Hodgkin, and diffuse large B-cell lymphoma), human papillomavirus (other head and neck cancers), and hepatitis B and C viruses (hepatocellular and intrahepatic bile duct cancers).
Table or Figure
