JMIR Publications

ABSTRACT

Background:

Widespread antiretroviral therapy (ART) has converted HIV into a chronic, manageable condition and markedly reduced AIDS-related mortality. Consequently, non-AIDS-defining cancers (NADCs) have emerged as a leading cause of death among people living with HIV (PLWH). Conventional Cox models treat competing events (non-cancer deaths) as censored observations and therefore overestimate absolute cancer mortality risk. Fine-Gray competing-risk methods are required for unbiased estimation of the cumulative incidence of cancer-related death.

Objective:

This study aimed to identify independent predictors of cancer-related mortality among PLWH using Fine-Gray subdistribution hazard regression and to develop and internally validate a prognostic nomogram that estimates individualized 3-, 5-, and 10-year absolute risks.

Methods:

We performed a retrospective cohort study of all PLWH diagnosed and reported in Ningbo, China, between 2006 and 2025 (n=7595). Cancer-related death was the event of interest and non-cancer death the competing event. Participants were randomly split 7:3 into training (n=5317) and validation (n=2278) cohorts. Multivariable Fine-Gray models yielded subdistribution hazard ratios (sHRs). A nomogram incorporating the selected predictors was constructed. Discrimination was assessed by the concordance index (C-index) and time-dependent area under the receiver-operating-characteristic curve (AUC); calibration was evaluated graphically. Risk-stratification performance was examined with cumulative incidence functions and Gray’s test.

Results:

During a median follow-up of 77 months (IQR 37–118), 262 cancer deaths occurred (193 NADCs, 73.7 %). Eight variables remained independently associated with cancer-related mortality: later calendar period of diagnosis (2021–2025 vs 2006–2010: sHR 7.8, 95 % CI 3.6–16.7), male sex (sHR 1.7, 95 % CI 1.0–2.9), older age (≥60 vs <20 years: sHR 6.4, 95 % CI 1.5–27.3), unmarried status (sHR 1.9, 95 % CI 1.0–3.7), history of sexually transmitted infection (sHR 5.4, 95 % CI 3.8–7.7), delayed or no ART initiation (sHR 3.2, 95 % CI 2.2–4.7), lower baseline CD4 count (CD4 ≥500 vs <200 cells/µL: sHR 0.3, 95 % CI 0.2–0.4), and hepatitis C virus coinfection (sHR 50.4, 95 % CI 34.5–73.6). The nomogram showed excellent discrimination (C-index 0.927 training, 0.898 validation) and time-dependent AUCs of 0.892–0.958. Calibration plots demonstrated close agreement between predicted and observed risks. The model stratified patients into high- and low-risk groups with markedly different 10-year cumulative incidences (approximately 10.6–11.0 % vs 0.3–0.5 %; Gray P<.001).

Conclusions:

Cancer-related mortality, predominantly from NADCs, constitutes a substantial long-term burden for PLWH. The Fine-Gray nomogram provides accurate, individualized absolute-risk estimates that can support risk-adapted cancer surveillance and timely clinical intervention.