Abstract PR-18: Risk prediction model for high-grade squamous intraepithelial lesions in persons living with HIV in Puerto Rico

• Published in: Cancer epidemiology, biomarkers & prevention biomarkers & prevention, 2022-01, Vol.31 (1_Supplement), p.PR-18-PR-18
• Main Authors: Soto-Salgado, Marievelisse, Suárez, Erick, Deshmukh, Ashish, Viera, Tariz, Ramos-Cartagena, Jeslie, Pericchi, Luis R, Ortiz, Ana Patricia
• Format: Article
• Language: English
• ISSN: 1055-9965; 1538-7755
• DOI: 10.1158/1538-7755.DISP21-PR-18

Background: Persons living with HIV (PLWH) with concurrent HR-HPV infection are vulnerable to the development of anal high grade squamous intraepithelial lesions (HSIL), which greatly increase their risk of anal cancer. Despite anal cancer shares biological similarities with cervical cancer, there is no evidence-based screening recommendations. Clinical studies are underway evaluating clinical performance of anal cancer screening and optimal anal HSIL management approach. Due to the lack of evidence-based screening strategies for anal HSIL; we proposed to construct and validate a risk prediction model for anal HSIL among PLWH. Methods: We used data from the Anal Neoplasia Clinic (ANC) of the UPR Comprehensive Cancer Center from May, 2016 to April, 2021. From 688 HIV+ patients evaluated in the clinic during this period, 397 (57.7%) had information on biopsy-confirmed anal HSIL and thus were included in the analysis. The primary outcome was histologically confirmed anal HSIL. Potential predictors included demographic characteristics (e.g. sex at birth, age, and marital status), clinical characteristics (e.g. years living with HIV, last CD4 Nadir count, last HIV viral load, HPV vaccine, previous abnormal anal cytology, anal symptoms, HR-HPV result), and lifestyle risk factors (e.g. lifetime tobacco use, age at first sexual intercourse, lifetime number of sexual partners, history of anal sex, and history of STIs). For the prediction analysis, univariate logistic regression models (ULRM) were used to assess potential predictors of HSIL. Then, a multivariate logistic regression model (MLRM) was used to predict HSIL. The inclusion criteria for the MLRM were: (a) known or hypothesized risk factors for HSIL according to the scientific literature, (b) p-values < 0.25 in the ULRM, and (c) lower BIC and AIC values based on our data. After computing the probability of HSIL, we categorized these probabilities in deciles, to set the scale boundaries. Receiver-operating curves and the Area under the Curve (AUC) were constructed to summarize the performance of the scoring system. Results: Median age of patients was 45.8±12.6 and 72.0% were men. 47.6% had biopsy-confirmed anal HSIL and 52.4% had a benign lesion or LSIL; 37.5% had HR-HPV infection. Sex at birth, marital status, previous anal cytology, HIV diagnosis before HAART introduction, BMI, age at fist sexual intercourse, and HR-HPV were associated (p