Performance of screening tools for cervical neoplasia among women in low- and middle-income countries: A systematic review and meta-analysis

Objective: To evaluate the performance of visual inspection with acetic acid (VIA) testing, visual inspection with Lugol’s iodine (VILI), primary HPV testing, and conventional Pap smear in detecting CIN2+ among non-pregnant women aged 30–65 in LMICs between 1990 and 2020. Design: Systematic review and meta-analysis. Setting and participants: Low- and middle-income countries, non-pregnant women aged 30–65. Methods: CENTRAL (Cochrane Library), CINAHL, Embase, Global Health, PubMed, and Web of Science databases were systematically searched to identify studies evaluating the performance of cervical cancer screening methods in LMICs. A diagnostic test accuracy meta-analysis was conducted to evaluate the performance of 4 screening methods in detecting CIN2+ relative to biopsy or cytology reference standards. Pooled statistics for sensitivity, specificity, diagnostic odds ratios, and summary receiver operating characteristic curves were determined for each method. Subgroup analyses were performed to examine whether there was variation in performance based on different reference standards for defining CIN2+, specifically: colposcopy-directed biopsy, biopsy alone, colposcopy alone, or liquid-based cytology. Results: Eighteen studies were identified through systematic review. Twelve studies were included in meta-analysis; 11 were cross-sectional and 1 was a randomized controlled clinical trial. The remaining six of the eighteen studies were inclided in a narrative syntehsis. Pooled estimates for sensitivity for VIA, VILI, primary HPV testing, and conventional Pap smear were 72.3%, 64.5%, 79.5%, and 60.2%, respectively; pooled estimates for specificity were 74.5%, 68.5%, 72.6%, and 97.4%, respectively; the diagnostic odds ratios were 7.31, 3.73, 10.42, 69.48, respectively; and the area under the summary receiver operating characteristic curves were 0.766, 0.647, 0.959, and 0.818, respectively. Performance of the screening method varied based on the reference standard used; pooled estimates using either colposcopy-directed biopsy or biopsy alone as the reference standard generally reported lower estimates; pooled estimates using either colposcopy alone or liquid-based cytology as references reported higher estimates. Conclusions and implications: This meta-analysis found primary HPV testing to be the highest performing cervical cancer screening method in accurately identifying or excluding CIN2+. Further evaluation of performance at different CIN thresholds is warranted.