Diagnostic accuracy of human papillomavirus testing in primary cervical screening: A systematic review and meta-analysis of non-randomized studies

Objective. This is a meta-analysis of studies comparing HPV testing to cytology with regard to their accuracy in the detection of underlying high grade cervical intraepithelial neoplasia in primary cervical cancer screening. Methods. A systematic review was conducted following the Cochrane Collaboration Guidelines. A systematic search was performed in 8 electronic databases. Strict selection criteria were applied in terms of types of participants, types of interventions and methods to limit verification bias. Where possible we calculated the sensitivity, specificity, positive and negative predictive value of cytology and the HPV test, as well as sensitivity and specificity ratios for the detection of CIN2 or worse. Random effect models were used for pooling accuracy parameters. The results were displayed using forest plots. Results. We identified 25 studies fulfilling the inclusion criteria. The pooled sensitivity of HC2, PCR, cytology (ASCUS or worse) and cytology (LSIL or worse) was 90%, 80.9%, 72.7% and 61.6%, respectively, and the pooled specificity was 86.5%, 94.7%, 91.9% and 96.0%, respectively. The ratio of the sensitivity of HC2 to cytology (ASCUS) was 1.25 (95% CI = 1.20-1.29), and the corresponding specificity ratio was 0.97 (95% CI = 0.96-0.98). The ratio of the sensitivity of combination of HC2 and cytology (ASCUS) to HC2 alone was 1.05 (95% CI = 1.04-1.06) and the ratio of the specificity 0.95 (95% CI=0.94-0.96). For women over 30 years, the sensitivity of HC2 was 94.8% and the specificity 86.0%. Conclusions. Compared to cytology, the HC2 and PCR are substantially more sensitive for prevalent CIN2 or worse but significantly less specific. The combination of HC2 and cytology has the highest sensitivity and lowest specificity. However, reduction of the incidence of or mortality from invasive cervical cancer among HPV screened subjects compared to cytologically screened subjects has not yet been demonstrated. (c) 2006 Elsevier Inc. All rights reserved.