Accuracy of combined molecular and morphology-based triage for HPV-positive women in routine cervical cancer screening services from Colombia

WHO cervical cancer elimination goals comprise 70% of highly-sensitive screening coverage and 90% treatment of precancerous lesions. Triage for HPV-positive women may challenge sensitivity of screening algorithms and women's follow-up, particularly in low- and middle-income countries (LMIC) where screening quality and protocol adherence are frequently deficient. We aimed to determine the accuracy of triage for HPV positive women in routine screening services from Colombia by a prospective cross-sectional study. Consecutively, HPV DNA-positive women underwent six triage strategies (conventional cytology, two methods of visual inspection, HPV16/18/45-genotyping, telomerase, and HPV mRNA). Positive triage results underwent regular colposcopy/biopsy in public hospitals. Adjusted sensitivity, specificity, and predictive values for CIN2+/CIN3+ were estimated for stand-alone and combined tests. We explored the impact of triage strategies on referral rates and the complete screening algorithm (screening plus triage). Overall 16,242 women underwent HPV screening and 1789 (11.0%) were HPV-positive. In total, 20.1% of women were lost to follow-up. mRNA showed the highest positivity rate (0.64 among HPV-positive and 0.05 among the total screened cohort), the highest sensitivity (0.94 95%CI 0.75-0.96), and the lowest specificity (0.36 95%CI 0.29-0.43). Parallel testing with HPV-mRNA revealed the highest increase in sensitivity for all triage strategies. Accuracy of cytology and visual inspection differ between screening units but parallel testing with HPV16/18/45 genotyping significantly increased their sensitivity (over 0.80). Morphology-based triage for HPV-positive women remains a suitable alternative for routine practice in LMIC if combined with HPV16/18/45-genotyping; however, point-of-care triage would be preferable to reduce losses to follow-up. HPV-mRNA triage deserves cost-benefit analyses.