Method: We developed a probabilistic decision tree simulation model to estimate outcomes associated with different screening algorithms. The model uses epidemiologic data from the Cancer Registry of Norway and follows individual women attending primary screening at baseline through subsequent six years, allowing for loss-to-follow-up and spontaneous regression of CIN2+. We compared the current Norwegian strategy (Strategy 1) entailing primary cytology with co-testing (HPV and cytology) for delayed triage of atypical and low-grade cytology results (ASC-US/LSIL), with two alternative HPV strategies. Both HPV strategies involve primary HPV testing followed by reflex cytology for any HPV-positive result. For Strategy 2, the threshold for diagnostic colposcopy/biopsy is high-grade atypical cells or worse (ASC-H+), while Strategy 3 involves a lower threshold on the cytology result, i.e., any atypical cells (ASC-US+). Primary outcomes were CIN2+ detected and number of colposcopies/biopsies performed.
Result: Among 100,000 women, we project the three strategies would detect 503, 338 and 715 CIN2+, and require respectively 2685, 1925 and 3696 colposcopies. Consequently, the number of colposcopies per CIN2+ detected were 5.34, 5.70 and 5.17 for Strategy 1, 2 and 3, respectively. Strategy 2 results in a 32% reduction in detected CIN2+ while simultaneously increasing the number of colposcopies per CIN2+ by 7% compared to the current strategy, thus providing a less efficient algorithm. Conversely, Strategy 3 may increase detection of CIN2+ by 42% at a cost of 38% more colposcopies, and detect CIN2+ more efficiently than the current strategy (i.e., decreasing the colposcopy to CIN2+ ratio by 3%).
Conclusion: There is a potential for improving the current screening algorithm by implementing screening with primary HPV-testing. However, unless the lower threshold for referral to colposcopy/biopsy is utilized, primary HPV-testing may detect fewer cancer precursors and require more colposcopies per detected case. The differential effectiveness of the three algorithms in terms of preventing invasive cancer will depend on the extent to which cancer precursors regress or progress into cancer.