TRADE-OFFS IN CERVICAL CANCER SCREENING – BALANCING DETECTED CANCER PRECURSORS AND RESOURCE USE
Candidate for the Lee B. Lusted Student Prize Competition
Method: We developed a probabilistic decision tree model to estimate age-stratified (ages 25-33 and 34-69) outcomes for different screening algorithms. The model uses epidemiologic data from the Cancer Registry of Norway and follows a cohort of women attending primary screening through one screening round (i.e., 3 years), allowing for loss-to-follow-up and spontaneous regression of CIN2+. We compared the current cytology-based strategy (involving co-testing (HPV and cytology) for delayed triage of low-grade cytology results), with nine alternative strategies that allow reflex HPV-testing from the primary screening sample for women with low-grade and/or inadequate cytology results. Primary outcomes included CIN2+ detected, total costs, total number of tests (cytology, HPV) and consultations, and number of colposcopies/biopsies performed. We calculated the incremental cost-effectiveness (ICER) and harm-benefit ratios (IHBR) for each strategy compared with the next most costly/harmful strategy, defined as the additional cost and colposcopies required, per additional CIN2+ detected, respectively.
Result: Among 100,000 women, we projected that the current Norwegian strategy would detect 1,325 and 473 CIN2+ for ages 25-33 and 34-69, respectively. For the alternative strategies, these outcomes ranged from 1,365 to 1,725 and 462 to 525, respectively. The current strategy was one of the least efficient strategies (i.e., strongly dominated) with respect to both ICER and IHBR, for both age groups. For the alternative strategies and ages 25-33 and 34-69, respectively, the ICERs ranged from NOK7,729 to 68,059 and NOK16,544 to 180,483, while the IHBRs ranged from 3.75-12.22 and 0.67-28.93 colposcopies, per additional CIN2+ detected.
Conclusion: Reflex HPV-testing allows for improvements in both effectiveness and cost-effectiveness of the current screening algorithm. Candidate strategies can detect a larger number of precancers while using fewer resources compared to current practice; however, the optimal strategy depends on society’s willingness to pay costs and accept harms. Ultimately, the down-stream effectiveness of the alternative algorithms will depend on the extent to which precancers regress or progress into cancer.