Systematic review of model-based cervical screening evaluations

Abstract

Background: Optimising population-based cervical screening policies is becoming more complex due to the expanding range of screening technologies available and the interplay with vaccine-induced changes in epidemiology. Mathematical models are increasingly being applied to assess the impact of cervical cancer screening strategies.

Methods: We systematically reviewed MEDLINE®, Embase, Web of Science®, EconLit, Health Economic Evaluation Database, and The Cochrane Library databases in order to identify the mathematical models of human papillomavirus (HPV) infection and cervical cancer progression used to assess the effectiveness and/or cost-effectiveness of cervical cancer screening strategies. Key model features and conclusions relevant to decision-making were extracted.

Results: We found 153 articles meeting our eligibility criteria published up to May 2013. Most studies (72/153) evaluated the introduction of a new screening technology, with particular focus on the comparison of HPV DNA testing and cytology (n = 58). Twenty-eight in forty of these analyses supported HPV DNA primary screening implementation. A few studies analysed more recent technologies - rapid HPV DNA testing (n = 3), HPV DNA self-sampling (n = 4), and genotyping (n = 1) - and were also supportive of their introduction. However, no study was found on emerging molecular markers and their potential utility in future screening programmes. Most evaluations (113/153) were based on models simulating aggregate groups of women at risk of cervical cancer over time without accounting for HPV infection transmission. Calibration to country-specific outcome data is becoming more common, but has not yet become standard practice.

Conclusions: Models of cervical screening are increasingly used, and allow extrapolation of trial data to project the population-level health and economic impact of different screening policy. However, post-vaccination analyses have rarely incorporated transmission dynamics. Model calibration to country-specific data is increasingly common in recent studies.

Figure 1

PRISMA Flow diagram of study selection process. *Articles published in journals not included in the British Library catalogue or Thompson Reuters Impact Factor (IF) list.

Figure 2

Characteristics of included studies. *Exclusively these technologies; AFR, African Region; Auto; automated cytology; HPV, HPV DNA testing; LMIC, low and middle income countries; VIA, VIA vs HPV DNA testing and cytology; WPR, Western Pacific Region.

Figure 3

Number of single-country studies per country.

Figure 4

Number of studies by analysis and prevention type over time. Dark blue, Economic Screening; Light blue, Economic Screening & Vaccination; Orange, Epidemiological Screening; Yellow, Epidemiological Screening & Vaccination.