OBJECTIVES: To critically appraise published cost-effectiveness analyses (CEAs) of human papillomavirus (HPV) testing in cervical screening regarding the appropriateness of comparisons between strategies and the usefulness of the interpretation of cost-effectiveness estimates.  METHODS: The PubMed database was searched for relevant CEAs of cervical screening using HPV testing. The identified CEAs were carefully appraised for their quality of analyses, reporting and interpretation of results. Specific examples of modelling shortcomings were selected as illustrations of what to avoid when estimating the cost-effectiveness of HPV-based screening.  RESULTS: The review identified 29 relevant CEAs. Regarding basic errors, 11 of the 29 calculated the incremental cost-effectiveness ratios (ICERs) either partly or completely incorrectly. Ten studies failed to fully report costs and effects; either simply reporting ICERs or depicting a cost-effectiveness plane. Regarding more fundamental errors, 23 failed to include sufficient screening interval comparators against which to meaningfully estimate ICERs; effectively leading to average cost-effectiveness ratios being mistakenly identified as ICERs, which biases cost-effectiveness ratio estimates downwards. Finally, none of the studies gave specific consideration to the magnitude of the change in costs and effects of adding HPV testing to a given strategy, either with a simple graphical interpretation or with a formal interpretation using the net benefit framework.  CONCLUSIONS: Model specification is typically the most difficult part of a model-based CEA, whereas simulating relevant strategies is relatively straightforward once the model is built. Similarly, once results have been generated, their correct presentation and interpretation is relatively straightforward. However, this analysis shows that these relatively easy aspects of CEA are being performed poorly in the HPV screening literature. Consequently, a few simple improvements to basic aspects of CEAs of HPV-based screening could greatly enhance the usefulness of such analyses to decision makers.