Purpose: Current policy regarding hepatitis A (HA) vaccination in high-endemicity regions and among high-risk groups is influenced by published cost-effectiveness analyses (CEA); however results of these CEA vary widely across studies. A systematic review was conducted to explore the effects of methodological quality on such variation. Key aspects of vaccination, such as adequate representation of vaccine-induced herd-immunity and accurate estimation of the degree of under-reporting (approximately 1:10) were assessed.

Methods: Cost-effectiveness studies of HA vaccine were identified (MEDLINE, EMBASE, HSTAR, and SSCI; MeSH “cost-effectiveness” AND “hepatitis A”), and included if they were cost-effectiveness/utility studies of HA vaccine. Citations and full-text articles were reviewed independently by two reviewers. Back referencing, author searches, and expert consultation ensured literature saturation.

Quality of reporting was assessed using a 21-item quality tool (Neumann 2000). Methodological issues specific to vaccine evaluations were appraised by Beutels' 2002 guidelines. Key modeling issues were examined using Sculpher's 2000 framework.

Results: Thirty-four cost-effectiveness studies were included from full-text-article review (n=95) and citation-screening (n=700). Nine conducted an additional cost-utility analysis and six a cost-benefit analysis. All were model-based studies, thirteen utilizing Markov models. One used a dynamic model, capturing effects of herd immunity. Strategies included mass childhood/adult (n=11/n=2) and selective childhood/adult vaccination (n=24/n=8). Populations assessed were healthy participants (n=31), and patients with chronic liver disease (n=3).

The median quality using Neumann's tool was 62% (13/21 range [14-90%]), similar to the median score of 58% reported for CEA studies in other clinical areas. Cost-utility studies attained higher quality (n=9, median 76% [57-90%]) than cost-effectiveness (n=25, median 52%[14-71%]).

Cost-utility studies assessed using Beutels' guidelines clearly stated model assumptions (7/9), utilized proper time span (7/9), and captured relevant costs (7/9). Many did not discuss alternative modeling approaches (7/9).

Using Sculpher's framework, cost-utility studies specified model assumptions (9/9), and most justified model parameters (7/9). Some did not discuss implications of relaxing model assumptions (3/9). None adjusted for herd immunity, though four acknowledged this as a study limitation. One study adjusted for under-reporting, though five studies discussed it.

Conclusions: Methods used to evaluate the cost-effectiveness of HA vaccination were inconsistent, resulting in variable cost-effectiveness ratios across studies. Failing to adjust for herd-immunity and underreporting weakens scientific evidence and may result in biased estimates of the economic attractiveness of HA vaccination policies.