DISCUSSION

We reviewed the EEs done in Italy by Italian authors, following a common scheme to allow some comparisons and, thus, assess the “state of the art” of these studies. Like in other reviews, we have probably missed some significant studies. However, this probability was minimized by using systematic bibliographic search techniques. The extensive criteria used to avoid missing studies was the main reason for the large number of articles rejected after applying the inclusion criteria.

To assess the results of the review according to the attitude of the Italian pharmacoeconomic community, we used the most recent informal guidelines published in Italy (1), which stem from a workgroup of experts and, therefore, should be the result of a broad consensus. The results show a good agreement with regard to the type of costs to be investigated: nearly a third of the studies included indirect costs in addition to direct ones, whereas intangible costs (which the guidelines strongly recommend excluding) are almost ignored. However, despite the recommendations to use both the NHS and societal perspectives, always to do sensitivity analysis, and to rule out partial EEs, only four studies used the two viewpoints together, more than one-third of the full EEs did not assess the robustness of results with sensitivity analysis, and many partial EEs have been undertaken in Italy – according to the results of the statistical analysis economists even seem to prefer them.

Moving to information sources, clinical trials – as proposed by the guidelines – were used for efficacy in more than half of the full EEs, whereas only one-third estimated hospital unit costs through microcosting studies – the source recommended by the Italian expert panel because it should reflect real costs better than tariffs and prices.

Comparing the present review to international ones (16;18), many findings are not directly comparable because these studies do not analyze the EE methods in detail. However, the comparable results are broadly similar. The technique most often used for EEs is CEA (85% in Nixon et al.; 70% in our review), the source of efficacy was most frequently the clinical trial (66% in Nixon et al.; 51% in our review), and pharmacoeconomic studies accounted for more than one-third of all EEs analyzed (36% in Pritchard; 38% in ours). These similarities are further confirmed by comparing our results to those of a recent Spanish methodological review (11). In Spain too, most EEs only considered direct costs (82% as in our study) and the technique most used was CEA (62%).

The methodological problems of EEs in Italy are probably similar to those in other countries (7). This problems may be explained by the difficulty of applying standard methods to all cases and gives further backing to the idea that experts should not set down rigid recommendations, if they cannot be followed by the health economists themselves. Many alternatives are available, but each has its own strengths and weaknesses. We share the opinion of scientists who maintain that standardization or regulation of methods may be counterproductive if guidelines attempt to set up too rigid an approach to evaluation (21). The adoption of a method as a standard should not mean that other methods can be ignored, nor should it constrict the scope of the analysis (20). On the other hand, methodological choices considerably affect the final results of EEs and their heterogeneity makes any kind of comparison difficult (12;19).

The most widely debated concern about potential biases relates to the influence of sponsors on the EE results (14;15;19). More than half of the full EEs focused on pharmacological therapy. Accordingly, we investigated the source of funding for pharmacoeconomic studies only. We found that twenty-five of the thirty-eight pharmacoeconomic full EEs were sponsored by companies (66%). The sponsor was clearly acknowledged in fourteen studies, two studies were co-authored by company employees, and we checked the sponsorship of the remaining nine studies directly with the authors. In sixteen of the twenty-five sponsored studies (64%) the sponsor's products were the dominant alternative (i.e., more effective and less costly); in five studies the sponsor's drug was recommended in specific conditions (e.g., selective use according to age or severity of patients), whereas in the four remaining studies, the authors outlined that the sponsor's drug was more effective but more expensive. It is worth mentioning that one sponsored study even demonstrated the dominance of a statin recently withdrawn for safety reasons all over the world.

These results in Italy confirm the legitimacy of the debate in the literature on possible bias in the results of published EEs sponsored by the pharmaceutical industry (15). To cope with it, Drummond suggested clarifying the contractual relationships between sponsors and researchers (7). More emphatically, a group of scientists claimed that researchers should sign contracts guaranteeing their right to control the design of the study, their access to relevant data sets, and the freedom to publish negative results (23). However, this seems difficult to achieve in practice and even more difficult to detect. Because the risk of bias cannot be easily ruled out, encouraging scientific journals to push for transparency in methodology seems more realistic (14).

In conclusion, the review showed that, in Italy, like elsewhere, there is a gap between theory and practice in EEs and sponsors can considerably affect the results of EEs. An awareness of these limits should lead the Italian health economics community to a more lively and constructive dialogue.