The cost-utility analysis of the BeSt trial: Is a camel in fact a horse with abnormalities in the distribution of dorsal fat? Comment on the article by van den Hout et al


  • Maarten Boers MSc, MD, PhD

    1. VU University Medical Center, Amsterdam, The Netherlands
    Search for more papers by this author
    • Dr. Boers is the inventor of the COBRA regimen and a part-time employee at VU University Medical Center, which participated in the BeSt trial.

The Cost-Utility Analysis of the BeSt Trial: Is a Camel in Fact a Horse with Abnormalities in the Distribution of Dorsal Fat? Comment on the Article by van den Hout et al

To the Editors:

I read with great interest the results of the cost-utility analysis of the BeSt (Behandelstrategieën voor Reumatoide Artritis) trial by van den Hout et al and the accompanying editorial by Bansback and Marra published recently in Arthritis Care & Research (1, 2). It is an excellent state-of-the-art study brimming with interesting data. My only concern is with the interpretation.

A casual reader might assume from the abstract and conclusion, and also from the editorial, that the case on the cost utility of biologic therapies as first-line therapy is far from closed, since scientists do not agree on the proper methodology (e.g., for attributing productivity costs). Not understanding the finesses of health economics, the reader cannot be blamed for concluding that it may be good value for the money to start patients with early rheumatoid arthritis on infliximab and high-dose methotrexate as first-line treatment. After all, in one of the analyses, this strategy (as given to group 4 of the trial) was calculated to cost a mean of €22,000 per quality-adjusted life year (QALY) compared with the costs and utility of an oral pulse of prednisolone plus methotrexate and sulfasalazine (the COBRA [Combinatietherapie Bij Reumatoïde Artritis] strategy as given in group 3 of the trial).

So is that the “BeSt” conclusion? That remains to be seen.

To start, the differences in QALYs between the groups are quite small, and most prominent in the first year of study. For example, for the British EuroQol Index (with a scale where 0 = dead and 1 = perfect health) the QALY at 1 year was 0.58 in the worst group versus 0.68 in the best group. In the second year the range was 0.68–0.73 and the differences were no longer statistically significant.

The simplest cost-utility analysis ignores societal costs and focuses on direct health care costs, i.e., drug and other medical expenditures. In this analysis, the incremental upfront cost per QALY of infliximab was €190,000. The authors say, with some understatement, that these are “costs that are generally considered too high.”

A more complex analysis is one that uses the societal perspective. Bansback and Marra note in their editorial that this perspective may be the most appropriate to choose since it includes costs of productivity losses that may be recouped by society due to effective therapy. However, it involves tough choices on how paid and unpaid labor are valued. In this study, the choice between the so-called friction cost method and the human capital method led to huge differences in the conclusions on the cost-utility, i.e., €22,000 or €130,000 per QALY, with even larger confidence intervals, which raised questions about the analysis itself. In other studies, such differences could be blamed on an unreasonably long horizon for extrapolation that favors the human capital method, but here the horizon was properly limited to the study duration of 2 years. Given that both the differences in measured utilities as well as productivity between the groups (especially groups 3 and 4) were relatively small, the only reason for these results that I can think of is in the baseline differences in productivity. At baseline, contract hours ranged between 14.1 in groups 1–3 to 16.7 in group 4. Differences were not significant due to highly skewed distributions. Contract hours decreased in all groups except group 3. Hours absent from work decreased in all groups, but if “loss” is related to contract hours then the baseline differences are important. Because worked hours are multiplied with wages to obtain the productivity “benefit,” small imbalances can then strongly influence the results. I find no mention that a correction for these imbalances was made in the analyses.

Another concern is that the editorialists assume there were no deaths in the first 2 years, but in my recollection there were. If so, how were these accounted for? Finally, 9% missing data in utility scores and 16% in cost diaries increase the uncertainty of the results, despite the use of state-of-the-art imputation techniques.

Looking beyond the study, probably the most cost-effective strategy is to start with the COBRA regimen of group 3, but switch to infliximab on nonresponse at 6 months. Thinking of the camel, and staying within the bounds and the uncertainties in the study outlined above, I suggest a more proper conclusion would read, “Initial combination therapy with methotrexate, sulfasalazine, and an oral pulse of prednisolone (COBRA regimen) appears to be the dominant cost-effective strategy as initial therapy in early active rheumatoid arthritis. Under extreme assumptions, starting combination therapy with infliximab might become cost effective at €22,000/QALY compared with the COBRA regimen.”

Maarten Boers MSc, MD, PhD*, * VU University Medical Center, Amsterdam, The Netherlands.