We thank our colleague, Dr. Boers, for his careful discussion of critical methodologic aspects of the economic evaluation of the BeSt study. We would like to respond to the concerns that he raises.
First, we agree that the QALY differences in our study were small, although statistically significant, and that tight disease control achieved similar long-term improvements in patients in all 4 adaptive treatment strategies. We think that other studies with fixed medication schedules or with optimistic mathematical models have tended to overestimate the impact of biologic therapies. Nevertheless, even the relatively small QALY gain that was obtained in our study by starting directly with infliximab would already justify the difference in medical costs if productivity is valued according to the human capital method, an established approach rather than an extreme assumption to calculate the economic consequences of illness.
Second, Boers suggests that the only reason for our observed (significant) differences in productivity costs must be the (nonsignificant) difference in initial labor participation, and he suggests that our statistical analyses should have corrected for these differences. Boers disregards the fact that cost-of-illness studies have invariably shown that rheumatoid arthritis (RA) has a large impact on productivity, and that therefore the observed differences in early effectiveness in our study are very likely to lead to differences in productivity. It is true that initial labor participation in groups 1 and 2 was relatively low (40% and 33%, respectively). This may have worsened the outcome in groups 1 and 2, which is why we explicitly reported and discussed the difference in initial labor participation. However, initial labor participation was very similar in groups 3 and 4 (both 46%). All cost-utility ratios quoted by Boers compare groups 3 and 4, so these cannot have been influenced by the initial level of labor participation. Moreover, with hindsight, random baseline differences should not be used to highlight or to remove statistical differences.
Third, Boers raises a concern about the number of deaths. Mortality in the 4 different study groups was not statistically significantly different (0, 1, 0, and 2 patients, respectively). This mortality was incorporated by using zero utility values, but QALY losses were not extrapolated beyond the 2-year study period. Patterns of infliximab use had not yet stabilized after 2 years, so extrapolating study results to a longer horizon could be inaccurate. Moreover, a proper long-term model would also have to incorporate reduced RA mortality due to effective treatment. Compared with many other economic evaluations, an important strength of our study is the use of primary data only, without modeling. Likewise, we do not speculate on strategies that were not included in the trial.
Camels and horses are both very useful animals. The economic and statistical analyses of the BeSt study were performed according to current state-of-the-art and most objections raised by Boers were already presented and discussed in our paper. His response confirms our conclusion that the crucial factor in the economic evaluation of biologic therapies is whether and how productivity is valued. The human capital method counts every hour lost as costs, whereas the friction cost method only takes into account the losses until an absent employee has been replaced. If we give more weight to productivity by using the human capital method, then productivity costs can compensate for medication costs in early-onset RA. If we give less weight to productivity by excluding productivity costs or by using the friction cost method, then currently, biologic therapies are probably too expensive.