Incremental or average cost-utility of routine cataract surgery? Erkki Soini, Department of Health Policy and Management, Department of Social Pharmacy, University of Kuopio 24 October 2006 I read with interest and enthusiasm the recently published article by Räsänen et al. that studied the cost-utility of routine cataract surgery in real life setting. The presented routine approach with everyday conditions is novel. Furthermore, the article is excellent and a well-written piece of hard work. Consequently, these comments are aimed as advises for future novel research. Unfortunately, the article raised few important methodological issues which should be discussed in details. Firstly, authors refer that "HRQoL gain was assumed to last till the end of the remaining statistical life" and "this approach is typically used for the calculation of QALYs gained by the medical intervention". I assume that the authors have applied direct extrapolation without any level of uncertainty to calculate the QALYs. Unfortunately, the golden standard for the QALY estimation seems to be Markov state transition models in which e.g. the risk of death can be properly handled.Secondly, authors do not point out any system (e.g. multivariate analysis) to handle the confounding which is typically present in routine-styled settings. This raises the question of comparability between one eye and two eye surgery even in the average cost per average QALY (i.e. cost-effectiveness ratio, CER) setting. This comparison seems to be present in e.g. figure 10 as the cost-effectiveness acceptability curves (CEAC).Thirdly, in addition to deterministic approach, authors use stochastic approach and use bootstrapping-based probabilistic sensitivity analysis (PSA) to present the sampling uncertainty as the cost-effectiveness planes (cf. figures 7-9). In their planes, axes are referred as "incremental". Unfortunately, these axes are not incremental as they compare subgroup results to "no treatment" option. Actually, the axes give distributions for average costs and average effectiveness. Thus, the decision is not based on the incremental cost-effectiveness ratio (ICER) and the consequent decision at margin is very different from the decision at average (e.g. CER: whether to operate any eye or not vs. e.g. ICER: whether to operate both eyes or just one eye). This difference in interpretation between CER and ICER as well as the use of generic health-related quality of life (HRQoL) and CER as the primary outcomes raise ethical questions which should have been addressed. For example, in novel cancer medication evaluations, the comparator is typically the best supportive care (BSC) introducing some costs and effectiveness, if no other treatment option is relevant.Fourthly, authors do not point out whether or not dependence between average costs and average effectiveness was present. This may have an impact on the shapes of CEACs and, thus, to the probabilities of average cost-effectiveness.Fifthly, the cost per QALY approach handled this way refers to the comparison of allocative efficiency (e.g. league tables). In economics, technical efficiency is a necessary condition for allocative efficiency. Unfortunately, we cannot be sure how technically efficient are the operations in the trial hospital and, thus, how generalizable are the cost per QALY results.Lastly, I suggest utilizing incremental net monetary benefit (INMB) regression method in depicting the CEACs and handling the subgroup heterogeneity in future routine research. In this case, INMB may have been obtainable through propensity score methods. Authors point out the importance of optimizing the custom in which cataract operations are carried out. This may have been obtainable through multivariate INMB regression methods including e.g. waiting time, type of cataract, visual acuity at baseline, subgroup and propensity score as the independent variables. Thank you. Competing interests None.