Given the increasing burden of diabetic disease worldwide and the high number of available medications, the objective of the research by Zhang et al. (1) is praiseworthy. The study design is straightforward and, in our opinion, there are no particular points of concern from a clinical perspective. However, we do believe the study suffers from a number of inconsistencies that might reflect methodological errors or require further elucidation to reinforce the scientific value of the results.
It is well known that pharmacoeconomic analyses using quality-adjusted life-years (QALYs) should have an incremental design. In 2012, Stein (2) observed that “the absolute cost-effectiveness ratio (cost divided by effectiveness) is meaningless and should never be used in scientific publications.” This point has been repeatedly stressed in numerous studies (3,4), some of which date back to the 1990s (4).
One point of controversy in study by Zhang et al. (1)—which needs to be clarified by the authors—is whether or not this analysis was incremental. For example, the (incremental) difference in expected QALYs between treatments of metformin, sulfonylurea, and insulin versus metformin and insulin was estimated to be 16.12 quality-adjusted days in women (i.e., 68.41–68.37 QALYs) and 14.20 quality-adjusted days in men (i.e., 64.38–64.34 QALYs). However, the authors did not explain if and how these incremental differences participated in the calculation of the final parameters. Our impression is that these differences were not involved in the final parameters. In such a case, the analysis was not incremental and would be therefore vitiated by the above methodological error pointed out in the previous literature.
Indeed, the final parameters of this study (particularly the “expected medication cost per QALY” [1]) raise a number of questions. Did the denominator of this ratio contain total QALYs or incremental QALYs? Unfortunately, the expected medication costs were not reported in Table 2, and so interpreting the ratio of an unknown parameter (expected medication cost) divided by an unclear parameter (total QALYs? or QALYs gained?) proves to be extremely difficult.
Most of the problems with this article deal with the results shown in Table 2. For example, the expected life-years (LYs) ranged from 65.54 to 68.66, but this information conflicts with the assumption that diagnoses were made at around 54 or 55 years. Were expected LYs calculated from birth or from diagnosis? As it is unlikely that, according to the latter hypothesis, life expectancy was projected until the age of 54 + 66 = 120 years, the model-simulated values of expected LYs were most probably calculated from diagnosis. In such a case, the model-predicted values covered an interval of approximately 12 years (i.e., from around 54 to around 66 years), but the results shown in Table 2 give the false impression that these simulations covered a total of about 66 years.
Finally, apart from the technical details of the mathematical calculations, the main question focuses on the practical meaning of “expected medication cost per QALY,” because most readers are not likely to grasp what information Zhang et al. (1) intended to convey with this parameter.
Article Information
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
