To the Editor:
We have read with interest the work of Shevitz and colleagues1 entitled “A Comparison of the Clinical and Cost-Effectiveness of 3 Intervention Strategies for AIDS Wasting” published in the April 1, 2005 issue. Their objective was to compare nutrition alone (NA) with oxandrolone plus nutrition (OX) or progressive resistance training plus nutrition (PRT) for AIDS wasting. They conclude that OX and PRT induce similar improvements in body composition but that PRT improves quality of life (QOL) more than NA or OX, particularly among patients with impaired physical function (PF). They further conclude that PRT was the most cost-effective intervention, whereas OX was the least.
We applaud the authors’ attempt to present health economics data for an area that has largely been neglected. Moreover, the issue of managing or preventing wasting in AIDS is pertinent, given its impact on patients' outcome and QOL.2 Based on our understanding and research work in this area, however, we have serious questions regarding the study design, and, more significantly, the methodology that the authors used to arrive at cost per quality-adjusted life-years (QALYs). Ultimately, we view their “societal perspective” cost utility analysis conclusion as seriously flawed.
Our rationale for these statements is the following. First, although this was a randomized study, certain baseline patient characteristics tended to benefit the PRT group and work against the OX group [body mass index [BMI]; 20.6 vs. 19.8 kg/m2), patients on highly active antiretroviral therapy (HAART; 69% vs. 88%), hepatitis C coinfection (20% vs. 30%), and baseline daily calorie and protein intake (2285 vs. 2623 kcal and 96.9 vs. 109.9 g). Recently published evidence suggests that hospital admission, length of stay, discharge to long-term care, and mortality risk increase with lower BMI, which, in turn, increases the intensity of health care resource consumption.3 Further, assessment of compliance with NA or OX interventions was based on patient report without assurances of pill or supplement count, whereas compliance with PRT was based on patient attendance in a supervised program. The fact that 72% of enrollees had household incomes less than $20,000 and 15% did not have secure housing should have led the authors to invoke added measures to ensure compliance. In addition, the primary end points measuring treatment success were functional changes, and the related specific assessment was again biased toward PRT (“learning effect”).
Second, looking at the presented results, the study undervalued the data associated with the clinical end points studied at 12 weeks: BMI, cross-sectional muscle area (CSMA), and fat-free mass (FFM). There were no significant changes in BMI in any of the treatment groups. In the CSMA, the OX group had the largest average increase (7% vs. 5% for the PRT arm and 1% for the NA arm) and was statistically significant for OX and PRT. In terms of FFM, the OX group had the largest average increase (1.72 kg vs. 0.9 kg for the NA arm and 1.17 kg for the PRT arm), which was significant, and the PRT arm did not show a significant increase. We think that the results of these clinical end points should have received greater scientific discussion in this article.
Finally, our most serious concern is with the methodology used by the authors to arrive at costs per QALY. The authors reported that they followed the method of Brazier and his colleagues to convert their physical functioning scores into a utility measure. According to Brazier et al,4,5 one needs 11 specific items covering 6 of the 8 health domains in the Short Form-36 (SF-36) to validly calculate a utility score. To determine physical function, the work of Hays et al6 was applied in the following manner. Hays used 9 items measuring PF, with only 5 of these coming from the SF-36. Hence, based on what Shevitz et al1 reported, there could be at most 5 of the 11 required items used in calculating the SF-6D. The authors did not refer to any special allowances made for them to use the Brazier method in this way, such as through special consultation with Brazier or John Ware (the major force behind the development of the SF-36). In addition, the Brazier method provides only a translation of the SF-36 into a preference-based utility score. To calculate a QALY further requires a period during which the patients are in a particular health state. How this was done for this study was not clear. Also, taking the QALY calculation as correct, the calculation of a cost-effectiveness ratio should have been reported as an incremental ratio of the added cost and added QALYs for each treatment. It was not clear how costs per QALYs were calculated in Table 3 of the article.
Notably, the authors stated that none of the treatment arms had a statistically significant impact on QALYs. Yet, cost utility ratios for each of the treatments were presented. At a minimum, given that the changes in QALYs were not significant, it should have been noted that any calculation of statistical confidence intervals for incremental cost utility ratios for the treatments would include a ratio of infinity. It does not make sense, especially considering the small sample size, to take a statistically nonsignificant change from a treatment and present a cost-effectiveness ratio as if the effect is known with certainty. The authors included a range of ratios, but this range seemed to be generated by varying costs alone.
Aside from the methodologic concerns, we believe that conservative conclusions should be put forth when attempting to influence societal choices for health care interventions based on health economics data. Given the presented results, an appropriate conclusion in our view would be as follows: “taking the societal perspective and looking at the estimated community-adapted model (ECM) costs for a relatively healthy population, the cost of PRT and of OX are approximately comparable per patient for the treatment period in question, with available imputed QALY changes that were insignificant for both.” Finally, and perhaps most importantly, it is not clear why exercise and OX should be considered as competing interventions.7
In summary, the study lacked the requisite objectivity and transparency needed to inform decisions about resource allocation and overstepped its bounds in forming conclusions.
Hind T. Hatoum, PhD
Surrey M. Walton, PhD
Department of Pharmacy Administration and Center for Pharmacoeconomic Research University of Illinois at Chicago Chicago, IL
1. Shevitz AH, Wilson IB, McDermott AY, et al. A comparison of the clinical and cost-effectiveness of 3 intervention strategies for AIDS wasting. J Acquir Immune Defic Syndr
2. Fields-Gardner C, Fergusson P. American Dietetic Association; Dietitians of Canada. Position of the American Dietetic Association and Dietitians of Canada: nutrition intervention in the care of persons with human immunodeficiency virus infection. J Am Diet Assoc
3. Van Nes MC, Herrmann FR, Gold G, et al. Does the mini nutritional assessment predict hospitalization outcomes in older people? Age Ageing
4. Brazier J, Usherwood T, Harper R, et al. Deriving a preference-based single index from the UK SF-36 Health Survey. J Clin Epidemiol
5. Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health. J Health Econ
6. Hays R, Cunningham WE, Sherbourne CD, et al. Health-related quality of life in patients with human immunodeficiency virus infection in the United States: results from the HIV Cost and Services Utilization Study. Am J Med
7. Strawford A, Barbieri T, Van Loan M, et al. Resistance exercise and supraphysiologic androgen therapy in eugonadal men with HIV-related weight loss: a randomized controlled trial. JAMA