Neurology -- Correspondence for Feldman et al., 63 (4) 644-650

Correspondence published:

Economic evaluation of donepezil in moderate to severe Alzheimer disease: Hein PJ van Hout, Judith E. Bosmans, and Wim A.B. Stalman (2 December 2004)

Reply to van Hout et al: Howard Feldman, Margaret Hux, Elliott Schwam (2 December 2004)

Economic evaluation of donepezil in moderate to severe Alzheimer disease 2 December 2004

Hein PJ van Hout,
VU University Medical Center, EMGO Institute
Van der Boechorststraat 7, 1081BT, Amsterdam, The Netherlands,
Judith E. Bosmans, and Wim A.B. Stalman
Send Correspondence to journal:
Re: Economic evaluation of donepezil in moderate to severe Alzheimer disease

Hpj.vanhout{at}vumc.nl Hein PJ van Hout, et al.

Although the authors reported a well-designed economic evaluation, they misleadingly concluded that treatment with donepezil is economically beneficial. [1] The authors state that they used standard X2 and t tests to compare costs. Nevertheless, no statistical test was presented to show the statistical significance of the cost saving.
We calculated the confidence intervals of the difference between placebo and donepezil for the costs categories (Table 1). The cost figures originate from table 4 of the original paper. The large standard deviations and the small difference in total costs in relation to the mean total costs in the treatment groups, already give an indication of insignificant differences. Confidence Interval Assessment (CIA) was used to calculate the 95% confidence intervals assuming normal distributed data. [2] A significant difference can be demonstrated when the interval does not comprise zero. Considering the skewed distribution of costs, the most appropriate method to analyze cost differences is bootstrapping. However, to do this we need individual patient data.
None of the cost differences was statistically significant. Therefore, the conclusion of the authors in the original article was incorrect and misleading. There is no evidence of economic benefit of donepezil compared to placebo in the treatment of moderate to severe Alzheimer disease.

References
1. Feldman H, Gauthier S, Hecker J, et al. Donepezil MSAD Study Investigators Group. Economic evaluation of donepezil in moderate to severe Alzheimer disease. Neurology. 2004;63:644-50.
2. Altman DG, Machin D, Bryant TN, Gardner MJ. Statistics with Confidence 2nd Edition. BMJ Books London 2000. http://www.medschool.soton.ac.uk/cia/main.htm

Reply to van Hout et al 2 December 2004

Howard Feldman,
University of British Columbia and Vancouver Hospital
S192-2211 Wesbrook Mall, UBC Hospital, Vancouver BC V6T 2B5, Canada,
Margaret Hux, Elliott Schwam
Send Correspondence to journal:
Re: Reply to van Hout et al

hfeldman{at}interchange.ubc.ca Howard Feldman, et al.

Van Hout et al have expressed their reservations over aspects of the methodology and the conclusions of our economic evaluation of donepezil in moderate to severe AD. They indicate that there was no statistical test presented to show the significance of the cost saving. We stated in the Methods that we used a direct statistical comparison of costs, and in Table 4 reported that there were no statistically significant differences at the 0.05 level between groups in mean costs over 24 weeks.[1]
Similar methodological approaches to ours have been used within the field of health economics.[3] Our analysis did not include patient-specific adjusted costs as would be needed to construct a bootstrapped confidence interval. In any event, we do not think that this would alter the lack of statistical significance between groups.
There are some discrepancies between the numbers in the Table provided by van Hout et al and the data reported in our study which makes comparisons difficult. It is not clear why their adjusted total costs were different from that in Table 4 of the manuscript. The difference in costs of unpaid caregiver time reported by van Hout et al should be a negative value.
With respect to our conclusion, the determination of health economic benefits has been previously recognized as an issue not readily approached using conventional tests of statistical significance because health resource utilization and cost data are typically much more variable than efficacy.[4] Achieving statistical significance would therefore require extremely large sample sizes and longer durations for adequate powering.[4] A recognized alternative is to piggyback onto the design of a trial of efficacy and safety, thus minimizing problems of internal validity and bias.[4]
Our economic analysis was conducted as an add-on study to the primary aim of the trial which investigated the efficacy and safety of donepezil in moderate to severe AD with a sample size determined on the basis of a CIBIC plus primary outcome. [5] We did not anticipate that we would achieve statistically significant differences. Nevertheless, we were able to show a net cost-savings in mean aggregate costs, independent of the previously demonstrated clinical benefits.[5]
It is important to understand that in cost evaluations designed to inform payers, it is total healthcare cost that is most relevant, and it is from this perspective that this analysis was performed.[3]
References
1. Feldman H, Gauthier S, Hecker J, et al. Donepezil MSAD Study Investigators Group. Economic evaluation of donepezil in moderate to severe Alzheimer disease. Neurology 2004;63:644-50.
2. Altman DG, Machin D, Bryant TN, Gardner MJ. Statistics with Confidence 2nd Edition. BMJ Books London 2000. http://www.medschool.soton.ac.uk/cia/main.htm.
3. Thompson SG, Barber JA. How should cost data in pragmatic randomised trials be analysed? BMJ. 2000;320:1197-200.
4. Drummond MF, O�Brien B, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programs. 2nd Edition. Oxford Medical Publications, NY 1997 pp256-9.
5. Feldman H, Gauthier S, Hecker J, Vellas B, Subbiah P, Whalen E. Donepezil MSAD Study Investigators Group. A 24-week, randomized, double- blind study of donepezil in moderate to severe Alzheimer's disease. Neurology 2001;57:613-20.