Archived Comments for:
Evaluating the cost effectiveness of donepezil in the treatment of Alzheimer's disease in Germany using discrete event simulation
Modelling cost offsets in a vacuum of evidence can be a costly mistake.
Tracy Comans, Griffith University
3 December 2013
We commend the authors for a conducting a cost-effectiveness analysis in the treatment of Alzheimer¿s disease - there exists a definite need for such work. However, we wish to query the costing used in the discrete event simulation model and believe that apparent errors may have a significant impact on the reported results and ultimate conclusions.
The authors state that the daily cost of donepezil 10mg is listed as ¿4.20 (p7). Accordingly, a yearly course would therefore cost ¿1,533 (i.e. 365 days × ¿4.20). The undiscounted survival in the model for a cohort of 1,000 (implied but not specified by the authors) is 4,870 years. We would assume that the drug cost should be a total of ¿7.47 million (i.e. 4,870 years × ¿1,533). However the reported drug cost is ¿4,625 (Table 4, p8) - a remarkable difference. The authors do specify patient discontinuation (Table 2) and of course mortality that would reduce the cost to some extent but unlikely to account for the 7 million Euro difference.
This potential error would drastically alter the conclusions reached in the article. For example for the base case versus untreated patients, the revised net difference in total costs would be ¿7.44 million (undiscounted) and the revised ICER would be in the region of ¿50,000. This would not be considered cost-effective according to the author¿s criteria.
The placebo adjusted change reported in the article for the MMSE for donepezil is 1.92 points on a scale of 0-30. This small effect, translates through in the model to an equally small effect on QALY change of 146 QALYs between groups or 0.14 per person (undiscounted p.9) - less than 8 weeks of life in full health per person. We believe this small effect is unlikely to justify the acquisition cost of donepezil of ¿1,533 per year per patient.
The majority of gains therefore are in cost savings driven by large changes in costs for moving from one MMSE category to another. For example a change from a MMSE of 10.5 to 9.8 results in an extra monthly cost of ¿1,023. As data on costs of management of dementia were not collected in the donezepil trials the authors rely on modeling to justify the cost-effectiveness of the medication. The assumptions made are not necessarily verifiable, nor reproducible. Specifically, it is difficult to comprehend that such small absolute changes in MMSE would likely be associated with any reduced propensity to be placed in an institution, or in significant reductions in other health and community care use.
The authors report that the costs of care are drawn from Teipel 2007. On investigation, this article draws the cost data from Hallauer et al 2000, ¿Costs of medical treatment of Alzheimer patients in Germany¿. Gesundheitsokonomie und Qualitatsmanagement 5: 73¿79. This article was unable to be sourced, however given that the cost data must be at least 15 years old, representing a generational change, the costs of treatment are unlikely to be similar today.
Ultimately, where cost offsets are likely to be the driving component of a cost-effectiveness analysis then the quality of such data should be assessed to the same scrutiny as clinical efficacy data. Costs or at least time associated with institutionalization can and should be collected within clinical trials when such importance is being placed upon them. It should also be noted that such potential savings in institutional or hospital costs are usually not able to be realized as in the short term the number of beds is fixed and another patient in the queue will take the place.
Finally, we ask how does this article increase scientific knowledge? Cost-effectiveness analyses primarily designed to lobby payers of health care and funded by those with most to gain ¿ the manufacturers; add little of interest to the literature and lack validity outside the country it is written for. Consequently we suggest that such work, if published at all, should be written as indicative of the author¿s expectations in a vacuum of evidence rather than as proof of cost-effectiveness.
Tracy A Comans and Joshua Byrnes,
Centre for Applied Health Economics and
Griffith Health Institute,
Griffith University,
Brisbane, Australia.
Modelling cost offsets in a vacuum of evidence can be a costly mistake.
3 December 2013
We commend the authors for a conducting a cost-effectiveness analysis in the treatment of Alzheimer¿s disease - there exists a definite need for such work. However, we wish to query the costing used in the discrete event simulation model and believe that apparent errors may have a significant impact on the reported results and ultimate conclusions.
The authors state that the daily cost of donepezil 10mg is listed as ¿4.20 (p7). Accordingly, a yearly course would therefore cost ¿1,533 (i.e. 365 days × ¿4.20). The undiscounted survival in the model for a cohort of 1,000 (implied but not specified by the authors) is 4,870 years. We would assume that the drug cost should be a total of ¿7.47 million (i.e. 4,870 years × ¿1,533). However the reported drug cost is ¿4,625 (Table 4, p8) - a remarkable difference. The authors do specify patient discontinuation (Table 2) and of course mortality that would reduce the cost to some extent but unlikely to account for the 7 million Euro difference.
This potential error would drastically alter the conclusions reached in the article. For example for the base case versus untreated patients, the revised net difference in total costs would be ¿7.44 million (undiscounted) and the revised ICER would be in the region of ¿50,000. This would not be considered cost-effective according to the author¿s criteria.
The placebo adjusted change reported in the article for the MMSE for donepezil is 1.92 points on a scale of 0-30. This small effect, translates through in the model to an equally small effect on QALY change of 146 QALYs between groups or 0.14 per person (undiscounted p.9) - less than 8 weeks of life in full health per person. We believe this small effect is unlikely to justify the acquisition cost of donepezil of ¿1,533 per year per patient.
The majority of gains therefore are in cost savings driven by large changes in costs for moving from one MMSE category to another. For example a change from a MMSE of 10.5 to 9.8 results in an extra monthly cost of ¿1,023. As data on costs of management of dementia were not collected in the donezepil trials the authors rely on modeling to justify the cost-effectiveness of the medication. The assumptions made are not necessarily verifiable, nor reproducible. Specifically, it is difficult to comprehend that such small absolute changes in MMSE would likely be associated with any reduced propensity to be placed in an institution, or in significant reductions in other health and community care use.
The authors report that the costs of care are drawn from Teipel 2007. On investigation, this article draws the cost data from Hallauer et al 2000, ¿Costs of medical treatment of Alzheimer patients in Germany¿. Gesundheitsokonomie und Qualitatsmanagement 5: 73¿79. This article was unable to be sourced, however given that the cost data must be at least 15 years old, representing a generational change, the costs of treatment are unlikely to be similar today.
Ultimately, where cost offsets are likely to be the driving component of a cost-effectiveness analysis then the quality of such data should be assessed to the same scrutiny as clinical efficacy data. Costs or at least time associated with institutionalization can and should be collected within clinical trials when such importance is being placed upon them. It should also be noted that such potential savings in institutional or hospital costs are usually not able to be realized as in the short term the number of beds is fixed and another patient in the queue will take the place.
Finally, we ask how does this article increase scientific knowledge? Cost-effectiveness analyses primarily designed to lobby payers of health care and funded by those with most to gain ¿ the manufacturers; add little of interest to the literature and lack validity outside the country it is written for. Consequently we suggest that such work, if published at all, should be written as indicative of the author¿s expectations in a vacuum of evidence rather than as proof of cost-effectiveness.
Tracy A Comans and Joshua Byrnes,
Centre for Applied Health Economics and
Griffith Health Institute,
Griffith University,
Brisbane, Australia.
Competing interests
None.