|
The cost of glucocorticoid-associated adverse events in rheumatoid arthritis |
Pisu M, James N, Sampsel S, Saag K G |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology Low doses of glucocorticoids (less than 15 mg/day prednisone or equivalent) were assessed. No other health technology was under investigation.
Type of intervention Treatment for inflammatory diseases.
Economic study type Cost-effectiveness analysis.
Study population The study population comprised hypothetical 60-year-old patients with early RA and no other prevalent conditions such as diabetes or hypertension. The control group comprised patients who were not on any RA-specific therapy.
Setting The setting was secondary care. The economic study was carried out in the USA.
Dates to which data relate The effectiveness data were collected from material published between 1960 and 2004. The resource use data was based on a decision analytic model. The unit prices were derived from literature published between 1993 and 2003. The price year was 2001.
Source of effectiveness data The effectiveness data were derived from a review and synthesis of published studies, supplemented with authors' assumptions.
Link between effectiveness and cost data The costing was carried out retrospectively on the basis of a hypothetical sample of patients.
Modelling A Markov model was developed to compare the health care costs. However, it also calculated the number of deaths that resulted from the use and non-use of glucocorticoids. The model incorporated seven health states. Specifically, no adverse events, fracture, hypertension, diabetes, fracture and hypertension, fracture and diabetes, and death. These were studied over a 2-year period with 6-monthly cycles. Half-cycle corrections were used where necessary. A total of 10,000 Monte Carlo simulations were run. The analysis was carried out using Data (version 3.9; Treeage Software, Cambridge). The authors defined a model that initially excluded myocardial infarction (MI) and stroke, but later included these events for completeness. The primary outcomes in the model were the total costs and number of deaths for users and non-users.
Outcomes assessed in the review The primary outcomes assessed were the probabilities (highest, lowest and average) of the following events:
osteoporotic fracture (hip, vertebral, pelvic);
gastrointestinal (GI) complications;
serious infections (pneumonia, urinary tract infection (UTI));
cataract;
MI;
stroke;
hypertension; and
diabetes mellitus.
The authors did not explicitly state that a systematic review of the literature was carried out, although they followed a systematic process.
Study designs and other criteria for inclusion in the review The authors searched for studies containing the terms "rheumatoid arthritis and glucocorticoid", "rheumatoid arthritis and adverse events" and "glucocorticoid and adverse events". Randomised controlled trials, open-label trials and observational studies were considered for inclusion. The authors searched for articles published between 1980 and October 2004. The inclusion criteria specified that patients in the treatment group received low-dose glucocorticoids (less than 15 mg/day prednisone or equivalent), patients in the control group did not receive any active therapy, and at least 6 months of glucocorticoid adverse events were reported. The authors included adverse events for which probability estimates could be reliably quantified.
Sources searched to identify primary studies MEDLINE and PubMed were searched. The reference lists of selected articles were also reviewed.
Criteria used to ensure the validity of primary studies Methods used to judge relevance and validity, and for extracting data Number of primary studies included A total of 36 studies were included in the review.
Methods of combining primary studies The authors used the highest and lowest probability values found in the literature to define the highest and lowest values of the triangular distributions used in the Monte Carlo simulations. The most common probability value was defined as the average probability value across all studies weighted by sample size.
Investigation of differences between primary studies The authors acknowledged that there were wide differences between the primary studies and used triangular distributions to assess the impact of this on the analysis. The results and differences between some studies were well discussed.
Results of the review The probability of osteoporotic hip fracture was 0.08 (0.00 - 0.36) in non-users and 0.14 (0.00 - 0.49) in users.
The probability of osteoporotic fracture was 0.04 (0.00 - 1.6) in non-users and 0.13 (0.12 - 5.56) in users.
The probability of osteoporotic pelvic fracture was 0.03 (0.01 - 0.06) in non-users and 0.18 (0.05 - 0.63) in users.
The probability of GI complications was 0.27 (0.05 - 1.24) in non-users and 0.90 (0.43 - 1.06) in users.
The probability of serious pneumonia infection was 1.66 (0.07 - 6.01) in non-users and 2.25 (0.29 - 7.72) in users.
The probability of serious UTI was 0.39 (0.07 - 1.24) in non-users and 0.44 (0.15 - 1.27) in users.
The probability of cataract was 0.73 (0.00 - 1.41) in non-users and 2.38 (0.63 - 3.0) in users.
The probability of MI was 0.14 (0.00 - 0.22) in non-users and 0.36 (0.21 - 0.63) in users.
The probability of stoke was 0.36 (0.07 - 1.24) in non-users and 0.28 (0.00 - 0.44) in users.
The probability of hypertension was 0.52 (0.00 - 3.88) in non-users. A probability of 0.96 (0.00 - 0.47) was reported in users, although this appears to be inconsistent.
The probability of diabetes mellitus was 0.34 (0.22 - 0.62) in non-users and 0.55 (0.00 - 1.27) in users.
Methods used to derive estimates of effectiveness The authors made some assumptions to support their model.
Estimates of effectiveness and key assumptions The authors assumed that patients could acquire only one of the relatively uncommon adverse events per 6-month cycle. They also assumed that GI complications, cataract, pneumonia, UTI, MI and stroke lasted only one cycle.
Measure of benefits used in the economic analysis Though the authors stated that their intention was to evaluate the costs, they also estimated the number of deaths over 2 years in users and non-users through the Markov model.
Direct costs The analysis was carried out from the perspective of the third-party payer, focusing on the direct medical costs. The costs were estimated over a 2-year period and discounted using the 6-month equivalent of an annual discount rate of 3%. The prices were measured in 2001 US dollars, and reflated using the medical care component of the Consumer Price Index. The unit costs for treating each of the adverse outcomes were obtained from a thorough review of the literature (publication dates ranged between 1993 and 2003), while the quantities were determined through the Markov model. The authors calculated the glucocorticoid-attributable cost factor (G-IFC) as the cost of purchasing the glucocorticoid plus the difference in costs between users and non-users, divided by the cost of purchasing the drug.
Statistical analysis of costs The unit costs were treated as having a triangular distribution. Standard deviations (SDs) for the cost results were calculated using this information.
Indirect Costs The indirect costs were not relevant to the perspective adopted.
Sensitivity analysis Extensive sensitivity analyses were carried out on both the effectiveness and cost outcomes. Sensitivity data for effectiveness were informed by Everdingen et al. (see 'Other Publications of Related Interest). Cost levels were altered to -25%, -50%, +25%, +50% and +100%, and the duration of therapy was increased to 10 years.
Estimated benefits used in the economic analysis The authors reported the probability of death for users and non-users of glucocorticoids. In the model without MI and stroke, the probability of death was 0.034 (SD=0.007) in non-users and 0.040 (SD=0.009) in users, a difference of 0.006 (SD=0.012). In the model with MI and stroke, the probability of death was 0.039 (SD=0.008) in non-users and 0.045 (SD=0.009) in users, a difference of 0.006 (SD=0.012).
Cost results In the model without MI and stroke, the mean cost was $696.80 (SD=155.0) in non-users and $1,142.2 (SD=165.9) in users, a difference of 445.4 (SD=227.4). The G-IFC was 1.46.
In the model with MI and stroke, the mean cost was $1,237.5 (SD=247.7) in non-users and $1,667.1 (SD=190.9) in users, a difference of 429.6 (SD=313.8). The G-IFC was 1.44.
The sensitivity analyses showed that GI complications and pneumonia had the largest impact on the costs.
The authors reported that if the model was run for 10 years then the costs would be between 5 and 6 times higher.
Synthesis of costs and benefits The costs and benefits were not combined.
Authors' conclusions 'On average for each dollar spent on purchasing these drugs an additional $2.1 was spent on treatment of adverse events. However, only $0.46 of this additional cost was strictly attributable to glucocorticoid use.' The authors also acknowledged that usage might lead to an additional 8 deaths per 1,000 users.
CRD COMMENTARY - Selection of comparators The authors compared glucocorticoid-associated morbidity in users and non-users. There was no comparator technology. The authors might have selected a commonly used drug, a relatively newly introduced drug, or considered drugs that are used in current practice in alternative settings, in order to obtain a comparator. A comparison would have enabled the reader to gain a better understanding of the relative costs and benefits associated with glucocorticoid usage.
Validity of estimate of measure of effectiveness The authors used systematic review methodology to undertake a thorough review of the current literature. The methods were well reported and justified, allowing relevant studies to be identified. The authors were able to exclude studies that did not relate to their own patient population and extract data relevant for their model. When combining data from published studies, the methods were well explained and included an adjustment for sample size.
Validity of estimate of measure of benefit The authors did not state as an objective the calculation of a measure of benefit. However, they derived the number of deaths in users and non-users as a summary measure of health impact. This measure was derived from the Markov model. Although informative, this measure does not enable the reader to understand the true health benefits associated with drug usage. Quality of life is obviously extremely important given the adverse events and the original condition. This paper would need to be considered in conjunction with a paper considering the full health benefits if decisions on the actual desirability of using glucocorticoids in current practice are to be formed.
Validity of estimate of costs The costing was carried out from the perspective of the third-party payer. It aimed to establish the cost of each adverse event. The authors were not able to break down the costs into their component parts as they were obtained from a review of the literature. The methodology behind the costing analysis was sound, with appropriate discounting and reflation carried out. Standard statistical analyses were carried out and sensitivity analyses established the impact of increasing and decreasing the costs. These efforts enable the reader to understand the robustness of the analyses, and also improve the generalisability of the results.
Other issues The authors highlighted that their study was one of very few to estimate the costs of glucocorticoid adverse events. They briefly summarised the results of other studies and discussed these results. The generalisability of the study was improved by identifying a range of studies for inclusion in the review and by carrying out extensive sensitivity analyses. The results were presented clearly and thoroughly, thus enabling the reader to draw their own conclusions. Several limitations were presented. For example, the potential overestimation of the relative impact of glucocorticoid use on health care costs, uncertainty in the correlation between adverse events, and not all possible adverse events were considered. The potential impact of these limitations on the study results was discussed. A further potential limitation was that there was no comparator technology in this analysis. Despite this, the study design and analysis carried out was consistent with and appropriate to the study question.
Implications of the study The authors did not make any recommendations for policy or practice following on from their study, although they identified research into the cost-utility associated with glucocorticoid use as an area for further work.
Source of funding Supported by the Agency for Health Care Research and Quality.
Bibliographic details Pisu M, James N, Sampsel S, Saag K G. The cost of glucocorticoid-associated adverse events in rheumatoid arthritis. Rheumatology 2005; 44(6): 781-788 Other publications of related interest Van Evergingen AA, Jacobs JW, Siewertsz Van Rcesema DR, et al. Low-dose prednisone therapy for patients with early active rheumatoid arthritis: clinical efficacy, disease modifying properties and side effects: a randomized double-blind , placebo-controlled clinical trial. Ann Intern Med 2002;136:1-12.
Indexing Status Subject indexing assigned by NLM MeSH Arthritis, Rheumatoid /drug therapy /economics /mortality; Cost of Illness; Gastrointestinal Diseases /chemically induced /drug therapy /economics; Glucocorticoids /adverse effects /economics; Humans; Markov Chains; Monte Carlo Method; Myocardial Infarction /chemically induced /drug therapy /economics; Pneumonia /chemically induced /drug therapy /economics; Stroke /complications /drug therapy /economics AccessionNumber 22005006439 Date bibliographic record published 30/04/2006 Date abstract record published 30/04/2006 |
|
|
|