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.

Several warfarin products with different bioavailability (F) for the prevention of thromboembolism in atrial fibrillation were examined.

Strategy 1 was warfarin with F=1 (brand warfarin).

Strategy 2 was warfarin with F=0.80.

Strategy 3 was warfarin with F=1.25.

Strategy 4 was warfarin F=1.25 and F=0.80, alternately every other month.

Secondary prevention.

Cost-effectiveness analysis.

The study population comprised a hypothetical cohort of patients with atrial fibrillation who were receiving warfarin.

The setting was secondary care. The economic study was carried out in Canada.

The effectiveness and some resource use data were estimated from studies published from 1991 to 1998 and from a series of patients followed in 1998. The price year was 1998.

The effectiveness evidence was derived from a synthesis of published studies.

A deterministic Markov model was constructed to examine the clinical and economic outcomes associated with the different formulations of warfarin in patients with atrial fibrillation. The time horizon of the model was one year with monthly cycles. The health states of the model were no event, bleed (serious or major haemorrhage), clot (ischaemic stroke) and death. A major bleed was considered either a gastrointestinal bleed or an intracranial haemorrhage requiring hospitalisation. The INR levels considered were INR < 2 (low), INR between 2 and 3 (therapeutic) and INR > 3 (high). Patients who experienced a serious haemorrhagic event discontinued anticoagulant therapy and were switched to treatment with aspirin (acetylsalicylic acid, ASA). A schematic representation of the decision tree was provided.

The outcomes estimated from the literature were:

the proportions of patients with therapeutic INR levels and with either low- or high-therapeutic INR levels with different levels of bioavailability;

the rates of clot and bleeding events;

the mortality rates; and

ASA outcomes such as the rate of bleed, rate of death and rate of ischaemic event.

An equation was used to transform yearly values into monthly values.

The authors did not state whether a systematic review of the literature had been undertaken to identify the primary studies. The information on the primary sources was limited. The proportion of patients with different therapeutic INR levels with F=1 were obtained from a series of 55 patients identified at the Sunnybrook and Women's College Health Sciences Centre in Toronto (ON) in 1998. The proportions of patients with different therapeutic INR levels with F=0.80 and F=1.25 were estimated using a published formula. Other clinical data were derived from two clinical trials.

Not stated.

The robustness of the primary studies was not examined, but the use of clinical trials as the main source of evidence ensured the validity of the estimates.

Not stated.

Three primary studies provided evidence.

Not stated.

Not stated.

The proportions of patients with low-therapeutic INR levels, therapeutic INR levels and high-therapeutic INR levels were, respectively:

with F=0.80, 46.5%, 50.2% and 3.3%;

with F=1, 18.8%, 66.2% and 15%;

with F=1.25, 9.4%, 38.5% and 52.1%.

The rate of clot was 0.015 per person/month (p/m) given a low INR (INR < 2), 0.0013 p/m given a therapeutic INR (INR 2 - 3), and 0.0061 p/m given a high INR (INR > 3).

The rate of bleed was 0.00 p/m given a low INR (INR < 2), 0.0009 p/m given a therapeutic INR (INR 2 - 3), and 0.0050 p/m given a high INR (INR > 3).

The rate of death from an ischaemic event was 0.024 p/m at low INR, 0.057 given ischaemic event, 0.057 p/m at therapeutic INR given ischaemic event, and 0.023 p/m at high INR given ischemic event.

The rate of death from a major bleeding event was 0.00 p/m at low INR given major bleeding, 0.00 p/m at therapeutic INR given major bleeding and 0.021 p/m at high INR given major bleeding.

The rate of bleed with ASA treatment was 0.00015 per month.

The rate of death from major bleeding on ASA given major bleeding was 0.00 per month.

The rate of ischaemic event on ASA was 0.0035 per month.

The rate of death from an ischaemic event with ASA given ischaemic event was 0.0109 per month.

The summary benefit measures were the expected clot rate and the expected bleed rate. Both measures were estimated using the decision model. Given the short time horizon, no discount rate was applied.

Discounting was not relevant since the costs were incurred during one year. The unit costs were not presented separately from the quantities of resources used for all items. The health services included in the economic evaluation were drugs (including different warfarin formulations, aspirin, mark-up, and dispensing fee), INR laboratory test, physician visits, and the treatment of ischaemic stroke, ischaemic stroke death, haemorrhagic stroke, haemorrhagic stroke death, and gastrointestinal bleed. The cost/resource boundary of the third-party payer (i.e. Canadian governmental provincial payer) was adopted in the study. The costs were estimated mainly on the basis of a government payer, a local accounting system and some published studies. Resource consumption was derived from the literature and from some assumptions. The price year was 1998. The costs estimated in previous years were inflated using the Canadian Consumer Price Index.

No stochastic analysis of the costs was carried out in the base-case analysis.

The indirect costs were not included in the economic evaluation.

Canadian dollars (Can$).

A sensitivity analysis was performed to examine the robustness of the model results. First, all of the drug costs were set equivalent to the cost of the generic agent, or alternatively to the cost of the brand-name agent. Second, the analysis was conducted without drug, hospital and INR costs. A Monte Carlo simulation of the cost of the different treatment strategies for the baseline analysis was also conducted, with 10 000 iterations.

The average expected clot rate was 0.055 with strategy 1, 0.058 with strategy 2, 0.056 with strategy 3, and 0.077 with strategy 4.

The expected bleed rate was 0.016 with strategy 1, 0.015 with strategy 2, 0.017 with strategy 3, and 0.021 with strategy 4.

The average annual per-patient costs were Can$1,361 with strategy 1, Can$1,350 with strategy 2, Can$1,334 with strategy 3, and Can$1,613 with strategy 4.

An incremental cost-effectiveness ratio (ICER; i.e. the incremental cost per clot avoided or per bleed avoided) was calculated to combine the costs and benefits of the alternative warfarin strategies. Strategy 1 was compared with all other options.

When the benefit measure was the clot rate, strategy 1 was the base strategy (most effective) and the ICER was Can$3,667 compared with strategy 2 and Can$27,000 compared with strategy 3. Strategy 4 was dominated.

When the benefit measure was the bleed rate, strategy 1 was dominated by strategy 2, the ICER was Can$27,000 compared with strategy 3, and strategy 4 was dominated.

When all drug costs in the model were equivalent to the cost of a generic agent, the results showed that strategy 1 was the dominant economic strategy in all cases but one: when the cost per bleed avoided was Can$30,000 compared with strategy 2. The same results were observed when all drug costs in the model were equivalent to the cost of the brand-name agent.

Overall, the base-case results of the cost analysis were robust to variations investigated in the sensitivity analysis. The hospital costs had the greatest impact on the overall costs.

The Monte Carlo simulation provided the following 95% CIs for the different treatment strategies: Can$693 to Can$10,879 for strategy 1, Can$652 to Can$10,838 for strategy 2, Can$652 to Can$10,838 for strategy 3, and Can$699 to Can$10,892 for strategy 4.

Strategy 1 (F=1) should be considered the dominant strategy for the treatment of patients with atrial fibrillation, given the substantial similarity in clinical outcomes and total costs. The difference in costs between generic and brand-name drugs could also play an important role.

The selection of the comparators was consistent with the objective of the study. The authors stated that strategy 4 (i.e. the strategy characterised by switching between different warfarin products) was only included in the model because of the concerns of drug substitution. You should decide whether they are valid comparators in your own setting.

The effectiveness evidence came mainly from published data. It was not stated whether a review of the literature was performed, and the primary studies appear to have been identified selectively. Some information on the primary sources was provided. The use of data derived from clinical trials ensured the validity of the clinical estimates. Some data were also obtained from a series of Canadian patients. The issue of uncertainty around the clinical estimates was not addressed and sensitivity analyses were not carried out on clinical inputs. It should be noted that the results for INR levels were obtained through an algorithm based on the authors' assumptions.

The summary benefit measures were specific to the interventions considered in the study and are not comparable with the benefits of other health care interventions. The impact of the interventions on quality of life was not investigated.

The costs included were consistent with the perspective adopted in the study. However, the unit costs were not clearly reported separately from the quantities of resources used, which limits the possibility of replicating the analysis in other settings. Several assumptions were made to derive resource consumption data. The costs were treated deterministically, but uncertainty around the economic estimates was investigated in the sensitivity analysis. The source of the data was reported. The price year was stated, which aids reflation exercises in other settings.

The authors reported the results of other studies that had examined the clinical and economic differences between alternative drug formulations. The issue of the generalisability of the study results to other settings was not explicitly addressed and only limited sensitivity analyses were carried out. These issues limit the external validity of the study. The authors noted some limitations of their analysis. First, there was variability between linking bioavailabilities to INR to clinical outcome. Second, the model might not have reflected actual clinical practice in all centres. Third, the long-term consequences were not modelled in this study. Finally, only major gastrointestinal and intracranial bleeds were considered in this analysis.

The study results suggested that the use of one warfarin agent within the range of acceptable bioavailability might be considered economically attractive. The authors highlighted some factors that should be considered when evaluating the relative value of warfarin products. First, new generic warfarin formulations may have lower drug acquisition costs than brand-name warfarin. Second, over the short term (1 month), excess costs may be associated with a generic preparation because of the additional INR monitoring required. Finally, alternating warfarin products may be associated with increased costs and worse clinical outcomes.

None stated.

Benson SR, Vance-Bryan K. In favor of coumadin over generic warfarin. American Journal of Health-System Pharmacy 1998;55:727-9

The European Atrial Fibrillation Trial Study Group. Optimal oral anticoagulant therapy in patients with non rheumatic atrial fibrillation and recent cerebral ischaemia. New England Journal of Medicine 1995;333:5-10.

Basskin L. A pharmacoeconomic analysis of Coumadin versus generic warfarin. Formulary 1998;33:573-87.