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.

Four treatment strategies for hypertension were compared:

initial angiotensin II receptor blocker (ARB) plus additional calcium antagonist if ARB alone was insufficient (A+C group);

initial calcium antagonist plus additional ARB if necessary (C+A group);

initial ARB with additional diuretic if necessary (A+D group); and

initial diuretic with additional ARB if necessary (D+A group).

Olmesartan medoxomil was the ARB used, azelnidipine was the calcium antagonist used, and trichlormethiazide was the diuretic used.

Primary prevention.

Cost-effectiveness analysis.

The target patient population studied comprised male and female Japanese patients, aged 55 years at entry to the model. The patients had moderate essential hypertension (diastolic blood pressure, DBP, of 100 mmHg) with or without diabetes. Diabetic patients were assumed to be free of complications.

The setting was primary care. The economic study was conducted in Japan.

The effectiveness evidence came from a range of studies published between 1974 and 2004. The foundation of the current model was the authors' previous model (Saito et al. 2003). Resource use was mainly drawn from the authors' previous model and from another study, both published in 2003. The drug prices were from April 2004.

The effectiveness data were derived from a review or synthesis of published studies.

A Markov model was used to establish the cost-effectiveness of medical management of hypertension over the patients' lifetime. Observation periods in large-scale clinical trials are typically 5 years, whereas metabolic changes, complications and cardiovascular events may occur further into the future. Hence, evidence from clinical trials may be inadequate for a true assessment of cost-effectiveness, so a model is required to explore the association between extension of survival and the costs associated with antihypertensive drug treatment. The health states of the model were coronary heart disease (CHD), stroke, relapse of stroke, cardiovascular disease (CVD) death and non-CVD death. Two models were added to map the development of diabetes and progression of diabetic nephropathy. These included microalbuminuria, apparent proteinuria, chronic renal failure and haemodialysis. The main assumptions of the model were as follows

the treatment was initiated with the target DBP of 90 mmHg;

the DBP of patients who achieved the target value after 3 months of treatment could be maintained at 90 mmHg by single-drug therapy throughout the patients' lives; and

the target value could be controlled by adding another antihypertensive drug to the original treatment regimen.

The outcomes included:

the probabilities of mortality and morbidity (CHD, stroke, diabetes and nephropathy),

the incidence of hyperuricemia,

the antihypertensive effect of therapies, and

discontinuation of treatment.

A systematic review was not described. The authors referred to their previous model (Saito et al. 2003), and also drew on a number of randomised controlled trials and a compliance study with an implicit aim of using primarily Japanese studies.

Not reported.

Not reported.

The main judgement criterion appears to have been that the studies were Japanese.

Twenty-one primary studies were included in the review.

The authors did not formally combine the results of the individual primary studies. However, they did describe applying adjustments based on the results of Japanese studies to event risks from studies in other populations, to account for differences in Japanese patients. In the absence of direct comparisons, the authors also used several assumptions of equivalent effectiveness between two or more treatments.

No differences between the primary studies were investigated.

Mortality and risk of CVD (CHD and stroke) were estimated using the Framingham equations (Anderson et al. 1991, see 'Other Publications of Related Interest' below for bibliographic details), crudely adjusted by 20% for Japanese patients.

The authors estimated that the response rate (antihypertensive effect) was 80.2% for ARB, 83.4% for calcium antagonist, 83.4% for diuretic, 100% for ARB + calcium antagonist, and 100% for ARB + diuretic.

Discontinuation was reported cumulatively over 5 years, with rates at 5 years of 7.7% for ARB, 12.9% for calcium antagonist, 18.1% for diuretic, 12.9% for ARB + calcium antagonist, and 18.1% for ARB + diuretic.

Incidence of diabetes was reported cumulatively over 20 years, with rates after 20 years of 17.1% for ARB, 20.7% for calcium antagonist, 24.2% for diuretic, 18.9% for ARB + calcium antagonist, 20.7% for ARB + diuretic, and 20.7% for untreated patients.

Annual probability of progression to each of the four stages of nephropathy was reported. ARB and ARB + calcium antagonist were assumed to be equivalent.

All other treatments were assumed to be equivalent, with a higher probability of progressing to each stage.

The incidence of hyperuricemia was assumed to be zero for all patients except those receiving a diuretic (8%).

Expected survival was the outcome measure used in the economic analysis.

The costs were presented with and without discounting at 3% per year. The quantities and the costs were analysed separately though not presented separately. The costs of CHD, stroke, diabetes, diabetic nephropathy, death, chronic care, outpatient management, antihypertensive drugs, uric acid lowering drugs and laboratory tests were all included. Most of the costs were based on the authors' previous model and other reported studies, so it was unclear whether the estimation was based on actual data or derived using modelling. Some assumptions were applied (e.g. blood test monitoring of patients receiving diuretics). Drug prices related to April 2004, while other costs were drawn from studies published in 2003.

The costs were treated deterministically.

Not relevant to the perspective of a general practice.

Japanese yen (JPY).

Five parameters were varied in the range +/- 10% in the sensitivity analyses. These were antihypertensive effect, discontinuation of treatment, incidence of diabetes, inhibition of progression of diabetic nephropathy, and drug price. The incidence of hyperuricemia was also varied in the range 3 to 13%. A sensitivity analysis adjusting risk for diabetic patients was conducted. In the estimation of CHD risk, the authors assumed that CHD risk associated with concomitant diabetes was 2.02 times higher than that in patients without diabetes, based on a published Japanese study. The estimation of stroke used a published Japanese estimate of relative risk of stroke in patients with diabetes compared with those without. A final sensitivity analysis assumed that ARB + diuretic would successfully inhibit the progression of diabetic nephropathy.

The analysis was conducted by gender and whether or not the patients had diabetes, and was presented without discounting over the patients' lifetime.

For male patients without diabetes, expected survival was 25.48 years in the A+C group, 25.41 years in the C+A group, 25.47 years in the A+D group, and 25.37 years in the D+A group.

For male patients with diabetes, expected survival was 24.81 years in the A+C group, 24.38 years in the C+A group, 24.71 years in the A+D group, and 24.28 years in the D+A group.

For female patients without diabetes, expected survival was 31.26 years in the A+C group, 31.13 years in the C+A group, 31.24 years in the A+D group, and 31.06 years in the D+A group.

For female patients without diabetes, expected survival was 29.91 years in the A+C group, 29.09 years in the C+A group, 29.73 years in the A+D group, and 28.91 years in the D+A group.

When discounting of 3% was applied, the expected survival of patients without diabetes was longest in A+C group (male, 17.38 years; female, 20.06 years) and shortest in the D+A group (male, 17.33 years; female, 19.97 years).

Among patients with diabetes, expected survival was longest in the A+C group (male, 17.07 years; female, 19.47 years) and shortest in the D+A group (male, 16.81 years; female, 19.03 years).

An incremental analysis was not performed.

The analysis was conducted by gender and whether or not the patients had diabetes, and was presented without discounting over the patients' lifetime.

For male patients without diabetes, the expected cost was JPY 9.65 million in the A+C group, JPY 9.57 million in the C+A group, JPY 9.67 million in the A+D group, and JPY 9.44 million in the D+A group.

For male patients with diabetes, the expected cost was JPY 17.00 million in the A+C group, JPY 19.52 million in the C+A group, JPY 17.61 million in the A+D group, and JPY 19.68 million in the D+A group.

For female patients without diabetes, the expected cost was JPY 10.65 million in the A+C group, JPY 10.65 million in the C+A group, JPY 10.69 million in the A+D group, and JPY 10.52 million in the D+A group.

For female patients without diabetes, the expected cost was JPY 20.58 million in the A+C group, 23.90 million in the C+A group, JPY 21.39 million in the A+D group, and JPY 24.11 million in the D+A group.

When discounting of 3% was applied, the expected cost of patients without diabetes was highest in the A+C group (male, JPY 5.59 million; female, JPY 5.68 million) and lowest in the D+A group (male, JPY 5.21 million; female, JPY 5.28 million).

Among patients with diabetes, the expected cost was highest in the C+A group (male, JPY 10.57 million; female, JPY 12.01 million) and lowest in the A+C group (male, JPY 9.41 million; female, JPY 10.54 million).

An incremental analysis was not performed.

The estimated benefits and costs were not combined, so a cost-consequences analysis was presented. Readers can calculate incremental cost-effectiveness ratios for themselves from the results provided.

The sensitivity analysis showed that changes in parameters had larger effects on patients with diabetes. However, the effect of changes in parameters related to diabetes was larger than the effects of changes in drug price on the expected costs. Inhibition of nephropathy had a large effect on the expected costs and survival with concomitant diabetes: in the A+C group, its effects on survival and costs were 0.24 years and JPY 2.11 million, respectively. In male diabetic patients, the assumption that ARB + diuretic inhibited progression of nephropathy resulted in almost equal costs and survival for the A+C and A+D groups.

There was little difference in cost-effectiveness of the four treatments for hypertensive patients without diabetes, so appropriate combination therapy for these patients should be chosen on an individual basis. However, for hypertensive patients with concomitant diabetes, the authors concluded that cost-effectiveness in the A+C group was noticeably high in terms of both the costs and survival, suggesting that this regimen was superior to the others.

The authors noted that US and World Health Organization guidelines recommend diuretics as first-line therapies for hypertension, while European and Japanese guidelines recommend a number of different drugs first-line. They did not explicitly justify their choice of the comparators within the article. However, in a preceding introductory editorial, a commentator noted that calcium antagonists and ARBs are the most used antihypertensive drugs in Japan, while diuretics are the first-line drugs used in the USA. You should decide whether these are relevant comparators in your own setting.

The authors did not state that a systematic review of the evidence was undertaken. It was unclear whether the review was conducted in such a way as to minimise bias and identify all relevant research. The clinical evidence was mainly taken from Japanese clinical trials, although the quality of the trials was not discussed. The authors used data from the included studies selectively and also applied some unsubstantiated assumptions and modifications to the data. They did not consider differences between the primary studies when estimating effectiveness, except to use some crude adjustments on non-Japanese data with the intention of adapting estimates to the Japanese population. The authors did not sufficiently justify assumptions with reference to the medical literature. Limited one-way sensitivity analyses were conducted, which were unlikely to quantify the true levels of uncertainty in the model. Probabilistic sensitivity analysis might have been more effective in incorporating uncertainty into the model results.

The estimation of benefit (survival) was modelled using a Markov state transition model. This was an appropriate approach to take in view of the short-term clinical trial data and the long-term outcome of interest. The choice of life expectancy rather than quality-adjusted life expectancy was not justified.

The categories of costs relevant to the perspective adopted were included, as were all relevant costs for each category. Most of the costs and quantities were not reported separately, making it difficult to reproduce the calculations for another setting. The resource quantities and costs were drawn from published studies and, in particular, the authors' previous model (Saito et al. 2003). A sensitivity analysis was conducted for drug prices only, and this may limit the interpretation of the study findings. The sources of the unit costs or charges were not documented, and it was unclear whether they were inflated to the same price year using an appropriate health price index. Discounting was conducted, which was appropriate given the long time horizon of the study.

The authors did not compare their results with the findings of other cost-effectiveness studies, except for their own (Saito et al. 2003). The issue of generalisability to other settings was implicitly irrelevant, given the authors' aim of localising the model to the Japanese setting. The authors did not present their main results selectively. The authors undertook appropriate discounting of the costs and benefits, but applied it only in a sensitivity analysis, from which only some results were reported. The decision not to present incremental cost-effectiveness ratios was noted, but not explained or justified. However, this is not a major difficulty as the reader could calculate the incremental cost-effectiveness results themselves from the data provided. The authors' conclusions reflected the scope of the analysis but did not adequately acknowledge or explore the uncertainty inherent in the model, which arose from the multiplicity of data sources and ad hoc combination of clinical data.

The authors reported a number of limitations to their study. They noted that the CHD risk equation was an adjusted form of the Framingham equation (Anderson et al. 1991) and that data from Japanese clinical trials would provide more accurate estimations for Japanese patients. The transition probability of progression of nephropathy may have been overestimated because the data were taken from a study in which patients were treated with insulin. The authors also noted that the model might not have included all the factors associated with antihypertensive treatment, e.g. any renoprotective effects of drugs. The indirect costs and utility were not included on the basis of difficulty of estimation. Finally, different targets for blood pressure and different drug combinations could also have been examined.

The authors stated that among combination therapies with ARB, calcium antagonist and diuretic, there is little difference for hypertensive patients without concomitant diabetes. However, the most suitable combination therapy for hypertensive patients with concomitant diabetes would appear to be ARB + calcium antagonist, based on the pharmacoeconomic analysis. The authors noted that the limitations of their model provide scope for future investigation.

Supported by a grant from Sankyo, Japan.

Anderson KM, Odell PM, Wilson PW, et al. Cardiovascular disease risk profiles. Am Heart J 1991;121:293-8.

Saito I, Kobayashi K, Saruta T. Economic analysis of antihypertensive agents in treating patients with essential hypertension. J Clin Therap Med 2003;19:777-88 (Japanese).