The Cost Effectiveness of B‐Type Natriuretic Peptide Measurement in the Primary Care Setting—A UK Perspective

According to the European Society of Cardiology (ESC) task force on cardiac failure,1 the prevalence of cardiac failure is estimated at 0.4%–2% of the total population and increasing. The disease carries a poor prognosis, with a 50% mortality rate at 3.2 to 4 years.2 The survival rate of women with a diagnosis of heart failure is worse than that of breast cancer; only lung cancer carries a worse prognosis. Treatment with angiotensin-converting enzyme inhibitors, diuretics, and β blockers improves both quality and quantity of life.1 They are most effective when given early, therefore the challenge is to provide early and accurate diagnosis of cardiac failure. The clinical diagnosis of cardiac failure is difficult because the symptoms are nonspecific.3 Many patients present to their general practitioner (GP) with a history that could be due to a variety of causes, ranging from hypothyroidism to cardiac failure. A common presentation is a complaint of dyspnea, but other presentations include nonspecific fatigue and ankle swelling. It has been estimated that the diagnosis of cardiac failure is accurate in only 30%–60% of cases referred from primary to secondary care. Conversely, the incidence of a missed cardiac failure diagnosis in the community is unknown but is estimated to be at least as great as that of the incidence of symptomatic disease (1.5%) .4 The current preferred method of investigation of suspected cardiac failure is echocardiography. The “gold standard” for diagnosis is a reduced ejection fraction, although there is little agreement on the actual value that constitutes definite impaired systolic function. Therefore, in theory, the strategy would be to refer all patients with nonspecific symptoms for echocardiography. This is an unrealistic expectation in the UK health economy. The present system refers as few as 30% of potential patients with potential heart failure. The thought of increasing current echocardiography services by some 200% would cause major logistical difficulties for most providers. There is a significant shortage of trained echocardiographers, which is unlikely to be met in the short (or even medium) term. Even if echocardiography could be expanded, the current prevalence estimates of 30%–60% are likely to fall if there were an increased rate of referral. At the present rate of diagnostic accuracy, the amount of inappropriate testing is 40%–70%, which is clearly a waste of resources. For an estimated cost of echocardiography of $200 per procedure (an optimistic cost), this would yield a loss of $8000–$14,000 per 100 patients investigated. Therefore, there is a need to improve the initial patient selection process. The measurement of natriuretic peptides, especially B-type natriuretic peptide (BNP) and the N-terminal prohormone of BNP (NTproBNP), have been shown to be clinically effective5 as well as cost-effective6 in the acute care setting and have been accepted into routine clinical practice. Studies have also shown that they are diagnostically efficient in the primary care setting.7-10 They can be measured accurately and reliably.11-13 To analyze the potential value of BNP and NTproBNP measurement in primary care, a sequential decision model has been developed and an economic analysis performed. The comparison of the two strategies was performed with cost-minimization analysis. This is the simplest form of cost-economic analysis because it directly compares like with like and does not require cost/quality considerations. The cost effectiveness of a sequential testing strategy is well suited to cost-minimization analysis. The premise is that two of the investigation pathways should have the same outcome and total costs. If one is more expensive than the other, it is cost-inefficient. The two pathways compared were echocardiography for all referrals and an initial investigation to exclude patients without cardiac failure (true negatives) followed by echocardiography in those patients who initially tested positive (true positives and diagnostic false positives). The decision model (Figure 1) illustrates a process map or decision algorithm showing the decision steps used in current clinical practice and the effect of interposition of a preliminary screening test as recommended by the ESC task force on cardiac failure and the UK National Institute for Clinical Excellence (NICE). The cost model (Figure 2) represents an elaboration of the decision algorithm to include disease prevalence and the sensitivity and specificity of the testing strategies. Cost neutrality will occur when the cost of echocardiography (C1 + C2) equals the cost of the initial testing (C3) plus the cost of echocardiography in the diagnostic true-positive cases plus false positives (C3 + C4). This can be illustrated for the simple case of a perfect test with 100% sensitivity and specificity with a disease prevalence of 20%. The initial screening test costs $20 and the echo-cardiogram costs $200. If 100 subjects are considered, the cost of the echocar-diographic arm will be $20,000, comprising the cost of echocardiography in the true positives, C1, $4000 (20 subjects ×$200) plus the cost of echocardiography in the true negatives, C2, $16,000 (80 ×$200). The two-test model has a screening cost, C3, of $2000 (100 ×$20) plus the cost of echocardiography in the true positives alone, C4, of $4000 (20 ×$200). The saving, C5, $16,000, is based on the number of echocardiograms not performed. According to this simplistic analysis, it is apparent that the argument in favor of prescreening from a cost-benefit perspective is very compelling. Decision algorithm showing the decision steps used in current clinical practice and the effect of interposition of a preliminary screening test as recommended by the European Society of Cardiology and the National Institute of Clinical Excellence Cost model representing an laboration of the decision algorithm to include disease prevalence and the sensitivity and specificity of the testing strategies The determinants of the cost-minimization analysis will be disease prevalence, the ratio of the costs of the initial test to the definitive test (echocardiography) , and the specificity of the initial test used. The other issue that must be addressed is test sensitivity. Clearly, to be effective, the initial test will have to be maximized for sensitivity to achieve as close to 100% sensitivity as is feasible. This will again be influenced by disease prevalence since the absolute number of cases missed relates to the prevalence. A test with 95% sensitivity will miss 0.5 cases per 100 at 10% prevalence and one case per 100 at 20% prevalence. Model validation was performed using two datasets. A simulated data-set was constructed based on a disease prevalence of 20%, with a test sensitivity of 90% and a specificity of 80%. The general case for a single test vs. a two-test model was also developed. Cost neutrality (where the single test strategy is no more expensive than two sequential tests) will occur when the total cost of initial testing plus the cost of the diagnostic false positives is equivalent to the cost of performing the diagnostic test alone (C1 + C2 = C3 + C4). The model was solved to test the cost-efficiency when maximized for sensitivity to judge the relative contributions of test costs and test specificity at different disease prevalence. The ESC has proposed a strategy of prescreening based on electrocardi-ography (ECG) and/or measurement of BNP ECG has a reported sensitivity of 0.81–0.93 and a specificity of 0.26–0.63. However, these studies are not representative of the accuracy of reporting in general practice, where diagnostic accuracy may be more problematic. In a prospective observational study undertaken at Northwick Park Hospital in London (Drs. T.K. Lim and R. Senior), data were obtained that compared GP ECG interpretation, cardiologist ECG interpretation, and BNP measurements, with a final diagnosis based on echocardiography (written communication, 2005). This produced real data for model evaluation. The simulation data based on a test sensitivity of 90% and a specificity of 80% is mapped to the terms used in the model and is illustrated in Table I. This showed that an initial test strategy is cost-effective but at a detriment of two missed diagnoses. Solving the model for cost neutrality by setting sensitivity to 100% reveals that because of the relative cost differential between the cost of the initial screening test and the definitive procedure, the specificity of the initial test does not have to be high. This is illustrated in Figure 3, where the ratio between the cost of the echocardiogram and the screening test is plotted against test specificity for different degrees of disease prevalence. The lines correspond to the point of cost neutrality. Any screening test above the line will be cost-effective; anything below the line will be more expensive. As the two tests become closer in cost, the number of echocardiograms that can be avoided rises, but it can be seen that cost-efficiency is more likely than extra cost. Cost-minimization analysis showing specificity of initial test required at different ratios of echocardiography/initial screening test cost. Individual lines represent different rates of disease prevalence. Cost benefit is seen in all values of specificity above the line, indicating cost neutrality for any prevalence rate. This was validated using the Northwick Park Hospital dataset (written communication, 2005). This study showed that the ECG had a sensitivity of 0.53 vs. 0.95 and a specificity of 0.63 vs. 0.47 (GP vs. cardiologist). For natriuretic peptide measurement, sensitivities of >99% could be achieved at a decision threshold of 169 pg/mL for NT pro BNP by receiver operator characteristic curve analysis, although a specificity of 56% is seen. However, Figure 3 illustrates that this would still be expected to be cost-efficient. This is the case when calculated. The data are shown for both ECG and BNP in Table II using the findings from the Northwick Park Hospital study. It can be seen that GP-read ECG is apparently more cost-effective but misses significantly more cases than natriuretic peptide measurement. The use of a prescreening test before referral for echocardiography is cost-effective by a cost-minimization strategy provided the ratio of costs is at least two where a specificity of only 55% is required for cost neutrality. The ECG is not a suitable initial test, as neither sensitivity nor specificity is adequate. The problem with ECG interpretation is that the result is either normal or abnormal. Natriuretic peptides have shown sensitivity and specificity for the diagnosis of cardiac failure. They have the advantage that they are numeric values so the decision threshold can be optimized for sensitivity to avoid the problem of missed cases. It is apparent that even at relatively low specificity, cost-efficiency is likely because of the relatively high cost of echocardiography. It can be argued that there are other reasons for ordering an echocardiogram than the detection of systolic dysfunction. Certainly, no single test can be used without the appropriate clinical context, but in view of the high diagnostic efficiency of BNP measurement when applied judiciously, it will certainly enhance decision making. In primary care, BNP measurement has been shown to have a very high negative predictive value for a range of cardiovascular disorders other than systolic dysfunction.14 In conclusion, the provision of an initial test, as recommended by the ESC and NICE makes clinical and economic sense. BNP can be optimized to give maximal sensitivity and is readily available on a range of platforms, avoids the vagaries of ECG interpretation, can be easily ordered from primary care, and is relatively inexpensive. BNP measurement is, therefore, the test of choice for initial testing in the primary care setting.