Statistics from Altmetric.com
Chronic heart failure (CHF) is a major cause of hospitalisation and mortality.1 ACE inhibitors (ACEIs), β-blockers (BBs) and aldosterone antagonists (AAs) significantly reduce hospitalisation and improve survival of patients with CHF due to left ventricular systolic dysfunction (LVSD).1 ,2 Optimisation of medical therapy achieves the best outcomes, but this is often not achieved. The 2010 UK National Heart Failure Audit showed that almost 50% of patients with CHF are not receiving the target doses of their drug treatments.3 Furthermore, the progressive nature of CHF syndrome dictates treatment modification, which is best facilitated through monitoring to detect changes in the clinical condition.
In this context, there has been considerable interest in the potential role of brain natriuretic peptides (BNP and NT-proBNP). Natriuretic peptides (NPs) are neurohormones secreted primarily from the left ventricle in response to left ventricular wall stretch, and are predictors of left-ventricular function and of prognosis. The plasma NP level is of value in the diagnosis of CHF,4 and most drugs used to treat CHF significantly reduce NP level.5–9 Treatment optimisation for patients with CHF is usually based on clinical assessment and patient tolerance. Monitoring NP levels has been proposed as an alternative. There has been one cost-effectiveness analysis of the management of CHF using serial measurement of circulating NP concentration by a specialist compared with clinical assessment by a specialist.10 This analysis was based on one trial11 and showed that NP monitoring was cost-effective. However, this analysis was developed from a US perspective, thus it is of limited relevance to a European context. There is now more evidence on the impact of NP monitoring. We have therefore developed a new cost-effectiveness model to assess, from a UK NHS perspective, the cost-effectiveness of NP monitoring. This work was undertaken initially as part of the process to update the National Institute for Health and Clinical Excellence (NICE) clinical guideline on CHF, published in August 2010.1 Since the publication of the cost-effectiveness assessment by NICE,1 the results of additional trials assessing NP monitoring have been published. We have thus updated the original cost-effectiveness model with this new available evidence.
The aim of this economic evaluation was to assess the cost-effectiveness of three strategies for optimising medical therapy in patients with CHF: management guided by serial measurement of circulating NP concentration by a specialist (NP monitoring); management guided by clinical assessment by a specialist (clinical assessment); and usual care in the community (usual care).
Overview of the model
The model was run in two patient populations: CHF due to LVSD; and patients with CHF from any cause. We looked at two age groups: ≤75 and >75 years. It was developed from the perspective of the NHS and personal social services in England and Wales (ie, costs to patients were not taken into account). We used the quality-adjusted life-year (QALY) as the measure of health outcome. A lifetime horizon was modelled. An annual discount rate of 3.5% was applied to both costs and health outcomes incurred after 1 year, as is standard practice for health economic evaluations conducted for the NHS.12 Costs presented are in 2011 British pounds.
This economic evaluation was conducted on MS Excel 2003, and probabilistic analyses were modelled using a Monte Carlo approach.13 Probability distributions were applied to each parameter (γ for unit costs and standardised mortality ratios (SMRs); β for utility scores and probabilities; log normal for risk ratios, mean drug dosage and mean number of outpatient visits), and 5000 Monte Carlo simulations were computed for every analysis, with all model parameters set simultaneously, each selected at random from their respective distribution. This approach allows us to estimate the uncertainty around the cost-effectiveness results.
Identification and summary of clinical evidence
A systematic search of the literature was undertaken on Medline, Embase (OVID), the Cochrane Library and Cinahl (EBSCO).12 An initial literature review for NICE showed no published study on this topic before 2000.2 As part of the process to update the NICE clinical guideline on CHF (published in 20101), our search covered from 2000 to October 2009. From this search, four clinical trials were identified that assessed NP monitoring in patients with CHF due to LVSD.11 ,14–16 One of these was excluded since it was small, reported limited outcomes, and assessed the use of NP monitoring to guide BB uptitration alone (Beck-da-Silva et al 16), as opposed to all drug therapy as was used in the other trials. We were aware of a completed, but at that time unpublished, trial, BATTLESCARRED, which assessed NP monitoring in patients with CHF of any cause. The BATTLESCARRED research group gave us advance access to their paper, which has since been published.17 The search which covered until October 2009 was then updated to cover until September 2012. Two new trials were found to be eligible, one of NP monitoring in patients with CHF due to LVSD18 and one assessing patients with CHF of any cause,19 which also reported results for the subgroup of patients with CHF due to LVSD. Two trials were excluded, since the NP arm in these studies also received specialist nurse input.20 ,21 A third was excluded, the SIGNAL-HF trial, since it assessed NP monitoring in primary care as opposed to NP monitoring in secondary care by specialists.22
Five trials (Troughton et al,11 TIME-CHF,14 STARS-BNP,15 PROTECT,18 PRIMA subgroup19) that compared NP monitoring and clinical assessment by a specialist in patients with CHF due to LVSD were combined in meta-analysis for this economic evaluation. BATTLESCARRED,17 which compared NP monitoring, clinical assessment and usual care and was conducted in patients with CHF of any cause, was used for this economic evaluation with PRIMA main analysis,19 comparing NP monitoring and clinical assessment by a specialist in patients with CHF of any cause. Meta-analyses were conducted using Cochrane Review Manager (RevMan5) software. Fixed-effects (Mantel–Haenszel) techniques were used to calculate risk ratios (relative risk) for the binary outcomes. Cost-effectiveness analyses were also conducted on age subgroups for patients with CHF due to LVSD (<75 years/≥75 years) and patients with CHF of any cause (≤75 years/>75 years).
Estimation of survival and QALYs
All-cause mortality was taken from the clinical trials. Life-years were calculated from Kaplan–Meier survival curves when available, using the life table method.23 Survival curves were available from BATTLESCARRED17 and TIME-CHF14 for all patients and age subgroups (<75 years/≥75 years for TIME-CHF14 and ≤75 years/>75 years for BATTLESCARRED17). Where survival curves were not available, life-years were calculated by combining the baseline risk from the clinical assessment cohort with risk ratios (table 1).
After the trial end date, it was assumed that there were no differences in mortality between the trial groups. Post-trial mortality estimates were taken from a UK-based study24 that followed up 6478 patients with definite CHF for 3 years from 1991 and reported sex- and age-group SMRs. This was the only source of evidence found in the literature for CHF-specific SMRs. We adjusted the SMRs to account for the effect of ACEIs and BBs on survival.25 ,26 Then, we estimated life expectancy beyond the trials’ follow-up using the official life tables for England and Wales,27 but adjusting the mortality using the CHF-specific adjusted SMRs.
To estimate mean utility scores for each of the trials, we used the mean utility scores stratified by New York Heart Association (NYHA) class reported by Gohler et al,28 weighting the scores by the proportion of patients in each NYHA class at trial baseline. In the absence of evidence to the contrary, we assumed that the mean utility scores stayed constant over time and were the same for each intervention.
Estimation of costs
Resource use was taken from the clinical trials and combined with standard UK unit costs.29 ,30 Resource use components considered were hospitalisation, drug usage, outpatient visits, NP measurements and tests of renal function. For the post-trial period, a yearly cost per patient was applied.
We used the risk ratio for number of patients admitted to hospital from final trial follow-up to estimate probability of hospitalisation and assumed admissions occurred evenly over the follow-up period (table 1). Each admission was costed at £1812.29
Drug usage was calculated for each clinical trial, except for PROTECT because the reported data did not allow an appropriate calculation. Drugs were costed using British National Formulary prices.31
Four outpatient visits were planned after baseline in Troughton et al,11 STARS-BNP,15 TIME-CHF14 and PROTECT,18 and nine in BATTLESCARRED17 and PRIMA.19 Additional (unplanned) outpatient visits were reported by Troughton et al 11: mean of 0.9 per patient in the NP-monitoring cohort and 0.3 per patient in the clinical assessment cohort. PROTECT also reported these data as a median of one additional (unplanned) outpatient visit for the NP-monitoring cohort. We used this greater difference of one additional visit for the NP-monitoring and none for the clinical assessment cohorts, and assumed no additional visits took place for the usual care cohort in BATTLESCARRED. Specialist outpatient costs were estimated to be £116 per visit.29 An additional £28 was added to each outpatient visit in the NP-guided management group to cover costs of NP testing. In the BATTLESCARRED usual care cohort, it was conservatively assumed that all attendances were with the primary care physician. The mean cost per primary care physician consultation has been estimated nationally to be £53.30
When initiating or modifying dosages of ACEIs, Angiotensin Receptor Blockers (ARBs), diuretics and AAs (spironolactone/eplerenone), biochemistry testing for renal function is necessary. Using available data, we calculated probabilities of treatment modifications, and used a cost of £1.26 for each test.29 In the absence of data for the BATTLESCARRED usual care cohort, we assumed no biochemistry testing for this group.
The same yearly cost per patient was assumed for each intervention after the trial period, £1827. This was derived from a yearly cost of heart failure from the year 200031 adjusted for inflation to 2011.30
Patients with CHF due to LVSD
Table 2 shows the breakdown of cost components, life-years and QALYs for the cost-effectiveness analysis developed in patients with CHF due to LVSD. The NP-monitoring option is more effective in terms of life-years and QALYs, but more costly. However, using a cost-effectiveness threshold of £20 k per QALY gained, NP monitoring is highly cost-effective compared with clinical assessment (table 3). The results for age subgroups are similar (<75 years, ≥75 years) (table 3).
Patients with CHF of any cause
Table 4 presents the breakdown of cost components, life-years and QALYs for the cost-effectiveness analysis developed in patients with CHF of any cause. NP monitoring was the most effective and most costly option. Specialist clinical assessment is cost-effective compared with usual care, and NP monitoring is cost-effective compared with specialist clinical assessment (table 3). For those older than 75 years, NP monitoring was ruled out, because specialist clinical assessment was more effective in this group and less costly. Specialist clinical assessment was cost-effective compared with usual care for this subgroup (table 3). For those 75 years and younger, specialist clinical assessment was excluded on the basis of extended dominance. NP monitoring was cost-effective compared with usual care (table 3).
We found that NP monitoring was cost-effective using a threshold of £20 k per QALY in patients with heart failure due to LVSD and in patients with heart failure of any cause aged 75 or under, but not in people aged over 75.
The cost-effectiveness results for patients 75 years and older differed using outcomes from the BATTLESCARRED trial17 conducted in patients with CHF of any cause, or from the TIME-CHF trial14 conducted in patients with CHF due to LVSD. Compared with clinical assessment, NP monitoring was associated with improved survival in TIME-CHF,14 but poorer survival in BATTLESCARRED.17
The objective of this economic assessment was to measure the use of NP monitoring in secondary care, as informed by the available evidence from the literature. Therefore, we excluded the SIGNAL-HF study22 based in primary care, which compared management guided by NT-proBNP monitoring to a structured treatment plan based on clinical guidelines. In this study, there were no differences in drug use between the two groups or in the primary composite end point of days alive, days out of hospital and symptom score from the Kansas City Cardiomyopathy Questionnaire, or in its individual components.
On the basis of a review of our initial health economic analysis, and the clinical effectiveness data, the 2010 NICE guideline for the diagnosis and management of CHF recommended that NP monitoring be considered by the specialist in some patients (eg, those in whom uptitration is problematic or those who have been admitted to hospital).1 The specification of specialist was because the studies used to inform this analysis were all set in secondary care. The guideline development group noted that the impact of medical therapy guided by NP monitoring on outcomes was mediated by intensifying medical therapy, and avoiding admissions through earlier intervention when there was a change in clinical status.1 The guideline did not recommend routine use of NP monitoring to guide CHF management because of the uncertainties of the impact of this strategy in people aged over 75 years, who represent the majority of people hospitalised with heart failure.
Our current evaluation is based on international clinical trial data, but costs have been based on the UK's NHS. Costs and budgets do vary considerably between countries; however, given the results of the different analyses performed—which used, when appropriate, the most conservative estimates and assumptions (against NP-monitoring strategy), the relatively low incremental cost-effectiveness ratios, and the similarity of the conclusions of all conducted analyses (except for patients >75 years)—we believe that monitoring of NP levels probably represents good value for money for most European countries. Furthermore, these results are in agreement with a US-based cost-effectiveness study10 based on the trial by Troughton et al.11
We used a lifetime horizon to estimate the impact of the different monitoring strategies. To do this, we used data from a cohort recruited in 1991, which were published in 2005.24 It is likely that survival from heart failure has improved since then because of the advent of ACEIs and BBs. Therefore, we modified the survival to take account of these advances, using data from the relevant trials.25 ,26 If we have overestimated survival, we have overestimated the benefits of NP monitoring.
We did not find NP monitoring-guided management of CHF of any cause by the specialist to be cost-effective for patients older than 75 years. This reflects the negative results of some of the trials in this subgroup. It may be that increased prevalence of comorbidities in this age group, in particular renal dysfunction, increases the risk of adverse effects of uptitration of medical therapy and limits clinician choice. Secondly, the prevalence of heart failure with preserved ejection fraction (HFPEF) increases with age, and the evidence base for uptitration of therapy beyond diuretics for people with HFPEF has not been established. We were not able to tease out the relative importance of age, comorbidity or HFPEF. It is possible that an individual patient data analysis would throw additional light on the relative contributions of these as to why it was found that NP monitoring was not cost-effective in patients older than 75 years (unless they had LVSD). Nevertheless, the lack of effectiveness of NP monitoring in patients over the age of 75 with HFPEF appears plausible given that increasing age is associated with poorer renal function and thus lower tolerance to higher doses of diuretics.
In conclusion, optimisation of medical therapy in CHF guided by serial NP measurements by a specialist is cost-effective (at a threshold of £20 000 per QALY) compared with both medical therapy guided by specialist's clinical assessment and usual care in the community, if CHF is caused by LVSD or in patients <75 years old with CHF of any cause.
The authors acknowledge the contributions of the members of the Chronic Heart Failure guideline development group to the development and interpretation of this model.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.