Skip Navigation

European Journal of Heart Failure 2007 9(12):1186-1195; doi:10.1016/j.ejheart.2007.10.003
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (1)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Goode, K. M.
Right arrow Articles by Cleland, J. G.F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Goode, K. M.
Right arrow Articles by Cleland, J. G.F.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© 2007 European Society of Cardiology

Screening for left ventricular systolic dysfunction in high-risk patients in primary-care: A cost-benefit analysis

Kevin M. Goode*, Andrew L. Clark, Janet A. Bristow{maltese cross}, Kim B. Sykes and John G.F. Cleland

Department of Cardiology, Castle Hill Hospital Kingston-upon-Hull, United Kingdom

* Corresponding author. Department of Cardiology 3rd Floor Haughton Building, Hull Royal Infirmary, Anlaby Road, Kingston-upon-Hull, HU3 2JZ, United Kingdom. Tel.: +44 1482 675006; fax: +44 1482 675922. kevin.goode{at}hey.nhs.uk (K.M. Goode).


   Abstract
Top
 Notes
 Abstract
1. Introduction
 2. Methods
 3. Results
 4. Discussion
 5. Conclusions
 References
 
Background: Appropriate screening strategies are needed to cost effectively identify patients with undiagnosed and untreated left ventricular systolic dysfunction (LVSD).

Aim: To investigate the cost-benefit of screening high-risk patients in primary-care for LVSD (EF<40%) using various screening strategies.

Methods: Patients considered at high-risk of developing LVSD were recruited from three primary-care practices. Patients with known LVSD were excluded. Echocardiography, electrocardiography and blood tests were performed blinded to an NT-proBNP result. Logistic regression (LR) and receiver-operating characteristic analysis were used to assess the univariate and multivariable utility of NT-proBNP, QRS duration, symptoms and evidence of myocardial infarction (MI) to detect LVSD.

Results: 427 patients were assessed. 7.5% had undiagnosed LVSD. NT-proBNP, QRS, symptoms and MI were independent predictors of LVSD (p<0.014) and the resultant LR-model had an area-under-the-curve of 0.89 (0.84–0.94) and specificity of 54% (51–79%) at a sensitivity of 100%. The LR-model avoided 24.1% (18.1– 48.3%) of the cost and 50.1% (44.1–74.3%) of the echocardiograms compared to screening using echocardiography alone.

Conclusions: Screening high-risk groups in primary-care increases the pick-up rate for undiagnosed LVSD and using an LR-model combining NT-proBNP, QRS, symptoms and evidence of MI has significant cost benefits.

Key Words: Screening • Primary-care • Left ventricular systolic dysfunction • Natriuretic peptides • Electrocardiogram • Cost-benefit analysis

Received March 9, 2007; Revised September 24, 2007; Accepted October 11, 2007


   1. Introduction
Top
 Notes
 Abstract
 1. Introduction
2. Methods
 3. Results
 4. Discussion
 5. Conclusions
 References
 
Left ventricular systolic dysfunction (LVSD) is common, affecting perhaps 2% of the population [1,2]. In patients with known risk-factors such as previous myocardial infarction (MI), angina, hypertension or diabetes this increases to 7-9% [3,4]. Effective treatments exist and can approximately double life expectancy [5]. However, many patients with LVSD are asymptomatic, undetected and untreated [1,2]. Even when a diagnosis has been made, inadequate medical treatment is common [6]. At present, the standard way to diagnose LVSD is by cardiac imaging, typically using echocardiography [7]. Whole population screening for LVSD using echocardiography is impracticable both in terms of cost [8] and access to trained echocardiographers. However, it has been argued that screening for LVSD in high-risk populations may prove no more costly than existing screening programmes such as those for cervical and breast cancer [9,10]. Furthermore, it has been shown that pre-screening of similar populations using natriuretic peptides, electrocardiography (ECG) and hand held echocardiography prior to traditional echocardiography further reduces the cost burden of screening [4].

Amino-terminal pro-brain natriuretic peptide (NT-proBNP) has been used to screen for definite LVSD (EF≤35%) but not mild LVSD (EF 35-45%) in the general population [11] and to screen high-risk populations (previous MI, angina, hypertension and diabetes) for heart failure [12].

A normal 12-lead ECG is a powerful predictor of normal left ventricular function in some populations [13] and may improve general population screening when combined with NT-proBNP [11,14]. QRS width measured manually from a surface ECG is a crude measure of the severity of left ventricular (LV) dysfunction [15,16] and has the benefit of being automatically measured on most modern ECG recorders. The combination of QRS width and natriuretic peptides may improve the accuracy of diagnosing impaired LV-function in an outpatient HF-clinic setting [17] and in dyspnoeic patients presenting to an emergency department [18].

The incidence of LVSD in those with a recent MI is as high as 40% [19] and therefore a history of prior MI may be a useful predictor of LVSD in patients with as yet undiagnosed LVSD. Similarly the presence of breathlessness or peripheral oedema in patients presenting to primary-care raises the clinical suspicion of heart failure.

We sought prospectively to demonstrate the utility of screening high-risk patients for previously undiagnosed LVSD. We also wanted to explore whether using widely available non-echocardiographic observations and tests, alone or in combination, could reduce the number of referrals for echocardiography and provide cost avoidance when compared to screening using echocardiography alone.


   2. Methods
Top
 Notes
 Abstract
 1. Introduction
 2. Methods
3. Results
 4. Discussion
 5. Conclusions
 References
 
2.1. Study population
We identified those patients at high risk of having LVSD from the disease registers of three primary-care practices. To be considered high risk, a patient had to be registered with one or more of the following: ischaemic heart disease (IHD), previous myocardial infarction (MI), atrial fibrillation (AF), diabetes for at least 10 years, hypertension for at least 10 years; or currently taking a loop diuretic. Patients with previously documented evidence of LVSD were excluded.

The practices were selected to be socially and economically diverse with one from an affluent commuter village, a second from an inner-city low income area and a third from a rural market town with a mixed distribution of income.

Patients were then contacted by letter from the primary-care centre, and were reviewed in the community where a blood sample was drawn for NT-proBNP. Samples were taken using EDTA Vacutainers and centrifuged at 4 °C. The resultant plasma was analysed using an Elecsys 1010 analyser (Roche diagnostics; Mannheim, Germany) within 6-h of being drawn.

2.2. Clinical assessment
Patients were then invited to attend for half a day at the hospital-based community heart failure service, where they underwent standardised clinical examination, 12 lead ECG, echocardiography and blood tests for urea and electrolytes and full blood count. All investigations were blind to the NT-proBNP result. The clinical examination was made by the same person that reported on the ECG and echocardiogram.

Patients were deemed to have evidence of a prior MI if there was any documented history of MI or automatically reported MI from the ECG recorder. QRS duration and ECG evidence of MI were reported automatically using a GE Marquette MAC 1200 or MAC 5000 resting ECG system with automated interpretation software.

Their symptom status was graded according to the severity of breathlessness and peripheral oedema. Those with either NYHA>II or oedema above the ankles were classified as being ‘markedly symptomatic’, those with NYHA=II or ankle oedema alone and not already classed as markedly symptomatic were classed as ‘mildly symptomatic’. Those with NYHA class I and only trivial or no peripheral oedema were classed as ‘asymptomatic’.

Echocardiography was carried out by one of three trained operators. Left ventricular function was assessed by attempted measurement of ejection fraction (EF) using Simpson's biplane method (possible in 65% of subjects); and in all subjects by estimation on a scale of normal/mild/mild-to-moderate/moderate/moderate-to-severe/severe left ventricular impairment. Left ventricular impairment of greater than mild (EF<40%) was deemed sufficient to warrant treatment for heart failure and is denoted as clinically significant left ventricular systolic dysfunction (CS-LVSD). Left ventricular function was assessed by a second operator blind to the assessment of the first; where there was disagreement on the severity of LV dysfunction, the echo was reviewed jointly with the third operator and a consensus reached.

2.3. Analysis
2.3.1. Univariate analysis
Logistic regression (LR) analysis was used to find significant univariate predictors of CS-LVSD (p<0.05). The variables were ranked according to how well they could differentiate between those with and without CS-LVSD, as given by the change in –2 log-likelihood ({Delta}-2LL). Candidate predictors included all routine biochemistry, full blood count, examination, clinical history and automated ECG variables. All variables with p<0.05 on univariate analysis were entered into a multivariable LR-model to test for independence.

2.3.2. Model development
Those variables found to be independently predictive of CS-LVSD (namely NT-proBNP (log-transformed), QRS duration, symptom status and evidence of prior MI) were combined in a multivariable LR-model. The predicted probabilities generated by this model were saved for comparison using ROC-curve analysis.

2.3.3. ROC-curve analysis
The diagnostic performance of each independent predictor and the resultant LR-model were compared using receiver-operating characteristic (ROC) curve analysis. The diagnostic utility of NT-proBNP was assessed using age and sex specific cut-offs as established by Galasko et al. [20] and using a single cut-off of 150 pg/ml (17.7 pmol/l). The diagnostic utility of QRS width was assessed at a cut-off of 91 ms. The single cut-offs for NT-proBNP and QRS width were chosen to have the same sensitivity as that achieved using age and sex specific cut-offs for NT-proBNP, thus enabling a comparison of test specificities at a pre-specified sensitivity. Analysis of the ROC-curve for the LR-model yielded a probability (p) at which the specificity of the model was greatest whilst maintaining a sensitivity of 100% (i.e. no patients missed by the screening test).

2.3.4. Cost avoidance/echocardiogram reduction
A recent paper by Collinson [21] outlined a model for estimating the cost avoidance and reduction in number of required echocardiograms that accounts for the effects of test sensitivity, specificity and disease prevalence. We have elaborated on this idea by using bootstrapping [22] to estimate the 95% confidence intervals (CI) for specificity and disease prevalence in order to estimate CIs for cost avoidance and echo reduction. The equations used to calculate cost avoidance and reduction in required echocardiograms have been described previously [23]. We have assumed the cost of echocardiography (including interpretation) was {euro}150, of NT-proBNP was {euro}22.50 and of an ECG was {euro}16.50 [4]. ECG costs were based on automated computer interpretation and not expert clinical evaluation.

Since many screening programs only achieve a sensitivity of between 70-80% we estimated the cost avoidance, reduction in required echocardiograms and minimum cost-ratio at various test sensitivities using the LR-model.

Differences in central tendency were analysed using the Mann-Whitney U-test to avoid making assumptions about normality and categorical data tested using Pearson's {chi}2 statistic. Standard statistical comparisons and logistic regression were performed using SPSS v13. ROC-curve analysis and bootstrapping were performed using MATLAB– programs written in-house. An arbitrary level of 5% statistical significance (two-tailed) was used throughout.


   3. Results
Top
 Notes
 Abstract
 1. Introduction
 2. Methods
 3. Results
4. Discussion
 5. Conclusions
 References
 
We identified 654 patients on the high-risk registers, of which 427 consented to take part in the study and attended for assessment. Those patients that did not participate in the study were older than those that did (mean age 71.6 years vs 70.0 years, p=0.0001) and were less likely to be male (43.9% vs 57.1%, {chi}2=14.9, p=0.0001). A third of the patients were in more than one high-risk group with hypertension being the commonest single group to appear in, see Table 1. The practices selected for this study served different population sizes. There were no differences in age (p=0.88) or proportion of men (p=0.15) between the patients recruited from the village or market town practices. However, the patients from the inner-city practice were older than both the village practice (p=0.014) and the market town practice (p=0.001).


View this table:
[in this window]
[in a new window]

  		Table 1 Demographic and registry distribution by practice

 
CS-LVSD was present in 7.5% (5.1-10.0%) of patients. Compared to those without CS-LVSD, patients with CS-LVSD were more breathless and more likely to have angina, a previous MI, be on loop diuretics, have a broader QRS width and a higher NT-proBNP (see Table 2).


View this table:
[in this window]
[in a new window]

  		Table 2 Characteristics of the study population and a comparison between patients with and those without clinically significant left ventricular systolic dysfunction (CS-LVSD)

 
3.1. Univariate and multivariable analysis
Only 11 predictors were found to be significantly (p<0.05) predictive of CS-LVSD (see Table 3) and of these only log-transformed NT-proBNP, QRS width, evidence of prior MI and symptom status remained independently predictive of CS-LVSD on multivariable analysis. Individually, these four variables were only moderately predictive, with a best area-under-the-curve (AUC) of 0.72 (0.65-0.84) achieved with NT-proBNP (see Fig. 1a). More importantly none of the variables could achieve good specificity whilst maintaining a sensitivity of 100% (i.e. without missing any patients with CS-LVSD).


View this table:
[in this window]
[in a new window]

  		Table 3 Univariate and multivariable analysis of predictors of clinically significant left ventricular systolic dysfunction (CS-LVSD)

 


Figure 01
View larger version (13K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1 (a) Comparison of ROC-curves for NT-proBNP, QRS width, evidence of prior MI and symptom severity. (b) ROC-curve for multivariable logistic regression model with bootstrap confidence intervals compared to NT-proBNP gives AUC values.

 
Using age and sex specific NT-proBNP cut-offs a good specificity was achieved (SP=52.9%), but this was at the expense of sensitivity (SN=84.4%). A single NT-proBNP cut-off of 150 pg/ml gave identical sensitivity but lower specificity (SP=44.5%), see Table 4. Similarly, at a QRS width cut-off of 91 ms the specificity was lower still (SP=42.6%) for the same level of sensitivity. Symptom status and evidence of MI had higher test specificities but much lower sensitivities.


View this table:
[in this window]
[in a new window]

  		Table 4 Comparing the diagnostic performance of independent predictors of clinically significant left ventricular systolic dysfunction (CS-LVSD) and the resultant multivariable logistic regression model

 
3.2. Logistic regression model
The multivariable relationship between the log-odds of disease and the explanatory variables included in the final LR-model was calculated to be;


Formula 1

(1)

Where BNP is log-transformed NT-proBNP (pmol/l), QRS is QRS width (ms), Mildly is mildly symptomatic, Markedly is markedly symptomatic (see Methods for definitions) and MI is evidence of prior MI. The resulting probabilities of the LR-model to predict CS-LVSD is calculated by;


Formula 2

(2)

The probability p derived using the LR-model was diagnostically more powerful than NT-proBNP alone (AUC: 0.89 vs 0.72), see (Fig. 1b). Assuming that no patients should be missed when screening (i.e. sensitivity of 100%) the LR-model achieved a specificity of 54.2% (50.4-79.0%) compared to 5.4% (3.8-46.3%) using NT-proBNP alone. At a sensitivity of 84.4% (i.e. that achieved using age/sex specific NT-proBNP cut-offs) the LR-model gave a much higher specificity (SP=81.4%) than either NT-proBNP or QRS width alone, see Table 4.

3.3. Short-term cost effectiveness
The use of the LR-model would achieve cost avoidance over screening using echocardiography alone of 24.1% (18.1-48.3%) and reduce the number of echocardiograms required by 50.1% (44.1-74.3%) at a sensitivity of 100%. By lowering the test sensitivity it was possible to achieve greater cost and echocardiogram avoidance. For example, at a test sensitivity of 90% the cost avoidance increased to 40.1% (56.5-24.2%) and the echocardiogram avoidance to 65.4% (49.2-82.0%), see Fig. 2.


Figure 02
View larger version (7K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 2 Estimated (a) cost saving and (b) reduction in required echocardiograms at different sensitivities, using the LR-model. Solid line is actual value and dashed lines are 95% bootstrap confidence intervals.

 
3.4. Best registries to screen
Patients with CS-LVSD occurred in all of the high-risk groups but more frequently in the IHD, previous MI and "diuretic use" groups (75%, 75% and 41% of total CS-LVSD cases respectively). It may be a more effective use of resources to screen only those high-risk groups that have the best pick-up rate for CS-LVSD. By combining the previous MI and "diuretic use" groups, 30 out of a possible 32 patients (94%) with CS-LVSD would have been identified.

This would require the screening of only 198 patients or 46.4% of the original cohort size and would increase the proportion with CS-LVSD from 7.5% to 15.2%, but would result in 2 cases being left unidentified.


   4. Discussion
Top
 Notes
 Abstract
 1. Introduction
 2. Methods
 3. Results
 4. Discussion
5. Conclusions
 References
 
Patients with LVSD are at high risk of sudden death and of having or developing heart failure. Appropriate treatment can improve life expectancy and remove or reduce the symptoms of heart failure. However, patients with LVSD often have few symptoms and may be difficult to identify by clinical means alone. Clearly, a screening strategy is necessary. Screening the general population might be feasible but is likely to be expensive because of a high rate of false positive tests that would require further investigation. Even though the pick-up rate would be far higher than in cancer screening programmes, such a programme would be expensive, time consuming and labour intensive. Most cases might be identified by screening only high-risk groups and eliminating a large proportion of false positive tests.

4.1. Register screening
Overall, we found that screening of the disease registers held in primary-care, increased the proportion of patients having CS-LVSD from 2% seen in the general population to 7.4%, a level similar to that observed by Hobbs et al. [3]. However, screening the hypertension registers saw a pick-up rate of only 3% (even though we only screened those with hypertension for at least 10 years), whilst screening the MI register increased the pick-up rate to almost 18%, a level at which screening seems almost obligatory.

Of those with a diagnosis of CS-LVSD, 78% were symptomatic with breathlessness or lower-limb oedema. Of these, 56% were taking no loop diuretic, thiazide or spironolactone; 60% were not taking an ACE inhibitor or ARB; 60% were not taking a beta-blocker; and 28% were taking none of these therapies. Clearly registers by themselves do not improve patient care unless linked to a strategy to improve patient management.

4.2. Pre-screening strategies
We hypothesised that a strategy of using NT-proBNP alone or in combination with QRS duration could be used to reduce the numbers of patients needing echocardiography. Previous work has suggested that the negative predictive value of a normal NT-proBNP to screen for heart failure in high-risk populations is very high [12]. However, when screening for LVSD (with and without symptoms), much lower negative predictive values have been seen [24,25].

On its own, we found that NT-proBNP was an unsatisfactory tool for predicting or excluding CS-LVSD, confirming previous findings [25]. Adding QRS width, history/evidence of MI and symptom status to the diagnostic pathway using a logistic regression model greatly improved screening utility. This would have reduced the number of echocardiograms required by at least 44% and provided a cost avoidance of at least 18% compared to screening high-risk registers using echocardiography alone. Lowering the test sensitivity further improved the specificity of the LR-model. Given that manual screening for cervical cancer only has a sensitivity ranging from 50-75% and up to 94% using Autopap-directed re-screening [26], this may be ethically acceptable.

4.3. Detecting asymptomatic LVSD
Assuming that patients on disease registers were to get regular review, those with persistent symptoms would hopefully raise enough concerns for the primary-care practitioner to refer them for echocardiography. However, asymptomatic patients with undiagnosed LVSD are unlikely to raise enough suspicion for further investigation. Unfortunately, these very patients generated lower LR-model probabilities than both mildly and markedly symptomatic patients with CS-LVSD (median (IQR): 0.0044 (0.0081-0.0323) vs 0.0331 (0.0870-0.2093) and 0.0420 (0.1486-0.4989), p<0.001). Consequently, if we try to improve the cost avoidance of pre-screening by lowering test sensitivity this may only serve to miss those patients that are unlikely to be flagged for echocardiography.

4.4. Echocardiography post-MI
We found that 17.6% patients on the MI-registers had undiagnosed CS-LVSD and yet guidelines recommend that all patients presenting to hospital with an MI should have an echocardiogram as part of their diagnosis [27,28]. A contemporary epidemiological study of post-MI patients [29] found that 27% of patients surviving until discharge had not had LV-function assessed within 90 days of their admission. In addition, of those that did have LV-function assessed and were found not to have LVSD, 23% went on subsequently to develop HF within 1-6 years. It appears that guidelines are not always adhered to and that HF (and possibly LVSD) can develop sometime after an MI. Therefore, the regular screening of patients on primary-care MI-registers may be appropriate.

4.5. The role of the ECG
It has been shown that an ECG can be used to pre-screen for LVSD [14] when scored by a cardiology specialist. However, the measurement and interpretation of ECGs by primary-care physicians compares poorly against ECGs conducted by cardiology specialists [30]. Because of this, we have limited our choice of ECG measures to those that are automatically reported on most commercially available ECG recorders, and hopefully therefore less susceptible to inter- and intra-operator variability. Further studies will be needed to determine if this is indeed the case.

4.6. Cost to diagnose LVSD
The implementation of a screening program, should in principle, reduce the number of hospitalisations. Whether this is cost-effective in the long-term, compared to doing nothing, is beyond the scope of this study. However, it is simple enough to calculate the cost per diagnosis of the various screening strategies proposed.

If all patients on the MI register were referred for echocardiography, the cost per diagnosis would be {euro}850 and would capture 75% of the cases of CS-LVSD identified in this study. Screening both the MI and "diuretic use" registers would increase the cost marginally to {euro}990 per diagnosis, but capture 94% of the CS-LVSD cases identified. Introducing pre-screening of all the registers using NT-proBNP (whilst maintaining 100% sensitivity) almost tripled the cost to {euro}2424 per diagnosis. The LR-model faired better at {euro}1524 per diagnosis but was still almost double that of referring those on the MI register directly for echocardiography.

However, many of these screening strategies compare favourably to the cost per diagnosis of LVSD seen when patients with suspected heart failure are referred from primary-care. The prevalence of LVSD (defined as EF<40%) in patients referred from primary-care is between 9 and 13% [4,31] and this gives rise to an estimated cost per diagnosis of between {euro}1454-{euro}2124. Clearly, screening of the high-risk registers using the LR-model is similar in cost to routine referral from primary-care.

Galasko et al. [4] found that referring patients if they had either an abnormal NT-proBNP (refer to Table 4 for definition) or an abnormal ECG would cost {euro}1485 per LVSD diagnosis (sensitivity=94%; specificity=49%), a figure close to that achieved using our LR-model ({euro}1324; sensitivity=93.8%; specificity=68.6%). If we adjust Galasko's estimate using the prevalence seen in our study, their cost rises to {euro}1694 per diagnosis, suggesting that our strategy offers a cost improvement.

Galasko et al. argued that combining the ECG and NT-proBNP produced no additional cost savings. We have found the converse to be true. The lower cost per diagnosis seen by Galasko when using NT-proBNP alone occurred at the expense of test sensitivity (76%). This would have resulted in 27% of those with LVSD being missed by this screening strategy, a level that we consider to be too high.

4.7. Implications
The proposed LR-model does not use simple cut-off values for NT-proBNP or QRS width and this raises the question as to how it might be implemented in primary-care. Fortunately, it is relatively simple to construct a referral nomogram using the LR-model equation (Eq. (1)), see Fig. 3. This is done by establishing the LR-model cut-off that gives the desired sensitivity (in this case 100%) and drawing a line for all combinations of NT-proBNP and QRS width that result in this cut-off value using the LR-model equation. Separate lines are constructed for each combination of the categorical variables (i.e. evidence of MI and symptom status). NT-proBNP/QRS width value pairs that lay above the appropriate line (e.g. no evidence of MI and mildly symptomatic) should be referred for echocardiography. A prior MI and/or symptoms of breathlessness or lower-limb oedema, substantially lower the threshold for referral and in most cases should be referred irrespective of the NT-proBNP and QRS result.


Figure 03
View larger version (28K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 3 Referral nomogram. Patients with QRS width and NT-proBNP values that fall above the line for a given combination of prior MI and symptom status should be referred for echocardiography.

 
The proposed screening methodology is only presented as a one-off strategy. New patients would be added annually to the risk registers and there will be incident LVSD in those already tested. Therefore an appropriate interval between screening tests needs to be considered and its accumulative cost calculated. Short-term cost effectiveness presupposes that money will be saved by identifying patients with undiagnosed LVSD in the community or that the gain in quality life adjusted years (QALYS) would render it unethical not to screen. There also remains the ethical question of assigning additional disease status to people who were unaware that they might have it.

4.8. Study limitations
Much of this analysis depended on the accuracy of information held in disease registers. Practices may vary in their use and up-keep of such registries and therefore the observed utility of screening might be different than that observed. The patients that refused or were unable to participate in this study were older and more likely to be women and this may have resulted in some bias in the study.

We did not examine whether pre-screening could detect important cardiac disease other than CS-LVSD, such as left chamber dilatation, valve abnormalities and milder presentations of LVSD.

The ECGs were not conducted in primary-care and therefore we have not strictly validated the utility of QRS width as measured by the primary-care practitioner. However, since the QRS width and evidence of MI on ECG were established using automatic algorithms on the ECG recorder it is felt that these results should translate well into a primary-care setting.

The assessment of symptoms of breathlessness and lower-limb oedema were not made in primary-care and were not blind to the ECG and echocardiogram result. The primary-care practitioner may have rated the patient's symptoms differently and these could change between the time of referral and specialist assessment.


   5. Conclusions
Top
 Notes
 Abstract
 1. Introduction
 2. Methods
 3. Results
 4. Discussion
 5. Conclusions
References
 
Left ventricular systolic dysfunction identified by echocardiography was common in a group of patients identified as being at high risk by virtue of being registered in primary-care with either; ischaemic heart disease, prior myocardial infarction, atrial fibrillation, diabetes for at least 10 years, hypertension for at least 10 years or taking loop diuretics. Preliminary screening of this high-risk population using QRS width or NT-proBNP alone was insufficiently precise to be clinically useful. However, their inclusion in a logistic regression model, together with evidence of a prior myocardial infarction and symptoms of breathlessness or lower-limb oedema gave significant reductions in the number of false positive referrals and the cost per diagnosis of LVSD. This screening strategy compared favourably to the cost per diagnosis of LVSD seen when physicians refer patients with suspected heart failure from primary-care.


   Notes
Top
 Notes
Abstract
 1. Introduction
 2. Methods
 3. Results
 4. Discussion
 5. Conclusions
 References
 
{maltese cross} This paper is dedicated to the memory of Janet Bristow who sadly died in 2006. Back


   References
Top
 Notes
 Abstract
 1. Introduction
 2. Methods
 3. Results
 4. Discussion
 5. Conclusions
 References
 

  1. Davies M., Hobbs F., Davis R., et al. Prevalence of left-ventricular systolic dysfunction and heart failure in the Echocardiographic Heart of England Screening study: a population based study. Lancet (2001) 358:439–444.[CrossRef][Web of Science][Medline]
  2. McDonagh T.A., Morrison C.E., Lawrence A., et al. Symptomatic and asymptomatic left-ventricular systolic dysfunction in an urban population. Lancet (1997) 350:829–833.[CrossRef][Web of Science][Medline]
  3. Hobbs F.D., Davis R.C., Roalfe A.K., Hare R., Davies M.K. Reliability of N-terminal proBNP assay in diagnosis of left ventricular systolic dysfunction within representative and high risk populations. Heart (2004) 90:866–870.[Abstract/Free Full Text]
  4. Galasko G.I., Barnes S.C., Collinson P., Lahiri A., Senior R. What is the most cost-effective strategy to screen for left ventricular systolic dysfunction: natriuretic peptides, the electrocardiogram, hand-held echocardiography, traditional echocardiography, or their combination? Eur Heart J (2005) 27:193–200.[CrossRef][Medline]
  5. Cleland J.G., Clark A.L. Delivering the cumulative benefits of triple therapy to improve outcomes in heart failure: too many cooks will spoil the broth. J Am Coll Cardiol (2003) 42:1234–1237.[Free Full Text]
  6. Cleland J.G., Cohen-Solal A., Aguilar J.C., et al. Management of heart failure in primary care (the IMPROVEMENT of Heart Failure Programme): an international survey. Lancet (2002) 360:1631–1639.[CrossRef][Web of Science][Medline]
  7. Swedberg K., Cleland J., Dargie H., et al. Guidelines for the diagnosis and treatment of chronic heart failure: executive summary (update 2005): The Task Force for the Diagnosis and Treatment of Chronic Heart Failure of the European Society of Cardiology. Eur Heart J (2005) 26:1115–1140.[Free Full Text]
  8. Heidenreich P.A., Gubens M.A., Fonarow G.C., Konstam M.A., Stevenson L.W., Shekelle P.G. Cost-effectiveness of screening with B-type natriuretic peptide to identify patients with reduced left ventricular ejection fraction. J Am Coll Cardiol (2004) 43:1019–1026.[Abstract/Free Full Text]
  9. McMurray J.V., McDonagh T.A., Davie A.P., Cleland J.G., Francis C.M., Morrison C. Should we screen for asymptomatic left ventricular dysfunction to prevent heart failure? Eur Heart J (1998) 19:842–846.[Web of Science][Medline]
  10. Baker D.W., Bahler R.C., Finkelhor R.S., Lauer M.S. Screening for left ventricular systolic dysfunction among patients with risk factors for heart failure. Am Heart J (2003) 146:736–740.[CrossRef][Web of Science][Medline]
  11. Ng L.L., Loke I., Davies J.E., et al. Identification of previously undiagnosed left ventricular systolic dysfunction: community screening using natriuretic peptides and electrocardiography. Eur J Heart Fail (2003) 5:775–782.[Abstract/Free Full Text]
  12. Hobbs F.D.R., Davis R.C., Roalfe A.K., Hare R., Davies M.K., Kenkre J.E. Reliability of N-terminal pro-brain natriuretic peptide assay in diagnosis of heart failure: cohort study in representative and high risk community populations. BMJ (2002) 324. 1498-1500A.
  13. Mosterd A., de Bruijne M.C., Hoes A.W., Deckers J.W., Hofman A., Grobbee D.E. Usefulness of echocardiography in detecting left ventricular dysfunction in population-based studies (The Rotterdam Study). Am J Cardiol (1997) 79:103–104.[CrossRef][Web of Science][Medline]
  14. Hedberg P., Lönnberg I., Jonason T., Nilsson G., Pehrsson K., Ringqvist I. Electrocardiogram and B-type natriuretic peptide as screening tools for left ventricular systolic dysfunction in a population-based sample of 75-year old men and women. Am Heart J (2004) 148:524–529.[CrossRef][Web of Science][Medline]
  15. Murkofsky R.L., Dangas G., Diamond J.A., Mehta D., Schaffer A., Ambrose J.A. A prolonged QRS duration on surface electrocardiogram is a specific indicator of left ventricular dysfunction. J Am Coll Cardiol (1998) 32:476–482.[Abstract/Free Full Text]
  16. Khan N.K., Goode K.M., Cleland J.G., et al. Prevalence of ECG abnormalities in an international survey of patients with suspected or confirmed heart failure at death or discharge. Eur J Heart Fail (2007) 9:491–501.[Abstract/Free Full Text]
  17. Kruger S., Filzmaier K., Graf J., et al. QRS prolongation on surface ECG and brain natriuretic peptide as indicators of left ventricular systolic dysfunction. J Intern Med (2004) 255:206–212.[CrossRef][Web of Science][Medline]
  18. Sakhuja R., Chen A.A., Anwaruddin S., Baggish A.L., Januzzi J.L. Jr. Combined use of amino terminal-pro-brain natriuretic peptide levels and QRS duration to predict left ventricular systolic dysfunction in patients with dyspnea. Am J Cardiol (2005) 96:263–266.[CrossRef][Web of Science][Medline]
  19. Kober L., Torp-Pedersen C., Carlsen J.E., et al. A clinical trial of the angiotensin-converting-enzyme inhibitor trandolapril in patients with left ventricular dysfunction after myocardial infarction. Trandolapril Cardiac Evaluation (TRACE) Study Group. N Engl J Med (1995) 333:1670–1676.[Abstract/Free Full Text]
  20. Galasko G.I., Lahiri A., Barnes S.C., Collinson P., Senior R. What is the normal range for N-terminal pro-brain natriuretic peptide? How well does this normal range screen for cardiovascular disease? Eur Heart J (2005) 26:2269–2276.[Abstract/Free Full Text]
  21. Collinson P.O. The cost effectiveness of B-Type natriuretic peptide measurement in the primary care setting—a UK perspective. Congest Heart Fail (2006) 12:103–107.[CrossRef][Medline]
  22. Efron B., Tibshirani R. Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy. Stat Sci (1986) 1:54–77.[CrossRef]
  23. Goode KM, Cleland JG, Clark AL. Ruling out heart failure in primary-care: the cost-benefit of pre-screening using NT-proBNP and QRS width. Int J Cardiol in press.
  24. McClure S.J., Caruana L., Davie A.P., Goldthorp S., McMurray J.J. Cohort study of plasma natriuretic peptides for identifying left ventricular systolic dysfunction in primary care. BMJ (1998) 317:516–519.[Abstract/Free Full Text]
  25. Mueller T., Gegenhuber A., Poelz W., Haltmayer M. Head-to-head comparison of the diagnostic utility of BNP and NT-proBNP in symptomatic and asymptomatic structural heart disease. Clin Chim Acta (2004) 341:41–48.[CrossRef][Web of Science][Medline]
  26. Renshaw A.A. Measuring sensitivity in gynecologic cytology: a review. Cancer (2002) 96:210–217.[CrossRef][Web of Science][Medline]
  27. Van de W.F., Ardissino D., Betriu A., et al. Management of acute myocardial infarction in patients presenting with ST-segment elevation. The Task Force on the Management of Acute Myocardial Infarction of the European Society of Cardiology. Eur Heart J (2003) 24:28–66.[Free Full Text]
  28. Bertrand M.E., Simoons M.L., Fox K.A., et al. Management of acute coronary syndromes: acute coronary syndromes without persistent ST segment elevation; recommendations of the Task Force of the European Society of Cardiology. Eur Heart J (2000) 21:1406–1432.[Free Full Text]
  29. Torabi A., Khan N.K., Cleland J.G.F. Contemporary epidemiology of left ventricular systolic dysfunction after myocardial infarction. Eur J Heart Fail (2007) Supplements 6:A333. Heart Failure Congress, 9-12 June.
  30. Lim T.K., Collinson P.O., Celik E., Gaze D., Senior R. Value of primary care electrocardiography for the prediction of left ventricular systolic dysfunction in patients with suspected heart failure. Int J Cardiol (2006) 115:73–74.[Web of Science][Medline]
  31. Jeyaseelan S., Goudie B.M., Pringle S.D., Donnan P.T., Sullivan F.M., Struthers A.D. A critical re-appraisal of different ways of selecting ambulatory patients with suspected heart failure for echocardiography. Eur J Heart Fail (2007) 9:55–61.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (1)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Goode, K. M.
Right arrow Articles by Cleland, J. G.F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Goode, K. M.
Right arrow Articles by Cleland, J. G.F.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?