Heart failure (HF) is a major health burden for the Western world, with an increased prevalence in elderly individuals . HF represents an end-stage complication of several pathologies, including both cardiac and extracardiac pathologies. The most common HF diagnostic test is the measurement of reduced left ventricular ejection fraction (LVEF) using echocardiography . An evaluation of the circulating levels of brain natriuretic peptide (BNP) is considered to be a first-step triage test, and BNP measurements have been shown to have high specificity (Sp) but low sensitivity (Se) . Echocardiography has been used as a second-step triage test because it is considered to be more specific . Nevertheless, several recent studies indicate that approximately 50% of patients with HF have normal or near-normal LVEF or fractional shortening, although a variety of cut-off (CO) LVEF values have been applied unsuccessfully. Moreover, in these false-negative cases, additional diastolic dysfunction assessments do not provide additional clarity, because these readouts are weakly correlated with symptomatic or latent HF . Thus, it is clear that the currently available tests for HF greatly need improvement.
Using the most up-to-date laboratory methods to diagnose acute dyspnea, circulating BNP values can rule-out HF at values lower than 100 pg/mL . Only a few authors have diagnosed HF with BNP CO values between 300  and 400  pg/mL. However, both in acute cases and in those that fall within the BNP gray zone (100–400 pg/mL), which mainly correspond to patients with chronic, mild or latent HF, the LVEF and diastolic dysfunction assessments are reported to have a sensitivity of approximately 50%. Maisel et al. suggested that the presence of a large proportion of patients with acute breathlessness and BNP values in the gray zone could reduce the clinical utility of BNP; however, in practice, 75% of patients have values above 400 pg/mL or below 100 pg/mL . Nevertheless, these authors do not clarify what percentage of the BNP values were above the high CO point in acute HF breathlessness. HF can occur in the absence of cardiac ejection fraction reduction; in some subjects with true HF, such as in some cases of acute pulmonary edema, the presentation can be dramatic. These patients exhibit physiological and neuro-humoral patterns similar to those of patients with HF and reduced LVEF, including impaired oxygen consumption and elevated circulating levels of BNP . However, there is still controversy surrounding the underlying pathophysiology of these cases. Recently, the use of physiological descriptors such as “diastolic HF” has declined, and the more descriptive term “heart failure with normal ejection fraction (HFNEF)” has been adopted . The differential diagnosis between HF-associated acute dyspnea and other cardiac or non-cardiac dyspneas cannot be easily or quickly made at the point-of-care or in an emergency department . For HF diagnosis in emergency departments, echocardiography is challenging because it generally requires a longer turn-around time and advanced operator skills, which are disadvantages compared with the BNP test.
Thus, the objective of this meta-analysis of case-control and follow-up studies was to estimate a formula that can reliably predict HF in the widest possible range of levels. To achieve this goal, we utilized the New York Heart Association (NYHA) classifications and incorporated the simple BNP test. A similar approach was used in the study by Levey et al. , which opted not to utilize 24-h urine samples from nephrology patients to obtain creatinine clearance rates, as these measurements are time-consuming and difficult to obtain. Instead, these authors estimated the glomerular filtration rate (eGFR) using their Modification of Diet in Renal Disease (MDRD) formula, that was obtained by regressing serum creatinine levels according to age, gender and ethnicity in a sample of 1628 patients who had been randomized at 15 clinical centers. To test the data accuracy, these authors further compared the estimated values with the urinary clearance of I-125 iothalamate (reference method).
We used the meta-regression results of blood BNP as the most significant determinant of HFL. Traditionally, CO is determined from receiver operating characteristic (ROC) curve, that is a non-parametric model. As regression (and meta-regression) is a parametric linear model, it should obtain a higher Se and accuracy than non-parametric models. Using an estimation formula (eF) that was derived from the meta-regression, the current study aimed to determine whether and in which cases BNP should rule in the diagnosis of HF. This aim was successfully achieved due to the following factors: 1) the high number of subjects and the absence of evident biases in this meta- analysis; and 2) the use of BNP levels to provide diagnostic accuracy, which was shown to be reliable because it was assessed in both diseased and control subjects.
Studies published from 1998 to 2010 were selected from PubMed using the keywords “brain natriuretic peptide AND heart failure” and the limitations of “human species”, “age >18 years” and “English, French or Italian language”. Other publications were also identified in the references of the selected articles (Supplemental Data Figure 1, which accompanies the article at http://www.degruyter.com/view/j/cclm.2013.51.issue-6/issue-files/cclm.2013.51.issue-6.xml). The guidelines of the Meta-analysis of Observational Studies in Epidemiology (MOOSE)  were followed in performing this meta-analysis. Each study was reviewed twice, and the studies with discordant findings were reviewed a third time. The following inclusion criteria were applied: 1) studies that enrolled healthy subjects, and/or controls, and diseased subjects spanning the entire range of HF stages using the NYHA classification; 2) studies that used subjects without interfering pathologies (renal failure, pulmonary or systemic hypertension, obesity, and anemia); 3) studies that assessed HF levels (HFL) in diseased (HFLd) and control (HFLc) subjects and BNP Se and Sp in all patients; 4) studies that used whole blood or plasma BNP samples; 5) studies with a declared number of diseased and non-diseased subjects; 6) studies that used the maximum diagnostic accuracy to determine the CO; 7) studies that were not duplicates; 8) studies in which all of the subjects were present in no more than one sample group; 9) case-control or follow-up observational studies; 10) studies in which the subjects were consecutively enrolled; 11) studies that used BNP outcomes in a non-Gaussian distribution expressed in terms of median and inter-quartile ranges and that were reducible to a Gaussian distribution by transformation to the natural logarithm (ln); 12) studies with sample sizes of at least 30 patients; and 13) studies in which the right ventricle was not evaluated for HF.
First, the Gaussian distributions of BNP Se and Sp and the ln of the odds ratio (OR) were tested using the D’Agostino (1990) method . The OR was calculated as follows: (true positives/false negatives)/(false positives/true negatives). To normalize the distribution, the BNP outputs that were expressed as continuous data (i.e., the median, 25th percentile and 75th percentile BNP values for the diseased patients and controls, separately) were natural log-transformed to obtain the ln of the median BNP value for the diseased patients (lnmBNPd) and the controls (lnmBNPc). The extracted inter-quartile ranges were converted to one ln standard deviation by dividing them by 1.35 .
When the BNP blood values were expressed in pmol/L, they were multiplied by 3.46 so that all of the BNP values were uniformly expressed in pg/mL . A continuity correction was applied by adding 0.5 to the true-positive, true-negative, false-positive and false-negative data when the original values were equal to zero. HFLs were classified and subsequently scored according to the NYHA criteria [HFL0: subjects at risk, HFL1: latent dysfunction, HFL2: dyspnea from effort exceeding usual activities (mild HF), HFL3: dyspnea during usual activities (moderate HF), and HFL4: dyspnea at rest (severe HF)]. An additional HFL level, HFL(-1), was included in the meta-analysis to represent healthy subjects. When more than one HFL was considered in the study (with relative percentages), the mean value was obtained, and this value resulted in the generation of continuous HFL values (Supplemental Data Table 1). These HFL values were also categorized using the meta-ANOVA.
Subsequently, a subgroup random-effects meta-analysis and summary ROC curves, which were stratified by HFLd, were obtained to assess the Se, Sp, diagnostic BNP ln OR and the degree of heterogeneity across various subgroups (i.e., the variation in effect size that was attributable to heterogeneity [I2], with 0%–25% indicating homogeneity, 26%–50% indicating mild heterogeneity, 51%–75% indicating moderate heterogeneity, and 76%–100% indicating severe heterogeneity) .
Random-effects meta-regressions and meta-analyses of variance (meta-ANOVAs)
Compared with continuous lnmBNP data, binary BNP ln OR data are more reliable for detecting associations with HFLs in meta-regressions and meta-ANOVAs . Continuous lnmBNP data are not adjusted according to the BNP CO, which is specific to the study method and the laboratory method (LM) that was used. In contrast, the binary data are constitutionally adjusted for the CO values because the data represent the disease ratios, which reflect the frequencies of positive and/or negative results; therefore, these data are reliable outcomes for ascertaining the significance of HFLs as a disease-constitutive determinant or constitutive heterogeneity factor (HeF) in a first-order meta-regression and meta-ANOVA (Table 1, Step 1-A). Subsequently, all of the candidate first-order disease-conditioning HeFs (Table 1, Step 1-B), including age, gender, disease prevalence, LVEF, sample size, publication year, and study design, were assessed in relation to the BNP ln OR using the same statistics. With the meta-ANOVAs, significant heterogeneity among studies was considered as true heterogeneity only when the corresponding heterogeneity within the studies was insignificant . Only the concordant results of the meta-regression and the meta-ANOVA were accepted, because both meta-regressions and meta-ANOVAs can provide spurious results that become evident if they are compared with each other .
After the importance of HFL and the absence of disease-conditioning factors had been ascertained, the methodological factors were tested as candidate conditioning second-order HeFs; lnmBNPd was the response variable in the second-order meta-regression and meta-ANOVA (Table 1, Step 2). The LM levels were as follows: level 1, chemi-luminescent immunoassay (CLIA) produced by Biosite-Beckman; level 2, microparticle-enhanced immunoassay (MEIA); level 3, CLIA, using a fully automated ADVIA Centaur Bayer; level 4, immuno-radiometric assay (IRMA); and level 5, radio-immunoassay (RIA). In the LM scoring, the level functions as an indicator that has no ordinal meaning. If the lnmBNPd regression on the second- order HeFs was significant, further meta-regressions were performed to identify the homogeneous subgroup. Furthermore, with the meta-ANOVA, the mean values of the levels within the LM variable were compared with each other, using the tests of Bonferroni, Scheffé, and post-hoc comparisons with “dummy variables” . Next, only the homogeneous lnmBNPd and lnmBNPc data were combined and regressed on the corresponding HFLd and HFLc data. This step was performed to obtain a wide range of data to analyze in the subsequent third-order (or HF-scoring) meta-regression. Once the linear equation was obtained, the estimated HFL (eHFL) values were calculated (Table 1, Step 3).
Assessment of estimation formula (eF) performance
The following items were evaluated: 1) total bias (the mean difference between eHFL and HFL); 2) total precision [standard deviation (SD) of the total bias ; and 3) accuracy (the percentage of eHFL that was equal to HFL).
The bias assessment was conducted on the binary data set as follows: 1) the Begg and Egger tests were both applied  for the publication bias assessment; and 2) the methodological quality items were evaluated according to the recommendations of the Cochrane Collaboration  (Supplemental Data Figure 2). The bias estimation was accomplished as follows: 1) using a meta-regression of the BNP ln OR over every bias factor (bias score: 1=absent, 2=present, 0=undetectable); 2) repeating the meta-regression after excluding the biased studies (i.e., the studies with a significant meta-regression); and 3) using box-and-whisker plots to reevaluate the BNP ln OR after excluding the outliers (i.e., outcomes more extreme than 1.5 times the 25th or 75th percentile values) .
The meta-ANOVA was performed using SAS software version 9.1.3 (SAS Institute Inc., Cary, NC, USA). The other statistics and the graphics were obtained using Stata 10 (Stata Corp., College Station, TX, USA).
A total of 71 samples were included from 60 studies [4, 15–74]; the total meta-analysis included data from 26,485 subjects (Supplemental Data Figure 1; Supplemental Data Table 1). One study provided four samples , and seven studies each provided two samples [17–22, 49].
In the analysis, ln Se and ln Sp were not considered because of their non-Gaussian distributions (p<0.001 and p=0.006, respectively); conversely, ln OR (n=71, χ2=0.18, p=0.915), lnmBNPd (χ2=2.44, p=0.295) and lnmBNPc (χ2=1.41, p=0.493) were normally distributed.
In the subgroup analyses, which were stratified by HFLd for the random-effects meta-analysis, the BNP showed significant diagnostic power for each HFL (Figure 1). The overall OR was 24 (95% CI 18–32). Summary ROC curves and their diagnostic parameters are shown in Supplemental Data Figure 3 (sections A–E).
Random-effects meta-regressions and meta-ANOVAs
For the assessment of the disease-conditioning first-order HeFs (Table 1, Step 1-B), only disease prevalence demonstrated a concordant positive result in both the meta-regression and the meta-ANOVA. In addition, age was only significant in the meta-ANOVA, and EF was only significant in the meta-regression.
The constitutive first-order HeF HFL was a significant regressor (Table 1, Step 1-A) (n=71; slope 0.56, CI 0.24–0.88; residual I2=87%), and the LM served as the methodological second-order HeF (Table 1, Step 2). The regression was significant (n=71; slope −0.38, CI −0.59 to −0.16; residual I2=83.81%).
Within the LM, there was homogeneity among levels 1–3, In fact, using the meta-ANOVA mean comparison, a difference resulted between the first three levels and the last two. Furthermore, taking into account only the first three levels, lnmBNPd, regressed on HFLd, was homogeneous (n=48, slope 1, CI 0.7–1.32; residual I2=3%). When the meta-regression accounted for LM levels 4–5, the resulting model was heterogeneous and non-significant (n=23; slope 0.6, CI −0.1–1.3; residual I2=88%).
The third-order meta-regression was significant (n=94; slope 0.886, CI 0.72–1.05; residual I2=0.00%). The eF was determined as follows: eHFL=(lnmBNP–3.157)/0.886 (Table 1, Step 3). Consequently, the resulting BNP CO for HFL4 was 530 pg/mL. In fact, [ln(530)–3.157]/0.886=3.51=HFL4 CO (Table 2).
Assessment of eF performance
The total bias of 0.0997 was insignificant (n=94; SD=0.926, CI −2.84 to 3.06). For accuracy, the percentage of eHFL that was equal to HFL was 54%, and the percentage of eHFL that was equal to ±1 HFL was 35%; for HFL4, these values were 78% and 22%, respectively (Table 2).
Regarding publication bias, neither BNP ln OR nor lnmBNPd were significant in the Begg and Egger tests [n=71; z-test (1.14; p=0.255) and bias coefficient (1.36; CI −1.29 to 4) for BNP lnOR; z-test (1.06; p=0.29) and bias coefficient (−1.99; CI −10.39 to 6.41) for lnmBNPd]. No significant methodological bias was detected (Supplemental Data Figure 2). Regarding the excluded outliers, the BNP ln OR values of the studies [33, 34] and of the sample with severe systolic dysfunction in the study  were considered outliers. The complete results were confirmed after excluding the outliers.
Discussion and conclusions
In this random-effects meta-analysis, Table 2 shows that only HFL(-1) and HFL4 revealed good diagnostic performance for BNP. HFL(-1) showed total agreement between the real and estimated HFL results, and the BNP values remained below the negative CO values. HFL4 showed 78% reliability when comparing eHFL and HFL. Therefore, according to the three-step meta-regression method used here, HFL4 severe dyspnea can be more easily differentiated from non-HF acute dyspnea using eHFL, and HFL4 can be reliably “ruled in” using eF BNP CO (i.e., 530 pg/mL) and eHFL. Furthermore, BNP test is also useful as a point-of-care test in emergency departments because it is performed on whole blood. Conversely, NT-proBNP, a hormonal inactive form, works only on serum samples. In my previous meta-analysis , I found that, while BNP heterogeneity was explained by some heterogeneity factors (therefore making it useful), NT-proBNP was associated with unaccountable heterogeneity and therefore was not included in the current meta-analysis. However, NT-proBNP has been shown to have an excellent reliability in vitro ; in vivo this pro-hormone was sensitive in diagnosing acute heart failure, but less sensitive than BNP in detecting rapid variations in the pathophysiological changes in the patients [77, 78].
Table 2 shows that the BNP mean level at HFL4 was 665 pg/mL (CI 123–3604). By normalizing this outcome with the ln transformation, a value of 6.45 (4.8–8.2) was obtained, and 6 was the ln of 400 (i.e., the ln of the rule-in CO value by the method of the ROC curve); however, this value corresponds to approximately one-third of the half-standard deviation on the left-hand side of the mean value, meaning that it covers only the mean minus 14% of the possible BNP outputs of HFL4. Therefore, the rule-in ROC-curve CO has an accuracy of approximately 64% in detecting HFL4. With the three-step meta-regression method, 78% of the eHFL4 cases correctly estimated the corresponding HFL; this is also the accuracy of the CO found with eF. In previous studies and reviews [2, 4], only the rule-out CO (that optimized Sp) was adopted. Using the higher CO value assessed with the ROC curve, the Se and accuracy were unsatisfactory. The current study was the first to attempt to improve the HFL4 diagnostic accuracy of BNP using a meta-analysis. In this meta-analysis: 1) regression was used as a statistical model with parametric assumptions, which is more powerful to assess HFL4 than CO with the ROC curve; 2) eHFL was not determined using a single-study BNP CO and outcome value; and 3) concerning I2 and only considering the second-generation LM in the third-order meta-regression, the lnmBNP I2 was absent from this second-generation LM, and the study homogeneity contributed to the higher eF performance for HFL4. Furthermore, 1) the extreme eHFLs were one-side bounded, which contributed to the higher performance of the eF for these levels (Table 2); 2) among the negative and positive COs, the extreme HFLs were the farthest from the BNP “gray zone”. Other favorable conditions in this meta-analysis included the absence of relevant biases, the large sample size, and the wide observation range for BNP that covered both normal and diseased subjects. Diseased and control subjects were each considered at every possible HFL. In the Levey study , the mean difference between eGFR and GFR was 0.2, which is twice the mean difference between HFL and eHFL that was found in the current HF meta-analysis. Nevertheless, the Levey confidence interval was narrower (−0.08 to 0.48) because the trial samples in their study were more homogeneous (although multicentric) than ours. Conversely, in the current meta-analysis, the eHFL presented a higher accuracy of the mean but a wider standard deviation. One limitation of our study was that the highest and lowest HFL sample sizes were half the size of other HFLs, which spuriously reduced the subgroup I2 value (Figure 1) for HFL 1 and 4. However, the highest HFL minimized the confounding effect of the first-order conditioning factors on the BNP. The HFL4 eF can be usefully applied to the point-of-care, and in emergency departments it can quickly differentiate (using whole blood) HF-associated acute dyspnea from other severe forms of dyspnea that do not have an HFL4 component. The 22% incorrect diagnosis probability obtained in the current study is the lowest value for HFL4 diagnosis to date. The 78% accuracy of the eF needs to be confirmed in the Emergency Department practice. For example , dyspnea in a patient at rest could be caused by: 1) a significant pulmonary embolus or edema; or 2) a sudden HFL4 with another acute cause (that could be extra-cardiac). In the latter case, a normal ejection fraction could be present with the HF: if the extra-cardiac pathology has suddenly developed and the heart is not yet stressed enough to decrease its ejection fraction, but is operating insufficiently for an intrinsic reason (diastolic dysfunction ). In the condition described at the point 1) the heart could be a) normal; b) mildly or moderately insufficient; or c) suddenly severely insufficient in non-HF severe dyspnea, but this occurrence is very unlikely. In the cases of a) and b), we conclude that severe dyspnea is of non-HF origin. Only in HFL4 with severe dyspnea, can non-HF acute dyspnea be excluded. In practice, in non-HF severe dyspnea, a) and b) are always the cases. In mitral stenosis and/or insufficiency, that is a cause upstream of the left ventricle, BNP levels may not be very high despite severe symptoms . In fact, the mitral steno-insufficiency, if not yet complicated with HF, is another non-HF condition of severe dyspnea. The non-HF causes of acute dyspnea may be extra-cardiac (pulmonary embolus or edema) or cardiac (mitral non complicated steno-insufficiency). The rule remains that an eHFL outcome of <4 (i.e., <3.51) indicates non-HF acute dyspnea (Table 2). However, which is the “gray zone” of this new CO has to be determined in the future. The main exceptions to this rule appear to be when the pathologies of severe renal failure and obesity interfere . Obesity spuriously decreases the levels of BNP ; furthermore, in advanced renal failure, that interferes increasing the BNP outcome, the clinical assessment is crucial , as eHFL results are inaccurate in 22% of cases. A higher eF accuracy could be achieved by performing an individual patient data meta-analysis.
In this review, it was not possible to compare HFL4 patients with and without preserved ejection fractions, because subjects with normal ventricular ejection at this HFL were absent from the included studies. The reasons for this absence could be that the majority of patients with normal ejection fractions presented with HF of extra-cardiac origin that had not yet strongly compromised the cardiac structure. The second inclusion criterion of this review was ”studies that considered subjects without interfering pathologies”; the “interfering” pathology could be the pathology causing the HFNEF, particularly in the very old subjects with significant co-morbidities. Furthermore, there should be a different distribution of the HFNEF across HFLs, and its prevalence at HFL4 should be significantly lower than in HFL2 and HFL3. In fact, HFNEF is synonymous with diastolic HF, that is characteristic of non-acute HF stages. Consequently, the roles of advanced age and co-morbidities require special attention in the future, as both of these factors are likely to occur.
Intermediate eHFLs appeared to have a trend of decreasing accuracy toward the middle level of eHFL2 (Table 2). The low percentage of reliable eHFLs at levels 0–3 (the “gray zone” of diagnostic outcomes) likely depends on the cardiac and/or extra-cardiac BNP conditioning factors, such as arrhythmias, coronary ischemia, valvular dysfunctions, age, gender, disease prevalence, renal failure, pulmonary or systemic hypertension, anemia, and body mass index . These first-order conditioning HeFs can be more exhaustively explored trough a large, multicenter follow-up study or an individual patient meta-analysis .
In conclusion, the eF CO, which was determined using the BNP ln OR, lnmBNP and HFL values with a three-step meta-regression method, was shown to “rule in” a severe HF diagnosis. Compared to the 50% accuracy of LVEF measured with echography and the 64% accuracy of the “rule in” ROC BNP CO, the HFL CO estimation procedure by the three-steps meta-regression performed significantly better by reliably estimating approximately 78% of severe HF cases. HFL eF could therefore be helpful in the differential diagnosis of severe dyspnea from HFL4 or other acute cardiac or non-cardiac pathologies. Concerning the large “gray zone” between the negative and positive COs, further research is needed on the intermediate HFL HeFs.
Conflict of interest statement
Authors’ conflict of interest disclosure: The authors stated that there are no conflicts of interest regarding the publication of this article.
Research funding: None declared.
Employment or leadership: None declared.
Honorarium: None declared.
Paulus WJ, Tschope C, Sanderson JL, Rusconi C, Flachskampf FA, Rademakers FE, et al. How to diagnose diastolic heart failure: a consensus statement on the diagnosis of heart failure with normal left ventricular ejection fraction by the Heart Failure and Echocardiography Associations of the European Society of Cardiology. Eur Heart J 2007;28:2539–50.CrossrefGoogle Scholar
Logeart D, Saudubray C, Beyne P, Thabut G, Ennezat PV, Chavelas C, et al. Comparative value of Doppler echocardiography and BNP assay in the etiologic diagnosis of acute dyspnea. J Am Coll Cardiol 2002;40:1794–800.CrossrefGoogle Scholar
Maisel A, Mueller C, Adams K Jr, Anker SD, Aspromonte N, Cleland JG, et al. State of the art: using natriuretic peptides in clinical practice. Eur J Heart Fail 2008;10:824–39.PubMedCrossrefGoogle Scholar
Clerico A, Passino C, Emdin M. Clinical considerations and applications in cardiac diseases. Clinical applications in extra-cardiac diseases. In: Clerico A, Emdin M, editors. Natriuretic peptides. Trento: Spinger, 2006:91–160.Google Scholar
Levey AS, Coresh J, Greene T, Stevens LA, Zhang YL, Hendriksen S, et al. Chronic Kidney Disease Epidemiology Collaboration. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med 2006;145:247–54.CrossrefPubMedGoogle Scholar
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology. A proposal for reporting. J Am Med Assoc 2000;283:2008–12.Google Scholar
Higgins JP, Deeks JJ. Selecting studies and collecting data. In: Higgins JP, Green S, editors. Cochrane handbook for systematic reviews of interventions. Chichester: Wiley, 2008:151–86.Google Scholar
Prontera C, Emdin M, Zucchelli GC, Ripoli A, Passino C, Clerico A. Analytical performance and diagnostic accuracy of a fully-automated electrochemiluminescent assay for the N-terminal fragment of the pro-peptide of brain natriuretic peptide in patients with cardiomyopathy: comparison with immunoradiometric assay methods for brain natriuretic peptide and atrial natriuretic peptide. Clin Chem Lab Med 2004;42:37–44.Google Scholar
Borenstein M, Hedges LV, Higgins JP, Rothstein HR. Subgroup analyses. Meta-regression. In: Borenstein M, Hedges LV, Higgins JP, Rothstein HR, editors. Introduction to meta-analysis. Chichester: Wiley, 2009:149–203.Google Scholar
Wang MC, Bushman BJ. Random effects models in meta-analysis. Reporting the results of sensitivity analysis. In: Integrating results through meta-analysis review using SAS sofftware. Cary, NC: SAS Institute, 1999:273–346.Google Scholar
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;8476:307–10.Google Scholar
Steichen TJ. Tests for publication bias in meta-analysis. Harbord RM, Whiting P. Meta-analysis of diagnostic accuracy using hierarchical logistic regression. In: Sterne JA, editor. Meta-analysis in Stata: an updated collection from the Stata journal. Lakeway Drive: Stata Press, 2009:151–99.Google Scholar
Reitsma JB, Rutjes AW, Whiting P, Vlassov VV, Leeflang MM, Deeks JJ. Assessing methodological quality. In: Deeks JJ, Bossuyt PM, Gatsonis C, editors. Cochrane handbook for systematic reviews of diagnostic test accuracy Version 1.0.0. The Cochrane Collaboration, 2009. Available from: http://srdta.cochrane.org/. Accessed 21 October, 2012.
Bassan R, Potsch A, Maisel A, Tura B, Villacorta H, Nogueira MV, et al. B-type natriuretic peptide: a novel early blood marker of acute myocardial infarction in patients with chest pain and no ST-segment elevation. Eur Heart J 2005;26:234–40.Google Scholar
Emdin M, Passino C, Prontera C, Fontana M, Poletti R, Gabutti A, et al. Comparison of brain natriuretic peptide and amino-terminal pro-BNP for early diagnosis of heart failure. Clin Chem 2007;53:1289–97.CrossrefGoogle Scholar
Bibbins-Domingo C, Ansari M, Schiller NB, Massie B, Whooley MA. Is B-type natriuretic peptide a useful screening test for systolic or diastolic dysfunction in patients with coronary disease? Data from the heart and soul study. Am J Med 2004;116:509–16.CrossrefGoogle Scholar
Hammerer-Lercher A, Ludwig W, Falkensammer G, Müller S, Neubauer E, Puschendorf B, et al. Natriuretic peptides as markers of mild forms of left ventricular dysfunction: effects of assays on diagnostic performance of markers. Clin Chem 2004;50:1174–83.CrossrefGoogle Scholar
Redfield MM, Rodeheffer RJ, Jacobsen SJ, Mahoney DW, Bailey KR, Burnett JC Jr. Plasma brain natriuretic peptide to detect preclinical ventricular systolic or diastolic dysfunction. A community-based study. Circulation 2004;109:3176–81.CrossrefGoogle Scholar
Mueller T, Gegenhuber A, Poelz W, Haltmayer M. Biochemical diagnosis of impaired left ventricular ejection fraction – comparison of the diagnostic accuracy of BNP and NT-proBNP. Clin Chem Lab Med 2004;42:159–63.Google Scholar
Lukovicz T, Fischer M, Hense HW, Döring A, Stritzke J, Riegger G, et al. MONICA Investigators. BNP as a marker of diastolic dysfunction in the general population: importance of left ventricular hypertrophy. Eur J Heart Fail 2005;7:525–31.CrossrefGoogle Scholar
Mikkelsen KV, Bie P, Møller JE, Videbaek L, Villadsen HD, Haghfelt T. Neurohormonal activation and diagnostic value of cardiac peptides in patients with suspected mild heart failure. Int J Cardiol 2006;110:324–33.CrossrefGoogle Scholar
Nielsen OW, McDonagh TA, Robb SD, Dargie HJ. Retrospective analysis of the cost-effectiveness of using plasma BNP in screening for left ventricular systolic dysfunction in the general population. J Am Coll Cardiol 2003;41:113–20.CrossrefGoogle Scholar
Maisel AS, Koon J, Krishnaswamy P, Kazenegra R, Clopton P, Gardetto N, et al. Utility of BNP as a rapid, point-of-care test for screening patients undergoing echocardiography to determine left ventricular dysfunction. Am Heart J 2001;141:367–74.CrossrefGoogle Scholar
Conen D, Zeller A, Pfisterer M, Martina B. Usefulness of BNP and C-reactive protein in predicting the presence or absence of left ventricular hypertrophy in patients with systemic hypertension. Am J Cardiol 2006;97:249–52.CrossrefGoogle Scholar
Funck F, Jourdain P, Guillard N, Fulla Y, Sadeg N, Bellorini M, et al. Utilité du dosage du BNP aux urgences chez les patients suspects d’angor instable avec ECG normal. Arch Mal Cœur 2003;96:181–5.Google Scholar
Hamilton AJ, Swales LA, Neill J, Murphy JC, Darragh KM, Rocke LG, et al. Risk stratification of chest pain patients in the emergency department by a nurse utilizing a point of care protocol. Eur J Emerg Med 2008;15:9–15.CrossrefGoogle Scholar
Win HK, Chang SM, Raizner M, Shah G, Al Basky F, Desai U, et al. Percent change of BNP levels during treadmill exercise as a screening test for exercise-induced myocardial ischemia. Am Heart J 2005;150:695–700.CrossrefGoogle Scholar
Lubien E, DeMaria A, Krishnaswamy P, Clopton P, Koon J, Kazanegra R, et al. Utility of BNP in detecting diastolic dysfunction. Comparison with Doppler velocity recordings. Circulation 2002;105:595–601.CrossrefGoogle Scholar
Karaca I, Gülcü E, Yavuzkir M, Dağli N, Ilkay E, Ozbay Y, et al. BNP level in the diagnosis of asymptomatic diastolic dysfunction. Anadolu Kardiyol Derg 2007;7:2662–7.Google Scholar
Prontera C, Storti S, Emdin M, Passino C, Zyw L, Zucchelli GC, et al. Comparison of a fully automated immunoassay with a point-of-care testing method for BNP. Clin Chem 2005;51:1274–6.CrossrefGoogle Scholar
Rogers RK, Collins SP, Kontos MC, Zuber M, Arand P, Michaels AD. Diagnosis and characterization of left ventricular hypertrophy by computerized acoustic cardiography, BNP, and electrocardiography. J Electrocardiol 2008;41:518–25.CrossrefGoogle Scholar
Seino Y, Ogawa A, Yamashita T, Fukushima M, Ogata K, Fukumoto H, et al. Application of BNP and NT-proBNP measurements in cardiac care: a more discerning marker for the detection and evaluation of heart failure. Eur J Heart Fail 2004;6:295–300.CrossrefGoogle Scholar
Sutton TM, Stewart RA, Gerber IL, West TM, Richards AM, Yandle TG, et al. Plasma natriuretic peptide levels increase with symptoms and severity of mitral regurgitation. J Am Coll Cardiol 2003;41:2280–7.CrossrefGoogle Scholar
Catuzzo B, Ciancamerla F, Bobbio M, Longo M, Trevi GP. In patients with severe systolic dysfunction, only BNP is related to diastolic restrictive pattern. J Card Fail 2003;9:303–10.CrossrefGoogle Scholar
Clerico A, Iervasi G, Del Chicca MG, Emdin M, Maffei S, Nannipieri M, et al. Circulating levels of cardiac natriuretic peptides (ANP and BNP) measured by highly sensitive and specific immunoradiometric assays in normal subjects and in patients with different degrees of heart failure. J Endocrinol Invest 1998;21:170–9.CrossrefGoogle Scholar
Vasan RS, Benjamin EJ, Larson MG, Leip EP, Wang TJ, Wilson PW, et al. Plasma natriuretic peptides for community screening for left ventricular hypertrophy and systolic dysfunction. The Framingham heart study. J Am Med Assoc 2002;288:1252–9.CrossrefGoogle Scholar
Lim P, Monin JL, Monchi M, Garot J, Pasquet A, Hittinger L, et al. Predictors of outcome in patients with severe aortic stenosis and normal left ventricular function: role of BNP. Eur Heart J 2004;25:2048–53.CrossrefGoogle Scholar
Maeder MT, Brutsche MH, Staub D, Morgenthaler NG, Bergmann A, Noveanu M, et al. BNP and C-terminal-pro-endothelin-1 for the prediction of severely impaired peak oxygen consumption. J Intern Med 2009;265:604–15.CrossrefGoogle Scholar
Krishnaswamy P, Lubien E, Clopton P, Koon J, Kazanegra R, Wanner E, et al. Utility of BNP levels in identifying patients with left ventricular systolic or diastolic dysfunction. Am J Med 2001;111:274–9.CrossrefGoogle Scholar
Wieczorek SJ, Wu AH, Christenson R, Krishnaswamy P, Gottlieb S, Rosano T, et al. A rapid BNP assay accurately diagnoses left ventricular dysfunction and heart failure: a multicenter evaluation. Am Heart J 2002;144:834–9.CrossrefGoogle Scholar
Epshteyn V, Morrison K, Krishnaswamy P, Kazanegra R, Clopton P, Mudaliar S, et al. Utility of BNP as a screen for left ventricular dysfunction in patients with diabetes. Diabetes Care 2003;26:2081–7.CrossrefGoogle Scholar
Aspromonte N, Ceci V, Chiera A, Coletta C, D’Eri A, Feola M, et al. Rapid BNP test and Doppler echocardiography for early diagnosis of mild heart failure. Clin Chem 2006;52:1802–8.CrossrefGoogle Scholar
Aspromonte N, Feola M, Milli M, Scardovi AB, Coletta C, Carbonieri E, et al. Prognostic role of BNP in patients with diabetes and acute decompensated heart failure. Diabet Med 2007;24:124–30.CrossrefGoogle Scholar
Dieplinger B, Gegenhuber A, Haltmayer M, Mueller T. Evaluation of novel biomarkers for the diagnosis of acute destabilised heart failure in patients with shortness of breath. Heart 2009;95:1508–13.CrossrefGoogle Scholar
Zhao SQ, Hu YM, Li Q, Liu XR, Wang M, Zhang WY, et al. The clinical value of rapid assay for plasma BNP in differentiating congestive heart from pulmonary causes of dyspnoea. Int J Clin Pract 2008;62:214–20.Google Scholar
Malfatto G, Branzi G, Giglio A, Villani A, Facchini C, Ciambellotti F, et al. Transthoracic bioimpedance and BNP levels accurately indicate additional diastolic dysfunction in patients with chronic advanced systolic heart failure. Eur J Heart Fail 2010;12:928–35.CrossrefGoogle Scholar
Shah KB, Kop WJ, Christenson RH, Diercks DB, Kuo D, Henderson S, et al. Lack of diagnostic and prognostic utility of circulating plasma myeloperoxidase concentrations in patients presenting with dyspnea. Clin Chem 2009;55:59–67.Google Scholar
Zaphiriou A, Robb S, Murray-Thomas T, Mendez G, Fox K, McDonagh T, et al. The diagnostic accuracy of plasma BNP and NT-proBNP in patients referred from primary care with suspected heart failure: results of the UK natriuretic peptide study. Eur J Heart Fail 2005;7:537–41.CrossrefGoogle Scholar
Moyers B, Shapiro M, Marcus GM, Gerber IL, McKeown BH, Vessey JC, et al. Performance of phonoelectrocardiographic left ventricular systolic time intervals and BNP levels in the diagnosis of left ventricular dysfunction. Ann Noninvasive Electrocardiol 2007;12:89–97.CrossrefGoogle Scholar
Aspromonte N, Feola M, Scardovi AB, Coletta C, D’Eri A, Giovinazzo P, et al. Early diagnosis of congestive heart failure: clinical utility of BNP testing associated with Doppler echocardiography. J Cardiovasc Med 2006;7:406–13.CrossrefGoogle Scholar
Bhalla MA, Chiang A, Epshteyn VA, Kazanegra R, Bhalla V, Clopton P, et al. Prognostic role of BNP levels in patients with type 2 diabetes mellitus. J Am Coll Cardiol 2004;44:1047–52.CrossrefGoogle Scholar
Boldanova T, Noveanu M, Breidthardt T, Potocki M, Reichlin T, Taegtmeyer A, et al. Impact of history of heart failure on diagnostic and prognostic value of BNP: results from the BNP for acute shortness of breath evaluation (BASEL) study. Int J Cardiol 2010;142:265–72.CrossrefGoogle Scholar
Pizarro R, Bozzino O, Oberti PF, Falconi M, Achilli F, Arias A, et al. Prospective validation of the prognostic usefulness of BNP in asymptomatic patients with chronic severe mitral regurgitation. J Am Coll Cardiol 2009;54:1099–106.CrossrefGoogle Scholar
Maisel AS, Krishnaswamy P, Nowak RM, McCord J, Hollander JE, Duc P, et al. Breathing Not Properly Multinational Study Investigators. Rapid measurement of BNP in the emergency diagnosis of heart failure. N Engl J Med 2002;347:161–7.CrossrefGoogle Scholar
Antonini-Canterin F, Popescu BA, Popescu AC, Beladan CC, Korcova R, Piazza R, et al. Heart failure in patients with aortic stenosis: clinical and prognostic significance of carbohydrate antigen 125 and BNP measurement. Int J Cardiol 2008;128: 406–12.CrossrefGoogle Scholar
Chenevier-Gobeaux C, Delerme S, Allo JC, Arthaud M, Claessens YE, Ekindjian OG, et al. B-type natriuretic peptides for the diagnosis of congestive heart failure in dyspneic oldest-old patients. Clin Biochem 2008;41:1049–54.CrossrefGoogle Scholar
Lokuge A, Lam L, Cameron P, Krum H, de Villiers S, Bystrzycki A, et al. BNP testing and the accuracy of heart failure diagnosis in the emergency department. Circ Heart Fail 2010;3:104–10.CrossrefGoogle Scholar
Nishii M, Inomata T, Takehana H, Naruke T, Yanagisawa T, Moriguchi M, et al. Prognostic utility of BNP assessment in stable low-risk outpatients with nonischemic cardiomyopathy after decompensated heart failure. J Am Coll Cardiol 2008;51:2329–35.CrossrefGoogle Scholar
Lainchbury JG, Campbell E, Frampton CM, Yandle TG, Nicholls MG, Richards AM. BNP and NT-proBNP in the diagnosis of heart failure in patients with acute shortness of breath. J Am Coll Cardiol 2003;42:728–35.CrossrefGoogle Scholar
Arques S, Roux E, Sbragia P, Pieri B, Gelisse R, Luccioni R, et al. Comparison of BNP with left atrial enlargement by echocardiography for the diagnosis of new-onset congestive heart failure with a preserved left ventricular systolic function in the setting of longstanding hypertension. Int J Cardiol 2008;128:123–5.CrossrefGoogle Scholar
Dao Q, Krishnaswamy P, Kazanegra R, Harrison A, Amirnovin R, Lenert L, et al. Utility of BNP in the diagnosis of congestive heart failure in an urgent-care setting. J Am Coll Cardiol 2001;37:379–85.CrossrefGoogle Scholar
Chung T, Sindone A, Foo F, Dwyer A, Paoloni R, Janu MR, et al. Influence of history of acute heart failure on diagnostic performance and utility of BNP testing for acute dyspnea in the emergency department. Am Heart J 2006;152:949–55.CrossrefGoogle Scholar
Dokainish H, Zoghbi WA, Nasser M, Lakkis NM, Al-Bakshy F, Dhir M, et al. Optimal non-invasive assessment of left ventricular filling pressures: a comparison of tissue Doppler echocardiography and BNP in patients with pulmonary artery catheters. Circulation 2004;109:2432–9.CrossrefGoogle Scholar
Mastandrea P. Some heterogeneity factors afffecting the B-type natriuretic peptide outcome: a meta-analysis. Clin Chem Lab Med 2008;46:1686–95.Google Scholar
Prontera C, Zaninotto M, Giovannini S, Zucchelli GC, Pilo A, Sciacovelli L, et al. Proficiency testing project for BNP and NT-proBNP immunoassays: the CardioOrmocheck study. Clin Chem Lab Med 2009;47:762–8.Google Scholar
Di Somma S, Magrini L, Ferri E. In-hospital BNP and NT-proBNP variations are predictors of short-term and long-term outcome in acute decompensated heart failure. Crit Care 2011;15:116.CrossrefGoogle Scholar
Di Somma S, Magrini L, Tabacco F, Marino R, Talucci V, Marrocco F, et al. Brain natriuretic peptide and N-terminal Pro-B-type natriuretic peptide show a different profile in response to acute decompensated heart failure treatment. Congest Heart Fail 2008;14:245–50.CrossrefGoogle Scholar
Deeks JJ, Higgins JP, Altman DJ. Analysing data and undertaking meta-analysis. In: Higgins JP, Green S. editors. Cochrane handbook for systematic reviews of interventions. Chichester: Wiley, 2008:243–96.Google Scholar
About the article
Dr. Paolo Mastandrea is a Clinical Pathologist and Biostatistician, and Assistant Director in the Laboratory Medicine Department, Azienda Ospedaliera S. G. Moscati- Avellino (Italy). He is particularly experienced in the area of the Laboratory Emergency Medicine; other areas of his interest are the Clinical Biochemistry and Metabolism. He is author of several publications, particularly in the field of the laboratory test meta-analysis, cardiac markers, biochemistry of metabolism and laboratory test diagnostic accuracy. He is active member of the SIBioC (Itallian Society of Clinical Biochemistry) group of study in Biostatistics.
Published Online: 2012-11-05
Published in Print: 2013-06-01