This individual patient data meta-analysis concluded that use of age-stratified NT-pro-BNP decision limits considerably improved the performance of the test for ruling out heart failure in symptomatic patients. However, the data suggested that results with the new thresholds were very similar to those obtained with existing thresholds so these conclusions do not appear reliable.

To evaluate the use of age-related decision limits for N-terminal pro-B-type natriuretic peptide (NT-proBNP) for ruling out suspected systolic dysfunction in symptomatic patients in primary care.

PubMed and EMBASE were searched (dates not reported), researchers in the field were contacted and reference lists of retrieved papers were screened. No publication restrictions were applied.

Studies that evaluated the Roche Diagnostics NT-pro-BNP assay for the Elecsys immunoassay analyser in patients derived from primary or community care and that reported echocardiographic estimation of ejection fraction (reference standard) were eligible for inclusion. Studies had to be able to supply individual patient data to be included.

Included studies were conducted in UK and elsewhere in Europe, New Zealand and USA. The median age of included patients was 62 years (interquartilee range 54 to 72 years) and 40% were male. The prevalence of systolic dysfunction (left ventricular ejection fraction <0.40) was 18% (range 2% to 50%). The proportion of patients with ischaemic heart disease ranged from 8% to 62%, the proportion with hypertension ranged from 21% to 57% and the proportion with diabetes mellitus ranged from 4% to 31%.

The authors did not state how studies were selected for inclusion.

The authors report neither a formal quality assessment of the included studies nor checking of individual patient data and resolution of any discrepancies with the original investigator.

Individual patient data on NT-proBNP measurements, demographics, ejection fraction and clinical data were obtained for each included study. An ejection fraction of less than 40% was considered as indicative of left ventricular systolic dysfunction.

Data sets were analysed individually and then pooled. Individual receiver operating characteristic (ROC) curves were calculated for each study, data were then pooled across studies and the analyses repeated. Separate ROC curves were estimated based on age group and gender and the optimal threshold selected based on a specificity of 50%. Estimates of accuracy were calculated based on these thresholds and based on thresholds recommended in European and United States guidelines (<125ng/L at all ages, <400ng/L at all ages and <125ng/L in those aged up to 75 years and >450 in those aged over 75 years). The area under the curve (AUC) together with 95% confidence intervals (CIs) was used to quantify the overall accuracy of the test. Estimates of sensitivity, specificity and negative predictive value together with 95% CIs were reported.

Ten studies (5,508 participants) were included.

AUCs for the individual studies ranged from 0.83 to 0.97. The overall summary AUC was 0.89 (95% CI 0.88 to 0.90). Optimised age-dependent thresholds for a specificity of 50% were 41ng/L in people aged under 50 years, 73ng/L in people aged 50 to 75 years and 223ng/L in people aged over 75 years. These values were rounded to 50ng/L, 75ng/L and 250 ng/L. Based on these thresholds, the sensitivity and specificity of NT-proBNP in patients aged up to 50 years were 99.2% (95% CI 95% to 100%) and 57.2% (53% to 61%), in patients aged 50 to 75 years were 95% (95% CI 93% to 96%) and 51% (49% to 53%) and in those aged over 75 years were 88% (95% CI 84% to 92%) and 54% (95% CI 50% to 57%). Results for other thresholds recommended in guidelines were reported. These showed that sensitivity was generally lower although specificity was higher compared with the optimised age-dependent thresholds. Negative predictive values based on a pre-test prevalence of 18% were reported. These ranged from 92% to 99.7% for the age standardised thresholds and from 91% to 98% based on existing guidelines.

Exclusion of asymptomatic patients or those not receiving treatment with a vasoactive drug did not affect the conclusions. Division according to gender gave slightly different AUC but confidence intervals overlapped.

In a large population of patient in primary care, use of age-stratified NT-pro-BNP decision limits considerably improved performance over current standards, with an excellent negative predictive value for exclusion of reduced left ventricular dysfunction.

The review addressed a clear question. Inclusion criteria were defined. It would have been helpful if the authors had provided information on how many, if any, studies were excluded where they could not obtain individual patient data. Details on the search were limited with no details on search terms or dates and so the possibility of missing studies could not be ruled out. Details on the review process were not reported and so it was not possible to determine whether appropriate steps were taken to minimise bias and errors. Study quality was not formally assessed and so the risk of bias in the included studies was unclear. The robustness of the evidence was unclear as the authors did not report whether they checked the individual patient data.

It appeared that a sophisticated analysis based on individual patient data was carried out, but lack of details on exact methods used made it difficult to determine the robustness of the analysis. The authors' conclusions appeared somewhat strong and did not appear to be directly supported by the data. Although the negative predictive values for the age-derived thresholds were high, these did not differ greatly from those obtained using the thresholds recommended in existing guidelines.

The authors' conclusions regarding their age-derived thresholds do not appear to be reliable, although their recommendation that this test can be used to exclude reduced left ventricular dysfunction appears reliable.

Practice: The authors stated that NT-proBNP had sufficient supporting evidence to be implemented as an initial test for the presence of significant reduced left ventricular systolic function in patients with symptoms that suggested heart failure; further diagnostic tests such as echocardiography should be reserved for patients with elevated levels or continuing strong suspicion of reduced left ventricular systolic function. Decision values of 50ng/L (age <50 years), 75ng/L (50 to 75 years) and 250ng/L (>75 years) could be recommended as optimised rule-out thresholds.

Research: The authors did not report any implications for research.

Not stated.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.