Natriuretic peptide testing to detect left ventricular systolic dysfunction in the OxVALVE cohort: diagnostic accuracy study

Problem

Over 40 million people worldwide have a diagnosis of heart failure (HF). This clinical syndrome has a worse prognosis than common cancers. However, there is increasing evidence that HF can be prevented through intervention at a pre-clinical stage. Therefore, the identification of patients at risk of developing HF in the community is a research priority.One group at high risk of developing clinical HF are people with Left Ventricular Systolic Dysfunction (LVSD), defined as left ventricular ejection fraction (EF) below 50% on echocardiography. However, identifying people with LVSD is challenging. There is a paucity of evidence to demonstrate how accurately biomarkers such as brain natriuretic peptide (BNP) correlate with LVSD in community populations.We aimed to determine how accurately BNP detects LVSD in the Oxford Valvular Heart Disease population cohort study (OxVALVE); and the impact of differing LVSD definitions and BNP positivity thresholds on test performance. We also explored diagnostic accuracy by New York Heart Association (NYHA) class, which describes limitations during physical activity.

Approach

We conducted a retrospective diagnostic accuracy study using data collected as part of the prospective OxVALVE population cohort study. This study recruited 4009 community participants aged 65 years and older between August 2009 and May 2016, through seven general practices in Oxfordshire. Participants who had undergone both the index test (BNP) and the reference standard (echocardiography) were included (termed the OxVALVE-NP cohort). The main outcome measures of sensitivity, specificity, negative predictive value, positive predictive value were calculated for BNP compared to echocardiography.

Findings

Of 1398 patients included in the analysis, the diagnosis of LVSD (EF<50%) was confirmed in 46 participants and excluded in 1352. At a BNP threshold of ≥50 pg/ml, sensitivity was 21.74% (95% confidence interval [CI] 10.95-36.36) and specificity 96.89% (95% CI 95.82-97.75) for the detection of LVSD (EF<50%), compared to echocardiography. At BNP≥ 35 pg/ml, sensitivity was 46.67% (95% CI 21.27-73.41) and specificity 93.06% (95% CI 91.59-94.34) for the detection of EF<40%. For NYHA class II, at BNP≥ 35, the sensitivity was 66.67% (95% CI 22.28-95.67) and specificity 88.49% (84.14-91.99).

Consequences

In a well-phenotyped community cohort with a low prevalence of LVSD, BNP measurement had a low sensitivity and high specificity for detecting LVSD. Performance improved at lower BNP thresholds and with lower EF cut-offs for LVSD definition. Accuracy also improved in participants who described a slight limitation in physical exercise (NYHA II). Our results do not support widespread community screening in low prevalence settings. However, there might be some value in measuring BNP in people who describe even a subtle reduction in exercise tolerance. This group may not present to primary care so more proactive strategies to identify them might be warranted.