Diagnostic accuracy of natriuretic peptide screening for left ventricular systolic dysfunction in the community: systematic review and meta‐analysis

AIMS: Heart failure (HF) is a global health burden and new strategies to achieve timely diagnosis and early intervention are urgently needed. Natriuretic peptide (NP) testing can be used to screen for left ventricular systolic dysfunction (LVSD), but evidence on test performance is mixed, and international HF guidelines differ in their recommendations. Our aim was to summarize the evidence on diagnostic accuracy of NP screening for LVSD in general and high‐risk community populations and estimate optimal screening thresholds. METHODS: We searched relevant databases up to August 2020 for studies with a screened community population of over 100 adults reporting NP performance to diagnose LVSD. Study inclusion, quality assessment, and data extraction were conducted independently and in duplicate. Diagnostic test meta‐analysis used hierarchical summary receiver operating characteristic curves to obtain estimates of pooled accuracy to detect LVSD, with optimal thresholds obtained to maximize the sum of sensitivity and specificity. RESULTS: Twenty‐four studies were identified, involving 26 565 participants: eight studies in high‐risk populations (at least one cardiovascular risk factor), 12 studies in general populations, and four in both high‐risk and general populations combined. For detecting LVSD in screened high‐risk populations with N‐terminal prohormone brain natriuretic peptide (NT‐proBNP), the pooled sensitivity was 0.87 [95% confidence interval (CI) 0.73–0.94] and specificity 0.84 (95% CI 0.55–0.96); for BNP, sensitivity was 0.75 (95% CI 0.65–0.83) and specificity 0.78 (95% CI 0.72–0.84). Heterogeneity between studies was high with variations in positivity threshold. Due to a paucity of high‐risk studies that assessed NP performance at multiple thresholds, it was not possible to calculate optimal thresholds for LVSD screening in high‐risk populations alone. To provide an indication of where the positivity threshold might lie, the pooled accuracy for LVSD screening in high‐risk and general community populations were combined and gave an optimal cut‐off of 311 pg/mL [sensitivity 0.74 (95% CI 0.53–0.88), specificity 0.85 (95% CI 0.68–0.93)] for NT‐proBNP and 49 pg/mL [sensitivity 0.68 (95% CI 0.45–0.85), specificity 0.81 (0.67–0.90)] for BNP. CONCLUSIONS: Our findings suggest that in high‐risk community populations NP screening may accurately detect LVSD, potentially providing an important opportunity for diagnosis and early intervention. Our study highlights an urgent need for further prospective studies, as well as an individual participant data meta‐analysis, to more precisely evaluate diagnostic accuracy and identify optimal screening thresholds in specifically defined community‐based populations to inform future guideline recommendations.