Design: Prospective diagnostic accuracy study with invasive reference standard and international multicentre validation
Population: 482 patients with chronic unexplained dyspnoea; HFpEF confirmed invasively in 386, non-cardiac dyspnoea in 96
Primary outcome: Sensitivity and accuracy for invasively confirmed HFpEF
Effect: Current approach: sensitivity 55–60%, accuracy 61–67%. Two-threshold scheme with left atrial compliance: sensitivity 95–99% among definitively classified patients; invasive testing needed in ~30% rather than ~60%

Patients with chronic unexplained dyspnoea underwent invasive haemodynamic exercise testing with simultaneous echocardiography; 386 of 482 had heart failure with preserved ejection fraction. Using the H2FPEF, HFA-PEFF or HFpEF-ABA scores with exercise echocardiography as currently recommended, sensitivity was 55–60% and accuracy 61–67%. On those numbers, a normal non-invasive study means very little.

Adding resting left atrial compliance — left atrial reservoir strain divided by E/e' — raised sensitivity to 84–85% but pushed the false-positive rate to 31–43%. The fix was to stop using one threshold. With separate cut points optimised in opposite directions (exercise E/e' ≥ 13.8 or resting left atrial compliance ≤ 1.6% to rule in; both exercise E/e' < 7.2 and compliance > 4.4% to rule out), sensitivity among definitively classified patients reached 95–99%, and the share of patients still needing invasive exercise testing fell from about 60% to about 30%. The findings replicated in an international validation cohort.

The honest conclusion the authors draw is the one worth carrying into clinic: exercise echocardiography cannot replace invasive testing. It can sort patients into rule-in, rule-out, and a genuinely indeterminate middle who should go on to catheterisation rather than be discharged with a normal-looking echo.

Report left atrial reservoir strain alongside E/e' on any exercise echo done for unexplained dyspnoea.
Use two thresholds, not one — a value between the rule-in and rule-out cut points is an indeterminate result, not a negative one.
Tell the patient with an indeterminate study that invasive exercise testing is the next step, rather than that their echo was normal.
Audit whether your lab's strain software is validated for reservoir strain before adopting the published cut points.
In centres without invasive exercise haemodynamics, this framework at least identifies who needs referring rather than reassuring.

Why it matters

A normal exercise echo has been treated as reassurance in suspected HFpEF, and at 55–60% sensitivity it is not.

The statistics, in plain English

Sensitivity of 95–99% sounds close to perfect, but it applies only to patients the two-threshold scheme classified definitively — about 70% of them. The remaining 30% have no answer from the echo at all, which is the point: the gain comes from admitting uncertainty rather than from a better single cut point.