Acarix reports 92% specificity for HFrEF algorithm in 218-patient feasibility study

Acarix AB (Nasdaq First North Premier Growth Market: ACARIX) has disclosed peer-reviewed exploratory results showing that seismocardiography technology underlying its CADScor System may be able to identify heart failure as well as its existing coronary artery disease application. The Aalborg University-led study enrolled 218 patients, including 198 referred from general practice with suspected heart failure and 20 additional patients with confirmed heart failure with reduced ejection fraction. Researchers developed an AnyHF-score for any form of heart failure and an HFrEF-score focused on reduced-ejection-fraction disease, comparing performance with established clinical diagnostics including NT-proBNP. The results are potentially commercially significant for Acarix, but the heart-failure algorithm remains investigational and is not covered by the CADScor System’s current FDA indication.

The strongest result came from the HFrEF algorithm. It produced an area under the receiver-operating-characteristic curve of 92.9% compared with 94.7% for NT-proBNP, while specificity reached 92% compared with 68.8% for the blood test. Positive predictive value was 57.7% versus 27.1%, meaning the acoustic algorithm generated substantially fewer false-positive classifications in this study population.

Why does higher specificity matter if NT-proBNP had slightly better overall AUC?

A screening or triage tool is useful not only when it identifies disease but when it helps clinicians avoid unnecessary downstream testing in people who do not have the condition. NT-proBNP can be elevated for reasons other than heart failure, including age and other medical conditions, creating referrals for echocardiography that eventually prove negative.

The Acarix HFrEF-score’s 92% specificity compared with 68.8% for NT-proBNP suggests fewer patients without HFrEF were incorrectly classified as positive in this exploratory cohort. The positive predictive value more than doubled from 27.1% with NT-proBNP to 57.7% with the acoustic score.

That does not establish that CADScor should replace the blood test. NT-proBNP delivered slightly higher overall AUC, and diagnostic performance depends heavily on patient population, disease prevalence and thresholds.

The more plausible future role could be triage. A rapid acoustic measurement might help clinicians decide which patients need urgent echocardiography or additional testing, potentially working alongside rather than replacing biomarkers.

What did the AnyHF-score reveal about potentially avoidable referrals?

The broader AnyHF-score detected heart failure with 90.9% sensitivity and an 87.5% negative predictive value. Researchers also found that 71 of 139 referred patients who ultimately did not have heart failure were classified by the acoustic approach as minimal risk.

That represents approximately 51% of the non-heart-failure referral group. If a similar result were reproduced prospectively, an acoustic triage tool could potentially reduce a meaningful number of echocardiograms ordered for patients who ultimately do not have heart failure.

The word “potentially” matters. This feasibility study did not randomize clinics to use CADScor versus standard care and measure how many imaging procedures were actually avoided, so a 51% minimal-risk classification cannot be translated directly into a 51% reduction in echocardiography.

Health-economic validation would therefore require a separate workflow study. Physicians would need evidence that withholding or delaying imaging in acoustically low-risk patients does not miss clinically important disease.

Why can Acarix not commercialize the heart-failure algorithm immediately?

The existing CADScor System has FDA De Novo clearance and CE marking for use as a diagnostic aid in symptomatic patients suspected of stable coronary artery disease. Its current intended use does not include heart-failure diagnosis.

Acarix itself describes the new study as exploratory and says the algorithm requires external validation in larger independent cohorts. The company expects a larger validation programme would also be needed before seeking FDA approval for the new application.

This creates both regulatory risk and commercial optionality. A successful new indication could allow the same underlying hardware platform to address another large cardiology workflow without requiring Acarix to develop an entirely unrelated device.

Failure to reproduce the result would leave CADScor’s existing coronary-disease business unchanged. That asymmetry makes the heart-failure programme potentially valuable but still far from a booked revenue opportunity.

How important could a second indication be for Acarix’s current financial scale?

Acarix remains a small commercial medtech company. Second-quarter revenue was approximately SEK2.31 million, up 102% from SEK1.14 million a year earlier, while the company recorded a net loss of approximately SEK9.48 million.

United States revenue increased more than 200% to around SEK1.9 million during the quarter, and the company delivered 29 CADScor systems, including 26 in the United States. Acarix also reported a June cash position of roughly SEK29 million and continued to prioritize higher utilization of installed devices.

Those numbers show why indication expansion could matter disproportionately. A larger addressable clinical workflow can increase the potential utilization of each installed platform, which is particularly valuable when consumables and repeat testing contribute recurring revenue.

The company first needs the current CAD business to scale more meaningfully. A new heart-failure indication would add opportunity, but development and regulatory spending can also increase cash requirements before incremental revenue appears.

Why should investors be cautious about extrapolating a 218-patient feasibility study?

The study was conducted at two centers and included a relatively small population, while 20 confirmed HFrEF patients were added separately to strengthen algorithm development. That makes the dataset appropriate for feasibility work but different from the broad, prospectively enrolled population needed to support general clinical use.

Machine-learning algorithms can also perform better in development cohorts than in independent external datasets if they capture characteristics specific to the original sample. External validation is therefore essential before the reported specificity and positive predictive value can be assumed to transfer to other health systems.

The study was also funded by Acarix, which does not invalidate the peer-reviewed result but reinforces the importance of independent replication. The authors themselves explicitly describe the work as exploratory.

The commercially meaningful sequence is therefore clear: independent validation, prospective workflow testing, regulatory clearance and then reimbursement and physician adoption. Skipping any of those stages would overstate how close the heart-failure opportunity is to revenue.

What would turn the CADScor heart-failure signal into a meaningful second franchise?

A larger external study reproducing approximately 90% specificity for HFrEF would be an important first step. Acarix would then need a prospective trial designed around the exact algorithm and intended-use population proposed to regulators.

FDA clearance or approval for the heart-failure indication would create the legal commercial pathway in the United States. Reimbursement would then determine whether physicians have a practical financial reason to incorporate the test into routine evaluation.

Finally, utilization has to increase. A diagnostic platform creates stronger economics when installed devices perform more reimbursed tests rather than simply increasing the number of boxes shipped to clinics.

The feasibility study gives Acarix a credible new research direction. It does not yet give the company a heart-failure business, and that distinction is precisely what makes the next validation study important.

Key takeaways on Acarix’s heart-failure feasibility study

• The study enrolled 218 patients.
• Acarix’s AnyHF-score achieved 90.9% sensitivity.
• The AnyHF-score had an 87.5% negative predictive value.
• The HFrEF-score produced 92.9% AUC versus 94.7% for NT-proBNP.
• HFrEF specificity was 92% versus 68.8% for NT-proBNP.
• Positive predictive value was 57.7% versus 27.1%.
• The algorithm classified 71 of 139 non-HF referred patients as minimal risk.
• The study was exploratory and conducted at two centers.
• The heart-failure algorithm is not currently FDA-approved.
• Acarix reported second-quarter revenue of SEK2.31 million and a SEK9.48 million net loss.