CA-4F version 1.4
78-84% sensitivity, 77-88% specificity
28.0 patients recommended for follow-up / cardiologist / year (Whole Database screening)
Canada (Class I MDSW), United States (CDSS)
This page was last updated 03 November 2022
CA-4F generates probabilistic estimates of transthyretin amyloid cardiomyopathy ("ATTR-CM") on technetium-99 m pyrophosphate ("Tc-99 m PYP") scintigraphy using data extracted from electronic health records. It categorizes patients as low, moderate, high or very-high pre-test likelihood of ATTR-CM on Tc-99 m PYP by comparing parameters reported in their most recent echocardiograms and ECGs to those of patients with and without the disease.
The algorithm is applied to input parameters extracted by members of the Ensho Health Clinical Operations Team as a service using the Apollo aEDC system. Patients with prior investigations for ATTR-CM, familial amyloid polyneuropathy, AL amyloidosis, sarcoidosis, Fabry disease, obstructive hypertrophic cardiomyopathy, Gaucher disease, hereditary hemochromatosis, glycogen storage disease, Hurler syndrome, Hunter syndrome and Neimann-Pick syndrome are excluded from analyses.
Very High Risk
≤ 4.0 / year
≤ 7.0 / year
≤ 10.0 / year
1,729 / year
* simulated using retrospective data
CA-4F was trained using the complete de-identified electronic health records of 199 patients of which Wild Type ATTR-CM was confirmed in 150 (75%) and ruled out in 49 (25%) by technetium-99 m pyrophosphate (Tc-99 m PYP) scintigraphymay or endomyocardial biopsy. The area under its receiver operating curve is 0.86 and it has been calibrated to achieve a sensitivity of 78% at a specificity of 77%.
The algorithm has been retrospectively validated against the de-identified electronic health records of 176 cardiology patients including 51 with Tc-99-m-PYP- and biopsy-confirmed ATTR-CM and 125 in which ATTR-CM was ruled out in which it achieved a sensitivity of 84% and a specificity of 88%. Its positive and negative predictive values in patients without prior clinical suspicion for ATTR-CM is not known.
A multi-centre post-market prospective study to evaluate its practical utility and positive predictive values when screening consecutive community cardiology patients is underway.
CA-4F analyses can be requested for individual patients ("Single Record") or whole EMR databases ("Whole Database"). Whole Database screens include all patients for which an office encounter was billed in the previous 90 days and recur every 90 days until terminated.
Requests are accepted for cardiologists who self-attest to timely access to Tc-99 m PYP scans when:
Placed through Compass ↗︎
Received from within an EMR system that supports direct ordering
Sent by email or efax (using a PDF requisition ↓)
All requests placed between the hours of 8:00am and 8:00pm Monday through Friday Eastern Time are reviewed by a member of the Clinical Operations team within 48 hours. Requests for Single Record analyses are fulfilled within 72 hours. Requests for Whole Database screens are fulfilled within 10 business days for specialists with an active integration.
Comprehensive reports ↗︎ are prepared and distributed for each patient with a moderate, high or very high pre-test likelihood of ATTR-CM. Each report includes:
A summary of the results
A clinical interpretation of the findings
A recommended course of action
Corroborating or conflicting signs and symptoms
Findings from any relevant investigations
Reports are made available online in HTML and PDF formats through Compass. Requesting physicians and office staff are notified of new results by email.
Contact firstname.lastname@example.org for pricing
CA-4F is intended for use as a decision support tool to aid qualified healthcare professionals in identifying individuals in which follow-up investigation for ATTR-CM by Tc-99 m PYP may be clinically appropriate. It is not diagnostic and does not replace the independent judgment of the treating physician. Any medical diagnosis and all decisions related to patient care and treatment choices should be based on the independent judgment of the treating physician and should take into account all information related to the patient, including without limitation, the patient and family history, direct physical examination and diagnostic tests.
The CA-4F algorithm is a process for converting input parameters to a likelihood estimate. It is deployed at the Toronto, Canada data lab of Ensho Health through the CA-4F CDL Module ("CDL Module"). The CDL Module is comprised of the CA-4F Likelihood Estimator ("Likelihood Estimator") which encodes it in software and a graphical user interface called the CA-4F Controller ("Controller"). The CDL Module is deployed exclusively at the Toronto Data Lab of Ensho Health where it is applied to the data of requesting physicians as a service. The CDL Module was developed to the ISO 13485:2016 standard for medical devices in compliance with Ensho’s Quality Management System.
The CA-4F CDL Module is classified as Clinical Decision Support Software according to Food and Drug Administration guidance FDA-2017-D-6569 ↗︎ on the interpretation of Section 3060(a) of the 21st Century Cures Act (Cures Act) since its input parameters are clinical characteristics recorded in patient charts at a single point in time (criterion 1-2) and it provides recommendations to health care professionals about diagnosis whose basis are clearly communicated (criterion 3-4).
The CA-4F CDL Module is registered as medical device software in Canada under the Medical Device Establishment License of Ensho Health (license 16208).
Rest of World
CA-4F is not yet available in Australia, European Union member states or Israel.