Introduction
Fabry Disease (FD) is an X-linked recessive, lysosomal storage disease resulting from mutations in the GLA-encoded α-galactosidase A (α-Gal A) enzyme. The clinical manifestations of FD are a consequence of accumulation of glycosphingolipids in various tissues throughout the body with subsequent cellular dysfunction and end-organ damage. The reported prevalence of FD varies widely, although recent data suggests a prevalence estimate up to 1% amongst a referral cohort with LVH. Moreover, cardiac-limited phenotypes have been described in association with specific GLA mutations. Early recognition of FD is important because enzyme replacement therapy (ERT) with agalsidase beta has been shown to reduce the incidence of renal, cardiovascular, and cerebrovascular events. Deep-learning artificial intelligence applied on the standard 12-lead electrocardiogram (AI-ECG) has emerged as a useful tool for the early detection of a plethora of cardiovascular conditions, including HCM. Due to the shared phenotypic characteristics of FD and HCM, we sought to evaluate the ability of an AI-ECG HCM algorithm to detect potential cardiac involvement in patients with FD.
Methods
Patients diagnosed with FD at Mayo Clinic between 1997 and 2023 were identified from the electronic medical record. The inclusion criteria required patients to have a genetically confirmed FD diagnosis and a digital 12-lead ECG on record within 4 weeks of the initial FD diagnosis or within 4 weeks of latest follow-up visit. Participants were classified into three distinct groups based on their cardiac involvement over the study period: Group 1, no cardiac involvement at baseline or during follow-up; Group 2, no cardiac involvement at the time of FD diagnosis but present during follow-up; Group 3, cardiac involvement at the time of FD diagnosis. To determine cardiac involvement, manual review of ECGs, echocardiography, and cardiac magnetic resonance imaging (MRI) was conducted. Echocardiograms were specifically reviewed for left ventricular (LV) wall thickness >12 mm, sex-specific increases in LV mass index, abnormal global longitudinal LV strain, and right ventricular (RV) hypertrophy (>7 mm). Additionally, cardiac MRI reports were reviewed for evidence of reduced T1 (compared to native values for specific scanner), LV wall thickness >12 mm, RV hypertrophy, and abnormal late gadolinium enhancement in a nonischemic pattern (e.g., midmyocardial distribution). Individuals with LGE in a subendocardial distribution (ischemic pattern) and/or LGE involving only the RV insertion sites were not classified as having myocardial involvement. ECGs were reviewed for evidence of short PR interval, bundle branch block, and other conduction disease.
The AI-ECG HCM algorithm was applied to each patient’s baseline and/or follow-up ECG (as available) as previously described. The previously determined cutoff for this AI-ECG algorithm to be considered positive is 0.11 (i.e. value >0.11 suggests high probability of HCM). The AI-ECG algorithm output is expressed as a probability for that particular ECG belonging to a patient with HCM. An analysis of variance test was used to compare the mean AI-ECG probabilities among the three patient groups. Within each group, matched pair analyses were performed to assess the progression of cardiac involvement from diagnosis to latest follow-up in patients with ECG readings at both timepoints. Statistical significance was set at p < 0.05.
Results
A total of 99 patients with FD were included (Group 1, n =40; Group 2, n =31; Group 3, n =28). Table 1 outlines pertinent demographic and other characteristics. A total of 54 patients had only either a baseline or follow-up ECG while 45 patients had both. At baseline, Group 1 had a mean age of 38 years, with 53% female, and a mean LV mass index of 85 g/m². Group 2 had a mean age of 47 years, with 52% female, and a mean LV mass index of 152 g/m². The mean duration between diagnosis and cardiac involvement for patients in group 2 was 9.1 years. Group 3 had a mean age of 55 years, with 50% female, and a mean LV mass index of 185 g/m². Given that early initiation of enzyme replacement therapy (ERT) may significantly slow or even halt disease progression we assessed for prevalence of ERT use amongst the groups. In group 1, 65% were on ERT (approximately 75% patients in all groups treated with agalsidase beta, and the remaining with migalastat). The mean duration of enzyme replacement therapy in groups 1, 2, and 3 was 5.1±1.1 years, 6.2±1.0 years, and 6.3±1.0 years, respectively ( p value>0.05 amongst the 3 groups).
| Group 1 | Group 2 | Group 3 | P -Value | |
|---|---|---|---|---|
| ( n =40) | ( n =31) | ( n =28) | ||
| Age, years | 38 | 47 | 55 | 0.0002 * |
| Sex | 59% females | 51.6% females | 50% females | 0.7626 |
| ECG baseline, n | 28 | 16 | 26 | |
| ECG follow up, n | 30 | 26 | 17 | |
| On ERT (FZ%, GF%), % | 65% (76%; 24%) | 70% (76%, 24%) | 79% (74%, 26%) | 0.3282 |
| ERT duration, years | 5.1 | 6.1 | 6.3 | 0.658 |
| PR interval BL, ms | 145.1 | 146.9 | 163.8 | 0.1088 |
| PR interval FU, ms | 149.2 | 152.7 | 169.5 | 0.2077 |
| AF BL, n | 0 | 2 | 8 | 0.0003 * |
| AF FU, n | 3 | 14 | 7 | 0.0009 * |
| VT BL, n | 0 | 1 | 1 | 0.5066 |
| VT FU, n | 0 | 3 | 1 | 0.1223 |
| Non-Smoker, % | 85% | 67.7% | 75% | 0.5093 |
| CAD, % | 0% | 12.9% | 14.2% | 0.0513 |
| VHD, % | 7.5% | 29% | 17.8% | 0.0577 |
| Hypertension, % | 30% | 54.8% | 75% | 0.0011 * |
| Hyperlipidemia, % | 27.5% | 38.7% | 57% | 0.0481 * |
| Diabetes Mellitus, % | 5% | 6.4% | 21.4% | 0.0621 |
| CKD/ESRD, % | 35% | 35.4% | 39.2% | 0.9297 |
| Stroke/TIA, % | 7.5% | 29% | 10.7% | 0.0318* |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
