Daily Cardiology Research Analysis
Three impactful cardiology studies stood out: a randomized AF screening trial showing that machine learning risk selection plus 14‑day patch ECG substantially increases new AF detection; a large JACC analysis establishing allometrically correct indexing (area/height) for ascending aorta size with prognostic value; and a multicenter JACC Heart Failure study combining SCAI shock stages with machine learning phenotypes to improve early risk stratification in cardiogenic shock.
Summary
Three impactful cardiology studies stood out: a randomized AF screening trial showing that machine learning risk selection plus 14‑day patch ECG substantially increases new AF detection; a large JACC analysis establishing allometrically correct indexing (area/height) for ascending aorta size with prognostic value; and a multicenter JACC Heart Failure study combining SCAI shock stages with machine learning phenotypes to improve early risk stratification in cardiogenic shock.
Research Themes
- AI- and ML-enabled cardiovascular risk stratification
- Imaging-derived standardization and prognostic indexing
- Population screening with wearables and precision selection
Selected Articles
1. Systematic, randomized atrial fibrillation screening using detailed phenotyping with a risk prediction model combined with patch electrocardiogram in a Swedish population aged 65 years or older: the CONSIDERING-AF trial.
In a randomized, intention-to-invite trial of 2960 adults ≥65 years, machine learning risk selection plus 14-day patch ECG increased AF detection compared with standard care (3.8% vs 0.7%; RR 5.6; NNI=32). The enrichment strategy outperformed general invitations, supporting targeted screening.
Impact: Demonstrates a scalable, precision-screening pathway that substantially improves AF detection, a prerequisite for stroke prevention. Integrates ML risk models with wearable diagnostics in a randomized design.
Clinical Implications: Health systems could deploy ML-enriched invitations with 14-day patches to prioritize high-yield AF screening in adults ≥65, improving detection while conserving resources; pathways should address participation and linkage-to-treatment.
Key Findings
- AF incidence over 6 months was higher with ML risk model + invitation vs general control (3.8% vs 0.7%; RR 5.6; NNI=32).
- ML-enriched invitation detected more AF than general invitation (P < 0.001).
- No difference between general invitation and general control, highlighting the need for risk enrichment.
Methodological Strengths
- Randomized, intention-to-invite design with clear primary endpoint (6-month AF incidence).
- Integration of ML risk prediction with objective 14-day continuous patch ECG monitoring.
Limitations
- Participation rate in invitation arms was 43%, potentially limiting yield.
- Not powered for downstream clinical outcomes; comparison vs RPM control showed a nonsignificant difference.
Future Directions: Evaluate cost-effectiveness, stroke prevention impact, and linkage-to-therapy; test engagement strategies to improve participation; validate in diverse health systems.
AIMS: Atrial fibrillation (AF), often asymptomatic and underdiagnosed, is an independent risk factor for ischaemic stroke. A knowledge gap remains regarding the optimal target population and method to use for AF screening. We aimed to test whether screening for AF using a machine learning-based risk prediction model (RPM) and 14-day continuous patch electrocardiogram (ECG) (Philips ePatch) in high-risk individuals ≥ 65 years is more effective than standard care. METHODS AND RESULTS: Individuals ≥ 65 years were assigned to general or RPM cohort. The general cohort was randomized to control or invitation. In the RPM cohort, high-risk individuals, identified by RPM, were randomized to control or invitation. The primary outcome was 6-month AF incidence, analysed as intention-to-invite, comparing RPM + invitation with general + control. Of the 2960 randomized individuals, participation was 43% (632/1480) in invitation arms. Atrial fibrillation incidence was higher in RPM + invitation than in general + control arm (3.8%, 28/740 vs. 0.7%, 5/740; P < 0.001), yielding a risk ratio of 5.6, [95% confidence interval (2.2, 14.4)], and a number needed to invite of 32. Atrial fibrillation was more often detected in RPM + invitation than in general + invitation arm (1.1%, 8/740; P < 0.001), but not more often than in RPM + control arm (2.2%, 16/740; P = 0.07). No difference was found between general + invitation and general + control arms (1.1%, 8/740 vs. 0.7%, 5/740; P = 0.40). CONCLUSION: Among high-risk individuals ≥ 65 years, the combination of a machine learning-based RPM and long-term ECG recording was superior to standard care in identifying new AF cases.
2. CSWG-SCAI Stages Combined With Machine Learning-Based Phenotypes for Serial Risk Stratification in Cardiogenic Shock.
In 7,716 cardiogenic shock patients, combining SCAI stages with ML phenotypes applied serially in the first 72 hours markedly improved prognostic discrimination. Early transitions were common within 6 hours, and cardiometabolic phenotype (III) strongly predicted progression and mortality.
Impact: Provides an actionable framework to move beyond treatment intensity labels toward mechanism-informed stratification, enabling earlier targeted interventions in a high-mortality syndrome.
Clinical Implications: Use combined SCAI stage + ML phenotype early (within 6 h) to triage, escalate support, and tailor therapies; integrate into shock teams’ workflows and decision support for dynamic risk updates.
Key Findings
- Serial application showed 78% stage and 77% phenotype transitions within 6 hours, then relative stability to 72 hours.
- Subclassifying by SCAI + phenotype discriminated mortality (e.g., D-III: ~37–40% vs C-I: ~12–14%).
- Admission phenotype III (cardiometabolic) strongly increased odds of progression to SCAI D/E or death (OR ~11.4).
Methodological Strengths
- Large multicenter registry with granular serial (6–12 h) phenotyping across 72 hours.
- Integration of mechanistic ML phenotypes with established SCAI staging; outcome-anchored analysis.
Limitations
- Retrospective analysis with potential residual confounding.
- External generalizability and operational thresholds for interventions require prospective validation.
Future Directions: Prospective trials to test phenotype-guided escalation (e.g., MCS timing) on survival; embed real-time CDS integrating SCAI+phenotype; validate in non-tertiary settings.
BACKGROUND: Cardiogenic shock (CS) severity can be defined by the SCAI (Society for Cardiovascular Angiography and Interventions) stages (A to E), or machine learning-based phenotypes (I: noncongested, II: cardiorenal, III: cardiometabolic). OBJECTIVES: This study aims to evaluate sequential applicability and prognostic relevance of combining SCAI stages and ML-based phenotypes for risk stratification of patients with CS. METHODS: The authors retrospectively applied both classification systems at 6- to 12-hour intervals for the first 72 hours to patients from the multicenter CSWG (Cardiogenic Shock Working Group) registry. The primary outcome was in-hospital mortality. RESULTS: A total of 7,716 CS patients were included (admission CSWG-SCAI stages A to E: n = 1,526, n = 1,602, n = 838, n = 2,445, and n = 1,305, respectively; phenotypes I to III: n = 2,963, n = 3,266, n = 1,487, respectively). Within 6 hours from admission, CSWG-SCAI stages and phenotypes changed in 78% and 77% of patients, respectively, then remained relatively unchanged throughout the first 72 hours. Combining ML-based phenotypes with CSWG-SCAI stages to subclassify patients improved risk stratification (mortality for stages C-I: 12%-14%, C-II: 22%-26%, D-I: 21%-23%, D-II: 31%-34%, and D-III: 37%-40%). Admission phenotypes II and III strongly increased the odds of CS progression from stages A-C to D-E or death within 72 hours of admission (phenotype II: OR: 1.2 [95% CI: 0.99-1.39]; P = 0.051; phenotype III: OR: 11.4 [95% CI: 3.50-36.95]; P < 0.0001). CONCLUSIONS: Most patients with CS reached phenotype I and stage D within 6 hours after admission. Combining ML-based phenotypes with CSWG-SCAI staging may facilitate the transition from typical treatment intensity-based approaches to mechanistic classification that reflects the heterogeneity within CS populations.
3. Ascending Aortic Dimensions and Body Size: Allometric Scaling, Normative Values, and Prognostic Performance.
Across two large biobanks using AI-based aortic segmentation, conventional ratio indices (diameter/height, diameter/BSA, area/BSA) showed nonlinear scaling, whereas area/height was allometrically correct (exponent ~1) and prognostically informative without residual body size bias.
Impact: Establishes a simple, biologically consistent indexing (area/height) that removes residual body size bias and links to adverse events, with immediate implications for diagnostic thresholds and surveillance of aortopathy.
Clinical Implications: Adopt area/height indexing for ascending aorta assessment in reports and guidelines to standardize risk thresholds, improve detection of pathologic dilation across body sizes, and refine surveillance intervals.
Key Findings
- In healthy reference groups from UKB and PMBB, diameter/height, diameter/BSA, and area/BSA displayed highly nonlinear scaling.
- Area/height had allometric exponent near 1 (UKB 1.04; PMBB 1.13), supporting linear indexing without residual height association.
- Area/height showed consistent associations with adverse events in prognostic analyses.
Methodological Strengths
- Very large, independent cohorts with AI (CNN) image segmentation across MRI/CT.
- Log–log allometric modeling with external validation and prognostic assessment via Cox models.
Limitations
- Abstract truncation limits details on event definitions and follow-up duration.
- Residual confounding and modality differences (MRI vs CT) may influence measurements despite AI segmentation.
Future Directions: Define actionable area/height thresholds for intervention; test generalizability across ethnicities, ages, and device vendors; prospective validation for surveillance and surgical decision-making.
BACKGROUND: Ascending aortic (AscAo) dimensions partially depend on body size. Ratiometric (linear) indexing of AscAo dimensions to height and body surface area (BSA) are currently recommended, but it is unclear whether these allometric relationships are indeed linear. OBJECTIVES: This study aimed to evaluate allometric relations, normative values, and the prognostic performance of AscAo dimension indices. METHODS: The authors studied UK Biobank (UKB) (n = 49,271) and Penn Medicine BioBank (PMBB) (n = 8,426) participants. A convolutional neural network was used to segment the thoracic aorta from available magnetic resonance and computed tomography thoracic images. Normal allometric exponents of AscAo dimensions were derived from log-log models among healthy reference subgroups. Prognostic associations of AscAo dimensions were assessed with the use of Cox models. RESULTS: Among reference subgroups of both UKB (n = 11,310; age 52 ± 8 years; 37% male) and PMBB (n = 799; age 50 ± 16 years; 41% male), diameter/height, diameter/BSA, and area/BSA exhibited highly nonlinear relationships. In contrast, the allometric exponent of the area/height index was close to unity (UKB: 1.04; PMBB: 1.13). Accordingly, the linear ratio of area/height index did not exhibit residual associations with height (UKB: R CONCLUSIONS: Among AscAo indices, area/height was allometrically correct, did not exhibit residual associations with body size, and was consistently associated with adverse events.