Skip to main content
SCMR 29th Annual Scientific Sessions Main banner
Final Program


Program-At-A-Glance

Oral Abstracts Session

Virtual Recording

Session: Oral Abstracts Clinical Session 3

Improving Arrhythmia Risk Stratification with Sex-Specific Cardiac Magnetic Resonance Modeling in Non-ischemic Cardiomyopathy

Friday, February 6, 2026
9:50 AM - 10:00 AM  
Location: Asia IV and Hall 8 (3rd Floor)
Background: Current guidelines for primary prevention ICD in NICM rely on LVEF ≤35% and NYHA class II–III, yet their predictive value remains controversial, especially in patients with LVEF 35-50%. Sex differences in LGE prevalence and remodeling are well described, yet current risk models apply uniform thresholds without accounting for sex-specific variation. We hypothesized that sex-specific integration of CMR parameters would enhance arrhythmic risk prediction in NICM.

Methods: We retrospectively identified patients with NICM and LVEF < 50% who underwent CMR at Cleveland Clinic (2003–2021). Endpoints were appropriate ICD shock for ICD patients and sustained VT/VF/SCD for non-ICD patients. Cox models were developed using either guideline features (LVEF + NYHA), basic CMR model features (adding demographics, EKG variables, and CMR LVEF, LGE presence/absence), or advanced CMR features (adding comprehensive CMR features with continuous LGE). Forward feature selection was applied for the Advance models to find subset of features that consistently optimize validation C-index. Models were evaluated across 100 random training/validation splits, mean C-index and 95% CI computed. Feature effects were summarized as log-hazard ratios, where continuous features were scaled to 1SD and leaving binary features unscaled to enable cross-feature comparison. Kaplan–Meier curves with log-rank tests were generated for model-classified risk groups.

Results: Among 789 patients (39.7% women; mean age 53.4 years; mean LVEDVi of 131.0 mL/m²; mean LVEF of 31.1%; mean LGE of 3.0%), 226 (28.4%) had significant LGE (≥2%). Over 4.34 years of follow-up, 80 patients (10.1%) reached the primary endpoint (Table 1).

Model discrimination improved from the guideline model (C-index 0.596, 95% CI 0.586–0.605) to the basic CMR model (0.635, 0.624–0.646), with the advanced CMR model achieving the highest performance, particularly with sex-specific training (0.716, 0.707–0.725) (Fig. 1.1). Improvement was most pronounced in women, where the sex-specific model reached a C-index of 0.814 (0.802–0.826) compared with 0.533 for guideline and 0.623 for the basic model.

Feature importance (Fig. 1.2) showed LGE was a key predictor in both sexes. In men, HTN and QTc were most influential. In women, Relative Wall Thickness, LAVi, and RVEDVi carried greater weight, highlighting that the key predictors in women were CMR-derived.

In Kaplan–Meier analyses (Fig.2), guideline model stratified risk in men (p=0.002) but not in women (p=0.249). Basic CMR model improved discrimination in men (p < 0.001) and succeeded in stratifying women (p=0.034). The advanced CMR models also had significant risk separation in men (p=0.003) and provided the most significant stratification in women (p < 0.001).

Conclusion: Sex-specific CMR modeling outperformed guideline and basic models, with the largest improvements in women. These findings support sex-specific CMR criteria for arrhythmic risk stratification in NICM.

Table 1: Patient Demographics and baseline clinical and CMR characteristics.

  

  

Overall 

Male-ICD 

Male-No ICD 

Female-ICD 

Female-No ICD 

Female, n (%) 

  

316 (39.7%) 

0 (0.0%) 

0 (0.0%) 

134 (100.0%) 

182 (100.0%) 

Age, mean (SD) 

  

53.4 (15.4) 

54.9 (13.6) 

50.9 (17.2) 

57.8 (12.6) 

52.1 (15.7) 

BMI, mean (SD) 

  

28.9 (6.4) 

29.7 (5.7) 

28.6 (6.0) 

29.1 (7.2) 

28.2 (7.1) 

DM, n (%) 

  

137 (17.2%) 

34 (16.0%) 

44 (16.5%) 

29 (21.6%) 

30 (16.5%) 

HTN, n (%) 

  

389 (48.9%) 

114 (53.5%) 

131 (49.1%) 

68 (50.7%) 

76 (41.8%) 

HLP, n (%) 

  

327 (41.1%) 

114 (53.5%) 

87 (32.6%) 

68 (50.7%) 

58 (31.9%) 

AF, n (%) 

  

158 (19.8%) 

41 (19.2%) 

72 (27.0%) 

16 (11.9%) 

29 (15.9%) 

NYHA Class, n (%) 

1 

53 (6.7%) 

18 (8.5%) 

12 (4.5%) 

7 (5.2%) 

16 (8.8%) 

2 

415 (52.1%) 

134 (62.9%) 

126 (47.2%) 

76 (56.7%) 

79 (43.4%) 

3 

35 (4.4%) 

13 (6.1%) 

12 (4.5%) 

5 (3.7%) 

5 (2.7%) 

4 

293 (36.8%) 

48 (22.5%) 

117 (43.8%) 

46 (34.3%) 

82 (45.1%) 

Primary Endpoint, n (%) 

  

80 (10.1%) 

41 (19.2%) 

11 (4.1%) 

20 (14.9%) 

8 (4.4%) 

Time to event (years), mean (SD) 

  

4.33 (3.29) 

4.76 (3.22) 

3.99 (3.17) 

5.04 (3.51) 

3.82 (3.26) 

LVEDVi, mean (SD) 

  

131.0 (45.7) 

158.2 (49.5) 

118.0 (36.4) 

140.9 (47.2) 

110.8 (33.5) 

LVEF, mean (SD) 

  

31.1 (10.9) 

25.0 (8.4) 

35.5 (10.5) 

26.1 (8.7) 

35.4 (10.3) 

LV Mass Index, mean (SD) 

  

73.8 (24.7) 

84.8 (22.8) 

74.2 (25.8) 

72.0 (17.0) 

61.8 (24.1) 

RVEDVi, mean (SD) 

  

88.7 (28.5) 

98.5 (29.0) 

91.0 (26.6) 

83.4 (29.2) 

77.7 (25.5) 

RVEF, mean (SD) 

  

41.8 (12.3) 

37.7 (11.7) 

41.8 (11.2) 

42.6 (13.1) 

45.9 (12.4) 

LA EF, mean (SD) 

  

41.9 (15.5) 

41.0 (15.1) 

40.5 (16.8) 

44.7 (15.3) 

42.8 (13.6) 

LAVi, mean (SD) 

  

46.5 (19.4) 

55.2 (22.4) 

42.2 (15.3) 

47.5 (18.5) 

42.0 (18.2) 

LGE Burden, mean (SD) 

  

3.0 (7.0) 

3.6 (7.6) 

2.6 (6.0) 

4.2 (9.3) 

2.1 (5.5) 

MRF (%), mean (SD) 

  

14.4 (13.4) 

14.5 (14.0) 

13.5 (12.7) 

15.3 (13.2) 

14.8 (13.8) 

Sphericity Index, mean (SD) 

  

1.7 (0.2) 

1.5 (0.2) 

1.8 (0.3) 

1.5 (0.2) 

1.7 (0.2) 

Relative Wall Thickness, mean (SD) 

  

0.5 (2.1) 

0.3 (0.1) 

0.7 (2.7) 

0.2 (0.1) 

0.6 (2.8) 

MV Annular Diameter, mean (SD) 

  

3.5 (1.2) 

3.7 (0.4) 

3.5 (2.0) 

3.4 (0.4) 

3.4 (0.5) 

QRS, mean (SD) 

  

117.4 (33.9) 

130.6 (35.0) 

109.2 (26.2) 

129.5 (33.3) 

104.9 (34.9) 

QTc, mean (SD) 

  

449.5 (41.2) 

436.1 (42.1) 

454.4 (40.0) 

445.1 (39.7) 

461.4 (38.4) 

Cockcroft-Gault Creatinine Clearance (mL/min), mean (SD) 

  

104.5 (50.6) 

106.3 (43.5) 

114.3 (57.0) 

88.4 (45.5) 

99.6 (48.4) 

NYHA >2, n (%) 

  

328 (41.2%) 

61 (28.6%) 

129 (48.3%) 

51 (38.1%) 

87 (47.8%) 

Beta Blocker, n (%) 

  

617 (77.5%) 

202 (94.8%) 

164 (61.4%) 

128 (95.5%) 

123 (67.6%) 

ACEi/ARB/ARNi, n (%) 

  

569 (71.5%) 

199 (93.4%) 

156 (58.4%) 

121 (90.3%) 

93 (51.1%) 

Aldosterone Antagonist, n (%) 

  

293 (36.8%) 

120 (56.3%) 

57 (21.3%) 

71 (53.0%) 

45 (24.7%) 

AAD, n (%) 

  

92 (11.6%) 

20 (9.4%) 

44 (16.5%) 

12 (9.0%) 

16 (8.8%) 

CRT, n (%) 

  

178 (22.4%) 

96 (45.1%) 

6 (2.2%) 

76 (56.7%) 

0 (0.0%) 

ICD, n (%) 

  

347 (43.6%) 

213 (100.0%) 

0 (0.0%) 

134 (100.0%) 

0 (0.0%) 

CrCl>45, n (%) 

  

752 (94.5%) 

206 (96.7%) 

259 (97.0%) 

115 (85.8%) 

172 (94.5%) 

Significant LGE, n (%) 

  

226 (28.4%) 

78 (36.6%) 

73 (27.3%) 

36 (26.9%) 

39 (21.4%) 

Fig. 1.1: Comparison of model discrimination (C-index) across guideline, basic CMR, and advanced CMR models in the overall cohort, men, and women. Fig 1.2: Feature importance plots displaying Cox regression coefficients (log HR per SD) for the basic sex-agnostic CMR model and the advanced sex-specific CMR models in men and women. Features in each model: Guideline: LVEF, NYHA Class. Basic CMR Model: Female, Age, BMI, AF, Beta Blocker, CrCl > 45, LVEF, QTc, NYHA > 2, CRT, AAD, Significant LGE. Advanced CMR Models: Female, Age, BMI, DM, HTN, HLP, AF, NYHA Class, LVEDVi, LVEF, LV Mass Index, RVEDVi, RVEF, LA EF, LAVi, LGE Burden 5SD, MRF (%), Sphericity Index, Relative Wall Thickness, MV Annular Diameter, QRS, QTc, CrCl > 45.
Kaplan–Meier survival curves for model-classified risk groups, shown separately for men and women. Panels display (2.1) guideline model, (2.2) basic CMR model, and (2.3) advanced CMR model.