This retrospective study was conducted with approval from the institutional review board, which waived the requirement for written informed consent (IRB no. MR-36-24-013077). We consecutively enrolled 150 consecutive patients with HTN who received treatment at our center between June 2020 and December 2024. Inclusion Criteria: (1) documented HTN [≥2 office measurements of systolic blood pressure (SBP) > 140 mmHg and/or diastolic blood pressure (DBP) > 90 mmHg or current antihypertensive therapy]; (2) LVEF ≥50% (derived from CMR diagnostic reports); (3) absence of secondary LVH etiologies (such as moderate-severe valvular disease, alcoholic cardiomyopathy, and acquired or inherited cardiomyopathies confirmed by clinical or paraclinical investigations); (4) no structural heart disease (coronary artery disease, rheumatic heart disease, etc.); (5) normal hepatic and renal function and thyroid hormone levels. Exclusion Criteria: (1) failure to meet inclusion criteria; (2) inadequate image quality for analysis; (3) presence of focal LGE; (4) suspected hypertrophic cardiomyopathy (HCM) with diffuse ventricular wall thickening indistinguishable from HTN. Figure 1 shows the flow chart of the study.

HTN patients were stratified into two groups based on left ventricular mass index (LVMI): HTN-LVH group (n = 71, 54 males, mean age 51.03 ± 12.50 years) and HTN-LVN group (n = 79, 50 males, mean age 56.75 ± 11.38 years). LVH was defined as LVMI >81 g/m² for men and >61 g/m² for women (15). A healthy control group matched by age and sex (n = 60, 29 males; mean age 50.42 ± 12.84 years) was concurrently enrolled, with confirmed normal cardiac structure/function and absence of HTN or other cardiovascular diseases. Patients with arrhythmia on routine electrocardiogram (ECG) or recurrent arrhythmias that required follow-up underwent 24-hour dynamic electrocardiography (DCG) monitoring. DCG monitoring identified 42 HTN patients with VA (HTN-LVH:23, HTN-LVN:19). VA were classified per the 2022 ESC Guidelines as: non-sustained ventricular tachycardia (≥3 consecutive premature ventricular contractions, rate >100 beats per minute, duration ≤30 s), sustained ventricular tachycardia (>30 s or requiring urgent intervention), or ventricular fibrillation (9). Clinical data for all participants were systematically collected.

All examinations were performed on a 3.0 T MRI scanner (Skyra, Siemens Healthineers, Germany). Using an 18-channel body matrix coil with retrospective ECG gating and respiratory compensation. Cine imaging was acquired using balanced steady-state free precession (b-SSFP) sequence with the following parameters: repetition time (TR) 39.48 ms, echo time (TE) 1.43 ms, flip angle 47°, field of view (FOV) 340 × 285 mm², slice thickness 5 mm with no gap. Imaging planes included two-chamber, three-chamber, four-chamber views, and short-axis stacks covering the LV from mitral annulus to apex. For LGE imaging, Gd-DTPA (Beijing Beilu Pharmaceutical Corporation, China) was administered intravenously at 0.05 mmol/kg followed by 20 mL saline flush. LGE images were obtained 8–10 min post-injection using a 2D phase-sensitive inversion recovery sequence. The relevant parameters are as follows: TR 750 ms, TE 2.1 ms, flip angle 20°, FOV 350 × 262.5 mm², slice thickness 5 mm with no gap, and inversion time (TI) individually adjusted between 260 and 370 ms.

Conventional cardiac function parameters were analyzed using the Syngo.via post-processing workstation (Siemens Healthineers, Germany). All cine images were imported, and the software automatically delineated endocardial and epicardial contours at end-systole and end-diastole. Following manual adjustment to exclude papillary muscles and trabeculae, parameters were normalized to body surface area (BSA). The derived indices included: LVEF, LVMI, left ventricular end-diastolic volume index (LVEDVI), left ventricular end-systolic volume index (LVESVI), cardiac index (CI). Maximum left ventricular wall thickness (LVMWT) was measured at end-diastole from short-axis cine images. LGE images were independently evaluated by two experienced radiologists to identify focal fibrotic lesions.

LV myocardial strain measurements were performed using CVI42 (version 5.11.1, Circle Cardiovascular Imaging, Calgary, Canada). Following import of the cine sequences, endocardial and epicardial contours at end-diastole were automatically identified by the software. After manual correction, the following strain and strain rate parameters were calculated: global longitudinal strain (GLS), global radial strain (GRS), global circumferential strain (GCS), global peak systolic longitudinal strain rate (sGLSR), systolic radial strain rate (sGRSR), systolic circumferential strain rate (sGCSR), global peak early-diastolic longitudinal strain rate (eGLSR), early-diastolic radial strain rate (eGRSR), early-diastolic circumferential strain rate (eGCSR) (are shown in Figure 2).

To evaluate measurement reliability, we randomly selected CMR images from 20 HTN patients and 10 healthy controls for reproducibility validation. Two radiologists, each with over 2 years of experience in CMR post-processing, independently measured LV strain parameters to assess inter-observer agreement. Intra-observer reproducibility was determined by selecting the same radiologist repeat measurements on the same 30 subjects after a 3-month interval.

Continuous variables were tested for normality with the Kolmogorov–Smirnov test and expressed as mean ± standard deviation (normally distributed) or median (interquartile range) (non-normal). Categorical variables were reported as frequencies (%). For three-group comparisons, normally distributed data were analyzed by one-way analysis of variance (ANOVA) with Bonferroni post-hoc correction, and non-normally distributed data by the Kruskal–Wallis H test. HTN patients with and without VA were compared using independent t-tests (normal data) or Mann–Whitney U-tests (non-normal data). Categorical variables were analyzed with chi-square or Fisher's exact tests. Pearson correlation evaluated associations between strain parameters and other continuous variables. Variable selection was performed using least absolute shrinkage and selection operator (LASSO) regression, with subsequent multivariate logistic regression identifying independent risk factors for VA in HTN patients. Receiver operating characteristic (ROC) curves assessed the discriminative ability of strain parameters for VA in HTN patients. Intraclass correlation coefficients (ICC) quantified intra- and inter-observer reliability of LV strain measurements. LASSO regression was performed using the glmnet package (version 4.18) in R, with all other analyses conducted in SPSS 25.0 (IBM Corp., NY). A two-tailed P < 0.05 defined statistical significance.

Comparisons of clinical characteristics and conventional LV function parameters among the control, HTN-LVN, and HTN-LVH groups are summarized in Table 1. Compared with the control group, HTN patients (HTN-LVN and HTN-LVH) had significantly higher body weight, DBP, and SBP (all P < 0.05). Smoking prevalence was higher in the HTN-LVH group than in controls (P < 0.05). Notably, LVEDVI demonstrated an initial decrease followed by an increase across the control, HTN-LVN, and HTN-LVH groups (P < 0.05), aligning with hypertensive LV remodeling. Conversely, LVMI and LVMWT showed a progressive increase with statistically significant differences across all subgroups (all P < 0.05). Age, sex, height, alcohol use, BSA, heart rate (HR), stroke volume index (SVI), CI, and LVEF did not differ significantly among the groups. Furthermore, in comparisons between the two HTN subgroups, the HTN-LVH group exhibited significantly higher SBP and a greater proportion of patients with a history of multi-drug antihypertensive therapy (Hx Multi-Drug AHT) (both P < 0.05), while no significant differences were observed in HTN duration, diabetes prevalence, or hyperlipidemia prevalence (despite 3 missing hyperlipidemia cases, sensitivity analysis classifying missing data as “Unverified” confirmed result robustness: ΔP < 0.10 vs. primary analysis; Supplementary Table 1).

The LV strain parameters among control, HTN-LVN, and HTN-LVH groups are presented in Table 2 and Figure 3. GRS, GCS, GLS, and strain rate parameters (eGRSR, eGCSR, eGLSR) exhibited a progressive decrease across groups, with statistically significant differences observed in all intergroup comparisons (all P < 0.05). However, sGRSR and sGCSR were significantly reduced in the HTN-LVH group compared to controls (both P < 0.05), while no significant differences were observed between other subgroups (all P > 0.05).

Pearson correlation analysis revealed significant linear relationships between LV global strain parameters and cardiac functional indices in all participants (Figure 4). DBP showed inverse correlations with GRS (r = −0.28), GCS (r = −0.22), and GLS (r = −0.45). Similarly, SBP correlated negatively with GRS (r = −0.32), GCS (r = −0.29), and GLS (r = −0.58). LVESVI demonstrated negative correlations with GRS (r = −0.35) and GCS (r = −0.35), while SVI showed positive correlations with GRS (r = 0.24) and GCS (r = 0.27). LVMI correlated inversely with GRS (r = −0.29), GCS (r = −0.28), and GLS (r = −0.58). LVMWT exhibited negative correlations with GRS (r = −0.21), GCS (r = −0.19), and GLS (r = −0.57). All correlations were statistically significant (P < 0.05).

As shown in Table 3, HTN patients with VA demonstrated significantly higher LVEDVI and LVESVI, lower LVEF, and a higher proportion of Hx Multi-Drug AHT compared to those without VA (all P < 0.05). No significant differences were observed between the two groups in terms of height, weight, age, sex distribution, LVMI, LVMWT, SVI, CI, or HR. Notably, patients with VA exhibited reduced myocardial strain parameters including GRS, GCS, GLS, sGRSR, sGLSR, eGRSR, eGCSR, and eGLSR compared to VA-free patients (all P < 0.05), with the exception of sGCSR which showed no significant difference.

LASSO regression was employed for variable selection to construct a logistic prediction model. Figure 5 depicts the coefficients and mean standard error for 10-fold cross-validation. Following LASSO regression, 11 variables with nonzero coefficients were retained, as shown in Figure 6 and Table 4. Collinearity diagnostics revealed variance inflation factors (VIF) ranging from 1.089 to 1.897, indicating no significant multicollinearity among the variables. Multivariate logistic regression analysis demonstrated that only GCS, BSA, and Hx Multi-Drug AHT were independent predictors of VA in HTN patients (all P < 0.05), as presented in Table 4. ROC curve analysis (Figure 7) indicated high discriminatory value of GCS for distinguishing VA in HTN patients. Optimal predictive performance was achieved at a GCS cutoff of −17.005% [area under the curve (AUC) = 0.874, 95% CI 0.808–0.941; sensitivity 78.6%, specificity 88.9%]. The combined model integrating GCS, Hx Multi-Drug AHT, and BSA yielded an AUC of 0.923 (95% CI 0.868–0.960) with 85.71% sensitivity and 88.89% specificity.

Table 5 summarizes the intra-observer and interobserver variability of LV strain parameters. The results demonstrated good agreement for all strain parameters except sGLSR. The intra-observer ICCs ranged from 0.862 to 0.927, while interobserver ICCs ranged from 0.751 to 0.900.

Several limitations should be acknowledged in this study. First, as a single-center observational study in which all enrolled patients underwent CMR, the findings are subject to selection bias and should not be directly extrapolated to the general HTN population in routine clinical management. In addition, the relatively modest sample size inherent to this retrospective investigation, along with the lack of both external validation in independent cohorts and internal cross-validation, limits the robustness and generalizability of the findings. Second, The absence of longitudinal follow-up data limits prognostic assessment. Future investigations should specifically evaluate the prognostic value of CMR-FT-derived strain parameters in HTN patients, particularly those with concomitant VA, to better define their clinical significance for risk stratification. Third, while CMR-FT demonstrates significant clinical value, its dependency on specialized software and additional post-processing workflows currently limits widespread adoption. The integration of artificial intelligence may streamline strain quantification in future applications. Fourth, while T1 mapping and ECV fraction provide valuable assessment of diffuse fibrosis, technical constraints limited their systematic application in our early cohort. Future prospective studies will incorporate these techniques to examine their interactions with myocardial deformation parameters in VA risk evaluation. Finally, although CMR-FT and echocardiographic strain parameters show good agreement in literature (36, 37), the lack of contemporaneous echocardiographic data in our retrospective cohort precludes direct comparison, requiring verification in future multimodality studies.