This was a single-center retrospective cohort study. We searched inpatients from January 2021 to December 2024 from electronic medical records in our department. Patients were included meeting the following criteria: ① diagnosed with HCM (3), which was defined echocardiographically as a maximal left ventricular wall thickness ≥15 mm in the absence of other cardiac or systemic diseases that could account for the hypertrophy; ② ≥18 years old; ③ patients with both obstructive (left ventricular outflow tract pressure gradient ≥30 mmHg) and non-obstructive forms of HCM were eligible for inclusion.

Exclusion criteria were as follows: ① previously used SGLT-2i; ② combined with severe renal insufficiency (eGFR <30 mL/min/1.73 m²); ③ echocardiographic data were not available at 6 (±1) months after discharge; ④ STR was performed during the follow-up period.

Given that this was a retrospective observational study and no patient intervention was required, informed consent from subjects could be waived, which was approved by the ethics committee. In conducting the study, we maintained a strong focus on the protection of patients’ personal interests, rights, privacy, and image rights.

Our primary endpoints were the echocardiographic parameters and NYHA class, the former including mitral annular tissue velocity (e′), early diastolic mitral inflow velocity (E)/e′, and interventricular septal thickness (IVST) at 6-month (±1 month) follow-up.

Secondary endpoint was unplanned HF readmission (based on ICD-10 codes) by June 2025. Hospitalizations where HF was not the primary driver (e.g., for arrhythmia without acute HF, pneumonia, or elective surgery) were excluded. This endpoint was adjudicated by two independent cardiologists (F.L and L.W).

Safety endpoints included adverse reactions such as hypoglycemia and urinary tract infection.

Baseline data from HCM patients were collected, including medical history, concomitant medications, laboratory parameters—such as serum creatinine, HbA1c, fasting plasma glucose (FPG), B-type natriuretic peptide (BNP)—as well as echocardiographic measures at baseline and 6 months (±1 month) after discharge. We also recorded endpoint events including HF readmission, hypoglycemia and urinary tract infection.

For patients not followed up at our institution, telephone follow-up was made to obtain follow-up information from other healthcare facilities and to ascertain whether endpoint events had occurred. Those without any post-discharge follow-up data were excluded from the study.

We employed 1:1 PSM to balance baseline characteristics between the groups. The propensity score model incorporated clinically relevant confounding factors, including sex, age, body mass index (BMI), history of hypertension, and DM. Matching was performed using a 1:1 nearest-neighbor algorithm with a caliper width set to 0.2 standard deviations of the logit propensity score. Covariate balance was assessed by calculating absolute standardized mean differences (SMD) for all baseline variables.

Numbers (frequencies) were used to describe categorical variables, mean ± standard deviation to describe normal distribution continuous variables while median (interquartile range, IQR) to describe non-normal distribution. χ2 test was used for comparison between categorical variables, while T-test was used for normally distributed continuous variables and Mann–Whitney U test for non-normal continuous variables. Multivariable analysis was performed using linear regression or a bootstrap multiple linear regression model (1,000 repetitions), depending on whether the continuous variables were normally distributed or not. All fitted models were subjected to diagnostic checks, which confirmed the adherence to key statistical assumptions: residual independence (Durbin-Watson statistic), homoscedasticity, and approximate normality of residuals. Furthermore, the absence of significant multicollinearity was verified (all variance inflation factors <5). Model fit was reported using the adjusted R².

Kaplan–Meier curve was used to evaluate differences in HF readmission events. We also performed subgroup analysis of obstructive hypertrophic cardiomyopathy (oHCM) and non-obstructive hypertrophic cardiomyopathy (noHCM) at low-risk.

We used IBM SPSS version 26.0 (SPSS Inc., Chicago, IL, USA) and R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) to perform the statistical analyses described above.

HCM symptoms primarily stem from LV diastolic dysfunction due to hypertrophy and fibrosis, a pathology shared with HFpEF (3). Initial clinical evidence from Subramanian et al. (17) in non-obstructive HCM patients and supporting preclinical data (18) indicate that SGLT-2i can improve diastolic function, aligning with our findings. Furthermore, SGLT-2i may mitigate key pathological processes like myocardial fibrosis and promote reverse remodeling (10, 13–16). The DAPA-LVH trial (15) demonstrated that dapagliflozin significantly reduced left ventricular mass in patients with diabetes and LV hypertrophy, confirming its capacity to reverse this key structural abnormality.

This study provides evidence of a multi-dimensional and physiologically coherent benefit associated with SGLT-2i therapy in HCM (17). Although the absolute changes in some echocardiographic parameters were modest, their clinical relevance must be interpreted within the context of a chronic, slowly progressive disease where reversal of cardiac abnormalities is inherently limited (1, 3). The minimal change observed in the control group aligns with this expected natural history (1). In contrast, the SGLT-2i group demonstrated a concordant pattern of improvement: a trend toward structural change with reduced IVST (ΔIVST: −1.3 vs. −0.2 mm), a significant enhancement in early diastolic relaxation velocity (Δ septal e′: 0.7 vs. 0.04 cm/s), and a marked reduction in left ventricular filling pressure (ΔE/e′: −5.1 vs. 0.4). These parameters are physiologically interlinked, suggesting that attenuation of hypertrophy (15) may facilitate improved relaxation, which in turn lowers filling pressures (16, 17). Notably, the improvement in E/e′ was substantial (Cohen's d = 0.72), a parameter with direct clinical relevance to symptoms and prognosis in HCM (3). The coherent “structure-function-hemodynamics” improvement observed in this study suggests that SGLT-2i may modulate the pathological milieu in HCM through pathways such as improved myocardial energetics and attenuated fibrosis. This pharmacological strategy intriguingly resonates with fundamental discoveries in cardiovascular remodeling—for instance, recent work highlighting the pivotal role of signaling axes like MrgD/PIM1 in regulating pathological hypertrophy and fibrosis (19). While the specific molecular targets of SGLT-2i differ, they share the core biological principle of modulating the adverse cellular and extracellular matrix environment to facilitate functional recovery (19, 20).

A nuanced interpretation of the structural data remains warranted, however. Specifically, the between-group difference in IVST reduction did not reach statistical significance after multivariable adjustment. This likely reflects the biological reality that structural reversal lags behind functional improvement (16), especially in HCM where hypertrophy is a fundamental, genetically driven pathology and a particularly challenging therapeutic target (2, 3). Therefore, the consistent trend observed here should be regarded as an important preliminary signal, providing a rationale for investigating whether longer-term SGLT-2i treatment can translate these early functional benefits into clinically meaningful structural remodeling, thereby potentially interrupting the cascade from hypertrophy to overt HF.

Thus, our findings not only support the potential disease-modifying value of SGLT-2i in HCM but also situate it within the broader therapeutic endeavor to reverse maladaptive cardiac remodeling. This holistic evidence base strengthens the therapeutic rationale for SGLT-2i in HCM beyond isolated parameter changes.

Large real-world studies support SGLT-2i's benefits in HCM. A TriNetX analysis (21) of 872 matched patients showed a 76% lower all-cause mortality and 37% fewer cardiovascular events with SGLT-2i after two years. Similarly, a Korean cohort study (22) of 4,126 HCM patients with diabetes associated SGLT-2i use with significantly reduced risks of heart failure hospitalization (18%), all-cause mortality (45%), and sudden cardiac death (50%). These cohorts had high comorbidity rates, aligning with established high-risk profiles in HCM (23).

In our study, most patients initiated SGLT-2i due to comorbid DM or HF, which align with its approved clinical indications (8). It is noteworthy that the observed significant improvements in LV diastolic function and filling pressure did not translate into a significant reduction in HF hospitalization over a median follow-up of approximately 16.3 months in this study. This finding warrants careful interpretation within the context of the study design and disease pathophysiology. First, the follow-up duration may have been insufficient. In large randomized trials (11, 24) of SGLT-2i in HFpEF (a phenotype akin to HCM-related HF), the benefit on HF hospitalization typically requires more than 2 years to fully emerge. The early functional and structural improvements captured in our study may thus represent a potential harbinger of longer-term clinical benefit. Second, HF hospitalization is a multifactorial composite endpoint, particularly influenced in HCM by complex mechanisms such as dynamic outflow tract obstruction and arrhythmias (1, 2). The statistical power of our study, given its sample size and follow-up period, was limited for detecting potentially modest differences in this endpoint. Finally, this aligns with the recognized mode of action of SGLT-2i—the biological benefits mediated through improved myocardial energetics, reduced fibrosis, etc., are progressive and cumulative (10, 13, 14). Consequently, a temporal dissociation between short-term functional improvement (a “disease-modifying” signal) and a future reduction in hard endpoints is plausible. Future studies with longer follow-up, larger sample sizes, and possibly incorporating intermediate clinical endpoints such as exercise capacity, are needed to fully elucidate the long-term clinical value of SGLT-2i in HCM.

LVOTO is a key determinant of symptoms in oHCM (25). While first-line therapies like beta-blockers or myosin inhibitors (2, 5, 6, 26) primarily aim to reduce the gradient, our study found that SGLT-2i did not confer additional reduction in LVOT gradient compared to controls. This important finding helps define the therapeutic niche of SGLT-2i: its benefit in HCM appears largely independent of mechanical relief of obstruction.

Consequently, in oHCM, SGLT-2i should be viewed not as a treatment for obstruction, but as a potential adjunctive therapy targeting the underlying myocardial pathophysiology (e.g., diastolic dysfunction, fibrosis) (10, 13, 14), particularly in patients with concomitant HF or DM. For the substantial population with noHCM, where specific disease-modifying drugs are scarce, the improvements in diastolic function and structure observed in our study suggest SGLT-2i may hold particular promise, warranting dedicated investigation.

Arrhythmias, particularly AF, are major complications in HCM, driving HF and stroke risk (1, 2). The SHaRe registry (N = 4,591) reported high incidence rates of AF (20%) and ventricular arrhythmias (6%) in HCM cohort (27, 28). While guidelines recommend anticoagulation for all HCM patients with AF, SGLT-2i has been associated with a 26% reduction in ischemic stroke (26), though a large meta-analysis found it did not significantly lower AF incidence (29).

Ventricular arrhythmias, a leading cause of sudden cardiac death (SCD), stem from structural and electrical remodeling in HCM (1). SGLT-2i may mitigate this risk by attenuating fibrosis, reversing adverse remodeling, and stabilizing electrical activity through mechanisms such as improved myocardial energetics and calcium handling (10, 13–15, 30). Clinical trials in HF populations have demonstrated that SGLT-2i significantly reduces ventricular arrhythmias (31, 32), including sustained ventricular tachycardia/ventricular fibrillation recorded in patients with ICDs (32).

Critically, a recent meta-analysis (33), extended this benefit to SCD prevention. This meta-analysis by Matteucci et al., encompassing 58,569 patients across 8 RCTs, demonstrated that SGLT-2i significantly reduced the risk of SCD (OR: 0.82; 95% CI: 0.72–0.94; P = 0.0104) in broad populations with DM, HF, or CKD, thereby expanding the spectrum of their cardiovascular benefits. The prevention of SCD remains a major unmet clinical challenge, especially in individuals without high-risk features amenable to device therapy, as pharmacological strategies specifically approved for SCD primary prevention are currently lacking. In this context, the emerging evidence for SGLT-2i (33), alongside their proposed mechanisms of attenuating fibrosis, reversing adverse remodeling, and potentially stabilizing cardiomyocyte electrophysiology (10, 14), represents a paradigm-shifting advance in arrhythmic risk management. Although not yet studied specifically in HCM, SGLT-2i represents a particularly promising therapeutic option to address the residual risk of SCD in this population.

SGLT-2i demonstrate a favorable safety profile in heart failure and diabetes populations. While known class effects include urinary tract infections and diabetic ketoacidosis, a large meta-analysis (34) confirmed no significant increase in hypoglycemia or fracture risk. Real-world data from the TriNetX study (19) and our findings support its overall safety and good tolerability in HCM patients with clinical indications. Expert consensus (35) emphasizes patient evaluation and education prior to initiation.

This study provides real-world evidence for the benefit of SGLT-2i in HCM using conventional clinical imaging and functional biomarkers. Looking forward, advancing this field will require a deeper mechanistic understanding and dynamic monitoring of the myocardial microenvironment. Emerging biosensing technologies—such as electrochemical methods for the real-time detection of mitochondrial oxidative stress markers (36)—exemplify the next generation of tools for precise risk stratification and therapy monitoring in cardiomyopathy. Integrating such precision biomarkers with traditional endpoints in future research is a crucial step toward personalized and dynamic management of HCM.

Despite the findings presented, this study has several limitations that should be acknowledged: (1) Selection bias: As a single-center retrospective analysis, the study included a high proportion of patients with multiple comorbidities, many of whom were already on guideline-directed therapies for HF or diabetes. This may limit the generalizability of the results, particularly to low-risk HCM populations, which were underrepresented in our cohort. (2) Short follow-up duration: The relatively short follow-up period precludes definitive conclusions regarding the long-term efficacy and safety of SGLT-2i in HCM. (3) Unmeasured medication adherence: As a retrospective study, direct monitoring of adherence (e.g., via pill count) was not feasible. Adherence was assessed indirectly using available clinical data, including the frequency of prescription refills in outpatient records and physician documentation of patient-reported medication intake during follow-up visits. (4) Residual confounding: Despite statistical adjustments, unmeasured or unknown confounders inherent to observational designs may affect the results. (5) Data completeness: Retrospective data collection is susceptible to missing or inconsistently recorded information, potentially introducing bias. (6) Limited power for subgroup analyses: The small sample size in certain subgroups (e.g., low-risk patients) reduces the reliability of subgroup-specific conclusions.