This was a retrospective study of patients with HF who underwent successful CTO PCI at the First Affiliated Hospital of Zhengzhou University, Henan, China, between January 2016 and August 2019. The inclusion criteria were as follows: (1) at least one CTO lesion with a diameter of >2 mm located in the proximal or middle segment of the epicardial artery; (2) LVEF <50%, as confirmed by echocardiography, single-photon emission computed tomography (SPECT), or cardiovascular magnetic resonance (CMR); and (3) symptomatic angina and/or a functional ischemia test result. The exclusion criteria were as follows: (1) only medical therapy was administered to the patients, (2) CTO located in a branch vessel or a distal lesion, (3) CABG history, (4) an ST-segment elevation acute myocardial infarction (MI) within 48 h, and (6) the presence of a malignant tumor. The Scientific Research and Clinical Trial Ethics Committee of the First Affiliated Hospital of Zhengzhou University approved this study and waived the need to obtain informed consent from the patients, considering the retrospective nature of the study.

The predictors included in our study were as follows: demographic data (e.g., age and sex), comorbidities (e.g., hypertension, diabetes mellitus, atrial fibrillation, and renal insufficiency), laboratory results (e.g., fasting blood glucose, low-density lipoprotein, blood urea nitrogen, and serum creatinine), echocardiographic parameters (e.g., left ventricular ejection fraction, left ventricular end-diastolic dimension, and stroke volume), angiographic and procedural details (e.g., three-vessel disease, left main disease, and multiple CTOs), and medications used at discharge (e.g., statins, diuretics, aldosterone antagonists, and digoxin) at baseline. We performed the first echocardiography on admission. Similarly, we recorded the first laboratory test results within 48 h of admission. The primary endpoint was the occurrence of HFiEF 1 year after successful CTO PCI.

CTO was defined as a complete luminal obstruction with no antegrade flow or with antegrade or retrograde filling by collateral circulation for an estimated duration of at least 3 months, based on previous clinical data. HF was defined as an LVEF of <50% with dyspnea or equivalent symptoms. CTO PCI success was defined as the opening of CTO lesions and stent placement without immediate angiographic complications in the first or any subsequent attempts. For patients with multiple CTO lesions, CTO PCI success was defined as the opening of all CTO lesions. HFiEF was defined as an absolute LVEF improvement of ≥10% at 1 year. For those with a baseline LVEF of <40%, the LVEF had to improve by up to 40%.

All statistical analyses were carried out using R software and SPSS 22.0 (IBM Corp., Armonk, NY, USA). The mean ± standard deviation (SD) or median (interquartile range; IQR) has been used to express continuous variables. Categorical data are reported as counts or percentages. For data dimensionality reduction and variable selection, least absolute shrinkage and selection operator (LASSO) regression analysis was used. The lambda values were chosen after a 10-fold cross validation. Subsequently, multivariate logistic regression was used to identify independent factors influencing HFiEF. Then a nomogram was constructed. The concordance index (C-index) was calculated to assess the discriminative ability of the model via bootstrap resampling 1,000 times, and a receiver operating characteristic curve (ROC) was constructed. Calibration of the model was assessed according to a calibration curve, with 1,000 resamplings. A decision curve analysis (DCA) was performed to estimate the clinical benefit of the model. A two-tailed p < 0.05 was considered to denote statistical significance.

Figure 1 illustrates the study flowchart. A total of 638 patients diagnosed with HF underwent successful CTO PCI at our center between January 2016 and August 2019. One-year echocardiographic outcomes were available for 465 patients who were included in the final analysis. Among the 465 patients, 165 (35.5%) had HFiEF. These subjects were divided into two groups: HFiEF (n = 165) and non-HFiEF (n = 300) groups. The clinical characteristics of the two groups are summarized in Table 1. Compared with the non-HFiEF group, the HFiEF group had younger patients and a lower prevalence of previous myocardial infarction (MI), lower levels of N-terminal precursor B-type diuretic peptide (NT-proBNP), smaller left ventricular end-diastolic diameter (LVEDD), higher levels of estimated glomerular filtration rate (eGFR) and total cholesterol, as well as higher sodium glucose cotransporter 2 inhibitors (SGLT2i) use.

A total of 53 associated variables, including demographic data, past medical history, angiographic and procedural characteristics, echocardiographic parameters, laboratory results, and medications at discharge, were initially entered into the LASSO regression algorithm by 10-fold cross validation to identify the prognostic factors of HFiEF. Five potential variables with non-zero coefficients were selected: age (coefficient, −0.0080), previous MI (coefficient, −0.2080), eGFR (coefficient, 0.0007), LVEDD (coefficient, −0.0199), and SGLT2i (coefficient, 0.6890) (Figure 2A). Figure 2B depicts the changes in the LASSO coefficients.

The five predictors selected by LASSO were screened using multivariate logistic regression analysis to determine the independent factors predicting HFiEF at 1 year. Four variables were included in the final model. Table 2 shows the β-coefficients, standard error (SE), odds ratios (OR), 95% confidence intervals (CI), and p-values for each of the variables: age (OR: 0.969; 95% CI: 0.952–0.988; p = 0.001), previous MI (OR: 0.533; 95% CI: 0.357–0.796; p = 0.002), LVEDD (OR: 0.940; 95% CI: 0.910–0.972; p < 0.001), and SGLT2i (OR: 5.634; 95% CI: 1.756–18.080; p = 0.004).

A nomogram was constructed to predict HFiEF 1 year after successful CTO PCI in HF patients by incorporating the three predictive factors (age, previous MI, and LVEDD). The details of the scoring model are presented in Figure 3. The presence or level of each predictive factor was assigned one point, and the points for each individual factor were summed up to obtain the total.

We further assessed the discrimination and calibration of the model. The nomogram's C-index for predicting postoperative HFiEF was 0.666 (95% CI, 0.613–0.719) and was corrected to 0.656 by bootstrapping validation, which showed a moderate predictive accuracy. The ROC curves of the model and each index are presented in Figure 4A. The pooled area under the ROC of the nomogram was 0.666 (95% CI, 0.613–0.719), and the AUC was 0.582 (95% CI, 0.528–0.636) for age, 0.577 (95% CI, 0.530–0.624) for previous MI, 0.590 (95% CI, 0.535–0.646) for LVEDD, and 0.533 (95% CI, 0.511–0.554) for SGLT2i. Moreover, we compared the AUC between the nomogram and each predictive factor. The AUC of nomogram was significantly higher than that of age (p = 0.0028), previous MI (p = 0.0007), LVEDD (p = 0.0053), and SGLT2i (p < 0.0001) (Figures 4B–E). On the other hand, the calibration curve revealed that the predicted and actual probabilities for predicting HFiEF were in good agreement (Figure 4F).

DCA was performed to determine the net clinical benefit of the predictive model. It showed that compared to each factor, the nomogram predicted the occurrence of HFiEF 1 year after successful CTO PCI in HF patients more accurately (Figure 5).

In CTO patients with HF, prior studies have revealed that successful CTO PCI contributes to a statistically significant improvement in LVEF (15). However, the relationship between age and HFiEF in subjects with CTO and HF is unclear. Consistent with numerous reports, the current study identified age as an independent predictor of HFiEF in patients with HF (16–18). Li et al. reported that age was a powerful independent predictor of HFiEF over 6 months of follow-up among 447 patients with HF (16). Similarly, Park et al. found that patients with HFiEF were younger than those without (18). Furthermore, animal experiments suggested that aging animals showed persistent tissue fibrosis and cardiac dysfunction with impaired reverse remodeling, unlike younger animals that experienced complete cardiac function recovery (19). Older HF patients carry a higher risk of comorbidities such as diabetes mellitus, hypertension, and anemia, and are prone to significant calcification and multiple and diffuse lesions (11, 20). In addition, functions and physiological processes degrade to different degrees with age. The aging myocardium is more susceptible to pathological damage due to decreased mitochondrial function and myocardial regeneration (21–23).

MI is characterized by myocardial damage and apoptosis. MI triggers neuroendocrine regulations and an inflammatory response, resulting in changes in the morphological structure of the myocardium, ultimately affecting diastolic and systolic function (24, 25). Wilcox et al. determined that patients who suffered from MI were far less likely to experience HFiEF at 2 years (26). Choi et al. revealed that CTO PCI yielded favorable outcomes in non-MI HF patients (27). A good amount of dysfunctional, yet potentially salvageable myocardium may exist in these patients, thus being associated with the recovery of LV function after successful CTO revascularization and extensive stenotic lesion removal. However, for patients with MI, a large infarct size could result in adverse LV remodeling, leading to heart failure. Moreover, cardiac function recovery is unlikely to occur in patients with irreversible damage from MI (28). Importantly, the infarct size is closely related to the occurrence of adverse remodeling and functional impairment. Small infarcts are insufficient to lead to adverse remodeling (29).

Clinicians also pay close attention to LV size parameters, such as LVEDD, which largely reflect LV remodeling in HF. Previous reports have demonstrated that patients with a larger LV size have a higher risk of subsequent deterioration in LVEF (16, 17, 30). Park et al. identified that a larger LVEDD at baseline led to the recurrence of HF in individuals with idiopathic dilated cardiomyopathy (17). In the current study, LVEDD was the most significant predictor of HFiEF. After LASSO analysis, we found that smaller LVEDD at baseline was an independent factor that predicted a greater likelihood of improved LVEF than the other echocardiographic parameters, including LVEF at baseline. Our study underlined the clinical significance of LVEDD. Adverse remodeling persisted and steadily worsened due to ongoing myocardial injury, resulting in LV chamber dilatation. To some extent, LVEDD reflects the degree of LV remodeling. A larger LVEDD may intensify the disruption of the myocardial ultrastructure and represents more severe adverse cardiac remodeling, leading to a potentially lower cardiac function recovery (28). Therefore, it is essential to provide early revascularization, optimize medical therapy, and reinforce postoperative follow-up to reverse or alleviate cardiac remodeling and achieve good prognoses for these patients.

An important finding in our study was that patients prescribed with SGLT2i had a greater odds of being in the HFiEF group than the other drugs we collected, such as angiotensin receptor neprilysin inhibition (ARNI), beta-blocker, and ivabradine. SGLT2i worked by inhibiting SGLT2i in the proximal convoluted tubule to prevent glucose reabsorption and facilitate its excretion in urine (31). Recent studies have found that SGLT2i showed cardioprotective actions independent of the hypoglycemic effects in patients with and without type 2 diabetes mellitus (T2DM) (32–34). The DAPA-HF trial revealed that dapagliflozin significantly reduced HF hospitalization and cardiovascular mortality, independent of the presence of T2DM (32). However, the exact mechanism was not fully understood, and SGLT2i might have contributed to reducing generalized congestion and intravascular volume, thus decreasing cardiac afterload and preload (35). In contrast, previous studies have revealed that beta-blocker was a predictor of HFiEF (16, 18). Notably, details regarding the medications were not inadequate in many prior studies. Furthermore, the evidence of new classes of drugs for HF has been growing over the last few years. Therefore, we considered more details regarding medications, and demonstrated that SGLT2i use was a positive predictor of HFiEF in CTO patients with HF.

Otherwise, after LASSO analysis, although eGFR was selected as a candidate for further construction of the prediction model, it showed no statistical significance after multivariate logistic regression analysis. Considering the results of previous reports and the clinical significance of eGFR (14), we believed it may have a potential effect on the improvement of LVEF. Our follow-up duration may have been relatively short to draw concrete conclusions, and we plan to conduct a prospective and longer follow-up trial in the future.