This was an observational retrospective multicenter cohort study. We reviewed 12,956 consecutive subjects who underwent CCTA and echocardiography between 2002 and 2011 at the Seoul National University Hospital, Seoul National University Bundang Hospital, and Seoul National University Hospital Gangnam Center. We excluded individuals with significant LV systolic dysfunction (defined as ejection fraction < 40%), hypertrophic or infiltrative cardiomyopathy, severe valvular heart disease, and those with a history of previous myocardial infarction, coronary revascularization, or cardiac surgery.

A total of 5803 subjects were included in the final analysis, who had available echocardiographic data on the left ventricular mass index (LVMI) and relative wall thickness (RWT) and had no significant coronary artery stenosis (defined as luminal narrowing < 50%) on CCTA (Additional file 1: Fig. 1). The date of the final follow-up was March 18, 2016.

The study protocol was approved by the Institutional Review Board of Seoul National University Hospital (IRB No. J-1511-025-715), and was performed in accordance with the Declaration of Helsinki. Written informed consent was waived owing to the retrospective and observational nature of this study.

Electrocardiographic-gated CCTA images were acquired using a 64-slice multidetector scanner (SOMATOM Sensation 64, SOMATOM Definition, Siemens Medical Solutions, Forchheim, Germany; Brilliance 64, Philips Medical Systems, Best, The Netherlands). Before performing the scan, sublingual nitroglycerine (0.3 mg) was administered to provide transient coronary dilatation, and metoprolol (50 mg) was administered to subjects presenting with a heart rate of more than 60 beats per minute. The CCTA results were graded as none (luminal narrowing, 0%), nonobstructive (luminal narrowing, 1–49%), and obstructive (luminal narrowing ≥ 50%) respectively based on the severity of narrowing in any of the major epicardial coronary arteries.

All patients underwent comprehensive two-dimensional echocardiography within 3 months of initial CCTA. Hypertrophic alteration of the LV structure was quantified based on LVMI and RWT by echocardiography [14]. LVMI was estimated using a standard formula with LV cavity dimension and wall thickness at end-diastole and indexed to body surface areas. Linear internal measurement of the LV and its wall was performed in the parasternal long-axis view, and the values were obtained perpendicular to the long axis of the LV at or immediately below the level of the mitral valve leaflet tips [14]. LV mass index was calculated using the following equation:

RWT was calculated as two times the posterior wall thickness divided by the LV diastolic diameter. Increased LVMI was defined as LVMI > 95 g/m² in women and > 115 g/m² in men, and the cutoff for abnormal RWT was > 0.42, in both women and men [14]. Normal LV geometry was defined as normal LVMI and RWT. Abnormal LV geometry was defined as a composite of concentric remodeling (normal LVMI and increased RWT), eccentric hypertrophy (increased LVMI and normal RWT), and concentric hypertrophy (increased LVMI and RWT).

Using electronic medical records, we obtained laboratory information, including serum hemoglobin, total cholesterol, triglycerides, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, fasting blood glucose, creatinine, and estimated glomerular filtration rate (eGFR) accordingly for all the patients.

The primary outcome was all-cause mortality. Mortality data were obtained and verified via a centralized database of death records from the Korean Ministry of Security and Public Administration.

A two-tailed p value of < 0.05, was considered statistically significant. All statistical analyses were performed using SPSS version 23 (IBM Corp, Chicago, IL, USA) and R programming version 3.2.4 (http://​www.​R-project.​org; The R Foundation for Statistical Computing, Vienna, Austria).

A total of 5803 subjects without significant obstructive CAD consisted of 3884 (66.9%) men with a mean age of 56.6 ± 8.87 years (Table 1). Of them, 4045 (69.7%) subjects had normal LV geometry and 1758 (30.3%) had abnormal LV geometry. The subjects with normal LV geometry were younger than those with abnormal LV geometry (55.7 ± 8.68 vs. 58.5 ± 8.98 years, p < 0.001), and this was observed more frequently in men (69.6 vs. 65.9%, p < 0.001). There was no significant difference in the baseline laboratory results between the two groups, except for the decreased hemoglobin (14.4 ± 1.61 mg/dL in normal geometry vs. 14.6 ± 1.45 mg/dL in abnormal geometry, p < 0.001) and eGFR levels (77.2 ± 13.7 vs. 78.8 ± 13.3 mL/min/1.73 cm², p < 0.001). With regard to echocardiographic parameters, the abnormal LV geometry group had increased LV wall thickness and chamber size, but also larger left atrial size (36.5 ± 4.95 mm vs. 38.0 ± 5.48 mm, p < 0.001), higher pulmonary artery systolic pressure (27.0 ± 4.63 vs. 28.1 ± 5.06 mmHg, p < 0.001), and a higher ratio of peak early transmitral inflow velocity to early diastolic velocity of the mitral annulus (E/e’) (9.34 ± 2.84 vs. 11.4 ± 2.05, p = 0.003) than the normal LV geometry group. The baseline characteristics according to the four LV geometric patterns are shown in Additional file 1: Table 1.

During a mean follow-up of 6.2 ± 1.48 years, 84 (1.44%) subjects died in the study population. Of these, 56 subjects were from the normal LV geometry group (1.24%) and 28 were from the abnormal LV geometry group (2.32%). After adjustment for confounding factors, abnormal LV geometry was independently associated with an increased risk of all-cause mortality (adjusted HR 1.64; 95% CI, 1.02–2.67; p = 0.040) (Table 2). In the Kaplan–Meier survival analysis, the abnormal LV geometry group showed a worse prognosis than the normal LV geometry group (log-rank p < 0.001) (Fig. 1). In the subgroup analysis, the impact of abnormal LV geometry on all-cause mortality was directionally consistent with that in the main analysis (Additional file 1: Table 2).

We classified subjects into four groups according to the joint categories of LV geometry and CAD: 1) normal LV geometry with no CAD, 2) normal LV geometry with nonobstructive CAD, 3) abnormal LV geometry with no CAD, and 4) abnormal LV geometry with nonobstructive CAD. Compared to normal LV geometry without CAD, abnormal LV geometry without CAD (HR 2.38, 95% CI 1.41–4.03, p = 0.001) and those with nonobstructive CAD (HR: 4.54, 95% CI: 2.42–8.51, p < 0.001) had a higher risk of all-cause mortality (Table 3). In the Kaplan–Meier survival analysis, abnormal LV geometry had worse overall survival in both patients with no CAD (log-rank test, p = 0.024) and non-obstructive CAD (log-rank test, p < 0.001) (Fig. 2). After adjustment for confounding factors, abnormal LV geometry with nonobstructive CAD remained an independent predictor of all-cause mortality (adjusted HR 2.49; 95% CI, 1.24–4.99; p = 0.010) (Table 3).

When subjects were stratified into four groups according to LV geometric patterns, each abnormal LV geometric pattern was associated with an increased risk of all-cause mortality compared with normal geometry (log-rank, p < 0.001) (Additional file 1: Fig. 2). Eccentric hypertrophy carried the highest risk (HR 3.60, 95% CI 1.91–5.91, p < 0.001), followed by concentric hypertrophy (HR 2.34, 95% CI 1.05–5.23, p = 0.037), and concentric remodeling (HR 1.92, 95% CI 1.12–3.31, p = 0.019). After adjustment, the risk of all-cause mortality was significantly increased only in subjects with eccentric hypertrophy (adjusted HR 2.29; 95% CI, 1.25–4.18; p = 0.007) (Additional file 1: Table 3).

The most compelling advantage of our study is that it is a well-constructed, large imaging database study containing both CCTA and echocardiographic images from all participants. Furthermore, our study has an advantage over prior research owing to the inclusion of many Asians in whom the association of LV geometry with prognosis has not been extensively investigated. Specifically, in the landmark Framingham Heart Study (FHS) by Levy et al. [30], participants were predominantly White individuals of Western European descent. Although a more ethnically diverse group of individuals is reflected in the FHS OMNI cohorts, the proportion of the Asian population is only 28% [31], resulting in limited power to draw conclusive results. Additionally, considering the relatively low body mass index in our study population, our findings might be less confounded by overweight or obesity, which is an important confounding factor in studies investigating the association between LV geometry and prognosis [32].

However, several limitations should be considered when interpreting our findings. First, this was a retrospective observational study with inherent limitations, such as unmeasured confounders. Although we adjusted for a set of conventional cardiovascular disease risk factors, residual confounding cannot be completely excluded. Second, the exact clinical indications for CCTA scans were not documented, although most individuals underwent CCTA for the identification of CAD, owing to the presence of chest pain, dyspnea, and cardiovascular risk factors. However, since individuals with obstructive CAD confirmed by CCTA were excluded from our study, the difference in the clinical indications for CCTA scans might not substantially affect the results of this study. Third, since we focused on all-cause mortality as the primary outcome, data on specific causes of death were not available. In observational studies, all-cause mortality is generally considered a more robust and unbiased outcome than disease-specific mortality. However, data on the specific cause of death might have strengthened the association between LV geometry and cardiovascular prognosis in subjects with no or nonobstructive CAD. Fourth, there remains a possibility that certain conditions affecting both LV structure and mortality were not fully ruled out. To minimize this concern, we excluded individuals with cardiomyopathy and significant valvular heart disease from the analyses. Fifth, as Korean individuals were exclusively included in the present study, it is uncertain whether our findings can be generalized to other populations. Lastly, since there are no data on noninvasive stress testing, we could not entirely exclude the possibility of myocardial ischemia, which is clearly a confounding factor related to mortality.