Basic Clinical Characteristics
From March 2019 to July 2021, 4865 critically ill adult patients were consecutively admitted to SICU, and 656 of them were diagnosed with sepsis. Among these patients, 358 completed echocardiography examinations within the first 24h after admission to SICU. Of them, 215 patients were excluded mainly for the following reasons: poor echocardiographic image quality (n=87), incomplete echocardiographic data (n=21), incomplete clinical data (n=36), history of valvular heart disease (n=22), history of congestive heart failure (n=20), holder of implanted devices (n=12), atrial fibrillation (n=10), refusal to participate (n=4), and other reasons (n=3). Among the 87 patients with poor echocardiographic image quality, 35 patients had indistinct imaging of the entire endocardium (LVEF, GLS, and TMAD could not be measured accurately), and the other 52 patients had legible imaging of the apex and mitral annulus but indistinct imaging of the other endocardium parts, especially the lateral wall of the LV (TMAD could be measured, but LVEF and GLS could not be assessed) (Fig. 2). The acquisition ratio of TMAD was significantly higher than that of LVEF and GLS (84.8% vs. 62.2%, p < 0.001). Finally, 143 patients were included. The flow diagram of the study is shown in Fig. 3. Of all participants, 94 patients (65.7%) were male, mean age was 66.3 ± 16.6 years old, BMI was 22.9 ± 3.2, APACHE II score was 17.1 ± 6.2, and SOFA score was 6.8 ± 2.7. Baseline and clinical information of the patients were summarized in Table 1.
Incidence of SICM and survival analysis according to three different diagnostic criteria
The incidence of SICM in this study was 18.2%, 32.9%, and 38.5% according to the LVEF, GLS, and TMADMid criteria, respectively. When considering the speckle tracking technology criteria (GLS or TMAD), the incidence of SICM was significantly higher than that according to LVEF criteria (32.9% vs. 18.2%, p = 0.006; 38.5% vs. 18.2%, p < 0.001) (Fig. 4). The 28d mortality in the SICM group was 34.6% (LVEF criteria), 27.7% (GLS criteria) and 29.1% (TMADMid criteria), respectively, while the in-hospital mortality in the SICM group was 42.3% (LVEF criteria), 36.2% (GLS criteria) and 36.4% (TMADMid criteria), respectively. Kaplan-Meier curves demonstrated that, compared to non-SICM patients, SICM positives had significantly lower 28d (p = 0.011, LVEF criteria; p = 0.031, GLS criteria; p = 0.0064, TMADMid criteria) and in-hospital survival rates (p = 0.0097, LVEF criteria; p = 0.0157, GLS criteria; p = 0.0061, TMADMid criteria), regardless of the diagnostic criteria (Fig. 5).
Clinical information of SICM and non-SICM patients
Conventional and speckle tracking echocardiographic data of all patients were collected. According to the TMADMid criteria, all enrolled patients were divided into SICM or non-SICM groups. No significant differences were found in the APACHE II score, comorbidity, medication history, and sepsis source between the SICM and non-SICM. However, the SOFA score of the SICM group was significantly higher than that of the non-SICM group (7.8 ± 3.1 vs 6.1 ± 2.2, p < 0.001), and the mortality was significantly higher in those with SICM than in non-SICM patients (36.4% vs. 17.0%, p = 0.016). According to the laboratory results, patients in the SICM group had significantly higher cTnT, NT-proBNP, and CK-MB within the first 24 hours after admission, when compared to the non-SICM group (all p < 0.05). There was no significant difference in the length of ICU or hospital stay, nor in the duration of mechanical ventilation between the two groups (Table 1).
Echocardiographic data among patients with abnormal TMAD
All the echocardiogram data of SICM and non-SICM patients (TMADMid criteria) are shown in Table 1. LVEF, MAPSE, TAPSE, a’, MA Smax, LVOT VTI, SV, CO, TMAD 1, TMAD 2, TMADMid, and %TMAD were significantly lower in the SICM group than in the non-SICM group (all p < 0.05). Meanwhile, LVESV and GLS in those with SICM were significantly higher than in non-SICM patients (all p < 0.05).
On the other hand, correlations between TMAD and other LV systolic function echocardiographic parameters were analyzed. Positive correlations were detected between TMAD and LVEF, MAPSE, and MA Smax, respectively, while a negative correlation between TMAD and GLS was confirmed (all p < 0.001) (Fig. 6).
Furthermore, the diagnostic value of TMAD for SICM was evaluated using LVEF and GLS, consecutively, as diagnostic “golden standards”. According to the “LVEF standard”, the AUROC values of TMAD1, TMAD2, TMADMid, and %TMAD were 0.893, 0.887, 0.902, and 0.887 respectively. Based on the “GLS standard”, the AUROC values of TMAD1, TMAD2, TMADMid, and %TMAD were 0.885, 0.884, 0.911, and 0.908 respectively (Fig. 7). The cutoff value, sensitivity and specificity of each TMAD parameter were stated in Table 2.
Intra- and inter-observer variability
Adequateintra- and inter-observer reliabilities for LVEF, GLS, and TMAD were detected (all ICC > 0.75). Particularly, TMAD had the highest ICC value when compared to LVEF or GLS (Table 3). Regarding the required time for the LVEF, GLS, and TMAD offline analyses, respectively, the mean single measurement time for TMAD was significantly shorter than that for LVEF or GLS (40.5s ± 9.3s vs. 82.6s ± 15.2s, p < 0.001; 40.5s ± 9.3s vs. 70.2s ± 11.3s, p = 0.006, respectively).