Automated left atrial strain analysis for predicting atrial fibrillation in severe COVID-19 pneumonia: a prospective study

Adult patients (> 18 years of age) admitted to ICU at Amiens University Hospital for severe hypoxemic pneumonia related to SARS-Cov2 infection, with a TTE performed in sinus rhythm within 48 h of ICU admission, were prospectively included in the study. Exclusion criteria were patients with permanent AF, permanent atrial and/or ventricular pacing, patients under extracorporeal membrane oxygenation (ECMO), supraventricular tachycardia during the TTE exam and those with poor image quality for LA strain analysis. Patients were included on the day when TTE was performed.

This is an ancillary study of a prospective cohort study of patients with COVID-19 infection hospitalized in ICU at Amiens University Hospital (NCT04354558). This study was approved by the Amiens University Hospital IRB (Comite de Protection des Personnes Nord-Ouest II CHU–Place V. Pauchet, 80054 AMIENS Cedex 1, CNIL Number: PI2020_843_0026). In accordance with French law on clinical research for non-interventional studies, informed consent was waived but oral and written information as provided whenever possible to the patients and systematically to their families specifying that they could oppose the use of their data [14].

Data from electronical data, medical reports and biological values were collected prospectively. SARS-Cov2 infection was confirmed by a positive Reverse transcription polymerase chain reaction on nasopharyngeal swab or bronchoalveolar lavage on admission to our ICU. The severity of illness at the time of TTE exam was evaluated by the simplified acute physiology score (SAPS) II [15] and the Sequential Organ Failure Assessment (SOFA) score [16]. Vasopressor use was evaluated by the SOFA cardiovascular (SOFA cv) score [16]. Severity of COVID-19 pneumonia was defined according to the World Health Organization (WHO) case definition [17]. The severe group included patients with respiratory distress syndrome (respiratory rate ≥ 30 bpm) and/or oxygen saturation ≤ 93% at rest and/or ratio of arterial partial pressure of oxygen to fractional concentration of oxygen in inspired air < 300 mm Hg and/or > 50% lesion progression over 24–48 h by pulmonary imaging. The critical group included patients with respiratory failure requiring mechanical ventilation and/or with shock or organ failure [17]. The vital status at Day 30 was collected.

Occurrence of AF was defined by an AF episode lasting at least 30 s recorded by a 12-lead ECG or a single-lead ECG tracing [18] during ICU stay in patients with no prior history of persistent or permanent AF [4]. Patients with history of paroxysmal AF before ICU admission and/or with AF occurrence between ICU admission and time of TTE were not excluded.

All patients were monitored 24-h a day for all hemodynamic parameters including heart rate with 5-lead ECG. Twelve-lead ECG or offline electronic single-lead ECG tracing recorded were analyzed by a cardiac electrophysiologist blinded to the LAS analysis. The risk of ischemic stroke in patients with AF was assessed by the CHA₂DS₂-VASc score [18]. The primary endpoint was the occurrence of AF during ICU stay.

TTE was performed by trained operators in supine position, within 48 h of ICU admission. Standard echocardiography protocol were used in accordance with the American Society of Echocardiography guidelines [19] and the European Society of Cardiology [20]. Echocardiographic images were obtained by a high-quality commercially available ultrasound system (CX 50, Philips Healthcare). All operators had a level III competence of general adult TTE [21].

LAS analysis was obtained using an automated speckle tracking software (Auto-Strain QLAB 13.0, Philips Medical systems, Andover, MA, USA) with a LAS dedicated mode. The LAS was defined as the strain value in three phases: reservoir strain in systole (LASr), conduit strain in early diastole (LASct) and contraction strain in late diastole (LAScd) [10]. LASr was a positive value, while LASct and LACcd were negative values. LAS values, for each phase, were obtained from an optimized apical four-chamber view using an automated approach as recommended [10] (Fig. 1). The regions of interests (ROI) were generated automatically and LA endocardial border was manually adjusted when required. The QRS complex was used as initial zero-baseline strain electrocardiogram reference point as recommended [10]. All LAS measurements were performed by an experienced cardiologist blind to clinical data.

LAS values were automatically measured by the software from the LA longitudinal strain curve (Fig. 1). When using the QRS complex as a zero-reference point, the first peak positive deflection corresponds to the value of LA reservoir function. The value of LA contraction function was obtained at the beginning of the P wave contraction. The value of LAS conduit function was calculated as the difference between LASr and LASct values. LA phases definition and LAS measurement were performed according to the European Association of Cardiovascular Imaging (EACVI)/American society of Echocardiography (ASE) guidelines [10].

Data are expressed as mean ± standard deviation (SD), median [interquartile range] or numbers (percentage), as appropriate. Variables were compared between groups (AF and non-AF group) using Mann–Whitney or Chi-square tests, as appropriate. A receiver-operating characteristic curve (ROC) was built to assess the diagnostic performance of LASr, LASct, LAScd for prediction of AF occurrence. Area under ROC curves (AUC) of echocardiographic parameters were compared using Delong’s test. The Youden index was used to determine the optimal threshold of LAS parameters for the prediction of AF occurrence.

To evaluate independent factor associated with AF, univariate and multivariate logistic regression were performed. All factors with a P value < 0.10 in univariate analysis were included in the multivariate model. The calibration of the model was assessed by the Hosmer–Lemeshow goodness-of-fit statistic (good fit was defined as a p value of > 0.05) [22]. The C-statistic test was used to test the ability of the model to discriminate patients with and without AF [23]. Data are presented as odds ratio (ORs) and 95% confidence intervals (CIs). Cumulative risk curves, as function of time, were generated using the Kaplan–Meier method, and compared by the log-rank test. A statistical test was significant when P value was under 0.05. All P values are the results of 2-tailed tests. Statistical analyses were performed using SPSS software version 24 (IBM Corp, Armonk, NY).

To evaluate the intra-observer variability for offline LAS analysis, data of 10 patients were randomly selected and analyzed by two operators with at least a 1-week interval between the two analyses. Inter‐observer and intra-observer reproducibility of LAS measurements was assessed using intraclass correlation coefficient (ICC).

Between March 1st 2020 and February 15th 2021, 180 patients were admitted in our ICU for COVID-19 infection, 126 patients fulfilled the inclusion criteria and 47 patients were excluded. Especially, 6 patients (13%) were excluded for poor TTE image quality that did not allow LAS analysis. A total of 79 patients were included in the study. The study population was divided in 2 groups according to the presence of AF (AF group and no AF group) during ICU stay.

Medical history, chronic treatment, time to ICU admission (from first symptoms), biological investigations and hemodynamic parameters were comparable between the two groups. AF was documented in 16/79 patients (20%) with a median time of 7 [2–11] days (Table 1) from ICU admission. In the AF group, patients were older (73 [65–76] vs. 65 [59–70] years; P = 0.026) than in the no AF group. SAPS II score at ICU admission was higher for the AF group (58 [43–62] vs. 32 [21–49], P < 0.0001) than for the no AF group. In the AF group, there was significantly more critical patients (n = 11/16[69%] vs. n = 14/63[22%], P < 0.001) according to the WHO definition. Moreover, patients of the AF group had higher SOFA cv score than patients of the no AF group (4 [1–4] vs. 0 [0–1], P = 0.0001).

In the AF group, duration of mechanical ventilation was longer (28 [16–44] vs. 17 [10–24] days, P < 0.0001) as well as ICU length of stay (25 [14–33] vs. 12 [5–21] days, P < 0.001) compared to no AF group. However, there was no difference in mortality at 30 days between the two groups (n = 7/16 [43%] vs. 14/63 [22%], P = 0.11, respectively, for AF and no AF group).

Regarding echocardiographic parameters, only LAS parameters were significantly different between the 2 groups. In the AF group, LASr and LAScd were significantly impaired compared to the other group (− 20.2 [− 12.3; − 27.3] vs. -30.5 [− 23.8; − 36.2] %, P = 0.002 and − 8.1 [− 6.3; − 10.9] vs. − 17.2 [− 5.0; − 10.2] %, P < 0.0001, respectively)). LASct did not significantly differ between groups (P = 0.31).

In a multivariate model (Table 4), only LAScd and SOFA cv > 1 were independently associated to the occurrence of AF with an OR of 1.24 [95% CI 1.04 to 1.48] and 5.56 [95% CI 1.41 to 22.11], respectively. The discrimination ability of the model using C-statistics showed an AUC of 0.89 (95% CI 0.80 to 0.97) (Fig. 4). The 30-day cumulative risk of AF was 42 ± 9% with LAScd > − 11% and 8 ± 4% with LAScd ≤ − 11% (log rank test P value < 0.0001) (Fig. 5).

The reproducibility of LAScd measurement had a very strong correlation with an ICC of 0.86 (95% CI 0.52–0.96) for the inter-operator reproducibility and 0.94 (95% CI 0.74–0.98) for the intra-operator reproducibility.

Recent findings supported the higher likelihood of observing AF in COVID-19 patients admitted to the ICU [24]. Moreover, AF may worsen the clinical evolution of pneumonia in these patients [25]. Colon et al. reported an AF prevalence of 16.5% in ICU patient and showed that mechanical ventilation was strongly associated with AF [24]. Here, we reported comparable results as 20% of our patients developed AF during ICU stay and AF was strongly associated with a more critical state (69% vs. 22%; P < 0.0001). Pletzer et al. reported an in-hospital mortality of 39.2% (n = 65/166) in COVID-19 patients with AF and showed that AF was an independent predictor of in-hospital mortality [26]. However, in this report only 60% of patients with AF were hospitalized in ICU [18].

AF is a common complication of critical illness and is an independent predictor of mortality [27]. In septic patients, mechanical ventilation, organ failure and norepinephrine use were strongly associated with AF [4]. In our study, norepinephrine use was strongly associated with AF to the contrary to mechanical ventilation. However, AF was not associated with 30-day mortality (43% vs. 22%; P = 0.11) probably due to a lack of statistical power explained by the limited sample size.

AF is associated with adverse outcomes in COVID-19 patients [5, 17]. Therefore, the prediction of AF is of paramount clinical importance. In our study, LAScd was a strong predictor of AF and the identified cutoff value of − 11% was closed to that of previously observed cutoff in different cardiovascular disease. For example, in Chagas disease, the LA conduit function (− 12.6 ± 5.7%) was reported to be a strong predictor of AF [28] due to the depression of the LA conduit function [29]. In a cohort of ischemic stroke, Rasmussen et al., demonstrated that LAScd was worse for the AF group (− 12 ± 5 vs. − 16 ± 7%, P < 0.003) compared to patients without AF [12].

LAS reservoir parameter is also a prognostic factor for the occurrence of AF in ischemic stroke, heart failure or after cardiac surgery [12, 13, 30] and reflects LA compliance [11]. Several studies suggested that impaired LA reservoir function may be a sign of LA remodeling, caused by several cardiovascular conditions, such as hypertension, diabetes or ischemic heart disease [12]. In our study, LASr values were significantly impaired in the AF group 30.5 [23.8–36.2] % vs. 20.2 [12.3–27.3] %; P = 0.002). Goerlich et al. reported similar LASr values (30.4 [26.1–35.8] % vs. 22.3 [20.6–27.8] %; P < 0.001) and shown that LASr parameter was an independent factor of AF in COVID-19 patients [31]. However, in our study, only LAScd remained independently associated with AF probably due to the limited sample size of our cohort.

Clinical data on LA mechanistic dysfunction suggested a strong link between left ventricular diastolic dysfunction and risk of AF [32, 33]. LAS analysis, especially LAScd, has recently emerged as a powerful tool for left ventricular diastolic dysfunction evaluation [34] especially when left ventricular end diastolic pressure (LVEDP) was increased [35]. Severe hypoxemic COVID-19 pneumonia may be associated with diastolic dysfunction and/or increased LVEDP. Indeed, COVID-19 infection can lead to myocardial diastolic dysfunction [36] by direct virus related-myocardial injury, inflammation or cardiac fibrosis [37]. COVID-19 may unmask subclinical LA dysfunction or exacerbate preexisting LA dysfunction [38]. Moreover, recent findings suggested that COVID-19 patients with severe respiratory failure had a high prevalence of increased LVEDP [39]. All these elements may lead to AF. However, data about the potential effect of COVID-19 on LAcd function are currently lacking and further studies on the subject would be of great interest.

In this study, we found a high feasibility of LAS parameters in patients with respiratory failure as only 6 patients were excluded for poor image quality. Data on LAS analysis in ICU are scarce. Hence, the present study emphasizes the fact that LAS analysis can be easily performed in ICU patients using a dedicated mode for LAS analysis and an automated approach as recommended [10].