Soluble ST2 as a Useful Biomarker for Predicting Clinical Outcomes in Hospitalized COVID-19 Patients

Soluble suppression of tumorigenesis-2 (sST2) is an emerging biomarker for sepsis as well as for heart failure. We investigated the prognostic utility of sST2 for predicting clinical outcomes in hospitalized coronavirus disease 2019 (COVID-19) patients. In a total of 52 hospitalized COVID-19 patients, sST2 levels were measured using the ichroma ST2 assay (Boditech Med Inc., Chuncheon-si, Gang-won-do, Republic of Korea). Clinical outcomes included intensive care unit (ICU) admission, ventilator use, extracorporeal membrane oxygenation (ECMO) use, and 30-day mortality. sST2 was analyzed according to clinical outcomes. sST2, sequential organ failure assessment (SOFA) score, critical disease, and 4C mortality score were compared using the receiver operating characteristic (ROC) curve and Kaplan–Meier methods for clinical outcomes. The sST2 level differed significantly according to ICU admission, ventilator use, ECMO use, and 30-day mortality (all p < 0.05). On ROC curve analysis, sST2 predicted ICU admission, ventilator use, ECMO use, and 30-day mortality comparable to SOFA score but significantly better than critical disease. sST2 predicted ICU admission, ventilator use, and ECMO use significantly better than the 4C mortality score. On Kaplan–Meier survival analysis, hazard ratios (95% confidence interval) were 8.4 (2.7–26.8) for sST2, 14.8 (3.0–71.7) for SOFA score, 1.8 (0.5–6.5) for critical disease, and 11.7 (3.4–40.1) for 4C mortality score. This study demonstrated that sST2 could be a useful biomarker to predict ICU admission, ventilator use, ECMO use, and 30-day mortality in hospitalized COVID-19 patients. sST2 may be implemented as a prognostic COVID-19 biomarker in clinical practice.


Introduction
The coronavirus disease 2019 (COVID-19) pandemic has caused globally over six million deaths as of November 8, 2022, since its outbreak occurred in December 2019 [1,2]. The clinical spectrum of COVID-19 ranges from asymptomatic to critical disease [3]. Critical disease can present an acute respiratory distress syndrome (ARDS), sepsis, septic shock, thromboembolism, or multi-organ failure (MOF) [3][4][5]. MOF can occur in the lungs, heart, liver, kidney, and nervous system and may also present with hematological abnormalities [4]. The sequential organ failure assessment (SOFA) score is used as a severity index for assessing sepsis in COVID-19 [3,5]. Recently, 4C mortality score was developed as a prognostic index to predict the in-hospital mortality of COVID-19 [6]. 4C mortality score was calculated using age, sex, comorbidities, respiratory rate, peripheral oxygen saturation on room air, Glasgow Coma Scale, levels of urea, and C-reactive protein (CRP), and it predicted the in-hospital mortality better than the SOFA score [6].

Study Population
The enrollment and clinical outcomes of the study population are presented in Figure 1. From February to May 2020, a total of 396 COVID-19 patients at the Yeungnam University Medical Center (YUMC), Daegu, Republic of Korea, were studied. We excluded 255 patients who visited the outpatient clinic without routine blood tests in clinical practice or who were younger than 20 years. From 141 hospitalized patients over 20 years of age, we further excluded 89 patients without available residual ethylene-diamine-tetraacetic acid (EDTA) plasma samples after routine blood tests and known 30-day status. Finally, 52 patients were enrolled. Among the 52 patients, 44 patients were included in a previous study [5]. The study population had no specific limitations on care at enrollment. The symptom duration ranged up to 25 days. Therefore, we could not enroll the study population at the same point in the COVID-19 disease course. In a total of 52 patients, a follow-up period for 30-day status was five months from February to the end of June 2020. No additional blood sampling or intervention was performed. Medical records were reviewed to obtain demographic, clinical, and laboratory data. A total of 52 patients were treated with oxygen support, antibiotics, antiviral agents (Lopinavir/Ritonavir), or hydroxychloroquine.

sST2 Assay
A total of 52 residual EDTA plasma samples were collected from 52 patients at enrollment. The collected residual EDTA samples were aliquoted to avoid repeated freezing and thawing and were stored at −70 °C until measurement. Frozen samples were thawed at room temperature and gently mixed immediately before measuring sST2 levels. sST2 was assayed using the ichroma ST2 assay (Boditech Med Inc., Chuncheon-si, Gang-wondo, Korea) with an ichroma Ⅱ immuno-analyzer (Boditech Med Inc.) based on a fluorescence immunoassay. Both the ichroma ST2 assay and the ichroma Ⅱ immune-analyzer were approved by the Korean Ministry of Food and Drug Safety.

sST2 Assay
A total of 52 residual EDTA plasma samples were collected from 52 patients at enrollment. The collected residual EDTA samples were aliquoted to avoid repeated freezing and thawing and were stored at −70 • C until measurement. Frozen samples were thawed at room temperature and gently mixed immediately before measuring sST2 levels. sST2 was assayed using the ichroma ST2 assay (Boditech Med Inc., Chuncheon-si, Gang-won-do, Republic of Korea) with an ichroma II immuno-analyzer (Boditech Med Inc.) based on a fluorescence immunoassay. Both the ichroma ST2 assay and the ichroma II immune-analyzer were approved by the Republic of Korean Ministry of Food and Drug Safety.
The manufacturer's upper reference limit (URL) was 35 ng/mL, which was established as the cut-off value for predicting HF [13,15,18,21]. Analytical measurement intervals were from 3.1 to 200.0 ng/mL. sST2 levels were measured according to the manufacturer's instructions.

Statistical Analysis
Data were presented as the number (percentage) or median (interquartile range, IQR). The Shapiro-Wilk test was used to determine the normality of data distribution. The Mann-Whitney U test was used to compare continuous variables (sST2, SOFA score, and 4C mortality score) according to the clinical outcomes. Chi-squared test or Fisher's exact test was used to compare categorical variables (critical disease and 4C mortality risk group) With the receiver operating characteristic (ROC) curve analysis, the area under the curve (AUC), the optimal cut-off values, sensitivity, and specificity of sST2, SOFA score, critical disease, and 4C mortality were estimated to predict clinical outcomes [32]. A 95% confidence interval (CI) of AUC was calculated using the Mann-Whitney statistic approach, which was suggested as superior to others for the small sample size [32,33]. Kaplan-Meier survival analysis was used to estimate the hazard ratio (HR) with a 95% confidence interval (CI) for the 30-day mortality of sST2, SOFA score, critical disease, 4C mortality score, and 4C mortality score group. HRs (95% CI) of high and very high groups relative to low/intermediate groups were calculated.
The sample size for the Kaplan-Meier survival analysis was estimated based on the previous study [34]. The inputs were identical to those described in our previous study, except for the alternative survival probability; analysis time t = 1 month, accrual time α = 5 months, follow-up time b = 1-month, null survival probability, S 0 (t) = 0.013, 0.025, or 0.026, type I error rate (α) = 0.05, and the power (1 − β) = 0.8 [5]. The alternative survival probability was set to S 1 (t) = 0.231 based on the 30-day mortality of this study. Using log-minus-log transformation, which was suggested for improving the accuracy of the small sample size, the estimated sample size was between 11 and 15. Accordingly, the sample size of 52 was considered sufficient to perform the Kaplan-Meier survival analysis. MedCalc Software (version 20.111, MedCalc Software, Ostend, Belgium) was used for statistical analysis. p-value < 0.05 was considered statistically significant.

Results
The basic characteristics of the study population are summarized in Table 1. The median age (IQR) was 71.0 years (62.5-79.0), and males were 61.5% (n = 32). Among six patients with chronic cardiac disease, three patients had HF. The median sST2 level, SOFA score, and 4C mortality score were 46.6 ng/mL, 4.0, and 9.5, respectively. In WHO disease severity, critical disease was 71.2% (n = 37); sepsis and septic shock were 83.8% (n = 31) and 16.2% (n = 6), respectively. Eight sepsis and five septic shock patients had ARDS. Of a total of 52 patients, 26.9% patients admitted to ICU (n = 14). Among the 14 ICU patients, a ventilator was applied to 12 patients, and ECMO was applied to seven ventilated patients. The 30-day mortality was 23.1% (n = 12).
The sST2, SOFA score, critical disease, and 4C morality score according to clinical outcomes are presented in Table 2. The sST2 level and SOFA score differed significantly according to all clinical outcomes (all p < 0.05). The proportion of critical disease did not differ according to ventilator use and 30-day mortality, but the 4C mortality score differed significantly according to only the 30-day mortality. The sST2 level differed significantly among 4C mortality score risk groups (p = 0.042); however, the SOFA score and the proportion of critical disease did not (Table 3).       In the ROC curve analysis, the sST2 and SOFA score comparably predicted ICU admission, ventilator use, ECMO use, and 30-day mortality (sST2 vs. SOFA, all p > 0.05) (Figure 2). sST2 predicted ICU admission, ventilator use, ECMO use, and 30-day mortality better than critical disease (all p < 0.05). sST2 predicted ICU admission, ventilator use, and ECMO use better than the 4C mortality score (all p < 0.05). The sST2 and 4C mortality scores predicted comparably 30-day mortality (0.826 vs. 0.830, p = 0.963).

Discussion
This is the first study to explore the prognostic performance of the sST2, SOFA score, critical disease, and 4C mortality score simultaneously in hospitalized COVID-19 patients. In this study, most patients presented with critical disease. The median sST2 level of the total study population was 46.6 ng/mL, which was greater than the cut-off value for predicting HF, 35 ng/mL [13,15,18,21]. In this study, 25 of 29 patients with sST2 levels above 35 ng/mL had critical disease, and all 13 ARDS patients were included in these 25 patients. Among the 25 patients with critical disease, 20 patients presented dyspnea, which was known as the key symptom of HF [12]. It was not clear how many patients had HF because there was no further evaluation for the diagnosis of it. Thus, sST2 may have been elevated when reflecting HF, the degree of inflammation, or both. In the previous studies, the sST2 level was heterogeneous in COVID-19 [22][23][24][25][27][28][29][30]. In two previous studies, the overall sST2 levels were less than 10 ng/mL [22,23]. In other previous studies, median sST2 levels were from 48 ng/mL to 53.1 ng/mL in COVID-19 patients, similar to our data [24,28,29]. In this study, around 70% of patients had a 4C mortality score of nine or higher, belonging to the high and very high groups. Increasing age is the strongest variable in the 4C mortality score [6]. The age range from 70 to 79 is six points, and older than 80 is seven points. [6]. In this study, most of the patients were over the age of 70; therefore, most of them were given a score of six or higher.
Both the sST2 level and SOFA score were significantly associated with ICU admission, ventilator use, ECMO use, and 30-day mortality. Median sST2 levels were lower than the URL in GW patients, patients without a ventilator and/or ECMO use, and survivors; however, it was about five-to ten-fold higher in ICU patients, patients with ventilator and/or ECMO use, and non-survivors. Based on our data, sST2 seems to reflect clinical outcomes in COVID-19. Similar to our data, a higher sST2 level was significantly associated with ICU admission and mortality [24,25,28,30]. The median sST2 level in non-survivors was 107 ng/mL, and it was higher than that of survived ICU and GW patients; the

Discussion
This is the first study to explore the prognostic performance of the sST2, SOFA score, critical disease, and 4C mortality score simultaneously in hospitalized COVID-19 patients. In this study, most patients presented with critical disease. The median sST2 level of the total study population was 46.6 ng/mL, which was greater than the cut-off value for predicting HF, 35 ng/mL [13,15,18,21]. In this study, 25 of 29 patients with sST2 levels above 35 ng/mL had critical disease, and all 13 ARDS patients were included in these 25 patients. Among the 25 patients with critical disease, 20 patients presented dyspnea, which was known as the key symptom of HF [12]. It was not clear how many patients had HF because there was no further evaluation for the diagnosis of it. Thus, sST2 may have been elevated when reflecting HF, the degree of inflammation, or both. In the previous studies, the sST2 level was heterogeneous in COVID-19 [22][23][24][25][27][28][29][30]. In two previous studies, the overall sST2 levels were less than 10 ng/mL [22,23]. In other previous studies, median sST2 levels were from 48 ng/mL to 53.1 ng/mL in COVID-19 patients, similar to our data [24,28,29]. In this study, around 70% of patients had a 4C mortality score of nine or higher, belonging to the high and very high groups. Increasing age is the strongest variable in the 4C mortality score [6]. The age range from 70 to 79 is six points, and older than 80 is seven points. [6]. In this study, most of the patients were over the age of 70; therefore, most of them were given a score of six or higher.
Both the sST2 level and SOFA score were significantly associated with ICU admission, ventilator use, ECMO use, and 30-day mortality. Median sST2 levels were lower than the URL in GW patients, patients without a ventilator and/or ECMO use, and survivors; however, it was about five-to ten-fold higher in ICU patients, patients with ventilator and/or ECMO use, and non-survivors. Based on our data, sST2 seems to reflect clinical outcomes in COVID-19. Similar to our data, a higher sST2 level was significantly associated with ICU admission and mortality [24,25,28,30]. The median sST2 level in non-survivors was 107 ng/mL, and it was higher than that of survived ICU and GW patients; the median sST2 level was significantly higher in ventilated ICU patients than in GW patients [28].
In COVID-19, the mechanism for the release of sST2 is not fully understood. IL-33 is released mainly by injured epithelial alveolar cells and can be upregulated in COVID-19 [18,35]. COVID-19 induces IL-33 expression in the lungs, and activated IL-33 leads to the production of sST2 type II pneumocytes [8,35]. In COVID-19, ARDS, septic shock, or inflammatory mediators can lead to HF through multiple mechanisms [9], and HF can induce the upregulation of sST2 in the lungs leading to sST2 secretion by type II pneumocytes [36].
Our data demonstrated that the sST2 level reflected a 4C mortality score better than the SOFA score and critical disease. Unlike the SOFA score and critical disease, the median sST2 level increased significantly according to the 4C mortality score risk group. Inhospital mortality ranged from 0.0% to 1.7% in the low-risk group and 8.0% to 9.9% in the intermediate-risk group, which were substantially lower than that of the high-and very high-risk groups [6,37]. In line with the previous studies, the median sST2 level was normal in the low/intermediate risk group, and it was significantly lower than that of the high-risk group. Based on our data, the sST2 level seems to reflect the prognostic index for COVID-19 mortality.
In this study, the sST2 level and SOFA score comparably predicted ICU admission, ventilator use, ECMO use, and 30-day mortality. AUCs of the sST2 level was higher than that of the SOFA score for predicting ICU admission and 30-day mortality (0.878 vs. 0.865 in ICU admission; 0.826 vs. 0.716 in 30-day mortality). sST2 outperformed critical disease in ICU admission, ventilator use, ECMO use, and 30-day mortality. Although the sST2 level and 4C mortality score comparably predicted 30-day mortality, sST2 outperformed the 4C mortality score for predicting ICU admission, ventilator use, and ECMO use. HRs for the sST2, SOFA score, and 4C mortality score were significantly high except for critical disease. Based on our data, sST2 might be a useful biomarker for predicting clinical outcomes in COVID-19. Since COVID-19 patients could be aggravated rapidly, resulting in unexpected poor outcomes, it is important to predict COVID-19 prognosis earlier [31]. The WHO recommends monitoring vital signs, clinical warning scores, laboratory data, electrocardiogram, or chest imaging to detect a deteriorating COVID-19 patient or complications [3,31]. These signs may not be monitored in a timely manner. In addition, clinical assessments such as the SOFA score, WHO disease severity, and 4C mortality score need various clinical and laboratory data or complex calculations. A simple biomarker would be more objective and appropriate in the real hospital setting [31]. Based on our data, sST2 could be used as a simple, objective parameter for detecting patients with critical care demands, including ICU, ventilator, and ECMO, and predicting 30-day mortality in clinical practice [31].
sST2 was superior to CRP and IL-6 for predicting ICU admission and mortality [24]. sST2 and clinical scores comparably predicted ICU admission for ventilator use and inhospital mortality [29]. sST2 has a low biological variability and reference change values, and it is not affected by sex, age, body mass index, atrial fibrillation, renal function, or the prior diagnosis of HF [12]. In serial measurements, the sST2 level showed highly dynamic change according to the disease course in both COVID-19 and HF [23][24][25]36]. Accordingly, sST2 may be a reliable biomarker for the serial monitoring of hospitalized COVID-19 patients [12,38]. On the other hand, it is also necessary to consider the analytical performance and difference of sST2 assays to implement sST2 as a prognostic biomarker for COVID-19 [13,[38][39][40].
This study had several limitations. First, although a sample size of 52 was sufficient for the Kaplan-Meier survival analysis, it might not be sufficient for providing a meaningful result. In addition, a smaller sample size would result in a wide CI because of the margin of error. However, a previous study reported that larger sample sizes reduce variance but do not improve AUC regression [41]. Further studies with a larger sample size are needed to validate our findings. Second, the study population had a skewed distribution toward critical disease, and our data may be biased and not representative. Third, the duration from the symptom onset to routine blood tests after admission varied, and it was difficult to obtain blood samples at a fixed time during the first and second waves of the COVID-19 pandemic [5,42,43]. The heterogeneous disease course of COVID-19 may have affected our data. Fourth, we focused on the prognostic utility of sST2 for clinical outcomes in COVID-19. The prediction of HF due to COVID-19 was out of the scope of our study due to insufficient information on it.
In conclusion, this is the first study to explore the prognostic utility of sST2 for predicting clinical outcomes in hospitalized COVID-19 patients compared with SOFA score, critical disease, and 4C mortality score. sST2 predicted ICU admission, ventilator use, ECMO use, and 30-day mortality comparably to SOFA score but significantly better than critical disease. sST2 predicted ICU admission, ventilator use, and ECMO use significantly better than the 4C mortality score. sST2 could be a useful biomarker for predicting clinical outcomes in hospitalized COVID-19 patients. Further studies are needed to implement sST2 as a prognostic biomarker for COVID-19 in routine clinical practice.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.