Krebs von den Lungen 6 (KL-6) levels in COVID-19 ICU patients are associated with mortality

Background Krebs von den Lungen 6 (KL-6) is a high-molecular-weight mucin-like glycoprotein, which is also known as MUC1. KL-6 is mainly produced by type 2 pneumocytes and bronchial epithelial cells, and, therefore, elevated circulating KL-6 levels may denote disorders of the alveolar epithelial lining. The objective of this study is to verify if KL-6 serum level might support ICU physicians in predicting mortality, risk stratifying and triaging severe COVID-19 patients. Methods A retrospective cohort study, including all the COVID-19 patients who measured KL-6 serum values at least once during their ICU stay, was performed. The study sample, 122 patients, was divided in two groups, according to the median KL-6 value at ICU admission (median log-transformed KL-6 value: 6.73 U/ml; group A: KL-6 lower than the median and group B: KL-6 higher than the median). Results One-hundred twenty-two ICU patients were included in this study. Mortality was higher in group B than in group A (80 versus 46%; p < 0.001); both linear and logistic multivariate analyses showed ratio of arterial partial pressure of oxygen to fraction of inspired oxygen (P/F) significantly and inversely related to KL-6 values. Conclusion At ICU admission, KL-6 serum level was significantly higher in the most hypoxic COVID-19 patients and independently associated with ICU mortality.


Background/rationale
COVID-19 mainly affects the respiratory system by causing a low oxygenation index, even if clinical manifestations can be not very prominent in many hypoxic patients, with no complaint of dyspnoea, no significant increase in respiratory rate, and no respiratory distress (as in the "silent hypoxia" COVID-19 clinical scenario) [1]. It has been reported that about 33% of COVID-19 victims developed acute respiratory distress syndrome (ARDS) [2]. COVID-19 ARDS (CARDS) patients should be admitted to the intensive care unit (ICU), the elderly and those with comorbidities being at highest risk of death [3,4]. CARDS was recently classified into 3 different categories depending on hypoxemia [5]: mild hypoxemia (between 200 and 300 mmHg), mild-moderate (150 to 200 mmHg), and moderate-severe hypoxemia (< 150 mmHg). CARDS patients show some differentiations with ARDS caused by other factors [4], but type 2 pneumocytes damage or transformation is a very relevant element in both CARDS and ARDS by other viral causes. In fact, in case of alveolar type 2 cells injury, removal of alveolar oedema fluid is compromised. Moreover, type 2 cells harm reduces the production and turnover of surfactant, which is associated with poor outcome. Mortality of CARDS is evolving, thanks to the efforts of the international scientific community. Extracorporeal membrane oxygenation (ECMO) has been used to treat COVID-19 ARDS, even if there are still debates on ethical considerations and end-of-life decision-making [6]. Extracorporeal CO2 removal systems (ECCO2R), developed to correct hypercapnia and promoting ultra-protective ventilation protocols, have been applied to COVID patients too [7].
The Krebs von den Lungen 6 (KL-6) protein is a highmolecular-weight mucin-like glycoprotein, also known as MUC1 [5,8], produced by type 2 pneumocytes and bronchial epithelial cells, which may provide insight into the ARDS pathophysiology. KL-6 is known as an immunological biomarker reflecting the severity and progression of interstitial lung disease (ILD), a form of chronic fibrosing interstitial pneumonia with various aetiology [9]. In a recent systematic review [10], KL-6 concentration was higher in severe COVID-19 patients than in nonsevere patients (95% CI: 0.99-1.5; p < 0.001). However, in this systematic review, including 136 severe COVID-19 patients, the criteria for COVID-19 classifications as severe and non-severe were not homogeneous among the 6 covered studies.

Objectives
Since elevated circulating levels of KL-6 indicate disruption of the alveolar epithelial lining, we hypothesized that it could be advantageous analysing KL-6/MUC1 in a group of severe COVID-19-positive patients, to better risk stratify and triage them. Central question of this study was investigating the association between KL-6 serum levels and ICU COVID-19 mortality.

Study design and setting
This is a single-centre, retrospective observational study performed at the Salerno University COVID Hospital "G. Da Procida", which was designated as a COVID-only medical centre in Salerno, Italy, on the 8th of October 2020.

Participants
The ICU admission was reserved to severe COVID-19 patients. According to the National Institute of Health (NIH) guidelines [11], critical patients are identified as follows:

Severe illness
Individuals who have SpO2 < 94% on room air at sea level, a ratio of arterial partial pressure of oxygen to fraction of inspired oxygen (P/F) < 300 mm Hg, respiratory frequency > 30 breaths/min, or lung infiltrates > 50%.

Critical illness
Individuals who have respiratory failure, septic shock, and/or multiple organ dysfunction.
All the adult patients admitted in our COVID ICU from the 8th of October 2020 to the 15th of June 2021 were considered eligible if at least 1 KL-6 measurement was available at the time of their ICU admission. Terminal illness and pre-existent lung fibrosis were exclusion criteria.

Data sources/measurement
KL-6 was measured with commercial kit produced by Tosoh (ST AIA-PACK KL6) following producer instructions [12] on left over blood samples, on the 1st and 5th day of patients ICU stay.
The study sample was divided in 2 groups: group A with KL-6 serum concentration at ICU admission lower than the median KL-6 concentration and group B with KL-6 higher than the median value at ICU admission (median log-transformed KL-6 value: 6.73 U/ml). Imaging data were analysed by considering the number of interested pulmonary lobes, the presence of consolidations, crazy paving, emphysema, pneumomediastinum, and pneumothorax at CT scan.

Variables
The primary aim of this study was to assess the association between KL-6 serum levels and clinical severity scores in adult COVID-19 patients during the second pandemic wave, in order to improve early diagnosis and defining severity of COVID-19 and to provide references for clinical and laboratory research. In addition, we investigated the association between KL-6 serum levels with SOFA score and ICU COVID-19 mortality.

Statistical methods
The distribution of variables was evaluated by the Kolmogorov-Smirnov test and with graphical evaluation. Variable with positively skewed distribution were log transformed. At baseline, continuous variables were compared by student T-test or Mann-Whitney test and reported as mean ± standard deviation or median (interquartile range), according to the distribution. Categoric variable was reported as percentage, and comparison between groups was performed through the Pearson's chi-square test. Linear regression or logistic regression analyses were performed to assess the association between KL-6 and other variables. The Kaplan-Meier analysis and Cox regression were performed in the whole sample as well as patients with KL6 below and above the median. Youden index was used to compute the value which maximizes sensitivity and specificity. We used SPSS version 24, Chicago, IL, USA, and MedCalc.

Participants and descriptive data
One-hundred twenty-two severe or critically ill patients, out of 150 admitted to the Salerno University Hospital COVID ICU from the 8th of October 2020 to 15th of June 2021, were included in this retrospective cohort study; baseline characteristics of source population are listed in Table 1.

Main results
At ICU admission, KL-6 serum level was significantly lower (p = 0.00048) in the survivor's group (KL-6 median value 545 U/ml vs 1070 U/ml measured at admission in those patients who did not survive) (Figs. 1 and 2) and in those patients that were managed by noninvasive ventilation (NIV) for the whole length of their ICU stay (KL-6 median value 711 U/ml vs 1073 U/ml, measured in patients who required endotracheal intubation at admission or during their ICU stay) ( Table 2). At 5 ICU days, the difference between survivors and patients who later would die was still significant (p 0.029).
For research purpose, the study sample (122 patients) was divided in two groups, according to the median log-transformed KL-6 value (6.73 U/ml): group A (61 patients with KL-6 lower than the median) and group B (KL-6 higher than the median). Differences between the two groups, regarding clinical variables as SOFA score and comorbidities, and laboratory data collected at ICU admission as procalcitonin (PCT), white blood cells count, P/F are summarized in Table 3. Group B patients were more hypoxic at admission and aggravated by more organ failures. Mortality rate was higher in group B than in group A (80.3 versus 45.9%; p < 0.001). We did not detect significant differences in radiological features. Instead, group A and group B differed for P/F value (p 0.002) and SOFA score (p 0.005), while they were similar for age, sex, comorbidities distribution, WBC count, and PCT values, when evaluated by univariate linear and logistic regression analysis (Table 3). Pneumothorax incidence showed a trend of significance with KL-6 values in the linear regression, but this was not confirmed in the logistic analysis. P/F was significantly and inversely related to KL-6 (OR 0.99; 95% CI: 0.978/0.998; p 0.009) ( Table 4).

Outcome data
To understand the prognostic role of KL-6 in the subgroup of patients who were already sicker at ICU admission (as shown by their SOFA score, higher than the median value of SOFA calculated in the whole population, which was 5), a Kaplan-Meier analysis was performed. It showed a significant higher mortality in patients who had KL-6 value at admission > 6.73 U/ ml and SOFA score < 5.0 (median value of SOFA in this population), while in the subgroup with SOFA > 5.0 (higher than the median SOFA), the impact of KL-6 on mortality was not evident (Fig. 3). Accordingly, the Cox regression analysis showed a significant prognostic role of KL-6 on mortality in the whole sample (HR: 1.78, 95% CI 1.12-2.82, p = 0.02) as well as in the subgroup with SOFA lower than the corresponding median value (HR: 1.86, 95% CI 1.02-3.44, p = 0.045), but not in the subgroup of patients with SOFA above the median (HR 0.96, 95% CI 0.69-1.34, p = 0.96). Therefore, SOFA acted as a significant effect modifier of the KL-6/mortality link (p = 0.048).

Key results
In this retrospective prognostic study, we correlated KL-6 values at ICU admission with disease severity in COVID-19 patients during the second wave of the pandemic in a COVID-dedicated hospital. Our findings showed that elevated KL-6 (> 680 U/ml) was strongly associated with mortality in ICU, thereby contributing to the growing body of evidence for the utility of KL-6 in the context of COVID-19 infection.

Interpretation
KL-6 glycoprotein is mainly expressed on alveolar type 2 cells in the lung, and it is produced more prominently by proliferating, or regenerating, injured type 2 cells than by healthy type 2 cells. The presence of KL-6 has been used to monitor severity of disease in idiopathic pulmonary fibrosis [8]. Liu et al. [14], in a lung proteomics animal experimental research (rhesus monkeys), quantified the overall difference of protein expression pattern between control and COVID-infected groups. When compared to the control group, in the lung, 757 proteins were differentially expressed in the infected group. Our study did not check total protein expression in the lungs of COVID patients, but it aimed to identify an easy and available biomarker that physicians could use to risk stratify COVID-19 patients. The use of biomarkers to predict disease severity has proven essential for resource allocation, particularly for respiratory support needs. In a previous study on a population of 67 COVID survivors [15], median KL-6 was 365 U/ml (IQR 233-493), and the authors concluded that "high     [20], through a machine learning approach for predicting ICU mortality in COVID-19 patients, proved that age was the leading predictor, followed by total SOFA score at ICU admission, and the P/F used for SOFA calculation. In our previous experience [21] and in the present study too, patients' characteristics at ICU admission severely conditionate the success of advanced therapeutic strategies. In this paper, when SOFA score is higher than 5, prognosis is already compromised, and consequently, the prognostic power of KL-6 is reduced, while it can still be helpful to identify lung injury in patients with a lower organ failure score (SOFA < 5).
Ru et al. [22] reported that mean serum PCT levels were over eight times higher in critical COVID patients than in moderate patients. PCT levels may be associated with bacterial coinfection which might justify such an increase. In our experience, PCT was not a marker of severity of COVID lung failure at ICU admission. Nevertheless, we think that an increase in PCT should always be investigated for the presence of a concomitant bacterial infection.
Changes in KL-6 serum levels may be related to barotrauma and volutrauma too, even if it is very difficult to distinguish the origin of the biomarker release in a complex context as COVID ARDS.

Limitations
The main limitation of this research is that is an observational study with retrospective enrolment. Moreover, we think that KL-6 value as prognostic index should be tested in different COVID populations or timing, i.e. at emergency room admission, and the dynamic of its changes should be tested, to understand if it shows prognostic value for long-term COVID effects on the lung functions. In our retrospective study, in fact, multiple unmeasured variables may have affected the outcomes. Conclusions should be validated by larger, definite prospective studies in the future.

Conclusions
This paper supports the possibility to use KL-6 in a panel of biomarkers, which may help clinicians to stratify mortality risk and the need for ICU admission. It would be very advantageous to ascertain firm cutoff values for this purpose.