Serial measurement of circulating calprotectin as a prognostic biomarker in COVID-19 patients in intensive care setting

Introduction

Coronavirus disease 2019 (COVID-19), caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is characterized by a broad spectrum of clinical manifestations. The majority of patients are asymptomatic or have a mild disease, but some develop a respiratory failure requiring hospitalization, or even an acute respiratory distress syndrome and multiple organ failure leading to intensive care unit (ICU) admission or death [1, 2]. The varying disease severity significantly complicates patient care and puts enormous pressure on healthcare systems worldwide. In addition, the emergence of new variants of concern initiated new pandemic waves worldwide in both vaccinated and previously infected patients [3]. Although several risk factors and comorbidities related to a worse outcome have been described [2, 4], [5], [6], [7], early prediction of COVID-19 severity remains challenging [7, 8].

Calprotectin (CLP) is a heterodimeric complex formed by two calcium-binding proteins S100A8 and S100A9, also known as myeloid-related protein (MRP)-8 and MRP-14, and is part of the innate immune system. CLP constitutes more than half of the cytosolic protein content in neutrophils, monocytes and in activated macrophages [9]. Extracellular release of CLP is mediated through cell activation, neutrophil extracellular trap (NET) formation and NETosis. Here, CLP acts as an endogenous ligand of Toll-like receptor 4 (TLR4) and receptor for advanced glycation end products (RAGE) contributing to the inflammatory process through the modulation of leukocyte chemotaxis and expression of pro- and anti-inflammatory mediators [10, 11]. Whereas fecal calprotectin has already proven to be a reliable biomarker in the diagnosis and follow-up of inflammatory bowel disease [12], circulating calprotectin (cCLP) has recently gained attention as a systemic biomarker of neutrophil-related inflammation in both acute [13] and chronic inflammatory disorders [14].

Severe COVID-19 is associated with a hyperinflammatory response characterized by an atypical cytokine storm, lymphopenia and an increased count of aberrantly hyperactivated neutrophils. This hyperactivation is associated with increased liberation of neutrophil derived proteases, NETs and chemotactic proteins [15], [16], [17]. Two recent meta-analyses concluded that high levels of cCLP correlate with Covid-19 severity [17, 18]. Several studies have shown that cCLP levels at time-point of COVID diagnosis enabled prediction of ICU admission [19], [20], [21], need for mechanical ventilation [19, 22], development of multiple organ failure [23] and mortality [20, 24]. In contrast, studies on serial measurements of cCLP as a monitoring parameter of COVID-19 are limited [25], [26], [27].

The aim of this monocentric study was to investigate the prognostic value of serial measurements of cCLP in COVID-19 patients admitted to ICU in comparison to C-reactive protein (CRP).

Materials and methods

Results

Association of biomarkers levels with LOS-ICU

No significant interactions between time and LOS-ICU for both cCLP and CRP were reported (Table 1, Figure 1 [Panel A, B]). Significant main effects of LOS-ICU were reported for both biomarkers showing that longer LOS-ICU is associated with longitudinally higher biomarker levels (p-value LOS-ICU <0.001). These findings were observed whether deceased patients were included in- or excluded from the analysis.

Table 1:

Results of interaction model and main effects model of association biomarker level with length-of-stay at intensive care unit (LOS-ICU).

Association biomarker level with	p-Value interaction	Interaction model			Main effects model
		Group	Slope (95% CI)	p-Value slope	Doubling in LOS-ICU (95% CI)	p-Value LOS-ICU	Doubling in time (95% CI)	p-Value time
LOS-ICU
cCLP SERUM	0.924	/	/	/	3.3 µg/mL (1.5; 5.2)	<0.001	−1.7 µg/mL (−43.9; 0.6)	0.140
CRP	0.065	/	/	/	17.9 mg/L (9.7; 26.1)	<0.001	−24.0 mg/L (−31.7;-16.3)	<0.001

1. No significant interaction (p-value interaction) between time and LOS-ICU has been found, so the biomarker of interest does not evolve different over time for the quartiles of LOS-ICU (Q1, Median, Q3). As such, main effects of LOS-ICU time are reported using 1) the estimated difference in biomarker level for a doubling in LOS-ICU, averaged over time of measurement (“LOS-ICU”) and 2) the estimated change in biomarker level for a doubling in time of measurement, averaged over all levels of LOS-ICU (“Time”). A detailed description of the statistical model used is provided in Supplementary Data S2. Significant p-values are highlighted in bold. C.I., confidence interval; cCLP, circulating calprotectin; CRP, C-reactive protein.

Figure 1:

Model-predicted biomarker level over time. (A, B) Model-predicted evolution for patients with short/median/long length-of-stay at intensive care unit (LOS-ICU). (C, D) Model predicted evolution for patients with mechanical ventilation and without mechanical ventilation, with 95% confidence intervals. (E, F) Model-predicted evolution for deceased and non-deceased patients, with 95% confidence intervals. (A, C, E) circulating calprotectin (cCLP) levels measured in serum (EliA™ Calprotectin 2, Phadia™ 200, RUO serum/plasma protocol, Thermo Fisher Scientific, Phadia AB, Sweden). (B, D, F) C-reactive protein levels (cobas c501/503 analyzer, Roche Diagnostics, Mannheim, Germany). A detailed description of the statistical model used is provided in Supplementary Data S2.

Discussion

High cCLP levels at COVID-19 diagnosis have been suggested as an interesting biomarker predicting COVID-19 disease severity and poor outcome [17], [18], [19], [20], [21], [22], [23], [24]. However, studies that investigated the prognostic value of serial cCLP measurements in COVID-19 are limited. So far, we were able to identify only two studies that used longitudinally collected samples for cCLP analysis [22, 29]. Despite the relatively small sample size of 36 subjects and inconsistent time intervals between sequential sampling, Shi et al. demonstrated an upward trend of cCLP levels in six patients with worsening oxygenation. In this study, only two follow-up samples were included, and an objective definition of ‘worsening oxygenation’ is lacking [22]. The study of Chapuis et al. investigated the added value of longitudinal cCLP determinations in a large cohort (n=626) of COVID-19 patients with moderate disease severity, requiring hospitalization in a standard care unit. The authors were able to identify three possible trajectories, with increasing cCLP levels being predictive for a higher probability of a poor outcome (i.e. ICU transfer or death) [29].

The results of our study suggest that longitudinal follow-up of cCLP values in an ICU-setting may provide information regarding certain clinical outcomes. Despite the exclusion of deceased patients, cCLP (both matrices) and CRP levels were longitudinally higher in patients with longer LOS-ICU. This finding is consistent with earlier studies [19, 29]. No significant cCLP variation over time related to LOS-ICU was observed. However, this may be regarded as an advantage of cCLP over other more commonly determined biomarkers like CRP and procalcitonin, which can fluctuate strongly during the stay at ICU due to the influence of a potential concomitant bacterial coinfection.

At any time, cCLP levels were longitudinally higher in patients needing mechanical ventilation compared to the non-mechanical ventilated subgroup, however without a significant increase during ICU stay. Previous studies [19, 22] already indicated that systemic neutrophilic activation and subsequent calprotectin release is a major immunopathological feature of COVID-19 contributing to its severity and hence leading to a need for mechanical ventilation [16, 17]. Theoretically, there could be an impact of mechanical ventilation on cCLP levels since mechanical ventilation causes neutrophilic activation in the lungs and associated NET release in the alveolar space [30]. However, since no follow-up sample was collected in 13/25 patients requiring mechanical ventilation after its initiation, we were not able to investigate this possible interaction. In contrast, CRP values decrease over time for both ventilation patient subgroups.

Monitoring cCLP values during hospitalization at ICU may be particularly useful in predicting mortality: patients in the deceased group showed a significant increase in cCLP values. This increase was remarkable during the first two days after ICU admission. In contrast, in the non-deceased group, a non-significant decrease in cCLP values was observed. Regarding serial measurements of CRP, our results revealed a non-significant decrease over time in both the deceased and non-deceased groups. These results suggest that serial determinations of cCLP in the first few days after ICU admission could be a valuable strategy to provide the clinician quickly with useful prognostic information regarding mortality and at the same time limiting the financial impact by reducing the total number of determinations. Hence, cCLP has the potential to outperform CRP in predicting mortality in patients with COVID-19 in the ICU.

Finally, we did not find a statistically significant interaction between cCLP levels and the administration of medication. In our study population, almost all patients received corticosteroids, which did not seem to alter cCLP levels. The absence of a significantly different evolution of cCLP levels between subjects requiring antibiotics and those who did not is in line with the findings of two other recent studies. Cambier et al. demonstrated that the cytokine storm in patients with COVID-19 persisted or even increased during antimicrobial treatment for bacterial coinfection [15]. Silvin et al. on the other hand showed that the presence of a concomitant bacterial infection did not significantly alter the release of calprotectin [17]. This may be an advantage of calprotectin over other conventional biomarkers like CRP and procalcitonin, which are influenced by bacterial coinfection. CRP levels in our study were significantly longitudinally higher in the group that received antibiotic therapy. This can be explained by the fact that increasing CRP levels, together with clinical signs of coinfection and the identification of bacterial species in microbiological samples, encourage the physician to start antibiotic therapy.

Tocilizumab, an IL-6 receptor blocking drug, was administered to five COVID-19 patients in our study. Although treatment with IL-6 inhibitors rapidly blocks CRP secretion by the liver, we didn’t observe statistically significant differences in CRP nor cCLP levels between the medication subgroups, neither after correcting for longitudinal medication use in the various outcomes, potentially because of low sample size. However, in patients treated with IL-6 inhibitors, CRP should be regarded as futile for further patient follow-up and other biomarkers are warranted.

The use of antiviral agents in our study population was limited to the administration of remdesivir in 35/66 patients. Remdesivir was used in patients with a high viral load, believed to be in the early stage of the disease. We did not see a significant correlation between remdesivir administration and the evolution of cCLP values over time. However, in the group that received remdesivir, lower CRP values were observed.

The strength of our study is that it investigated the prognostic value of longitudinal results for two different biomarkers, in samples collected at fixed time intervals and for longer periods during patient stay at ICU. The study assessed the added value of serial cCLP measurements, in comparison to CRP, in predicting LOS at ICU, the need for mechanical ventilation and mortality.

Limitations include the lack of a control group and the relatively small sample size. In addition, for 13/66 (19.7%) patients, no follow-up samples were collected after initiation of mechanical ventilation. Next, as our study lacks complete data on hematological parameters, we were not able to correlate neutrophil counts nor neutrophil/lymphocyte ratios, integrated in several risk score formulations, with cCLP levels. Furthermore, besides CRP, no other biomarkers were taken into account. Further research should compare sequential analysis of cCLP against other markers of inflammation like IL-6 and procalcitonin [31], [32], [33]. In analogy with the recently published paper of Chapuis et al., an algorithm could be created based on longitudinal cCLP values to better predict outcomes in severe COVID-19 patients at ICU [29]. To further clarify the influence of medication therapy on cCLP levels, larger prospective studies are needed. At last, since CLP is being released from neutrophils locally, serial measurement of CLP in Bronchoalveolar Lavage liquid could be interesting in follow up of COVID-19 related lung injury.

Conclusions

Our study results suggest that the predictive value of baseline cCLP levels in COVID-19 patients could be refined by serial monitoring of the biomarker. Especially regarding mechanical ventilation and mortality, serial cCLP measurements have the potential to outperform CRP since confounding factors such as age and concurrent bacterial infection have only minor, non-significant effects on cCLP levels. Additionally, cCLP is synthesized and released locally by neutrophils whereas most of the other biomarkers are synthesized in the liver, and thus are prone to be affected by comorbidities and medication use [17].