Predictive markers related to local and systemic inflammation in severe COVID-19-associated ARDS: a prospective single-center analysis

Abstract

Background: As the COVID-19 pandemic strains healthcare systems worldwide, finding predictive markers of severe courses remains urgent. Most research so far was limited to selective questions hindering general assumptions for short- and long-term outcome.

Methods: In this prospective single-center biomarker study, 47 blood- and 21 bronchoalveolar lavage (BAL) samples were collected from 47 COVID-19 intensive care unit (ICU) patients upon admission. Expression of inflammatory markers toll-like receptor 3 (TLR3), heme oxygenase-1 (HO-1), interleukin (IL)-6, IL-8, leukocyte counts, procalcitonin (PCT) and carboxyhemoglobin (CO-Hb) was compared to clinical course. Clinical assessment comprised acute local organ damage, acute systemic damage, mortality and outcome after 6 months.

Results: PCT correlated with acute systemic damage and was the best predictor for quality of life (QoL) after 6 months (r = - 0.4647, p = 0.0338). Systemic TLR3 negatively correlated with impaired lung function (ECMO/ECLS: r = - 0.3810, p = 0.0107) and neurological short- (RASS mean: r = 0.4474, p = 0.0023) and long-term outcome (mRS after 6 m: r = - 0.3184, p = 0.0352). Systemic IL-8 correlated with impaired lung function (ECMO/ECLS: r = 0.3784, p = 0.0161) and neurological involvement (RASS mean: r = - 0.5132, p = 0.0007). IL-6 in BAL correlated better to the clinical course than systemic IL-6. Using three multivariate regression models, we describe prediction models for local and systemic damage as well as QoL. CO-Hb mean and max were associated with higher mortality.

Conclusions: Our predictive models using the combination of Charlson Comorbidity Index, sex, procalcitonin, systemic TLR3 expression and IL-6 and IL-8 in BAL were able to describe a broad range of clinically relevant outcomes in patients with severe COVID-19-associated ARDS. Using these models might proof useful in risk stratification and predicting disease course in the future. Trial registration The trial was registered with the German Clinical Trials Register (Trial-ID DRKS00021522, registered 22/04/2020).

Keywords: BAL; COVID-19; Carboxyhemoglobin; ICU; IL-6; IL-8; Outcome assessment; Quality of Life; TLR3.

Fig. 1

Schematic description of the study design. BAL bronchoalveolar lavage, ICU intensive care unit, QoL quality of life

Fig. 2

Heat map showing correlations of biomarkers vs. clinical and neurological short- and long-time outcome. Only r-values for significant correlations are shown. Depending on whether data showed normal distribution, Spearman nonparametrical or Pearson parametrical test was used. Cells with “X” did not reach significance. 1st PCT vs. 1st PCT not shown. For p-values, n and parameter statistics in multivariate analyses see Table 2. Correlations are visualised in Fig. 3. LOS ICU length of stay in intensive care unit, Ppeak mean peak ventilation pressure, NO nitric oxide, ECMO/ECLS extracorporeal membrane oxygenation/extracorporeal life support, RASS Richmond Agitation-Sedation Scale, mRS modified Rankin scale, CO-Hb carboxyhaemoglobin, BAL bronchoalveolar lavage

Fig. 3

Network graphics showing correlations of biomarkers vs. outcome. a 1st Leukos and 1st PCT vs. outcome. b TLR3 in blood and BAL vs. outcome. c Cytokine secretion in blood vs. outcome. d Cytokine secretion in BAL vs. outcome. Color code: grey boxes, biomarker, blue boxes, parameter for lung impairment, yellow boxes, parameter for neurological involvement and QoL, red arrow, positive correlation, green arrow, negative correlation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. LOS ICU length of stay in intensive care unit, Ppeak mean peak ventilation pressure, NO nitric oxide, ECMO/ECLS extracorporeal membrane oxygenation/extracorporeal life support, RASS Richmond Agitation-Sedation Scale, mRS modified Rankin scale, BAL bronchoalveolar lavage

Fig. 4

Multiple logistic regression models for local and systemic damage as well as quality of life. Model for local damage: a need for ECMO/ECLS and b need for dialysis. Model for systemic damage: c occurrence of thromboembolic events and d mortality. Model for quality of life: e EQ-5D-5L index. Comparison method: Akaike’s Information Criterion, Tjur’s R squared, Hosmer–Lemeshow hypothesis test. For Cut-off values and statistics see Table 3. ROC receiver operating characteristic

Fig. 5

Patients with higher CO-Hb more likely to die. a Increase in CO-Hb values from day one to day five. Wilcoxon matched-pairs signed rank test. Contingency diagrams showing distribution of alive/deceased between patient groups of high/low b CO-Hb max (Fisher’s exact test: p = 0.0102, Koopman asymptotic score: RR = 2.186, 95%-CI 1.300–3.502, Baptista-Pike: OR = 12.86 95%-CI 1.947–146.9, Cut-off: 3%) and c CO-Hb mean (Fisher’s exact test: p = 0.0079, Koopman asymptotic score: RR = 2.186, 95%-CI 1.295–3.686, Baptista-Pike: OR = 8.708 95%-CI 1.700–42.61, Cut-off: 2%). ****p < 0.0001. CO-Hb carboxyhemoglobin