Could serum total cortisol level at admission predict mortality due to coronavirus disease 2019 in the intensive care unit? A prospective study

ABSTRACT BACKGROUND: Critical diseases usually cause hypercortisolemia via activation of the hypothalamic-pituitary-adrenal axis. OBJECTIVES: To investigate the relationship between serum total cortisol level and mortality among coronavirus disease 2019 (COVID-19) patients in the intensive care unit (ICU), at the time of their admission. DESIGN AND SETTING: Prospective study developed in a pandemic hospital in the city of Şırnak, Turkey. METHODS: We compared the serum total cortisol levels of 285 patients (141 COVID-19-negative patients and 144 COVID-19-positive patients) followed up in the ICU. RESULTS: The median cortisol level of COVID-19-positive patients was higher than that of COVID-19 negative patients (21.84 μg/dl versus 16.47 μg/dl; P < 0.001). In multivariate logistic regression analysis, mortality was associated with higher cortisol level (odds ratio: 1.20; 95% confidence interval: 1.08-1.35; P = 0.001). The cortisol cutoff point was 31 μg/dl (855 nmol/l) for predicting mortality among COVID-19-positive patients (area under the curve 0.932; sensitivity 59%; and specificity 95%). Among the COVID-19 positive patients with cortisol level ≤ 31 μg/dl (79%; 114 patients), the median survival was higher than among those with cortisol level > 31 μg/dl (21%; 30 patients) (32 days versus 19 days; log-rank test P < 0.001). CONCLUSION: Very high cortisol levels are associated with severe illness and increased risk of death, among COVID-19 patients in the ICU.

Samples were collected in the morning, between 7 AM and 8 AM, of the day when the patient was admitted to the ICU.
These were used to evaluate the complete blood count, creatinine level, aspartate aminotransferase (AST) level, alanine aminotransferase (ALT) level, albumin level, C-reactive protein (CRP) level, D-dimer level and serum total cortisol level.
Cortisol levels were analyzed using the Cobas 6000 analyzer (Roche Diagnostics, Mannheim, Germany). The reference range for serum total cortisol measured using this device was 0.018-63.4 μg/dl.

Exclusion criteria
Patients with a history of previously known pituitary disorder, adrenal gland disorder, use of corticosteroids (inhaler, topical, oral or parenteral) or use of other drugs that could have disrupted the HPA axis during the previous 3 months, and those on mechanical ventilation at the time of sample collection were excluded from the study. All other patients aged > 18 years were enrolled.

Statistical analyses
The data were analyzed statistically using the Statistical Package for the Social Sciences (SPSS) for Windows software, version 22 (IBM, Chicago, Illinois, United States). Continuous variables were presented as medians with the corresponding interquartile range (IQR), and categorical variables were presented as frequencies and percentages. The chi-square test was used to analyze the categorical variables. Comparisons were made using the Mann-Whitney U test, Kruskal-Wallis test and one-way analysis of variance (ANOVA) test. Spearman's correlation test was used to examine the relationships between pairs of variables.
Univariate and multivariate logistic regression analyses were used to detect predictors of mortality. Receiver operating characteristic (ROC) analysis and the area under the curve (AUC) were used to examine the serum total cortisol levels with regard to predicting patient mortality. Sensitivity and specificity values were calculated, and the optimal cutoff levels for serum cortisol were defined. Kaplan-Meier survival curves and the log-rank test were used to examine overall survival. P < 0.05 was considered to indicate statistical significance.
While the proportion of the patients who died was 12% (n = 17) in group 1, it was 30.5% in group 2 (n = 44; P < 0.001).
The demographic, clinical and laboratory parameters of the groups are shown in Table 1.  Table 2.
The patients in the groups were divided into subgroups as survivors and non-survivors, and their cortisol levels were compared.
Univariate logistic regression analysis on the factors that influenced mortality among the COVID-19-positive patients in the ICU, including male sex, cortisol level, CRP level and albumin level, showed significant results ( Table 4). Multivariate logistic regression analysis on the factors affecting mortality showed that the cortisol level (odds ratio: 1.20; 95% CI: 1.08-1.35; P = 0.001) was a significant factor ( Table 5).

DISCUSSION
Despite all the support and treatments for COVID-19, the mortality rates among patients admitted to ICUs remain high across the world. In a study on 3,988 patients in Italy, the mortality rate in the ICU was 48.7%. 8 A study on 3,001 patients admitted to ICUs in the United Kingdom found a mortality rate of 31 %. 9 In a study on patients in 178 ICUs in Turkey, the mortality rate was 55.6%. 10 In our study, the mortality rate was 12% among COVID-19-negative patients who were admitted to the ICU, while it was 30.5% among COVID-19-positive patients. In stress-free individuals, cortisol is secreted in a daily pattern, with levels peaking in the early morning and dropping to their lowest level in the late evening. Any kind of acute illness or trauma leads to changes in daily cortisol secretion, 14,15 often accompanied by hypercortisolemia in proportion to disease severity, in cases of critical illness that causes severe acute physical stress. [16][17][18] The increase in cortisol release owing to acute stress in the ICU  is an adaptive mechanism of the body that triggers regulation of cardiovascular, immune and metabolic functions. An appropriate response from the HPA axis to the severe stress of critical illness is essential for survival because both very high cortisol responses and low responses (relative adrenal insufficiency) have been associated with higher mortality rates. [19][20][21] In our study, the total serum cortisol level on the first morning of hospitalization in the ICU was significantly higher in the COVID-19-positive patients than in the COVID-19-negative patients (21.84 versus 16.47 μg/dl; P < 0.001).
This may be attributable to the fact that COVID-19positive patients had more severe disease and higher CRP level.
Specific cytokines, such as CRP, which presents elevated concen- versus 519 nmol/l; P < 0.0001). 23 The COVID-19-positive group in that previous study had higher CRP levels, similar to those of our study; thus, we can conclude that cortisol elevation may be related to disease severity. In another study on 62 patients with severe sepsis and 63 with septic shock who were admitted to an ICU, the baseline total cortisol levels were 728 ± 386 nmol/l and 793 ± 439 nmol/l, respectively. Non-survivors had higher calculated total cortisol concentrations (980 ± 458 nmol/l) than the survivors (704 ± 383 nmol/l). 24 The lowest median cortisol level was in group 1 survivors, while the group 2 non-survivors had the highest cortisol level in the multiple-group comparisons that were performed by dividing the study subjects into subgroups of survivors and non-survivors. In the multivariate logistic regression analysis, in which we examined the effect of clinical parameters on mortality among COVID-19-positive patients, only the cortisol level was significant.
The cortisol cutoff point was 31 μg/dl (855 nmol/l) in the ROC curve analysis that was performed to predict mortality among the COVID-19-positive patients (sensitivity 59% and specificity 95%). It is challenging to predict, on the day of ICU admission, which patients are likely to die. In fact, it is difficult for clinicians to select patients for whom more time should be devoted, given the limited time and resources. In such cases, the serum total cortisol level could facilitate and guide the decision-making process for clinicians.      The graph was categorized as cortisol concentration ≤ 31 μg/ dl or > 31 μg/dl. marker that could be used to estimate mortality among COVID-19-patients admitted to the ICU.
The present study had certain limitations. First, we only performed analysis using a single baseline cortisol level that was measured on the first morning of admission to the ICU due to COVID-19. In our ICU, corticosteroid treatment is frequently applied from the first day of admission; therefore, measuring the serum cortisol levels under corticosteroid treatment may have provided inaccurate results. Second, we did not measure the level of adrenocorticotropic hormone (ACTH), cortisol-binding globulin or free cortisol. Therefore, it was not possible to comprehensively evaluate the effects of COVID-19 on the HPA axis.
Third, no ACTH stimulation test (synacthen test) was performed on the patients before starting this study. Therefore, an unknown state of adrenal insufficiency may have been overlooked. Lastly, this study was conducted in a single center. Larger, multicenter studies are required, in order to obtain more conclusive evidence.

CONCLUSION
Elevated cortisol level is an independent biomarker that enables prediction of adverse outcomes and mortality among COVID-19-positive patients admitted to the ICU. The ability to predict which patients in the ICU may deteriorate faster will help clinicians to allocate resources appropriately and raise the standard of patient care. Furthermore, we can consider a patient's cortisol levels while making a decision regarding the treatment approach.