Introduction

At the end of 2019, Global health is being challenged in an unprecedented way by the 2019 coronavirus disease (COVID-19) pandemic. The systemic inflammatory response to the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection is a hallmark of COVID-19, and most COVID-19 patients have abnormal inflammatory biomarkers1. C-reactive protein (CRP), an acute phase protein first described by Tillet and Francis, is synthesized by the liver in response to interleukin 6 (IL-6) and is a widely used biomarker of inflammation2. Elevated CRP concentrations are associated with cardiovascular disease and acute kidney injury3, and with inflammatory rheumatic diseases such as rheumatoid arthritis and gout4. CRP is also related to the severity of H1N1 influenza pneumonia patients5. Besides, most studies in recent years have pointed out that for COVID-19 patients, higher CRP concentrations are closely related to higher mortality risk6,7,8,9,10,11,12. However, some studies have pointed out that CRP may play a protective role in alveolitis, and in patients with acute lung injury, an increase in CRP levels is associated with a decrease in mortality13. Therefore, it is still controversial whether the increase of CRP in the early stage of COVID-19 can be used as a prognostic indicator. Moreover, most studies had small sample sizes and did not consider the curvilinear relationship between CRP and mortality. The aim of this retrospective cohort study was to determine whether elevated levels of CRP measured in the early stages of COVID-19 were associated with higher mortality in COVID-19 patients.

Methods

This is a retrospective cohort study of all patients admitted to 4 hospitals within the Montefiore Health System between March 1 and April 16, 2020, with SARS-CoV-2 infection. Information on demographics, comorbidities, admission laboratory values, discharge, and mortality was identifified through a health care surveillance software package (Clinical Looking Glass; Streamline Health, Atlanta, GA) and review of the primary medical records. The design details in this study have been previously reported14.

Data source

We downloaded the raw data free of charge from the DATADRYAD database provided by Eskandar et al. Source: Altschul, David (2021), Neurologic complications of COVID-19, Dryad, Dataset, <https://doi.org/10.5061/dryad.7d7wm37sz>. Under Dryad's Terms of Service, researchers can use this data for secondary analysis without infringing on the authors' rights.

Study population

All patients with real-time reverse transcriptase PCR positive assay testing for SARS-CoV-2 RNA were included.For patients with multiple admissions, only the last reported was considered for analysis.Data were collected on May 7, 2020.Exclusion criteria were as follows: (1) patients who were not admitted or died before admission were excluded because they rarely had a complete laboratory study panel and a complete neurological evaluation could not be assessed, (2) CRP = 0 mg/L or Patients with CRP ≥ 100 mg/L. The flow chart of the study is shown in Fig. 1.

Figure 1
figure 1

Flow chart of study population.

Full size image

Variables

The variable for this study was C-reactive protein, and the covariate was chosen based on our clinical experience, original studies, and other studies investigating risk factors for progression in COVID-19 patients. Based on the above principles, the following variables are therefore used as covariates: (1) Continuous variables: age, D-Dimer, temperature, oxygen saturation, mean arterial pressure, platelets, INR, BUN, creatinine, sodium, glucose, AST, WBC, ALT, lymphocytes, interleukin-6, ferritin, procalcitonin, troponin; (2) Categorical variables: ethnicity, myocardial infarction, peripheral vascular disease, congestive heart failure, cerebrovascular disease, dementia, chronic obstructive pulmonary disease, diabetes mellitus simple, renal disease, stroke.

Outcomes

Te outcome of the study was all-cause mortality.

Statistical analysis

Continuous variables were described as the means ± SD or median and interquartile ranges (IQR). Categorical data were presented as numbers and percentages. The diference according to the tertiles of the CRP was compared using one-way analysis of variance (ANOVA) for continuous data and chi-squared tests for categorical variables.

Calculations for Kaplan Meier survival were performed using Kaplan Meier analysis, and a Kaplan Meier curve was developed as a result.

We used a generalized additive model (GAM) to investigate the dose–response relationship between the CRP and mortality (Fig. 2). Adjusted for age, D-Dimer, temperature, oxygen saturation, mean arterial pressure, platelets, INR, BUN, creatinine, sodium, glucose, AST, WBC, ALT, lymphocytes, interleukin-6, ferritin, procalcitonin, troponin, ethnicity, myocardial infarction, peripheral vascular disease, congestive heart failure, cerebrovascular disease, dementia, chronic obstructive pulmonary disease, diabetes mellitus simple, renal disease, stroke. Estimating the hazard ratio (HR) of death was done using Cox proportional hazards modeling .

We then used a two-piece-wise linear regression model to examine the threshold effect of the CRP on mortality (Table 3). The turning point for the CRP was determined using “exploratory” analyses, which is to move the trial turning point along the predefned interval and pick up the one which gave maximum model likelihood. We also performed a log-likelihood ratio test and compared the one-line linear regression model with the two-piece-wise linear model. All the statistical analyses were performed using the EmpowerStats (www.empowerstats.com, X&Y solutions, Inc. Boston MA) and R software version 3.6.1 (http://www.r-project.org)15.

Ethics approval and consent to participate

The study received approval from Albert Einstein College of Medicine, Montefiore Medical Center ethical standards committee on human experimentation. Written informed consent was waived by the Albert Einstein College of Medicine, Montefiore Medical Center ethical standards committee given the retrospective design of the study. In addition to this, Methods of the study were performed in accordance with the relevant guidelines and regulations.

Results

Baseline characteristics

Data from 3545 patients were analyzed. The median age is 63.7 years (IQR54-76 years). Table 1 compares patient’s demographics, vital signs, laboratory results, and complications by tertile of CRP. Subjects in the highest tertile of CRP were older and had lower Oxygensaturation on admission than those in the lowest tertile of CRP (Table 1).

Table 1 Baseline characteristics and mortality according to the tertiles of the CRP (n = 3545).
Full size table
Figure 2
figure 2

Associations between the CRP and mortality in all patients with Covid-19. A threshold, nonlinear association between the CRP and mortality was found in a generalized additive model (GAM). Red rad line represents the smooth curve fit between variables. Blue bands represent the 95% of confdence interval from the fit. Adjusted for age, D-Dimer, temperature, oxygen saturation, mean arterial pressure, platelets, INR, BUN, creatinine, sodium, glucose, AST, WBC, ALT, lymphocytes, interleukin-6, ferritin, procalcitonin, troponin, ethnicity, myocardial infarction, peripheral vascular disease, congestive heart failure, cerebrovascular disease, dementia, chronic obstructive pulmonary disease, diabetes mellitus simple, renal disease, stroke.

Full size image

Mortality

The mortality rate was 918/3545 = 25.90% in our cohort. The mortality rate from the lowest tertile (0.50–5.50) to the highest (15.00–68.40) CRP was 141 (11.94%), 286 (24.32%), and 491 (41.33%) (Table 1).

Unadjusted association between baseline variables and mortality

In the univariate COX regression analysis model, CRP, age, D-Dimer, oxygen saturation, mean arterial pressure, BUN, creatinine, sodium, glucose, AST, WBC, lymphocytes, interleukin-6 and Procalcitonin are all correlated with mortality, and have statistical significance, as shown in Table 2.

Table 2 The unadjusted association between baseline variables and mortality (n = 3545).
Full size table

In addition, we can intuitively see from that the higher the CRP, the lower the survival rate. See Fig. 3 for details.

Figure 3
figure 3

Kaplan–Meier survival curves for all-cause mortality stratified by CRP tertile.

Full size image

Identifcation of nonlinear relationship

We observed a nonlinear dose–response relationship between the CRP and mortality (Fig. 2 and Table 3). When the CRP was < 15.6 mg/L, the mortality rate increased with an adjusted HR of 1.57 (95% CI 1.30–1.91, P < 0.0001) for every 10 mg/L increment in the CRP. When the CRP was ≥ 15.6 mg/L, the mortality rate increased with an adjusted HR of 1.11 (95% CI 0.99–1.24, P = 0.0819) for every 10 mg/L increment in the CRP (Table 3).

Table 3 Treshold efect analysis of the CRP and mortality.
Full size table

Using the generalized additive model, an inflection point between CRP and mortality was detected (Table 3). The linear regression model and a two-piece-wise linear regression model were compared, and the P value of the log-likelihood ratio test was 0.007. This result indicates that the two-piece-wise linear regression model should be used to fit the model.

Discussion

This retrospective cohort study found a curved relationship and saturation effect between CRP and the mortality risk in 3545 patients infected with SARS-CoV-2 admitted to four hospitals within the Montefiore health system between March 1 and April 16, 2020. When the CRP was small, the mortality rate increased significantly with the increase of CRP. When CRP > 15.6 mg/L, with the increase of CRP, the mortality rate increases relatively flat. Few studies have shown a curvy relationship between CRP and mortality in patients with COVID-19.

C-reactive protein is recognized as a marker of systemic inflammation and severe infection. As an acute phase reactant, CRP binds to phosphorylcholine in pathogen and host cell membranes and acts as an opsonin to enhance phagocytosis and facilitate clearance. Ligand-bound C-reactive protein also efficiently activates the classical pathway of the complement system, an essential component of natural host defense16. Prior to the COVID-19 global pandemic, up to 90% of significantly elevated C-reactive protein concentrations were attributable to infectious pathogens, most commonly bacterial pathogens17. Elevated C-reactive protein concentrations have also been reported in severe viral infections, including H1N1 influenza pneumonia and now SARS-CoV-2 infection5,18.In a previous study of 95 COVID-19 patients, CRP For every 10 mg/L increase, the patient's risk of death increased by 11%, and C-reactive protein concentration was associated with mortality9. Smilowitz et al. reported that patients with COVID-19 who had high levels of C-reactive protein on admission were associated with higher mortality, and that patients with both high D-dimer and CRP had a greater risk of death8. Similarly, among 281 hypertensive COVID-19 patients hospitalized with COVID-19, those who died had higher levels of C-reactive protein compared with survivors19. Although the studies collected patients over different time periods, the trends were similar. In addition, the high mortality rate of COVID-19 is mainly related to the poor control of the inflammatory cascade and the subsequent acute respiratory distress syndrome (ARDS)20. However, some studies have pointed out that elevated plasma CRP levels in patients with acute respiratory distress syndrome are associated with reduced mortality, which seems to be contrary to the established views of previous scholars13. Therefore, the correlation between CRP and the prognosis of COVID-19 infection is controversial, and most of the previous similar studies had small sample sizes, and rarely analyzed the curve fitting relationship between CRP and the mortality rate of COVID-19 patients. This is exactly the problem that this study strives to solve.

C-reactive protein concentration reflects the degree of acute inflammatory response. The role of systemic inflammation in the pathogenesis of COVID-19 is still not fully understood. A causal relationship between the inflammatory response measured by CRP and poor prognosis remains speculative. However, the deleterious inflammatory response observed in some patients with COVID-19 is similar to macrophage activation syndrome and may independently lead to multiple organ damage in COVID-1921,22. Studies have shown that the application of glucocorticoids Decreases C-reactive protein concentrations in patients with ARDS23. A recent randomized trial of immunosuppression noted that hormone therapy reduced mortality in critically ill COVID-19 patients. The clinical benefit of immunosuppression further supports the hypothesis that inflammation from viral infection contributes to adverse outcomes in COVID-1924. In addition, thrombotic inflammation has been implicated as a mediator of adverse events in COVID-19. In response to inflammatory stress, thrombus inflammation dysregulates the normal antithrombotic function of the endothelium, leading to leukocyte recruitment, complement and platelet activation, and enhanced microvascular coagulation25,26. This is supported by autopsy studies of COVID-19 patients. In these studies, platelets with fibrin microthrombi were found in the microcirculation of the lung, kidney, liver, and heart27. Among them, pulmonary embolism has a high mortality rate and is very common in new coronary pneumonia. Therefore, the risk of poor prognosis associated with SARS-CoV-2 infection increases when patients experience concomitant elevated CRP concentrations. However, what we did not expect is that there is a saturation effect between CRP and mortality. So far, the author has not found relevant literature to explain this phenomenon.

Among COVID-19 patients, higher C-reactive protein levels were associated with higher mortality risk with a saturation effect. This finding has interesting potential clinical and research implications. To the best of our knowledge, there are few clear references to the curvilinear relationship between CRP and mortality risk in the early stages of COVID-19. Compared with the linear regression relationship, the curve relationship in this study can more scientifically and accurately reflect the judgment value of CRP level on the mortality risk of patients with COVID-19.

Study limitations

A common problem in observational studies is unmeasured confounders. As shown in Table 1, subjects in the highest tertile of CRP were older and had lower Oxygensaturation compared with subjects in the lowest tertile of CRP. These differences may be indicative of unmeasured confounders, such as income and health insurance, which may affect the mortality risk. Although race, length of hospital stay, and age were included in the data set, the effect of unmeasured confounders on HR could not be estimated. In addition, the first measured CRP concentration on admission was used for the primary analysis. The admission physician's knowledge of the initial CRP concentration may have altered subsequent patient care and may have confounded the results. Although there are no data on steroid use, few patients may have received steroids prior to the initial CRP measurement during hospitalization. In some cases, steroids or the IL-6 inhibitor tocilizumab are used in critically ill patients. However, the timing of administration, duration of use, and dosage of these drugs were not recorded in this dataset.

In addition, the lack of information on interventions during the initial stabilization period may have influenced the relationship between CRP levels and survival. It is noteworthy that the potential of the intervention would be biased towards reducing mortality, leading to an underestimation of the association between CRP levels and mortality, but this reinforces the positive findings of this study.

Conclusions

The study identified a nonlinear dose–response relationship between CRP and mortality. For patients with COVID‐19, the association between the CRP and the mortality risk was curve and had a saturation effect. When the CRP was small, the mortality rate increased significantly with the increase of CRP. When CRP > 15.6 mg/L, with the increase of CRP, the mortality rate increases relatively flat.