HsCRP Variation is the Main Risk Factor for Clinical Outcome in COVID-19 Hospitalized Young and Middle-Aged Patients

Background: The clinical characteristics and risk factors of clinical outcomes of COVID-19 in elderly and non-elderly patients show great difference. We are the rst to explore the relationship between hsCRP variation and intensive care unit (ICU) admission in young and middle-aged COVID-19 patients compared with elderly patients. Methods: We included 273 hospitalized patients with conrmed COVID-19 from Tongji Hospital, Wuhan, China from Feb 10, 2020 to Mar 8, 2020. Clinical characteristics and risk factors of outcomes were compared between young and middle-aged patients with elderly patients. Results: Among young and middle-aged patients, hsCRP variation in those admitted to ICU was signicantly higher than that in discharged patients. Among patients admitted to ICU, hsCRP variation showed signicantly difference between young and middle-aged patients and elderly patients (median, 67.9 vs -10.2, P < 0.01). The hsCRP variation was an independent risk factor for ICU admission in young and middle-aged patients (OR = 1.068) and ROC curve revealed hsCRP variation signicant for the prediction of ICU admission (AUC = 0.925) with 92.9% sensitivity and 95.5% specicity. Conclusion: HsCRP variation is the major independent risk factors for ICU admission in young and middle-aged COVID-19 inpatients, but not in the elderly patients.

also studies that suggest the hsCRP variation as indicator of change in in ammatory stimulation intensity, which may be more substantial than hsCRP value itself (8,9).
HsCRP level elevates signi cantly in COVID-19 inpatients and shows difference in severe and critically ill patients (10), suggesting that the variation of C-reactive protein may affect the development of the disease. It is reported that hsCRP level shows difference in elderly patients and non-elderly patient (4).
However, there have been no studies on hsCRP variation in COVID-19 patients. Thus we are the rst to explore the relationship between hsCRP variation and prognosis in the young and middle-aged patients, compared with elderly patients.
As admission to intensive care unit (ICU) is an important factor affecting in-hospital mortality and patient prognosis, the purpose of this study is to investigate the relationship between hsCRP variation and ICU admission in young and middle-aged COVID-19 patients compared with the elderly patients.

Study Design and Participants
A retrospective study was performed on COVID-19 patients hospitalized from Feb 10, 2020 to Mar 8, 2020, at Tongji Hospital in Wuhan, China. All patients were diagnosed as severe cases according to WHO guidance. According to the age, patients were divided into 2 groups: the elderly (more than 60 years old) and the young and middle-aged (no more than 60 years old). Patients who had no clinical outcome observed or less than 18 years old were excluded in this study.

Data Collection
Data were extracted from the hospital's electronic medical record system, including patients' demographics, clinical characteristics, past history, laboratory examination on admission. Laboratory examination consisted of a complete blood count, blood chemical analysis, assessment of liver, renal and cardiovascular function, high sensitivity C reactive protein (hsCRP), procalcitonin (PCT). As disease progressed, maximum value of laboratory examination in hospital was also collected. HsCRP variation was obtained by calculating the difference between the maximum hsCRP value in hospital and hsCRP value on admission. Positive value indicated elevation of hsCRP, while negative value showed decline. Clinical outcomes of patients were observed and evaluated by admitted to intensive care unit (ICU) or discharged.

Statistical Analysis
Categorical variables were represented as frequencies and percentages, and compared using the chisquare and Fisher's exact test. Continuous variables were represented as medians and interquartile ranges or mean and SD, compared using Mann-Whitney U test and t-test as appropriate. The association between hsCRP variation and admission to ICU was evaluated by multivariable logistic regression. Three models were constructed to adjust potential confounding factors, including GFR, ALT and proBNP. The ROC curve was used to calculate the cut-off value of hsCRP variation. All statistical analyses and graphs were generated by using SPSS 25.0 and GraphPad Prism version 5, and p value < 0.05 was considered statistically signi cant.

Results
Demographic, clinical and laboratory characteristics of hospitalized patients Table 1 shows the demographic, clinical and laboratory characteristics of hospitalized patients. A total of 273 hospitalized patients diagnosed as COVID-19 were included in this study, with 188 of them were over 60 years old, with an average age of 69.5 years old, and 85 patients were no more than 60 years old, with an average age of 51.0 years old. About half of patients were male both in the elderly patients (50.4%) and the young and middle-aged (55.3%). The prevalence of hypertension in the elderly patients is 58.5%, which is signi cantly higher than that in the young and middle-aged (36.9%) (P < 0.01). There was no signi cant difference in the prevalence of diabetes between the elderly (21.7%) and the young and middle-aged patients (14.9%). The median systolic blood pressure of the elderly patients measured on admission was higher than the young and middle-aged patients (133 vs 123 mmHg, P < 0.01), which was consistent with the hypertension history. In addition, the median oxygen saturation showed no difference in two groups (95%). 4.3 vs 5.0, P < 0.01), creatinine (median, 64.5 vs 72.0, P = 0.01) were lower and GFR (median, 101.5 vs 85.7, P < 0.01) was higher in young and middle-aged patients, indicating better renal function reserve. HsCRP in young and middle-aged patients was lower than that in the elderly on admission (median, 9.4 vs 25.9, P = 0.02), but the hsCRP level in both the young and middle-aged patients and the elderly patients reduced signi cantly during hospitalization (P < 0.01). The maximum hsCRP value in hospital in the young and middle-aged patients was still lower than that in the elderly, but showed no signi cance (median, 3.4 vs 14.9, P = 0.02). As for cardiovascular function assessment, the levels of proBNP (median, 70 vs 184, P < 0.01) and cTnI (median, 2.6 vs 6.3, P < 0.01) in the young and middle-aged patients were lower than that in the elderly, indicating a greater impairment of cardiovascular function (Table 1).

Characteristics Of Young And Middle-aged Patients
In young and middle-aged patients, 16 patients were admitted to ICU and 69 patients were discharged. Compared with discharged patients, laboratory examination of patients admitted to ICU showed higher leukocytes count on admission (median, 7.4 vs 5.2, P = 0.01) and in hospital (median, 11.8 vs 5.5, P < 0.01), higher neutrophil percentage (median, 82.4 vs 58.1, P < 0.01), lower lymphocytes percentage (median, 6.9 vs 30.1, P < 0.01), and lower platelet count (median, 175.5 vs 278, P < 0.01). The total number of leukocytes in patients admitted to ICU increased signi cantly during hospitalization (P < 0.01).

Characteristics Of Patients Admitted To Icu
Among the patients admitted to ICU, characteristics are showed in Table 3. The mean age of young and middle-aged patients was 50.8 years old and that of the elderly patients was 75.7 years old. The comparison between young and middle-aged patients and the elderly found that the systolic blood pressure (median, 119 vs 137, P = 0.01) and diastolic blood pressure (mean, 73.9 vs 82.4, P = 0.03) of the young and middle-aged patients were signi cantly lower than that of the elderly. The young and middleaged patients showed higher ALT (median, 38 vs 20, P = 0.01), higher GGT (median, 54 vs 25, P < 0.01) and higher GFR (median, 89.7 vs 72.7, P = 0.01). The proBNP level of young and middle-aged patients was lower than that of the elderly (median, 408 vs 999, P < 0.01). Young and middle-aged patients showed higher PCT than elderly patients (median, 0.32 vs 0.18, P < 0.01), and there was no difference of hsCRP (median, 89.7 vs 103.7, P < 0.01) at the time of admission between the young and middle-aged patients and the elderly patients who were admitted to ICU. However, the maximum hsCRP value in hospital of young and middle-aged patients was signi cantly higher than the elderly patients (median, 153 vs 65.2, P < 0.01), as well as the hsCRP variation during hospitalization (median, 67.9 vs -10.2, P < 0.01).

Risk factors for ICU admission in the young and middle-aged and the elderly patients
In young and middle-aged patients, Table 4 shows that the increase of hsCRP variation was signi cantly correlated with the increased risk of ICU admission. After excluding the interference of GFR, ALT and proBNP by multivariate logistic analysis, hsCRP variation was still an independent risk factor for ICU admission, and OR value was 1.068. Table 4 also shows that the increase of proBNP was correlated with the risk of ICU admission in the elderly patients (OR = 1.026), while hsCRP variation was not related to the risk of ICU admission in the elderly. The OR values of different variables for ICU admission in the young and middle-aged people and the elderly people are shown in Fig. 1 respectively.  The area under the curve of hsCRP was 0.925 (P < 0.001) in young and middle-aged patients, which was statistically signi cant, indicating that hsCRP was signi cant for the prediction of ICU admission in young and middle-aged patients. The optimal cut-off value was 13.2 mg/L, with sensitivity of 92.9% and speci city of 95.5%. The area under the curve in the elderly was 0.528 (P = 0.632), which had no statistical signi cance. Therefore, the hsCRP variation in the elderly had no predictive signi cance. After comparing the AUC of the two curves, AUC in young and middle-aged patients was greater than that in the elderly patients (Z = 4.49, P < 0.001).

Discussion
The global outbreak of COVID-19 caused horrendous number of infection and death (5). The clinical characteristics of COVID-19 showed remarkable difference in young and middle-aged patients and elderly patients (4). However, there is a lack of research on the risk factors of disease progression to distinguish between the young and middle-aged patients and the elderly patients. This study is the rst study proposing the relationship between hsCRP variation and ICU admission in the young and middle-aged patients, compared with elderly patients. We lled the gap by comparing the characteristics of young and middle-aged patients with elderly patients in 273 hospitalized patients diagnosed with COVID-19 in Tongji hospital, Wuhan and evaluating the relationship between hsCRP variation and ICU admission by multivariable logistic regression.
Some studies suggested that hsCRP variation indicated the change in the intensity of in ammatory stimulation, which may be associated with disease progression and prognosis (8,9,11). CRP is an in ammatory biomarker produced by the liver, which is released into the blood during the acute phase of in ammation, while hsCRP is a sensitive indicator of disease activity and independent risk factor for a variety of diseases (7,12). However, as hsCRP is affected by a variety of factors, studies suggested that the hsCRP variation was also of important value, indicating that the change in the intensity of in ammatory stimulation may be more important than the hsCRP value itself (8,9,11).
Thus we focused on the association between hsCRP variation and ICU admission in the young and middle-aged patients, which was different from that in elderly patients. Baseline characteristics between the young and middle-aged patients and the elderly patients showed signi cant differences, indicating different factors affecting the ICU admission during hospitalization between two groups. On admission, the level of hsCRP in young and middle-aged patients was signi cantly lower than that in elderly patients, associated with lighter lung damage in COVID-19 (12), while they showed no signi cant difference in the maximum value of hsCRP during hospitalization, suggesting that hsCRP variation might be associated with disease progression and prognosis (13,14). Among young and middle-aged patients, we found hsCRP variation was the main independent risk factor for ICU admission in young and middle-aged patients. HsCRP increased signi cantly during hospitalization, indicating intense in ammatory response in young and middle-aged patients admitted to the ICU, while hsCRP decreased in the discharged patients during hospitalization, indicating the improvement of in ammation. There was a proposed in ammatory model which distinguished COVID-19 development into three stages to explain hsCRP variation as an independent risk factor of the ICU admission in young and middle-aged patients (15). The rst stage is the asymptomatic stage with virus incubation and then turns to the second stage, the direct toxicity and in ammatory activation of the lung, leading to aggravation of respiratory symptoms. In the third stage, patients experienced multi-system damage and hyperin ammatory state, which developed strong and lethal in ammatory response (15).
However, in the elderly patients, hsCRP variation was not the main risk factor of the ICU admission. At the time of admission, hsCRP level of elderly patients was signi cantly higher and suggested heavier basic infection due to the basic diseases and weaker defense barrier (4,12). Poor cardiovascular function in elderly patients contributed to high hsCRP level on admission (16,17). The hsCRP variation in the elderly patients admitted to ICU suggested a decrease of hsCRP during hospitalization, which showed signi cant difference to the young and middle-aged patients admitted to ICU. There may be two reasons for explaining the negative value of hsCRP variation in the elderly patients admitted to ICU. On the one hand, the elderly patients admitted to ICU experienced more severe dissociation of bilirubin and enzyme, suggesting more impairment of hepatic cells and lessor liver function reservation than young and middleaged patients, which resulted in reduced hsCRP production (18,19). In ICU patients, the elderly patients showed signi cantly lower ALT (median, 320 vs 38, P < 0.01) and higher TBIL (median, 11.2 vs 9.5, P < 0.01) than the young and middle-aged patients, suggesting more severe dissociation of bilirubin and enzyme in the elderly patients. Also among the elderly patients, TBIL was higher in patients admitted to ICU (median, 11.2 vs 7.8, P < 0.01) and there was a signi cant difference in TBIL/ALT values compared with discharged patients. On the other hand, there might be an increased blood volume and diluted plasma hsCRP level due to heart failure of elderly patients admitted to ICU (20). The elderly patients admitted to ICU showed higher proBNP than discharged elderly patients (median, 988 vs 156.5, P < 0.01) and young and middle-aged patients admitted to ICU (median, 988 vs 408, P = 0.01), suggesting more severe heart failure in elderly patients admitted to ICU. We also found that proBNP was an independent risk factor for ICU admission in elderly patients, while hsCRP was not, indicating worse cardiovascular function associated with poor prognosis in elderly patients. Elderly patients had higher prevalence of hypertension, which may also effected in the cardiovascular function storage. In addition to controlling in ammation, improving cardiovascular function should be emphasized in treatment of elderly patients.
More studies were needed to explore the role of aggravation of in ammation state in predicting the clinical outcome of elderly hospitalized patient.
Studies also showed hsCRP variation as a predictor of change in in ammation response and dynamic indicator of clinical outcome (21,22). In sepsis patients, hsCRP increased early in multiple organ injury stage (23) and hsCRP decrease in survivors was signi cantly greater than that of non-survivor (9), which demonstrated the low hsCRP variation would lead to better prognosis. There were also studies that used the reduction of hsCRP as an indicator of sepsis and SIRS severity to guide treatment (11,(23)(24)(25), indicating that the reduction of hsCRP was related to the reduction of patients' mortality (24,25). Studies on community acquired pneumonia patients suggested that compared to initial hsCRP itself, hsCRP variation was a better indicator of the prognosis, which could control the bias due to patients' confounding factors (25,26).
Other studies on COVID-19 were searched in PubMed and we implemented the following search strategy with these key words (in the title/abstract): "COVID-19" OR "coronavirus" AND"hsCRP" OR "CRP" OR "characteristics" OR "laboratory". After excluding irrelevant articles, 17 studies with observed hsCRP value were found, among those 4 studies contained hsCRP variation and disease outcomes, including 3 in COVID-19 and 1 in SARS. In those studies, higher hsCRP variation was found to suggest an increase in in ammatory response and was associated with disease prognosis in young and middle-aged patients (3,15). Young and middle-aged discharged patients showed decreased hsCRP with mean hsCRP variation as -18.6 mg/L (13,21,22), indicating a lighter in ammatory response in discharged patients consistent with our study. Also in SARS patients with mean age of 43, the hsCRP variation in patients admitted to ICU was signi cantly higher than that in discharged patients (122 vs 12 mg/L) (27). But in elderly COVID-19 patients with mean age of 69, hsCRP in those with poor prognosis as dead did not increase, but decreased with hsCRP variation as -27.65 mg/L instead (13).
Although in the studies above, none had clearly proposed the association between hsCRP variation and disease prognosis, the results still suggested that higher hsCRP variation was associated with the poor prognosis of young and middle-aged patients with coronavirus, but not in elderly patients, which proved the correctness of our nding. Furthermore, hsCRP variation was not only applied in indication of COVID-19 prognosis, but also could be extended to other coronavirus. In our study, the ROC curve suggested that hsCRP variation could be considered as an indicator of disease prognosis in young and middle-aged patients, and the cut-off value was 13.2 mg/L, indicating that young and middle-aged patients with the hsCRP variation over 13.2 mg/L compared with the day of admission might have a poor prognosis than patients with lower hsCRP variation.
There were still some limitations in our study. First, we included a relatively small sample size in this study, because we only covered COVID-19 hospitalized patients in one hospital and some of them didn't take the examination of hsCRP, which may cause bias in the results. Also, only the maximum value of hsCRP was available to collect during the hospitalization with no time span and dynamic curves.

Conclusion
In conclusion, this study found that hsCRP variation was the major independent risk factors for ICU admission in young and middle-aged COVID-19 inpatients, but not in the elderly patients. Early detection of hsCRP variation is a good indicator of clinical outcome, and early prevention of deterioration of in ammatory state may reduce the risk of ICU admission in young and middle-aged patients.

Availability of data and materials
The datasets used for the analysis in the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.