Introduction
Since December 2019, coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, has spread worldwide and become prevalent in more than 200 countries. On May 3, 2020, the World Health Organization reported more than three million confirmed cases of COVID-19 globally, with more than 230,000 deaths [1]. Because of the rapid viral transmission and the high mortality in critically ill patients, early and effective predictors of outcomes are needed to improve treatment stratification. Available laboratory and imaging indicators are limited owing to the characteristics of infectious diseases and the lack of health care workers and treatment resources. It is well known that inflammatory cytokine storms are prevalent in COVID-19 patients. After analyzing the most common inflammatory factors and other typical laboratory indicators, we found that the receiver operating characteristic (ROC) curve for serum procalcitonin (PCT) was the strongest predictor of the outcomes under study. The present study, therefore, retrospectively analyzed the prognostic value and the optimal cutoff value of the PCT level on admission to predict death in COVID-19 patients.
Methods
Study Participants
This single-center, retrospective cohort study included 129 COVID-19 patients treated at the Zhongfaxincheng campus of Tongji Hospital, Tongji Medical College, Huazhong University of Science & Technology (Wuhan, China). These patients were treated by the Peking University Medical Team between February 9, 2020, and March 20, 2020. The last enrolled patient was admitted on February 20, and end points within 30 days were observed for all patients until March 20. All patients had laboratory-confirmed SARS-CoV-2 infection (the High Pure Viral Nucleic Acid Kit was used to extract nucleic acids from clinical samples according to the kit instructions) and were transported from secondary or Fangcang hospitals to our ward because of disease aggravation and critical care demands. Diagnosis was done according to the interim guidance issued by the World Health Organization [2], and patients were clinically classified on the basis of the “Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia (Trial Version 7)” issued by the National Health Commission of the People’s Republic of China [3]. Severe cases were defined as having one of the following: (1) respiratory rate greater than 30 breaths per minute; (2) oxygen saturation of 93% or less; or (3) Pao2 /Fio2 ratio less than 300 mmHg. Critical cases were defined as having at least one of the following: shock, respiratory failure requiring mechanical ventilation, or extrapulmonary organ failure requiring intensive care. All definitions we used were for adults. The Ethics Committee of Peking University First Hospital approved this study (ethics approval number 2020-keyan-135).
Patient Clinical Data Collection
General clinical data for the enrolled patients were collected from the electronic medical record system. Included were demographic data (age and sex), clinical data (signs, symptoms, chronic medical illnesses, treatment, and clinical outcomes), and laboratory findings. Within 2 hours of admission, laboratory tests were performed to assess the levels of PCT, leukocytes, hemoglobin, neutrophils, and high-sensitivity C-reactive protein (hs-CRP), among other parameters. Fever was defined as an axillary temperature of 37.3°C or higher. PCT level was measured by an electrochemiluminescence method (Cobas e411, Roche, Germany), with the normal range defined as 0.02–0.05 ng/mL. All patients were treated according to the “Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia (Trial Version 5-7)” [3–5]. The clinical outcomes monitored were all-cause death and invasive and noninvasive ventilator use.
Statistical Analysis
All statistical analyses were conducted with IBM SPSS for Windows, version 22.0 (IBM Corp., USA) and R version 3.3.2 (https://www.r-project.org). Continuous variables were expressed as the median and interquartile range, and categorical variables were expressed as a number or percentage. The patient groups were compared via the Mann-Whitney U test for continuous variables and the chi-square or Fisher’s exact tests for categorical variables as appropriate. The optimal PCT level cutoff and the area under the concentration-time curve were calculated from the ROC curve. Survival curves were plotted with the Kaplan-Meier method and were used to compare patients in the two PCT groups. Because most patients were hospitalized for less than 1 month, we considered only incidents that occurred within 30 days of admission. All Kaplan-Meier curve analyses were based on a 30-day observation period. A two-tailed P<0.05 was considered to indicate statistical significance.
Results
Baseline Characteristics
Figure 1 presents the flowchart of the patient recruitment process. In total, 138 consecutive patients with confirmed COVID-19 were admitted between February 6 and February 20, 2020. Nine patients with missing laboratory results were excluded; thus, 129 patients were finally included in the study.
Table 1 presents the basic clinical characteristics of the patients, including age, sex, comorbidities, and routine laboratory results on admission. The median age of the 129 patients was 65 years (interquartile range 54–71 years); 51.9% patients (67/129) were male. The optimum cutoff value for the PCT level to predict all-cause death was 0.085 ng/mL from the ROC curve (Figure 2), with sensitivity of 95.7% and specificity of 72.6%. The area under the ROC curve for all-cause death was 0.876. ROC curves comparing PCT and other laboratory markers are provided in Supplementary Figures 1–3.
Baseline Characteristics and Clinical Outcomes of 129 Patients with Coronavirus Disease 2019 (COVID-19).
| Variable | All | PCT <0.085 ng/mL | PCT ≥0.085 ng/mL | P |
|---|---|---|---|---|
| Number of patients | 129 | 78 | 51 | |
| Age (years) | 65 (54-71) | 66 (54-71) | 64 (54-71) | 0.870 |
| Male | 67 (51.9%) | 35 (44.9%) | 32 (62.7%) | 0.014 |
| Chronic medical illness | ||||
| Coronary artery disease | 26 (20.2%) | 18(23.1%) | 8 (15.7%) | 0.306 |
| Hypertension | 59 (45.7%) | 36 46.2%) | 23 (45.1%) | 0.906 |
| Chronic pulmonary disease | 20 (15.5%) | 12 (15.4%) | 8 (15.7%) | 0.963 |
| Chronic renal disease | 9 (7.0%) | 4 (5.1%) | 5 (9.8%) | 0.308 |
| Diabetes | 24 (18.6%) | 16 (20.5%) | 8 (15.7%) | 0.491 |
| Signs and symptoms | ||||
| Spo2 at admission (%) | 95 (89–97) | 96 (94–98) | 91 (83–95) | <0.001 |
| HR at admission (bpm) | 96 (83–111) | 94 (81–108) | 101 (86–117) | 0.016 |
| SBP at admission (mmHg) | 133 (117–149) | 132 (119–145) | 136 (113–152) | 0.931 |
| DBP at admission (mmHg) | 83 (73–94) | 82 (72–94) | 85 (75–95) | 0.585 |
| Fever | 112 (86.8%) | 67(85.9%) | 45 (88.2%) | 0.366 |
| Cough | 109 (84.5%) | 64 (82.1%) | 45 (88.2%) | 0.343 |
| Sputum production | 80(62.5%) | 46(59.7%) | 34 (66.7%) | 0.428 |
| Shortness of breath | 87 (67.4%) | 48 (61.5%) | 39 (76.5%) | 0.077 |
| Chest pain | 24 (18.6%) | 15 (19.2%) | 9 (17.6%) | 0.821 |
| Sore throat | 27 (20.9%) | 16(20.5%) | 11 (21.6%) | 0.885 |
| Diarrhea | 68 (52.7%) | 44(56.4%) | 24 (47.1%) | 0.298 |
| Nausea | 45 (34.9%) | 29 (37.2%) | 16 (31.4%) | 0.499 |
| Vomiting | 27 (20.9%) | 14 (17.9%) | 13 (25.5%) | 0.303 |
| Stomach ache | 25 (19.4%) | 14 (17.9%) | 11 (21.6%) | 0.611 |
| Headache | 45 (34.9%) | 29 (37.2%) | 16 (31.4%) | 0.499 |
| Muscle ache | 62 (48.4%) | 38 (49.4%) | 24 (47.1%) | 0.799 |
| Fatigue | 81 (63.3%) | 50 (64.1%) | 31 (62.0%) | 0.810 |
| Laboratory test | ||||
| WBCs (×10/L) | 5.54 (4.38–7.73) | 5.17(4.27–6.27) | 7.33(4.74–11.54) | <0.001 |
| Lymphocytes (×10/L) | 0.90 (0.65–0.90) | 1.09 (0.73–1.61) | 0.72 (0.47–1.12) | 0.001 |
| Neutrocytes (×10/L) | 3.96 (2.65–3.96) | 3.42 (2.47–4.41) | 6.34 (3.34–9.48) | <0.001 |
| Hemoglobin (×109/L) | 124 (115–138) | 122 (114–132) | 129 (116–144) | 0.478 |
| Albumin (g/L) | 34.4 (31.0–37.8) | 35.2 (32.1–38.7) | 32.4 (29.6–35.9) | 0.002 |
| Creatinine (μmol/L) | 73.0 (58.0–91.0) | 69.0 (57.0–83.8) | 85.0 (63.0–99.0) | 0.024 |
| D-dimer (μg/mL) | 1.27 (0.53–2.49) | 0.68 (0.40–1.84) | 1.95 (1.25–6.79) | 0.002 |
| IL-6 (pg/mL) | 18.8 (4.93–50.1) | 7.78 (2.77–21.49) | 51.3 (19.6–90.0) | <0.001 |
| IL-8 (pg/mL) | 10.5 (5.5–25.0) | 7.40 (5.00–15.35) | 23.9 (10.4–34.3 | 0.001 |
| IL-10 (pg/mL) | 5.00 (5.00–7.40) | 5.00 (5.00–5.00) | 7.10 (5.00–10.4) | 0.001 |
| TNF (pg/mL) | 8.30 (5.70 –12.0) | 7.00 (4.65–9.40) | 12.4 (8.18–14.3) | <0.001 |
| ESR (mm/hr) | 37.5 (21.3–59.5) | 32.0 (20.0–43.3) | 55.5 (32.5–79.3) | 0.001 |
| Ferritin (μg/mL) | 662.7 (382.1–1325.5) | 478.4 (288.1–752.3) | 1383.3 (692.2–2110.1) | 0.001 |
| hs-CRP (mg/L) | 4.3 (2.15–10.65) | 3.3 (1.9–6.18) | 11.0 (3.6–32.5) | <0.001 |
| Treatment | ||||
| Noninvasive ventilation | 28 (21.7%) | 2 (1.6%) | 26 (51.0%) | <0.001 |
| Invasive mechanical ventilation | 11 (8.5%) | 1(1.3%) | 10 (19.6%) | <0.001 |
| Clinical outcomes | ||||
| Death | 23 (17.8%) | 1 (1.3%) | 22 (43.1%) | <0.001 |
Values are given as the median and the interquartile range or the number of patients and the percentage.
DBP, diastolic blood pressure; ESR, erythrocyte sedimentation rate; HR, heart rate; hs-CRP, high-sensitivity C-reactive protein; IL, interleukin; PCT, procalcitonin; SBP, systolic blood pressure; Spo2, pulse oxygen saturation; TNF, tumor necrosis factor; WBCs, white blood cells.
PCT Level Cutoff Selection and Grouping
Among the routinely tested laboratory parameters, PCT level had the highest value for predicting in-hospital death in COVID-19 patients. The areas under the curves indicated that levels of hs-CRP, neutrocytes, and D-dimer were also strong predictors in this patient cohort (Table 2). Overall, 78 patients had a PCT level that was lower than the optimum PCT cutoff value of 0.085 ng/mL, and 51 patients had PCT levels of 0.085 ng/mL or greater. There was a higher proportion of men in the group of patients with PCT levels of 0.085 ng/mL or greater than in the group with PCT levels less than 0.085 ng/mL (32 [62.7%] vs. 35 [44.9%], P=0.014). Additionally, patients with higher PCT levels had lower lymphocyte (P<0.001) and albumin (P=0.002) levels and higher levels of white blood cells (P<0.001), neutrocytes (P<0.001), creatinine (P=0.024), hs-CRP (P<0.001), D-dimer (P=0.002), and interleukin-6, interleukin-8, interleukin-10, tumor necrosis factor, erythrocyte sedimentation rate, and ferritin (all P≤0.001).
Areas Under the Concentration–Time Curve (AUCs) for Routine Laboratory Tests to Predict Death in Patients with Coronavirus Disease 2019 (COVID-19).
| Routine laboratory test | AUC | 95% CI |
|---|---|---|
| Procalcitonin | 0.876 | 0.806–0.947 |
| hs-CRP | 0.781 | 0.687–0.875 |
| Neutrocytes | 0.777 | 0.687–0.896 |
| Leukocytes | 0.743 | 0.614–0.872 |
| D-dimer | 0.734 | 0.625–0.843 |
| IL-6 | 0.828 | 0.727–0.928 |
| IL-8 | 0.759 | 0.601–0.917 |
| IL-10 | 0.772 | 0.638–0.907 |
| TNF | 0.694 | 0.541–0.846 |
| ESR | 0.627 | 0.453–0.800 |
| Ferritin | 0.705 | 0.547–0.863 |
CI, confidence interval; ESR, erythrocyte sedimentation rate; hs-CRP, high-sensitivity C-reactive protein; IL, interleukin; TNF, tumor necrosis factor.
End Points
During the 30-day observation period, 28 of the 129 patients (21.7%) were treated with noninvasive ventilation, and 11 (8.5%) were treated with invasive mechanical ventilation. Twenty-three patients (17.8%) died; mortality was significantly higher in patients with high PCT levels than in patients with low PCT levels (43.1% vs. 1.3%, P<0.001; Table 1).
The Kaplan-Meier survival curves (Figures 3 and 4) showed that the risks of death and ventilator use (including invasive and noninvasive ventilators) were significantly increased in patients with PCT levels of 0.085 ng/mL or greater (risk of death, log rank χ 2=38.935, P<0.001; risk of ventilator use, log rank χ 2=15.670, P<0.001).
Discussion
The results showed that a PCT level of 0.085 ng/mL or greater can effectively predict the adverse outcomes of in-hospital death and ventilator use in severe and critical COVID-19 patients. This powerful prognostic factor is proposed, which uses a single laboratory test and its well-established cutoff value.
PCT level is widely used to assess the risk of bacterial infection and progression to severe bacterial sepsis and septic shock. For patients with suspected or confirmed lower respiratory tract infections, including community-acquired pneumonia, acute bronchitis, and acute exacerbation of chronic obstructive pulmonary disease, the PCT level can aid in decision-making regarding antibiotic treatment of inpatients and emergency department patients [6, 7].
As the COVID-19 epidemic has spread, PCT level has been shown to be a valuable tool for the early identification of patients at risk of bacterial infections and adverse outcomes. A recent analysis of 1099 COVID-19 patients from multiple medical centers in China revealed that patients with low PCT levels (<0.5 ng/mL) had a lower incidence of end point events (ICU admission end points, invasive ventilation, death) [8]. Most COVID-19 patients have a PCT level of less than 0.25 ng/mL, and even less than 0.1 ng/mL is often seen; these levels are similar to those seen in previous viral epidemics (H1N1 influenza [9], SARS [10], and Middle East respiratory syndrome [11]). According to a recently published meta-analysis of COVID-19 patient data, PCT level greater than 0.5 ng/mL corresponded to a risk of severe SARS-CoV-2 infection that was almost five times higher than in patients with lower PCT levels [12].
Elevated PCT level on hospital admission suggests that the viral infection may be accompanied by a bacterial infection. PCT level specifically reflects the severity of the overall systemic infection, while hs-CRP level may be affected by surgery, local bacterial infections, and other noninfective factors. While the levels of PCT and cytokines are all elevated in patients with bacterial infections, PCT remains in the plasma much longer than cytokines. Moreover, when the infection persists or develops into sepsis, the levels of cytokines such as tumor necrosis factor and interleukin-6 continue to decrease, while PCT level neither decreases nor increases but remains relatively stable. Furthermore, studies have suggested that cytokines such as interleukins can indicate the presence of only systemic inflammatory reactions, and they cannot be used to distinguish bacterial infections [6].
In our study, a clear PCT cutoff value of 0.085 ng/mL was established by ROC curve analysis. There was a significant difference in mortality but similarity in the symptoms and vital signs between the two groups of patients separated by this cutoff value. Therefore, early admission on the basis of PCT level could be used as a simple identifier of critical patients and would help to predict the occurrence of adverse events and to strategically conduct monitoring, intervention, and treatment to mitigate disease progression.
Limitations
Our study has several limitations. First, this was a single-center, retrospective study that included a small number of patients and was limited to severe and critical cases. Some patients remained under clinical observation and had not yet reached clinical end points beyond 30 days. Therefore, the results should be corroborated by further prospective analyses. Second, the duration from symptom onset to hospital admission differs among patients, and this might have influenced the PCT levels. Third, an adjusted regression model analysis was not conducted because of the low number of end points.