key: cord-0718049-9ee62g39
authors: Cai, Li; Zhou, Xi; Wang, Miao; Mei, Heng; Ai, Lisha; Mu, Shidai; Zhao, Xiaoyan; Chen, Wei; Hu, Yu; Wang, Huafang
title: Predictive nomogram for severe COVID-19 and identification of mortality-related immune features
date: 2020-11-04
journal: J Allergy Clin Immunol Pract
DOI: 10.1016/j.jaip.2020.10.043
sha: cca2c32b050499d1275b7b145429fb423db2e66e
doc_id: 718049
cord_uid: 9ee62g39

Background Severe COVID-19 patients have a high mortality rate. The early identification of severe COVID-19 is of critical concern. Additionally, the correlation between the immunological features and clinical outcomes in severe cases needs to be explored. Objective To build a nomogram for identifying severe COVID-19 patients and explore the immunological features correlating with fatal outcomes. Methods We retrospectively enrolled 85 and 41 patients with COVID-19 in primary and validation cohorts, respectively. A predictive nomogram based on risk factors for severe COVID-19 was constructed using the primary cohort and evaluated internally and externally. Additionally, in the validation cohort, immunological features in patients with severe COVID-19 were analyzed and correlated with disease outcomes. Results The risk prediction nomogram incorporating age, C-reactive protein, and D-dimer for early identification of severe COVID-19 patients showed favorable discrimination in both the primary (AUC 0.807) and validation cohorts (AUC 0.902) and was well calibrated. Patients who died from COVID-19 showed lower abundance of peripheral CD45RO+CD3+ T cells and natural killer cells, but higher neutrophil counts than that in the patients who recovered (P=0.001, P=0.009, and P=0.009, respectively). Moreover, the abundance of CD45RO+CD3+ T cells, neutrophil-to-lymphocyte ratio, and neutrophil-to-natural killer cell ratio were strong indicators of death in severe COVID-19 patients (AUC 0.933 for all three). Conclusion The novel nomogram aided the early identification of severe COVID-19 cases. Additionally, the abundance of CD45RO+CD3+ T cells and neutrophil-to-lymphocyte and neutrophil-to-natural killer cell ratios may serve as useful prognostic predictors in severe patients.

The novel nomogram aided the early identification of severe COVID-19 cases. Additionally, 47 the abundance of CD45RO + CD3 + T cells and neutrophil-to-lymphocyte and neutrophil-to-natural killer cell 48 ratios may serve as useful prognostic predictors in severe patients. China and rapidly became a global viral pandemic drawing international concern (1, 2). Severe acute 69 respiratory syndrome-coronavirus-2 (SARS-CoV-2) causes respiratory and intestinal symptoms similar to 70 other coronaviruses, such as severe acute respiratory syndrome-coronavirus (SARS-CoV) and Middle East 71 respiratory syndrome-coronavirus (3). In severe cases, shortness of breath rapidly develops into acute 72 respiratory distress syndrome (ARDS) combined with multiple organ dysfunction syndrome, which 73 increases the risk of mortality (4). Therefore, identifying possible predictors for severe COVID-19 74 outcomes is of critical concern. Importantly, the immune status of patients with COVID-19 closely 75

correlates with viral clearance and disease recovery (5). Several studies have reported diminished T 76 lymphocytes, including CD4 + and CD8 + T cells, and elevated circulating cytokines in patients with severe 77 COVID-19 compared with the levels in non-severe cases (6, 7). However, only a few studies have explored 78 the relationship between immune markers and clinical outcomes in severe patients. Thus, the aim of this 79 study is to identify risk factors for the severity of COVID-19-associated pneumonia, and build a predictive 80 model for the early identification of patients with severe COVID-19. Moreover, correlations between 81 immunological features and fatal outcomes in severe cases were explored. 82

We carried out a retrospective study to build a predictive nomogram for severe COVID-19. In one 85 COVID-19 treatment group of Wuhan Union Hospital, a total of 93 patients with COVID-19 were 86 consecutively admitted from January 27 to March 16, 2020. Of these patients, 85 were selected to form the 87 primary cohort according to the following inclusion criteria: (1) patients diagnosed with COVID-19; (2) 88 patient who had the laboratory test results (including blood routine, coagulation function, liver function, 89 kidney function, myocardial enzyme, C-reactive protein [CRP] ) available within the first week after 90 admission. In another COVID-19 treatment group of Wuhan Union Hospital, all 41 patients who were 91 admitted from February 6 to March 16, 2020, formed the validation group and used to validate the 92 nomogram. Before that, the 41 patients provided informed consent and were enrolled in the study for 93 immune analysis. The follow-up endpoint for the 41 patient was recovery from illness or death, and time 94 point for follow-up was set on the discharge day of the last patient. The diagnosis of COVID-19 was based 95 on the Guidelines for Diagnosis and Treatment of Novel Coronavirus Pneumonia (6 th version) released by 96 National Health Commission of China. A confirmed case was defined as an individual with laboratory 97 confirmation of SARS-CoV-2 which required positive results of SARS-CoV-2 RNA, irrespective of 98 clinical signs and symptoms. All patients were divided into non-severe and severe groups. For the 99 diagnosis of severe COVID-19, at least one of the following criteria should be met: [1] shortness of breath 100 with respiratory rate ≥30 times/min; [2] arterial oxygen saturation (resting status) ≤93%; and [3] arterial anticoagulant tubes. Surface markers were used to identify different immune cell subsets and gating 125 strategy was shown in Figure E1 (see Figure E1 in the Online Repository). All monoclonal antibodies used 126 have been listed in Table E1 (see Table E1 in the Online Repository). We incubated 100 µl of whole blood 127 with antibodies for 15 min at room temperature followed by incubation with 2 ml of BD lysing buffer at 128 room temperature for 10 min to lyse erythrocytes. After that, samples were washed once with 129 phosphate-buffered saline, resuspended in 300 μl of phosphate-buffered saline, and stored at 4℃ until 130 acquirement. Sample acquirement was performed on a BD FACSCanto flow cytometer and data were 131 analyzed using the Diva software. 132

Additionally, serum cytokines were analyzed from 21 patients; the whole blood was centrifuged at 300 ×g 133 for 15 min. The serum was extracted and stored at -20℃ before test. Multiple serum cytokines (IL-2, IL-4, 134 IL-5, IL-6, IL-8, IL-10, IL-1β, IL-17A, IL-12p70, TNF-α, IFN-α, and IFN-γ) were quantified using the 135 Human twelve Cytokine Kit (Jiangxi Saiji Bio-Tech) following the manufacturer's manual. 136

Categorical variables were expressed as frequency and percentage, and significance was detected using the 138 Chi-square or Fisher's exact test. The quantitative variables were expressed as mean±standard deviation, 139 and significance was evaluated using the Student's t-test. Non-normally distributed variables were 140 expressed in median and quartile intervals (IQR), and significance was determined using the 141

Mann-Whitney U test. For building a nomogram based on the primary cohort, variables with P<0.05 142 (univariate analysis) were selected as covariates for adjustment in the multivariate logistic regression 143 model to estimate odds ratio (OR) with 95% confidence intervals (95% CI). Afterward, a nomogram 144 predicting severe COVID-19 was formulated based on the results of multivariate analysis using the 145 package of rms in R version 3.6.3. The performance of the nomogram was validated internally (primary 146 cohort) and externally (validation cohort) by discrimination (a relatively corrected concordance index) and There were no significant differences in age, gender, baseline disease, or disease severity between the 157 patients included in the primary cohort and those excluded (see Table E2 in the Online Repository). 158

Additionally, no significant differences between the primary and the validation cohorts were observed 159 concerning age, gender, baseline disease, and disease severity according to the discharge diagnosis (Table  160 1). In the primary cohort, 37 (43.5%) patients were in the non-severe group, and 48 (56.5%) were in the 161 severe group. The median age of the severe group was higher than that of the non-severe group (median P=0.062). Compared with those in the non-severe group, patients in the severe group had worse laboratory 165 measurements, including elevated white blood cells, neutrophil%, D-dimer, aspartate aminotransferase, 166 blood urea nitrogen, and CRP as well as decreased lymphocyte% and albumin (Table 2) . were discharged after recovery from COVID-19 with a median hospital stay of 23.6 (4-56) days

In the primary cohort, univariable and multivariable analyses by Logistic regression model demonstrated 177 that old age and high levels of D-dimer and CRP were independent risk factors for predicting COVID-19 178 severity, with odd ratios (ORs) of 1.078 (95% CI: 1.015-1.114, P=0.014), 1.394 (95% CI: 1.027-1.893, 179 P=0.033) and 1.023 (95% CI: 1.002-1.045, P=0.035), respectively (Table 3) . 180

To provide a quantitative tool that predicts individual probability of developing a severe case of 181 COVID-19, a novel prognostic nomogram that incorporated age, D-dimer level, and CRP level was 182 established based on multivariable logistic analyses in the primary cohort ( Figure 1 ). The probability of 183 severe COVID-19 of each patient could be calculated by adding the scores for the three variables. 184

The nomogram was validated internally in the primary cohort and externally in the validation cohort. In the 186 primary cohort, the area under the ROC curve (AUC) was 0.807 (95% CI: 0.715-0.900) for the 187 age-D-dimer-CRP combination, which was superior to the AUC of any one of the three parameters 188 (ranging from 0.635-0.726), with a sensitivity of 72.9% and specificity of 74.8%. In the validation cohort, 189 the AUC was 0.900 (95% CI: 0.800-1.00) with a sensitivity of 86.4% and specificity of 89.5% ( Figure 2A 

All 41 patients had their blood collected for immune cell analyses. As shown in Table 4 , subsets of T cells 199 (CD4 + , CD8 + , CD45RA + , CD45RO + , and Treg) and B cells were lower in the severe group than in the 200 non-severe group. However, NK cell and neutrophil counts showed no differences between severe and 201 non-severe cases. Cytokine profiles were available for 21 patients (8 non-severe and 13 severe). There 202 were no significant differences in cytokine levels between the groups, except for IFN-α, which was higher 203 in the non-severe group than that in the severe group (P=0.019, see Table E3 in the Online Repository). 204

Among the 41 patients who underwent immune analysis, 22 patients with severe disease were further 206 divided into recovery (15 patients) and death groups (7 patients), according to their clinical outcome. As 207 shown in Table 4 , neutrophil counts were higher in the death group than that in the recovery group (9.6 vs. 208 3.5×10 9 /L, P=0.009). Moreover, the death group showed lower counts of CD45RO + CD3 + T cells 209 (including CD45RO + CD4 + and CD45RO + CD8 + T cells) and NK cells than the counts in the recovery were observed between the two groups (P=0.581, P=0.680). Cytokine analysis showed that IL-6 and IL-10 COVID-19 who had received or not received corticosteroid therapy. The results showed no significant 219 differences in immune cells between these two treatment groups, except for NK cells, which were lower in 220 patients who received corticosteroid therapy (156.1 vs. 251.5×10 6 /L, P=0.048). 221

To evaluate the predictive value of these immunological features toward a fatal outcome, ROC curves 222 were drawn and AUC values calculated. We also included the ratio of neutrophils to different lymphocyte 223 subsets as parameters. Our results identified CD45RO + CD3 + T cells, neutrophil-to-lymphocyte ratio 224 (NLR), and neutrophil-to-NK cell ratio (NNKR) as likely prognostic predictors with high AUC values 225 (0.933 for all three). Simultaneously, the cutoff values were calculated from the ROC curves, with a value 226 of 449.3 for CD45RO + CD3 + T cells (specificity: 86.7%, sensitivity: 100%), 5.0 for NLR (specificity: 227 93.3%, sensitivity: 85.7%), and 35.6 for NNKR (specificity: 86.7%, sensitivity: 100%; see Table E4 in the 228

In the first part of our study, old age and high levels of D-dimer and CRP were identified as independent 231 risk factors for severe COVID-19. We further constructed a nomogram using these clinical factors for the 232 sorting and monitoring of COVID-19 patients. The constructed nomogram, which was internally and 233 externally validated, showed good potential for discrimination and calibration and can be used for 234

identifying patients at a high risk of severe COVID-19. In the second part of our study, we observed significantly lower counts of CD45RO + CD3 + T and NK cells 246 in the death group than those in the recovery group. Notably, we found that CD45RO + CD3 + T cells 247 decreased stepwise in the non-severe, severe-recovery, and severe-death groups. Possible causes of these 248 CD4 + and CD8 + T cells were poorly generated to control the virus. However, the adoptive transfer of those 260

research, we suspect that low levels of total CD45RO + CD3 + T cells could imply a low level of activated or 262 virus-specific T cells against SARS-CoV-2 in the peripheral blood, resulting in a weak immune response 263 and persistent disease. Adoptive transfer of CD45RO + CD3 + T cells to control the virus may therefore 264 represent a potential therapeutic strategy in severe cases showing refractory COVID-19 pneumonia. 265 NK cells also play a crucial role against coronaviruses (21). The significantly lower NK cell counts in the 266 death group than that in the recovery group reflected a compromised innate immune system. Since our 267 results suggested that NK cells were affected by corticosteroid therapy, we believe that corticosteroids 268 should be used with caution during the treatment of severe cases of COVID-19. 269

We observed higher levels of IFN-α in the non-severe group than that in the severe group. Notably, 270 recombinant IFN-α is an approved therapeutic agent that inhibits viral replication (12). Our results may severity in previous studies (22, 23). The present study showed that IL-6 levels were significantly higher in 274 the death group than those in the recovery group, which indicates a persistent high inflammatory state in 275 the death group. 276

Liu et al. identified the neutrophil-to-CD8 + T cell ratio and NLR as powerful metrics for the early 277 identification of severe COVID-19 cases (24). We compared immunological features among severe cases 278 who had different outcomes and found that CD45RO + CD3 + T cells, NLR, and NNKR were useful 279 predictors of clinical outcomes. 280

COVID-19 patients, and their possible association to increase mortality in severe COVID-19 patients. 282

Determining the alterations in CD45RO + CD3 + T and NK cell populations, clinicians may be able to predict 283 poor prognoses of severe patients and follow potentially life-saving interventions. 284

However, this study has some limitations. First, the sample size enrolled in this study is relatively small. 285

Prospective multi-center studies with larger sample sizes are needed to verify the findings. Second, we 286

were not able to perform the immune cells analyses in the early stages of the illness. This restricted the All data take ×10 6 /L as the unit, otherwise was noted. 387 J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f IFN-α ≤7.9 9.2 (8.8-10.1) 7.7 (7.2-8.9) 0.019 7.9 (6.5-8.9) 7.6 (7.4-9.0) 0.606 TNF -a ≤5.5 8.1 (7.4-9.1) 7.7 (7.4-8.4) 0.645 8.0 (6.6-8.9) 7.9 (7.7-8.5) 0.940 IL-12p70 ≤3.2 10.4 (7.8-21.7) 9.7 (7.5-11.2) 0.301 9.8 (6.1-11.2) 8.7 (7.7-10.9) 0.918 IFN-γ ≤17.3 12.7 (10.9-15.1) 11.9 (9.8-13.1) 0.341 11.9 (10.3-13.1) 11.2 (9.8-14.1) 0.883 J o u r n a l P r e -p r o o f NLR: neutrophil-to-lymphocyte ratio; NCD3R: neutrophil-to-CD3 + T cell ratio; NCD3ROR:

neutrophil-to-CD45RO + CD3 + T cell ratio; NCD4ROR: neutrophil-to-CD45RO + CD4 + T cell ratio; NCD8ROR:

The continuing 2019-nCoV epidemic 307 threat of novel coronaviruses to global health -The latest 2019 novel coronavirus outbreak in Wuhan

International journal of infectious diseases : IJID : official publication of the International Society 309 for Infectious Diseases

Current epidemiological and clinical features of COVID-19; a 311 global perspective from China

Imaging and clinical features of patients with 313 2019 novel coronavirus SARS-CoV-2. European journal of nuclear medicine and molecular imaging

SARS-CoV-2: a storm is raging

The many faces of the anti-COVID immune response. The Journal of 318 experimental medicine

Characteristics and prognostic factors of disease 320 aminotransferase; ALT: Alanine transaminase; CK: Creatine kinase; BUN: Blood urea nitrogen; sCr: serum 379 creatinine