Prognostic Markers in Hospitalized COVID-19 Patients: The Role of IP-10 and C-Reactive Protein

Background SARS-CoV-2 is responsible for COVID-19, a clinically heterogeneous disease, ranging from being completely asymptomatic to life-threating manifestations. An unmet clinical need is the identification at disease onset or during its course of reliable biomarkers allowing patients' stratification according to disease severity. In this observational prospective cohort study, patients' immunologic and laboratory signatures were analyzed to identify independent predictors of unfavorable (either death or intensive care unit admission need) or favorable (discharge and/or clinical resolution within the first 14 days of hospitalization) outcome. Methods Between January and May 2021 (third wave of the pandemic), we enrolled 139 consecutive SARS-CoV-2 positive patients hospitalized in Northern Italy to study their immunological and laboratory signatures. Multiplex cytokine, chemokine, and growth factor analysis, along with routine laboratory tests, were performed at baseline and after 7 days of hospital stay. Results According to their baseline characteristics, the majority of our patients experienced a moderate to severe illness. At multivariate analysis, the only independent predictors of disease evolution were the serum concentrations of IP-10 (at baseline) and of C-reactive protein (CRP) after 7 days of hospitalization. Receiver-operating characteristic (ROC) curve analysis confirmed that baseline IP − 10 > 4271 pg/mL and CRP > 2.3 mg/dL at 7 days predict a worsening in clinical conditions (87% sensitivity, 66% specificity, area under the curve (AUC) 0.772, p < 0.001 and 83% sensitivity, 73% specificity, AUC 0.826, p < 0.001, respectively). Conclusions According to our results, baseline IP-10 and CRP after 7 days of hospitalization could be useful in driving clinical decisions tailored to the expected disease trajectory in hospitalized COVID-19 patients.


Introduction
In December 2019, many cases of pneumonia of unknown origin were reported in Wuhan (Hubei province, China).
Since then, this illness rapidly spread across China and all over the world. The causative agent of this new disease has been identified in SARS-CoV-2, a positive, single-stranded RNA virus with a genome length of less than 30 kb, belonging to the same β-coronavirus genus of SARS-CoV and MERS-CoV [1][2][3][4]. SARS-CoV-2 virus enters host's cells by binding the angiotensin-converting enzyme 2 (ACE2) expressed in nasal epithelium, low airways, and lungs [4][5][6][7]. This new viral agent is characterized by a high interhuman transmission rate: according to the current evidence, World Health Organization (WHO) reports that respiratory droplets and direct contact represent the most common routes of infection [4][5][6][8][9][10].
Patients affected by COVID- 19 show a wide range of clinical manifestations, ranging from a nearly asymptomatic or mild flu-like condition to severe interstitial pneumonia and acute respiratory distress syndrome. It is wellrecognized that severe clinical manifestations depend not only on viral infection but also on a heavy inflammatory response [1-4, 11, 12]. Pathogen recognition by antigen presenting cells activates both innate and adaptive immune cells to produce large amounts of proinflammatory mediators, in some cases leading to systemic spread of the aberrant immune response leading-in the most severe cases-to multiple organ failure and death [2,[13][14][15].
Reliable biomarkers allowing clinicians to identify, at an early stage, those patients whose condition will deteriorate would facilitate patients' stratification into risk groups for optimal resources allocation. To meet this need, many clinical research groups, including ours, analyzed their case series, identifying several demographic (i.e., age, gender, and comorbidities) and laboratory (i.e., CRP (C-reactive protein), IL-6, D-dimer, and creatinine) parameters that could be used as prognostic biomarkers [14,[16][17][18][19][20][21]. Unfortunately, the retrospective design of these studies has many potential biases.
The present prospective, observational monocentric study is aimed at investigating cytokine, chemokine, and growth factor expression in an Italian cohort of SARS-CoV-2 positive patients, by using a large and unbiased multiplex quantification approach. To this purpose, patients were studied at the time of hospital admission and along their hospital stay, during which they received per protocol corticosteroids and heparin, to identify possible biomarkers able to predict disease severity and evolution.

Patients.
We performed a prospective, observational cohort study during the third wave of COVID-19 epidemic in Italy: 139 consecutive patients were enrolled in non-ICU (intensive care unit) wards (including high-dependency/subintensive units) of "Maggiore della Carità" University Hospital in Novara, Italy, between January and May 2021. This study is part of the large multicenter observational study "BIAS" (Baseline Immunity status effect on sArs-cov2 presentation and evolution: comparison between immunocompetent and immunocompromised patients). The study protocol was approved by the local ethical committee (CE 7/21) and was conducted in strict accordance with the Declaration of Helsinki. Patients were selected according to precise inclusion and exclusion criteria and were asked to give a written consent. To be eligible for the study, patients should be adults (>18 years), hospitalized for confirmed SARS-CoV-2 infection (molecular RT-PCR or quick test positivity), with clinical symptoms not exceeding 12 days. Patients needing an immediate ICU admission or with a very severe clinical presentation (stage V renal failure, severe oncological condition) were excluded.
All patients meeting inclusion criteria for the study received treatment according to the "Maggiore della Carità" Hospital internal protocol. Specifically, all patients received oxygen supplementation, corticosteroids, and low molecular weight heparin (LMWH) unless contraindicated.

Endpoints
Definition. The expected endpoints were the following: (1) identification of biomarkers predicting at baseline and after 7 days of hospitalization an adverse disease evolution (death or ICU admission); (2) identification of clinical or immunological biomarkers predicting at baseline and after 7 days of hospitalization a rapid clinical recovery (discharge from hospital and/or National Early Warning Score 2 ðNEWS2Þ ≤ 2 for at least 24 hours within the first 14 days of hospitalization).

Blood Sample
Collection. Blood samples for routine analysis and for multiplex quantifications were collected by venous puncture using EDTA as anticoagulant at baseline (t0) and after 7 days of hospitalization (t7). Blood fractions were immediately separated by centrifugation and stored at -80°C until the time of analysis.

Routine Laboratory Evaluation.
For each patient, routine laboratory studies included a complete blood cell count, a common biochemistry panel (i.e., creatinine, alanine aminotransferase (ALT), and aspartate aminotransferase (AST)), as well as inflammatory markers (i.e., CRP and ferritin) and markers of coagulation and fibrinolysis (including Ddimer).

Results
From January until May 2021, 139 SARS-CoV-2-positive patients admitted to non-ICU wards of "Maggiore della Carità" Hospital in Novara meeting the study inclusion criteria were enrolled to the present study: 86 were males (61.9%), and 53 (38.1%) were females. The most common symptoms at hospital admission (t0) were dyspnea (62.6%) and dry cough (38.9%). Moreover, among the 139 patients, 83 started a COVID-19-related treatment before hospital admission (corticosteroids (52.5%), azithromycin (35.2%), and heparin (30.9%)). The majority of the enrolled patients at the time of hospital admission showed a moderate (74.1%) or severe (6.5%) respiratory failure (we defined the respiratory failure as moderate when 100 ≤ PiO 2 /FiO 2 < 200 and severe when Pi O 2 /FiO 2 < 100). Consistently median NEWS2 score recorded at the time of admission in our cohort was 5 (IQR: 4-6), a value indicating a potentially serious clinical deterioration in patient's conditions, requiring a close clinical monitoring [22]. The detailed demographical and baseline (t0) clinical description of the selected population is shown in Table 1.
Among the 139 patients initially enrolled to the study, 29 died during hospital stay or were transferred to ICU, while of the remaining 110 patients, 91 were discharged or reached a NEWS2 ≤ 2 for at least 24 hours within the first 14 days of hospitalization.

Outcomes
(1) We compared data obtained at baseline and at 7 days from the 29 patients who died or were transferred to ICU during hospital stay to those obtained from all the other patients Univariate statistical analysis is shown in Tables 2 and 3. At baseline, patients evolving toward a more severe form of the disease showed lower platelet and lymphocyte counts and glomerular filtration rates, in addition to higher RDW-CV (red cell distribution width-coefficient of variation), creatinine, LDH (lactic dehydrogenase), ferritin, D-dimer, IL-6,    (Table 2). After 7 days of hospital stay, patients with a worse disease evolution showed a statistically significant lower level in lymphocyte and platelet counts as well as in TNF-α levels and an increment in neutrophil to lymphocyte ratio, CRP, LDH, erythrocyte sedimentation rate, troponin I, ferritin, D-dimer, G-CSF, IFN-γ, IL-6, IL-8, IP-10, MCP-1, and MIP-1α values (Table 3).
Parameters being significant (p < 0:05) at univariate analysis were used to build multivariate analysis models to identify independent predictors of outcome (Tables 4 and  5). Higher RDW-CV, IP-10, and D-dimer values and lower platelet count at baseline predicted a worsening in clinical conditions (i.e., death or need to ICU admission) ( Table 4) also after correction for demographic and COVID-19 severity variables (Table 5), while after 7 days of hospitalization, the only biomarker with prognostic significance was the increase in CRP levels ( Table 6)  (2) We evaluated which biomarkers were useful to early identify patients who had a faster clinical recovery (hospital discharge or NEWS2 ≤ 2 for at least 24 hours within 14 days of hospitalization) with respect to all the other patients Univariate statistical analysis (Tables 7 and 8) highlighted a significant alteration in some parameters between the two considered populations. In particular, it has been observed that, at baseline, patients with a faster clinical resolution showed a reduction in RDW-CV, creatinine, CRP, neutrophil to lymphocyte ratio, troponin I, D-dimer, IL-4, IL-6, IL-8, IL-10, IP-10, MCP-1, and MIP-1α in addition to an increase in lymphocytes and platelets count and glomerular filtration rate (Table 7). At t7, indeed, it has been observed that patients with a more favorable prognosis showed an increase in platelets and lymphocytes counts, glomerular filtration rate, ALT, and albumin, in addition to a decrease in neutrophil to lymphocyte ratio, RDW-CV, neutrophil count, CRP, LDH, erythrocyte sedimentation rate, troponin I, D-dimer, G-CSF, IFN-γ, IL-6, IL-8, IP-10, MCP-1, and MIP-1α (Table 8).
Again, multivariate models were built to identify the laboratory findings better identifying patients with a faster recovery (hospital discharge or NEWS ≤ 2 within 14 days of hospitalization) (Tables 9 and 10). Such analysis highlighted that patients with a more favorable prognosis showed lower IP-10 and neutrophil to lymphocyte ratio values and higher glomerular filtration rate at the time of hospital admission (t0) ( Table 9). Interestingly, after correction for     age, gender, and severity of clinical presentation (NEWS2 score and PiO 2 /FiO 2 at baseline), only low IP-10 values appeared to be related to a favorable prognosis (Table 10). After 7 days of hospitalization, indeed, the only laboratory finding with prognostic significance for a more positive outcome was a low CRP value (Table 11)  Finally, to calculate the accuracy of the identified prognostic biomarkers, ROC curves were drawn, and cut-off values were identified according to the corresponding AUC (Figures 1-3). For IP-10 measured at baseline, the best cut-off value (0:7 < AUC ≤ 0:9) was 4271 pg/mL: IP-10 higher than 4271 pg/mL predicted a worsening in clinical conditions (87% sensitivity, 66% specificity), while IP-10 lower than 4271 pg/mL predicted a more favorable disease evolution (71% sensitivity, 78% specificity) (Figures 1(a) and  1(b)).
Therefore, in this observational prospective cohort, we screened a wide range of cytokines, chemokines, and routine laboratory markers to identify the best biomarkers predicting disease evolution and prognosis in hospitalized patients.
Since selected patients were hospitalized due to a moderate or severe respiratory failure, all were treated with a standardized protocol including steroids as dexamethasone or methylprednisolone and LMWH, aimed to manage the hyperinflammatory state and prothrombotic and hypercoagulable state observed in these patients [29].
By comparing the cytokine signature at hospital admission of the most severe patients to that of those experiencing milder forms of the disease, it appears that many bioactive molecules are involved (IL-4, IL-6, IL-8, IL-10, IP-10, MCP-1, MIP-1α, TNF-α, G-CSF, and IFN-γ); however, after multivariate statistical analysis, the only chemokine showing a clear predictive value at hospital admission is IP-10, showing both a positive association with greater disease severity and adverse prognosis, and an inverse association with faster recovery [30,31].
Additionally, among laboratory biomarkers tested at 7 days, only CRP was predictive either of a worse or a good prognosis.   10 Disease Markers IP-10 and CRP were the only laboratory parameters that retained a prognostic relevance even after correction for demographics (age and gender) and variables linked to disease severity (PiO 2 /FiO 2 and NEWS2 score either at baseline or at 7 days) [14,[32][33][34][35]. IP-10 is an IFN-γ induced protein released by a large number of cells, many of which involved in the immune response, such as T cells, neutrophils, monocytes, and endothelial cells [36,37]. IP-10 is a proinflammatory mediator involved in leukocyte homing to inflamed tissues and in the perpetuation of the inflammatory response, thus playing a pivotal role in inflammatory tissue damage [36][37][38]. Due to its known chemotactic action toward T cells, NK cells, monocyte/macrophages, and dendritic cells, this chemokine has been investigated as a potential biomarker for many pathological conditions (i.e., autoimmune diseases and viral infections) [36,37]. During the 2002 SARS outbreak, many studies highlighted that higher IP-10 blood levels predicted an adverse outcome, an immunological signature also found in SARS-CoV-2 patients [31,[38][39][40][41][42]. In the present study, IP-10 showed a strong predictive power, with a ROC curve-based cut-off of 4271 pg/mL that, if confirmed in other studies, could be used as biomarker to identify patients needing strict clinical and therapeutic monitoring or even to drive the decision to start anti cytokine treatment.
Among the routine laboratory parameters, CRP, an acute phase protein, is known to be a strong indicator of COVID-19 severity [14,32,35]. The majority of the available studies confirms CRP as prognostic marker in the early stages of COVID-19 infection [43][44][45][46], even before interstitial pneumonia signs become evident at computerized tomography [47,48]. In this study, CRP values after 7 days of hospitalization showed to have prognostic potential, with CRP higher than 2.3 mg/dL identifying patients undergoing clinical conditions worsening. Therefore, CRP value at 7 days seems to be the most reliable marker of treatment response in our population: we may speculate that the persistence of elevated CRP value at 7 days may be related to  11 Disease Markers several critical conditions like the lack of corticosteroid response or eventually the development of secondary infections [44,46,47,49]. Since biomarkers of treatment response in COVID-19 are lacking, to our knowledge, this result is novel, supporting the use of CRP measurement during the disease course to guide patient management.
As observed in many other studies, also for our study population, other baseline laboratory findings routinely assessed exist that contribute to a more accurate prognosis evaluation, such as RDW-CV, D-dimer levels, platelet count, and neutrophil to lymphocyte ratio at the time of hospital admission. RDW-CV is known to be a prognostic biomarker in several diseases [50][51][52] and also in COVID-19 patients [19,[53][54][55]; however, at ROC analysis, we obtained a very low sensitivity (RDW-CV values higher than 14.3% displayed 29.3% sensitivity and 89.4% specificity), therefore limiting its usefulness in clinical practice. Increased Ddimer levels and decreased platelet count were initially linked to the onset of disseminated intravascular coagulation (DIC), but today, thanks to the endlessly increase in scientific knowledge about COVID-19 disease, it is gaining attention the hypothesis of a coagulopathy with specific features different from the classical sepsis-related diffuse intravascular coagulopathy [56][57][58]. In this study, we observed that a baseline platelet count lower than 192 × 10 3/ μL predicted a worsening in clinical conditions, while D-dimer ROC curve analysis did not allow the identification of a sufficiently accurate cut-off.
Both neutrophilia and lymphopenia are associated to COVID-19 evolution [5,14,21,59,60]. Higher neutrophil count is generally a nonspecific marker of severity, as it is related to both thromboembolic complications and systemic inflammatory responses [21], while lymphopenia is associated with a dysregulation of immune response [14]. The neutrophil to lymphocyte ratio thus magnifies the prognostic role of both events, with the ability to predict disease severity [19,61,62]. We confirmed a role for this biomarker since a low neutrophil to lymphocyte ratio (ROC curvebased cut − off = 11:78 × 10 3 ) was associated with a shorter hospital stay.
Our study has several limitations: first, it was based on a single-center enrollment, so that a multicenter validation of our results is needed to make clinical practice recommendations. Moreover, the study was conducted in clinical practice so that slight differences in patients' treatment may have occurred; however, all patients were followed and treated according to a standardized treatment protocol issued at our center, guiding, in particular, but not limited to, steroids and heparin duration and dose, limiting the bias of different clinical approaches.

Conclusions
The increasing number of COVID-19 patients requiring hospitalization stressed the national health systems all over the world, thus highlighting the need of early predictors of disease evolution, to assist the medical staff in the patient management as well as in monitoring patients' conditions during hospitalization. In this prospective observational cohort study, we showed that, after a wide screening of different biomarkers and correction for demographic and disease severity variables, baseline IP-10 values and CRP values after 7 days hospitalization are independent predictors of patients' prognosis and in-hospital disease course and may help the physicians to stratify patients treatments.

Data Availability
Data are available upon reasonable request to be addressed to the corresponding author.