key: cord-1052325-2tf9v3ku authors: Soleimani, R; Khourssaji, M; Gruson, D; Rodriguez‐Villalobos, H; Berghmans, M; Belkhir, L; Yombi, J‐C; Kabamba‐Mukadi, B title: Clinical usefulness of fully automated chemiluminescent immunoassay for quantitative antibody measurements in COVID‐19 patients date: 2020-08-14 journal: J Med Virol DOI: 10.1002/jmv.26430 sha: 8bd35cca0bf6607603336100170da3e7e145daff doc_id: 1052325 cord_uid: 2tf9v3ku BACKGROUND: Since December 2019 we have been in the battlefield with a new threat to the humanity, known as SARS‐CoV‐2, which causes COVID‐19, characterized by viral pneumonia. It may be asymptomatic or cause various symptoms, ranging from flu like symptoms to ARDS and eventually death. At present, the only reliable test for COVID‐19 diagnosis is RT‐qPCR. Assessing the immune response against SARS‐CoV‐2 could increase the detection sensitivity of infected population. Hereby, we report the performances of a fully automated chemiluminescent immunoassay (CLIA) on 276 serum samples. METHODS: One hundred samples obtained from COVID‐19 negative subjects (COVID‐19 free) were analyzed to evaluate diagnostic specificity of antibody (Ab) detection. Thereafter, 176 samples obtained from 125 patients with confirmed COVID‐19 (COVID‐19 patients) were selected to assess the diagnostic sensitivity of the CLIA. All samples were analyzed on MAGLUMI(TM) 800 platform. RESULTS: All COVID‐19 free samples had Ab levels below the cut‐off values. Hence, the diagnostic specificity was estimated at 100% (95%CI= 96.3‐100.0; PPV= 100%). By the 18(th) day from the onset of symptoms we reached to an optimal diagnostic sensitivity (more than 95.0 %) In fact, the diagnostic sensitivity increased over time and between 15 to 25 days after symptoms onset, reached to 95.5% (95% CI= 84.9‐99.2). CONCLUSION: The new automated CLIA analyzer appeared to be a robust and reliable method to measure specific Ab against COVID‐19 at a high throughput. Our data suggest that combining Ab and nucleic acid detection could increase the diagnostic sensitivity. This article is protected by copyright. All rights reserved. specific Ab quantity in the sample. RLUs are then transformed to arbitrary units (AU/mL), using a computed 10-point calibration curve. A level > 1.00 AU/mL is interpreted as positive for both Ab. The claimed diagnostic sensitivity for IgG, IgM and combined IgG and IgM (IgG/IgM) were 91.21%, 78.65% and 95.60%, respectively. The manufacturer did not report any cross reactivity with other Ab against bacterial and viral respiratory infectious agents (including other CoVs). The claimed diagnostic specificity for IgG, IgM and IgG/IgM were 97.33%, 97.50% and 96.00%, respectively. All claimed within and between-run coefficients of variation (CVs) for positive quality controls were < 10.0%. The manufacturer did not claim within nor between-run CVs for negative quality controls [4] . For this purpose, we analyzed 100 residual samples (60 females (F); mean age= 37.2 years (95% confidence interval (95% CI) = 34.5-37.2)) obtained from COVID-19 negative subjects (COVID-19 free), and evaluated the method based on the guideline published by the "Haute Autorité de Santé in France" (reliable if diagnostic specifies are > 95%) [5] . Samples (serum samples were separated and frozen at the time of sampling) were divided in 3 subgroups. Forty samples with no known confounding factors and 40 others with supposedly confounding factors (cf. the supplementary table) known to interfere with serological assays such as autoimmune Ab (n=17) and infectious diseases Ab (23) . Samples were randomly selected among stored sera collected between October 2018 and February 2019. Another 20 samples were collected from asymptomatic subjects during overlapping period of Flu epidemic and COVID-19 outbreak in March 2020. All samples were tested in batch to detect any potential cross-reactivity. The specificity was defined as the percentage of true negative results (TN) correctly identified by the method and calculated by the formula Sp= TN/(TN+False positives (FP)). We also verified the cut-off of seropositivity for each Ab based on COVID-19 free sample levels (if < 1 AU/mL). For this purpose, we performed the immunoassay to measure Ab in all samples (n=176) collected from COVID-19 patients (n=125) at different stages of infection. COVID-19 was confirmed by positive RT-qPCR on nasopharyngeal swab and by radiographic criteria (bilateral chest involvement and/or GGO identified by X-ray or CT-scan). For each of these samples, the date symptoms onset was determined based on the clinical records. The diagnostic sensitivity was estimated based on the receiver operating characteristic (ROC) curves for 4 different time intervals depending on symptoms onset; 0-4 days after, 5-9 days after, 10-14 days after and 15-25 days after. We analyzed samples in batch and evaluated the method based on the guideline published by the "Haute Autorité de Santé in France" (reliable if diagnostic sensitivities are > 95%) [5] . At last, positive and negative predictive values (PPV and NPV) and test accuracy were statistically estimated to assess the proportion of true positive, true negative and accuracy of the CLIA method. The sensitivity was defined as the percentage of true positive results (TP) correctly identified by the method, and calculated by the formula, S= TP/(TP+False Negatives results (FN)). The positive predictive value (PPV) was defined as the probability of a positive result among true patients and calculated by the formula PPV= TP/(TP+FP). The negative predictive value (NPV) was defined as the probability of a negative result among healthy subjects and calculated by the formula PPV= This article is protected by copyright. All rights reserved. TN/(TN+FN). Accuracy was defined as the closeness of the measurements to an expected value and calculated by the formula ACC= (TP+TN)/(TP+FP+TN+FN). To verify the method accuracy, we used precision and trueness according to ISO 5725-2:2019 [6] . Precision was assessed by comparing the locally calculated withinrun and the between-run coefficients of variation (CVs) to those claimed by the manufacturer in package inserts. For this purpose, 2 control levels per Ab types (IgG= 0.29 & 3.92 AU/mL; IgM= 0.29 & 3.92 AU/mL) were analyzed according to CLSI EP15-A3. During 5 consecutive days, each control level analyzed 3 x/day. We considered the verification satisfactory if calculated CVs for positive control levels were < those claimed by the manufacturer. In the absence of standard material or proficiency testing programs for the moment, the positives control level (per Ab types) with known target value was analyzed 20x during 20 days to estimate the trueness. Trueness was calculated using the formula (bias %) = (M−V x 100) / V. With M as the mean value obtained on 20 replicates and V as known value for PQCs provided by the manufacturer's package inserts. [7, 8] . Limit of blanc (LOB), limit of detection (LOD), limit of quantitation (LOQ), Linearity and carry-over A) LOB was defined as the mean of blank replicates + (1.645 x SD blank replicates). For this purpose, negative samples with no measurand (NaCl 0.09% solution) analyzed 10x in row (IgG mean= 0.02 AU/mL; IgM mean= 0.44 AU/mL). LOD was defined as LoB + (1.645 x SD blank plasma pool replicates). LOQ were This article is protected by copyright. All rights reserved. verified by analyzing the results of the 3 x 3 serum pools with different concentrations of IgG (means at 191 [3, 12] . We verified the method if calculated carry-over were < 1% as an arbitrary criterion. Tukey tests were used to detect potential outliers [14] . Our study fulfilled the Ethical principles provided by the Declaration of Helsinki and was approved by the local Medical Ethics Committee (ref: 2020/06AVR/203). This article is protected by copyright. All rights reserved. All results related to baseline demographics, radiography, laboratory and clinical findings including symptoms comorbidities, and treatments of COVID-19 patients are summarized in Table 1 Table 2 ). The analysis on clinical features showed that the number of deceased (8/100 vs 9/25 COVID-19 patients) cases was statistically different between 2 latter subgroups (p< 0.001; Table 2) Diagnostic specificity The means of IgG and IgM levels and their 95% CI in COVID-19 free group (100 samples, 100 subjects) were 0.12 AU/mL (0.09-0.15) and 0.35 AU/mL (0.32-0.37), respectively. All COVID-19 free samples had Ab levels less than the seropositivity cut-off value claimed by the manufacturer (1 AU/mL). Hence, the diagnostic specificity was estimated at 100.0% (96.3-100.0). The PPV was also estimated at 100.0%. As none of these samples had Ab levels > 1 AU/mL, the seropositivity cut-This article is protected by copyright. All rights reserved. showed the time from symptoms onset to hospital admission (mean days= 6.8 vs 3.9), and to sera sampling (mean days= 12.8 vs 7.7) were significantly different between COVID-19 patients who developed Ab (80.0%; n= 125) and those without any immune response. All data regarding diagnostic sensitivities, NPVs, AUCs and test accuracies between 0 to 25 days after symptoms onset are summarized Table 3 . Precision, trueness, limit of blanc (LOB), limit of detection (LOD), limit of quantification (LOQ), linearity and carry-over The bias (%), within and between-run CVs to assess the method precision are summarized in table 4. Both negative control levels had target values less than estimated LODs. Hence, we did not take into account the highly elevated CV for negative control levels. Both positive control level CVs were less than 10%. The precision (< 10%) and trueness (< ±10%) assessments were satisfactory. The LOB and LOD of IgG and IgM were also showed in table 4. The LOQ of IgG was verified as all overall CVs/serum pool were less than 20%. The carry-over assessments were satisfactory as they appeared less than 1%. The linearity assessments were not satisfactory. The higher the dilution titer, the greater the result. Dilution testing using NaCl 0.09% solution were not also conclusive as we reached to more than 400% of recovery after 1:8 dilution titer/Ab types (data not shown). Hence, we propose to avoid any sample dilution in samples with high Ab concentrations. The comparison showed that the Euroimmun kit was less specific (100.0% vs 93.0%). Two false positives (FP) were observed, the first one was a sample from 2018, and anti-alanyl-tRNA synthetase (anti-PL 12) was presumed to be the cause of the cross reactivity. The other was a fresh sample collected from a healthy volunteer (HV) in Mars 2020. All examinations, including clinical, radiological and biological (RT-qPCR and serology) were negatives. However, MAGLUMI kit appeared to be less sensitive as it detected about 64.0% of COVID-19 patients (30/n= 47 vs 79.0% detected by Euroimmun kit), when combining IgG and IgM vs IgG and IgA couples. The degree of agreement estimated by the kappa was 0.77 (95% CI= 0.63 to 0.91) which could be considered as a substantial, to almost perfect agreement between the two methods. Here we have demonstrated that an immune response can be expected in almost all patients after more than 15 days from the onset of symptoms. The overall diagnostic sensitivity reached to 95.5%, about 18.0 days after symptoms onset. There were two patients who did not develop Ab in the subgroup of 15-25 days after symptoms onset: a 76-year-old man who showed an immune response at day of 25 after the onset of symptoms and died shortly after; and a 32-year-old woman on chemotherapy because of an active invasive ductal carcinoma. The latter patient fully recovered 19 days after symptoms onset (date of the last sampling). As there was not any statistical difference (p= 0.15) between immunocompromised patients who did not developed Ab to those who developed (including immunocompromised patients) an immune response, we did not exclude these data from the statistical analysis. More studies should be carried out on immunocompromised patients suffering from COVID-19 to fully understand how their immune system responds to SARS-CoV-2 infection. The statistical analysis showed age (p= 0.03) and time from symptoms onset to hospital admission (p= 0.001), and to sera sampling (p< 0.001) were significantly different between COVID-19 patients who developed Ab (80.0%; n= 125) to those without any immune response ( Table 2) . As the immune system takes several days (generally more than 4 days) to set in for developing Ab against a specific Ag, it was not odd to observe such differences [16] . Compared to our younger patients (mean= 64 years old), the more advance was the age (mean= 71 years old) the earlier was the hospital admission (mean= 5 days earlier). Other explanations could be the functional decline of the immune system in elderly patients and premature deaths [15] . The discrepancies between the diagnostic sensitivities could be explained by the number This article is protected by copyright. All rights reserved. of included patients in each study. One advantage of our study was the number of the patients (125) included and the number of the samples (176) tested. We did not use the immunosuppression criterion (caused by various diseases/therapies) to exclude 14 patients' data, as it appeared there was not any significant difference (p= 0.15) between immunocompromised and immunocompetent patient in our subcohorts. We also did not consider the fever as a sole criterion for the symptom onset as Padoan, et al. [15] opted such. As patients could present various symptoms, we considered all symptoms (fever, cough and dyspnea) to obtain the date of symptoms onset. A study using 535 plasma samples obtained from 173 confirmed COVID-19 patients on a ELISA platform supplied by Beijing Wantai Biological Pharmacy Enterprise Co.,Ltd, showed about 99.0% diagnostic specificity and 93.0% overall diagnostic sensitivity with a median of 11 days (IgG= 14 days & IgM= 12 days) after symptoms onset. The latter study also showed it may take more than a month to reach 100.0% seroconversion. Interestingly, the diagnostic sensitivity at all stages of infection increased and reached more than 99.0% (n= 173) using RT-qPCR (65.0%) and Ab measurement (94.0%) in combination [16] . Our results are in line with the latter report, we obtained an overall diagnostic sensitivity of 95.5%, 18 days after symptoms onset and a diagnostic specificity of 100.0% regardless of sampling period. These estimations should be nuanced as we included only hospitalized patients to assess the diagnostic sensitivities. Considering the current prevalence for seasonal coronaviruses, which is thought to be > 70.0% [17] , and provided data by the manufacturer in package inserts (using 104 sera with Ab against other bacterial and viral respiratory infectious agents, including CoVs), one could suggests that there is no evidence for cross-reactivities in our cohorts. This article is protected by copyright. All rights reserved. Another study with a home-brew designed ELISA kit using SARSr-CoV Rp3 N protein as an Ag, showed an increase of IgM levels from 50.0 to 81.0% (n= 16) and of IgG levels from 81.0 to 100.0%, from 0 to 5 days after hospital admission. Assuming it takes between 7 to 10 days from the onset of symptoms to hospital admission, we can conclude an optimal Ab sensitivities will be achieved after more than 12-15 days from symptoms onset [18] . Based on previous studies and our data, if the first test is negative but the clinical suspicion for COVID-19 remains, a second test should be carried out 2 weeks later to reach an optimal diagnostic sensitivity. Our data are preliminary, and we suggest to confirm the diagnostic specificity and PPV in more general subset. The claimed seropositivity cut-offs by the manufacturer appeared to be accurate. Hence, it could be used in all clinical laboratories. We have also verified the analytical performance of the CLIA method. All analytical verifications were successfully fulfilled beside the linearity assessments which did not confirm Padoan, et al.' report [15] . One explanation could be the latter report did not assess the linearity in higher IgG and/or IgM concentrations suggesting the lack of linearity in high concentrations. We assume that the lack of linearity/Ab types was not caused by the matrix effect as dilution tests using the NaCl 0.09% solution were not conclusive. We also successfully determined the LOB, LOD and carry-over [20] . These discrepancies could be explained by the number of tested samples, inclusion/exclusion criteria, sampling period, infection stages and age criteria in each study. More studies should be carried out to confirm these observations. Based on these studies the overall agreement between the 2 methods was more than 82.0% (Kappa's Cohen= 0.65), which could be considered a substantial agreement. We also gathered all patients' data to check whether they were in agreement with previous clinical/biological findings and if they were related to the seropositive status. We observed that the median time from symptoms onset to hospital admission, to hospital discharge and to death were 6. (Table 1) . Our results also suggested that all patient subgroups developed antibodies over the same period regardless of their outcome. Three studies with 453 (121 dead) patients reported the median time between symptoms onset and the hospital admission ranged from 7.0 to 10.0 [21] [22] [23] . The median times from symptoms onset to hospital discharge and death were 26.0 days (IQR= 22-29) and 16.0 days (12) (13) (14) (15) (16) (17) (18) (19) (20) , respectively comparing 113 deceased and 161 recovered cases [23] . These results were in agreement with our findings. The statistical analysis showed number of deceased cases was statistically different between patients who developed Ab and those without any immune response ( Table 2 ). In our cohort the fever and viral pneumonia were observed in 100.0%, dyspnea in 96.0%, cough in 95.0%, fatigue and/or myalgia in 38.0%, gastrointestinal symptoms in 24.0%, and nonspecific types including headache, confusion, anosmia or dysgeusia in 44.0% of COVID-19 patients ( Table 1 ). The analysis of 2 other studies with 412 hospitalized patients (n=412; 119 dead) showed patients had various symptoms such as fever (93.0%), cough (65.0%), fatigue (57.0%), dyspnea (40.0%), anorexia (29.0%), myalgia (26.0%) in the majority of cases [22, 23] . Our clinical findings on 125 patients were in agreement with these 2 studies beside dyspnea and cough which were more presents in our cohort. Based on our data, we propose to clinically/serologically monitor suspected cases at least for 2 weeks and to look for more specific symptoms such as fever, cough and dyspnea (Table 1 ). In our cohort, all patients were symptomatic and only 80.0% of them developed immune response. In our cohort, some symptomatic patients did not develop Ab (25/125) , and in addition, COVID-19 asymptomatic carriers were not included in our study. However, we could conclude that all symptomatic patients did not develop an immune response. This article is protected by copyright. All rights reserved. These latter studies show the majority of patients did not suffer from any comorbidities (52.0%; n=412 vs our data= 14.0%) and the more common comorbidities were arterial hypertension (AHT; 33.0% vs our data= 44.0%), followed by diabetes (15.0% vs our data= 26.0%) and cardiovascular diseases (12.0% vs our data= 20.0%) [22, 23] . We documented more comorbidities in our cohort and confirmed the age factor (but not the comorbidity) impacted drastically the final outcome. The mean age of deceased subgroup in our study was higher with a mean of 82.1 years old compared to the recovered subgroup, which had a mean of 62.5 years old (Table 1) . It appeared there was not any correlation between comorbidities and Ab development in our cohort ( Table 2) . Our data showed that other than radiology (100.0%), complete blood count (CBC) and blood biochemistry markers could be useful in COVID-19 patients diagnosis if they are available. Over 86.0% and 99.0% of our cohort experienced increased lactate dehydrogenase (LDH) and C-reactive protein (CRP) levels, respectively. These results are in agreement with previous studies, suggesting pathological radiology profile (98.0%; n= 552), skewed biochemistry (> 59.0%; n= 413) and CBC markers (> 58.0%; n= 414) [21] [22] [23] [24] . Hence, in addition to patient monitoring, CTscan, CBC (lymphocyte count) and blood biochemistry markers (LDH, CRP and other hepatic markers) could also be useful as a triage tool if are available (Table 1) . We could not any correlation between these biomarkers and Ab development ( Table 2 ). Studies should be carried-out among pauci-symptomatic, asymptomatic and hospitalized subjects to assess the correlation with biological profiles. Besides supportive care to monitor patient's state and increase oxygen saturation of organism by the means of non-invasive oxygen-therapy, intubation, extracorporeal membrane oxygenation (ECMO) and looking for concomitants or super-infections This article is protected by copyright. All rights reserved. (bacterial, fungal or viral), there is no specific treatment, nor effective vaccine for the treatment of COVID-19 [25] . Two recent reviews, in which a considerable number of articles were methodically analyzed, do not suggest any specific COVID-19 treatment [26, 27] . In our cohort 92.0% of patients received hydroxychlorquin, 43.0% antibiotics and only 8.0% corticosteroids in addition to supportive treatments (Table 1 ). Our data did not suggest that treatments were correlated to Ab development ( Table 2 ). The overall case fatality rate (OCFR), in our cohort was 13.6%. There appeared to be a positive correlation between age and the risk of death. Based on WHO daily situation reports by April 27, 2020 the OCFR was 7.0% globally (n= 3,090,445). Based on the daily situation report published on April 27, 2020, by Belgian center of Epidemiology of Infectious Diseases (Sciensano) the OFCR was 15.7% (n= 49,032) in Belgium. Discrepancies in OCFRs among countries, could be explained by different testing politics, hygienic measures, lockdown rules, case definitions, and age subgroups of infected population [28, 29] . Our study had limitations. First of all, all included patients were symptomatic and needed to be hospitalized. Hence, we can only suggest the same immune response in asymptomatic or pauci-symptomatic carriers. We did not perform any neutralizing antibodies assessment and it is unknown whether the CLIA method detects such antibodies. One could only suggest antibodies are neutralizing based on previous studies on animal and humans [16, 30] . Finally, we could not verify the upper LOQ of IgM as we did not have enough samples with sufficiently high concentrations. This article is protected by copyright. All rights reserved. The serology of COVID-19 patients alone could not allow a diagnosis without the clinical, biological and radiographic data of the patients. Our results showed we could expect a high number of clinically false negatives, particularly in the early stages of infection. On the other hand, clinical false positive results appeared to be rare. Hence, all negative tests should be interpreted cautiously. Our data on patients' characteristics showed, the age and the number of deceased cases were negatively related to Ab development, while the time from symptoms onset to sampling, and hospital admission, were positively related to Ab development. At present, RT-qPCR using nasopharyngeal swabs or lower respiratory tract samples is still the standard laboratory test for SARS-CoV-2 detection, despite the fact that the diagnostic sensitivity does not reach to 100.0%. Other methods which do not detect the virus but the disease (COVID-19), such as serology and radiological examination could be useful in the patient management. In fact, the radiological examination could allow the early detection and triage of patients. Therefore, a combination of RT-qPCR, serology and radiographic examination, would increase COVID-19 detection. As a result of these findings, we propose to check the immune response of suspected patients up at least 15 days after symptoms onset. The CLIA method appeared to be a robust and reliable to measure antibodies against COVID-19 at a high throughput. This article is protected by copyright. All rights reserved. Tables Table 1: Associations and discrepancies according to the diagnosis criteria (positive RT-qPCR and suggestive chest CT-scan and/or X-ray) used for including positive COVID-19 patients, their laboratory results, clinical information and treatments. Coronavirus disease (COVID-19) Pandemic Sample size calculations: basic principles and common pitfalls ICSH guidelines for the verification and performance of automated cell counters for body fluids Manufacturer's package inserts in 2019-nCoV IgG/M-en-EU Cahier des charges définissant les modalités d'évaluation des performances des tests sérologiques détectant les anticorps dirigés contre le SARS-CoV-2 Accuracy (trueness and precision) of measurement methods and results Bias in clinical chemistry Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures; Approved Guideline-Second Edition. CLSI document Limit of blank, limit of detection and limit of quantitation Evaluation of the Linearity of Quantitative Measurement Procedures: A Statistical Approach; Approved Guideline. CLSI document Interference Testing in Clinical Chemistry Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR Reference intervals: current status, recent developments and future considerations Analytical performances of a chemiluminescence immunoassay for SARS-CoV-2 IgM/IgG and antibody kinetics Antibody responses to SARS-CoV-2 in patients of novel coronavirus disease 2019 First infection by all four non-severe acute respiratory syndrome human coronaviruses takes place during childhood Molecular and serological investigation of 2019-nCoV infected patients: implication of multiple shedding routes Evaluation of two automated and three rapid lateral flow immunoassays for the detection of anti-SARS-CoV-2 antibodies Assessment of immune response to SARS-CoV-2 with fully automated MAGLUMI 2019-nCoV IgG and IgM chemiluminescence immunoassays Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study Lombardy Section Italian Society, I. and D. Tropical, Vademecum for the treatment of people with COVID-19 Current Drugs with Potential for Treatment of COVID-19: A Literature Review Pharmacologic Treatments for Coronavirus Disease 2019 (COVID-19): A Review Coronavirus disease (COVID-19) Pandemic. (2020) Reinfection could not occur in SARS-CoV-2 infected rhesus macaques Means of pool 4' Ab levels (AU/mL) x1.33; recovery % 128 Means of pool 5' Ab levels (AU/mL) x1 The authors would like to thank Oris Shenyan for proofreading. All p values were < 0.0001, except for IgG (p= 0.43), IgM (p= 0.77) and IgG/IgM (p= 0.50) obtained by computing 0 to 4 days results after the onset of symptoms.