key: cord-1050652-t5fw4a7i
authors: Poynard, Thierry; Deckmyn, Olivier; Rudler, Marika; Peta, Valentina; Ngo, Yen; Vautier, Mathieu; Akhavan, Sepideh; Calvez, Vincent; Franc, Clemence; Castille, Jean Marie; Drane, Fabienne; Sakka, Mehdi; Bonnefont-Rousselot, Dominique; Lacorte, Jean Marc; Saadoun, David; Allenbach, Yves; Benveniste, Olivier; Gandjbakhch, Frederique; Mayaux, Julien; Lucidarme, Olivier; Fautrel, Bruno; Ratziu, Vlad; Housset, Chantal; Thabut, Dominique; Cacoub, Patrice
title: Performance of serum apolipoprotein-A1 as a sentinel of Covid-19
date: 2020-11-20
journal: PLoS One
DOI: 10.1371/journal.pone.0242306
sha: d5487fdaf69ad870f55c924cf71d4a6240797406
doc_id: 1050652
cord_uid: t5fw4a7i

BACKGROUND: Since 1920, a decrease in serum cholesterol has been identified as a marker of severe pneumonia. We have assessed the performance of serum apolipoprotein-A1, the main transporter of HDL-cholesterol, to identify the early spread of coronavirus disease 2019 (Covid-19) in the general population and its diagnostic performance for the Covid-19. METHODS: We compared the daily mean serum apolipoprotein-A1 during the first 34 weeks of 2020 in a population that is routinely followed for a risk of liver fibrosis risk in the USA (212,297 serum) and in France (20,652 serum) in relation to a local increase in confirmed cases, and in comparison to the same period in 2019 (266,976 and 28,452 serum, respectively). We prospectively assessed the sensitivity of this marker in an observational study of 136 consecutive hospitalized cases and retrospectively evaluated its specificity in 7,481 controls representing the general population. RESULTS: The mean serum apolipoprotein-A1 levels in the survey populations began decreasing in January 2020, compared to the same period in 2019. This decrease was highly correlated with the daily increase in confirmed Covid-19 cases in the following 34 weeks, both in France and USA, including the June and mid-July recovery periods in France. Apolipoprotein-A1 at the 1.25 g/L cutoff had a sensitivity of 90.6% (95%CI84.2–95.1) and a specificity of 96.1% (95.7–96.6%) for the diagnosis of Covid-19. The area under the characteristics curve was 0.978 (0.957–0.988), and outperformed haptoglobin and liver function tests. The adjusted risk ratio of apolipoprotein-A1 for survival without transfer to intensive care unit was 5.61 (95%CI 1.02–31.0; P = 0.04). CONCLUSION: Apolipoprotein-A1 could be a sentinel of the pandemic in existing routine surveillance of the general population. NCT01927133, CER-2020-14.

a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

There is an urgent need to detect people in the general population who are at risk of being admitted to the hospital for Covid- 19 . Although viral nucleic acid testing and chest computed tomography are standard methods for diagnosing Covid-19 in patients with symptoms, these are time consuming. A review reported that there are only three models for predicting hospital admission in the healthy general population, and these are limited by a high risk of bias, and by using proxy outcomes [1] .

In February 2020 we observed in Pitié-Salpêtrière hospital Paris, France ("APHP-PSL"), that patients hospitalized for Covid-19 had a specific profile of apolipoprotein-A1 decrease and haptoglobin increase. This profile was never observed before in our prospective data on apolipoprotein-A1 from and since the first cohort of alcoholic liver disease in 1982 (S1 File) nor in patients at risk of liver fibrosis followed with FibroTest, which include apolipoprotein-A1 in its components. Therefore, in the large cohorts of patients followed by FibroTest for a risk of liver fibrosis (named serum-cohorts) we aim to compare the daily value of serum apolipoprotein-A1, of the ongoing year 2020 to previous years "Covid-19 Free".

Our hypothesis was supported by the literature reporting this profile of apolipoprotein-A1 in patients with viral pneumonia. In 1920, Harold A. Kipp found that a decrease in serum cholesterol was a marker of severe pneumonia [2] . One hundred years later, a meta-analysis confirmed that levels of high-density lipoprotein cholesterol (HDL) and a level below the median of its transporter, apolipoprotein-A1, were associated with a two-fold increase in mortality in patients with severe sepsis (S1 Fig, S1 Table) [3] . Its decrease seems to occur earlier than the increase in haptoglobin, another marker of sepsis, and independently of liver function tests [4, 5] . Recently, hypolipidemia was reported in Covid-19 patients with mild symptoms [6, 7] . Furthermore, unlike haptoglobin, which is synthesized mainly by the liver, apolipoprotein-A1 is synthesized by both the liver and the intestine [5] . This early decrease in apolipoprotein-A1 with normal liver function test and before the haptoglobin increase suggested an intestinal route of infection [8, 9] (S1 File).

Therefore, in the general population, apolipoprotein-A1 might be a sensitive marker of SARS-COV2 infection from apparently healthy carriers, to severe patients hospitalized with To support this hypothesis, we analyze apolipoprotein-A1 in large serum-cohorts and in Covid-19 patients.

In serum-cohorts, the first primary endpoint was to demonstrate a significant decrease of the mean daily value of apolipoprotein-A1 in the 34 weeks of the pandemic year 2020, vs. those of the year 2019. The second aim was to demonstrate the absence of significant confounding factors explaining such temporal association. Then, the third aim was to assess the dynamic temporal association of apolipoprotein-A1 values with the number of confirmed Covid-19 cases in SA and France.

In patients with Covid-19, the first aim was to evaluate the diagnostic performance of apolipoprotein-A1. For this purpose, the sensitivity was assessed in the hospitalized patients. The specificity was assessed in a previous prospective study performed in a representative sample of the French general population before the pandemic. Finally, the last aim was to evaluate the prognostic value of apolipoprotein-A1 in the hospitalized patients.

The prospective observational study in Covid-19 patients was approved by CER-Sorbonne University IRB, CER-2020-14, with a signed informed consent. All of the previously published patient analyses from retrospective databases were non-interventional studies, without supplementary blood samples, and were exempt from a review of the IRB (NCT01927133). The investigation was performed according to the principles of the Declaration of Helsinki. All authors had access to the study data and reviewed and approved the final manuscript.

We used three cohorts of serum from subjects at risk of liver fibrosis followed by FibroTest (FibroSure in USA) [10] . These serum-cohorts had three increasing levels of Covid-19 incidence in 2020. The large private-laboratories US cohort ("US-cohort"), has a lower risk of Covid-19 than the "(French-cohort") following patients in academic hospital treating Covid-19 patients and private-laboratories, and a high-risk cohort which included patients at the Pitié-Salpêtrière hospital Paris, France ("APHP-PSL"), which was a Covid center. The core temporal analysis compared the first 34 weeks of US consecutive anonymous serum 2020 vs. the serum 2019. Details of serum-cohorts were given in S2 File.

Decrease in apolipoprotein-A1 may be due to direct liver toxicity from SARS-CoV-2 [11] , but also to drug-induced liver disease (DILI) caused by medications (S2 Table) . We analyzed the kinetics of alpha2-macroglobulin (A2M) a specific marker of liver fibrosis [10] , and of haptoglobin, a sensitive biomarker of severe acute phase, as liver function tests, gammaglutamyl transpeptidase (GGT), alanine aminotransferase (ALT) and total bilirubin. In serum from subjects followed for non-alcoholic fatty liver disease (NAFLD) the kinetics of total cholesterol), triglycerides, fasting glucose, weight and height were also analyzed.

The number of confirmed Covid-19 cases in France and in the USA was assessed according to published data from the European Centre for Disease Prevention and Control (https:// ourworldindata.org/coronavirus-data) (S2 Fig).

Sensitivity and prognostic values were assessed in a prospective study of Covid-19 patients hospitalized in APHP-PSL. The primary endpoint was the survival without transfer in ICU at 28 days, adjusted on age, gender, haptoglobin, and liver tests.

We previously collected five cohorts, names here "specificity cohorts" which allowed us to retrospectively validate the specificity of apolipoprotein-A1 in a large group of subjects without Covid-19 [12] [13] [14] [15] [16] . The measurements were all performed on fresh prospectively collected serum and analyzed in the biochemistry unit of the APHP-PSL hospital, with the same methods as the Covid-19 cases. The core control population for specificity assessment was a group of healthy volunteers that was representative of the French population (CPAM: Caisse Primaire Assurance Maladie) [16] .

In order to identify a profile of patients with a possible intestinal route of infection, we compared the subsets of patients with or without diarrhea.

Apolipoprotein-A1, haptoglobin, A2M, GGT, ALT and bilirubin were assessed following Bio-Predictive (Paris, France) analytical recommendations [17] . The virological methods for the diagnosis used to diagnose SARS-CoV2 in respiratory samples, were detailed in S2 File.

The kinetics of apolipoprotein A1 and haptoglobin were assessed prospectively in patients with repeated serum during the hospitalization.

The first primary endpoint was to demonstrate a significant decrease of the mean daily value of apolipoprotein-A1 in the 34 weeks of the year 2020, vs. those of the year 2019. We defined a significant decrease of apolipoprotein-A1 as below 1.25 g/L, the optimal cutoff defined by the highest Youden index (sensitivity + specificity -1) using the hospitalized Covid-19 patients for sensitivity, and the CPAM population representative of the general population for the specificity [16] . The daily proportion of serum below this cutoff defined a significant risk of Covid-19. The mean of apolipoprotein-A1 on all the 2019 serum of the US-cohort was taken as the reference in a large population without Covid-19. This allowed to assess the proportion of low apolipoprotein-A1 during the spread of the pandemic, including the peak and the recovery periods. Thanks to the power of the sample size, we simply compared graphically the temporal trends between years for the study. A significant difference of the mean daily levels of apolipoprotein-A1, and the daily proportion of low apolipoprotein-A1 was defined as an absence of overlap of the 95% confidence interval (95%CI) between the curves, during the same 34 weeks periods per year.

The second analysis was to exclude confounding factors. The same graphical method was used. In the large US cohort this method permitted also to describe the trends of apolipoprotein-A1 after a triple stratification on age, gender and the cause of liver disease. The third primary endpoint was to assess the temporal association between the spread of Covid-19 and the decrease in apolipoprotein-A1. For this purpose, we compared graphically the daily mean number of confirmed Covid-19 cases and the daily proportion of low apolipoprotein-A1.

The diagnostic performance of biomarkers was assessed using non-parametric AUROCs. The prognostic values were assessed by survival curves, using Kaplan-Meier method, compared by Logrank test (cutoff being the median of this context of use) and adjusted by Cox model. The repeated serum assessments were compared by repeated ANOVA and Tukey-Kramer multiple-comparison test. R and NCSS-2020 were used as statistical software. 

The core temporal analysis compared the first 34 weeks (between January 1 st to August 20 th , 2020) of consecutive anonymous serum 2020 vs. the serum 2019 from the three routinely followed cohorts. The number for USA was 212,297 serum in 2020, 266,976 serum in 2019, then for the French cohort 20,652 serum in 2020, 28,452 serum in 2019 and for the APHP-PSL cohort 3,122 serum in 2020, and 3,928 in 2019, described in Table 1 . For the US-cohort in 2020, the median age was 54 years (IQR 41-63), 56% men, 80% followed for HCV and 20% for NAFLD, 71% with non-significant fibrosis (stage F0-F1 by FibroTest). During the same 34 weeks period in 2019 these characteristics were similar, 55 (IQR 42-63) years of age, 56% men, 80% HCV and 70% stage F0-F1. All the serum from 2018 were also analyzed for assessing the temporal variability of the bio- Due to the lockdown of populations during the pandemic, both in France and USA the number of the daily number of sera analyzed from January-August 20th, 2020 varied 

The reference daily proportion of low (<1.25 g/L) apolipoprotein-A1 assessed in the 2019 US-serum-cohort, was 27.3% (95% CI 27.2-27.5; 111,262 out of 407,138 serum). The mean daily levels of apolipoprotein-A1 decreased in the three serum-cohorts (Fig 1; all P<0.001), during the first 34 weeks of 2020. This was already highly significant in January 2020 in the US-cohort (lower panel), compared to first 34 weeks of 2019 and 2018. The mean daily proportion of serum with low apolipoprotein-A1 (<1.25 g/L) was 33.1% (95%CI 32.9-33.3 n = 70,212) on 34 weeks, that is 5.8% higher vs. the reference (P<0.001) in the same 34 weeks of 2019.

There was no significant difference between the years 2019 and 2018 for the daily mean by cohort (Fig 1) In the US cohort 2020, the proportion of low apolipoprotein-A1 was 37.0% on the first peak (April 7 th ), that is 9.7% greater than the reference proportion, 27.3%, in 2019. The proportion of low apolipoprotein-A1 was 40% on the second peak (August 18 th ), that is 10.7% Table 2 .

https://doi.org/10.1371/journal.pone.0242306.g002 greater than the reference proportion, 27.3%, in 2019. In France 2020, the increase reached 12.7% on April 14 th and dropped to 0% on June 20 th (Fig 2 and Table 2 ).

The same significant kinetics in apolipoprotein-A1 levels were observed after stratification of the temporal curves for gender and age in the three cohorts. The US-cohort was the only sample that had the necessary power to compare these two factors together between 2020 and 2019 and 2018, and in the subsets of patients with HCV or NAFLD (S3 Fig) . The higher drops were observed in August 2020 for subjects younger than 55 years both for male and female (Fig 3) .

Apolipoprotein-A1 decrease was similar during Covid-19 spread versus 2019 and 2018 in the US-cohort (S2 File, S3 Fig) . The kinetics of apolipoprotein-A1 were not associated with those of haptoglobin in the US-cohort. As expected, there was a significant haptoglobin increase at the peak of the pandemic in the French cohorts and particularly in the APHP-PSL cohort (Fig 4) 

For A2M, in the US-cohort, there was a significant lower mean serum value, when compared to 2019 and 2018, and detailed in S5 Fig. This significant decrease of A2M was regular since January 2018, persisted after stratification by age and gender, but was no longer significant in the serum of NAFLD, between 2020 and 2019 years. In HCV the significant decrease of A2M persisted only between the years 2019 vs 2018 after stratification by age and gender.

The other significant differences were, in the French-cohorts, GGT increased during the pandemic peak and returned to previous years' value thereafter (S6 Fig 1 Used as the reference population for the baseline mean apolipoprotein value. 2 In the APHP-PSL cohort, the proportion of low apolipoprotein-A1 in 2020 was 80.0% on the first peak, that is 52.7% greater than the reference control proportion, 27.3%, in 2019. 3 In the French cohort, the proportion of low apolipoprotein-A1 in 2020 was 40.0% on the first peak, that is 12.7% greater than the reference control proportion, 27.3%, in 2019. 4 In the US cohort, the proportion of low apolipoprotein-A1 in 2020 was 37.0% on the first peak, that is 9.7% greater than the reference proportion, 27.3%, in 2019.

https://doi.org/10.1371/journal.pone.0242306.t002

The daily mean number of confirmed Covid-19 cases paralleled the daily proportion of low apolipoprotein-A1, about 10 days after. In USA, the first ten Covid-cases were declared mid- January 2020, when the proportion of low apolipoprotein-A1 had already increased by several percent. A plateau around 34% of low apolipoprotein-A1 was reached on March 21st that is two weeks before the plateau of confirmed Covid-19 daily cases around 40,000 reached on April 7th (Fig 5, Table 2 ). In France, the first ten Covid-cases were declared first week of March 2020, when the proportion of low apolipoprotein-A1 started to increase by several percent (Fig 2) . The first peak of cases (n = 7,000) was reached mid-April, as well as the first peak of low apolipoprotein-A1 (40.0%) on April 14 th (Fig 2) . Details are given in S2A Fig. 

A total of 136 consecutive patients with severe Covid-19, but who did not require ICU were included. Their characteristics were similar to those published in such severity profiles (Table 3 , S3 File).

The characteristics of patients included in the five specificity-cohorts are presented in Table 4 , the median value of apolipoprotein at inclusion in Fig 6 and Table 5 and for haptoglobin in Fig  7 and Table 6 .

The area under the characteristics curve (AUROC;95%CI) in 136 Covid-19 cases and 7,481 core controls was 0.979 (0.959-0.989), which outperformed haptoglobin and liver function tests (Fig 8 and The adjusted predictive values according to prevalence predicted in the French population [16] , were detailed in S2 File. The specificity-sensitivity including blood donors were detailed in S9A Fig, including patients with rheumatological disease in S9B Fig, and including all the integrated six databases in S9C Fig. During the study period, in the same department 43 patients without suspected Covid-19 were excluded (S3 Table) . These patients were admitted for mixed severe diseases during the pandemic, and therefore could not be used to assess the specificity of apolipoprotein-A1 in the context of use of an early detection test in the general population. 

Apolipoprotein-A1 decrease in Covid-19

The prevalence of diarrhea on initial presentation was 29 out of 131 cases (22.1%;95%CI 15.4-30.2). In this subset, the only significant difference was a lower median number of polynuclear leucocytes (S4 Table, S2 File).

The prognostic value of apolipoprotein-A1 at inclusion for predicting the primary outcome was significant, risk-ratio (RR;95%CI) = 5.61 (1.02-31.0; P = 0.04), adjusted on age (1.04;1.01-1.07; P = 0.04), GGT (2.88;1.01-8.19; P = 0.04). The 71 patients with apolipoprotein-A1 value> = 0.84 g/L, the median value at inclusion, had a significant higher survival without ICU (93.0%;87.0-98.9) than the 65 patients with lower value (75.8%; 65.1-86.5; P = 0.02) (Fig 9) .

Repeated assessments of 256 sera among 136 patients showed the low level of apolipoprotein-A1 at baseline, and thereafter the significant increase already at 10 days in patients who survived without ICU (Fig 10) , as well as in the 16 patients who survived after their transfer to ICU (Fig 11) . Similarly, the mean haptoglobin decreased also significantly in these survivors (Figs 12 and 13 ).

Our study shows that apolipoprotein-A1 displayed a highly significant decrease in 2020 vs previous years, and a highly significant negative daily time-related association with the number of Covid-19 cases. Apolipoprotein-A1 decrease had a high sensitivity in prospective hospitalized patients, with a high specificity in retrospective controls, and an independent prognostic value for the survival without transfer to ICU. These results have certain strengths and limitations ( Table 8 ).

The decrease in apolipoprotein-A1 during the peak of the pandemic in April was not so surprising, as very low levels of HDL-cholesterol were already known in severe pneumonia since 1920 (S1 Table) [2] . More intriguing was the very early decrease observed since January 2020 

Apolipoprotein-A1 decrease in in the USA when the number of Covid-19 cases was unknown. The first known Covid-19 patient was detected on 27/12/2019 and 19/01/2020 in France and the USA, respectively (S3 File). The larger sample size of the US surveillance population, compared to the French, allowed detection of a significant 1% increase in the proportion of subjects possibly infected using the 1.25 g/L cutoff in January (Fig 2) , without any inflammatory signal as assessed by haptoglobin. We hypothesized that the SARS-CoV2 virus influenced the liver or intestinal synthesis of apolipoprotein-A1, in asymptomatic patients or in those with unusual mild symptoms. Indeed, there was a highly significant decrease in ApoA1 parallel (2 weeks before) to the confirmed Covid-19 cases. This decrease started few weeks before the incidence of confirmed cases, suggesting that apolipoprotein-A1 detected infected cases. Furthermore, the liver function biomarkers as well as the haptoglobin did not change so early.

The decrease in apolipoprotein-A1 in 2020 vs. previous years, as well as the time-related association of apolipoprotein-A1 in 2020 and Covid-19 might be due to numerous confounding factors. We acknowledge that consecutive sera were analyzed anonymously and therefore there was an unknown percentage of duplicated subjects with repeated sample over time. Since the database is anonymous, we don't know how frequently individuals were measured, as well as the average sample per person in a year. However, the routine of surveillance is FibroTest between 12 months to 4 years according to the initial stages. This reduced the risk of repeated samples from the same person. In the context of the pandemic we used cohorts of subjects requiring surveillance of liver fibrosis biomarkers which represent at least 30% of the general adult population in the USA and in France. In these cohorts 70% of the subjects had no or minimal fibrosis. The decrease in apolipoprotein-A1 in 2020 compared to 2019, and 2018 cannot be explained by bias due to gender, age, the cause of liver disease (Table 1 , S3 File), or the severity of liver diseases (S4 File). Despite the significant differences in several characteristics, none was associated with the decrease of apoplipoprotein-A1 in 2020 vs. 2019. The changes were observed both in HCV and NAFLD, after stratification on age and gender (S3 File). The prevalence of severe cases cannot explain the significant decrease in apolipoprotein-A1 already observed in January 2020. GGT a very sensitive liver biomarkers did not change during the first 3 months (S6 Fig). Finally, in these severe liver diseases, haptoglobin should also be significantly decreased, which was not the case (Fig 4) . Core population for assessing the specificity (1-% false positive) of apolipoprotein-A1 (ApoA1). 2 Other population for assessing the specificity (1-% false positive) of apolipoprotein-A1 (ApoA1). 3 Core population for assessing the sensitivity of apolipoprotein-A1 (ApoA1).

In the US-cohort the proportion of serum with NAFLD was increased by 2% in 2020 (20.0%) vs 2019 (18.1% Table 1 ). However, after stratification for age and gender, no significant changes were observed for all other biomarkers (S8 Fig). ALT was the only biomarker of the liver tests which increased significantly at the 13 th week of 2020, above the usual mean value observed in 2019 (S7 Fig). This increase in ALT was not associated with any other changes (S8 Fig). We have no clear explanation. We hypothesize that another confounding factor could be a non-severe DILI, including oral acetaminophen or hydroxychloroquine misuse during the pandemic. Such factor could explain an increase in ALT, without increase in haptoglobin, in subjects with mild symptoms.

The absence of haptoglobin change (Fig 4) associated with the linear apolipoprotein-A1 decrease (Fig 1) , has never been described before. It was known that in patients with severe fibrosis these two proteins decrease [10] . It was also known that in severe pneumonia the haptoglobin increase was associated to the apolipoprotein-A1 decrease (S1 Table) , as we observed in the cohorts with high prevalence of severe Covid-19. The 34 weeks followup permitted to see the return to normal values of haptoglobin in these cohorts, associated with the decrease of severe Covid-19 cases admissions (Fig 4) . Furthermore, the recovering patients followed by repeated serum in the prospective study, had both a significant increase of apolipoprotein-A1 (Figs 10 and 11 ) and a significant decrease in haptoglobin 10 days after inclusion (Figs 12 and 13 ).

Our results ( Table 2 , standardized proportion of low apolipoprotein-A1) suggest that the spread of the pandemic in the US-cohort would be around 5.8%, and 4.0% in the Core population for assessing the specificity (1-% false positive) of apolipoprotein-A1 (ApoA1). 2 Other population for assessing the specificity (1-% false positive) of apolipoprotein-A1 (ApoA1). 3 Core population for assessing the sensitivity of apolipoprotein-A1 (ApoA1).

https://doi.org/10.1371/journal.pone.0242306.t006 French-cohort. In France, this estimate does not differ from the recent French model that predicted a rate of infection between 2.8% to 7.2% in the general population [18] . This temporal association of low apolipoprotein-A1 with the number of confirmed cases, persisted in the two different pandemic changes. In France, after the national lockdown the proportion of low apolipoprotein-A1 returned to the cohort usual 2018-2019 values in June together with the dramatic regression of confirmed cases, still maintained in July. In the US, the apolipoprotein A1 and confirmed cases had the same kinetics with the dramatic increase in April and a plateau in June, still ongoing in August 2020 (Fig 5) .

There is a high risk of overestimating both the sensitivity and specificity of a test in Covid-19 when the participants enrolled in the studies might not be representative of targeted populations [1] . It was difficult to confirm the sensitivity for asymptomatic infection, due to the absence of validations of SARS-CoV2-antibodies. To validate the sensitivity of apolipoprotein-A1, a large number of asymptomatic "apparently healthy" subjects who are positive for 

Apolipoprotein-A1 decrease in Covid-19

SARS-CoV-2 viral nucleic acid testing is needed. Our patients included with Covid-19 symptoms had a median of 70 years of age and were severe enough to justify admission to hospital, but none of them required mechanical ventilation at admission, 86% survived, only 14.7% were transferred to the intensive care unit, and 11.8% died ( Table 3 ). The sensitivity in the prospective part of our study was similar in the 19 patients who were negative for viral nucleic acid testing (94.7%) to that in positive patients (89.9%). Thus, a simple measurement of apolipoprotein-A1 could be useful for clinicians due to the high percentage of false negatives in available viral nucleic acid testing [19] . Apolipoprotein-A1 or HDL-cholesterol are already being assessed in many large ongoing studies in patients with chronic diseases, which could rapidly validate our results.

Our previous cohorts of patients with severe liver diseases allowed us to identify the major risks of a significant decrease in apolipoprotein-A1 (false positives), mainly due to severe hepatic insufficiency and severe fibrosis. An impact of malnutrition on apolipoprotein-A1 values, was excluded as no significant changes in weight were observed. The controls not-exposed to SARS-Cov2 in the general population (CPAM) was prospective, but with possible limitations as not being contemporaneous (Table 8) .

The univariate prognostic value of apolipoprotein-A1 was confirmed (Fig 9) but to our knowledge, it was the first time that its prognostic value persisted after adjustment by age, a marker 

Among patients with recovery, there was a significant increase of apolipoprotein-A1 10 days after admission. Therefore, apolipoprotein-A1 measurement could help for the decision to transfer to ICU independently of other prognostic factors such as age and haptoglobin. Furthermore, this kinetics of apolipoprotein-A1, a decrease during the hospitalization and an increase to usual reference values during recovery of Covid-19 cases is a strong argument for a causal interpretation of the temporal association observed also prospectively in the French cohort. The ongoing followup in the US cohort could permit to validate this normalization of apolipoprotein-A1.

Mechanisms of the early decrease in apolipoprotein-A1 before recognition of the pandemic. We never observed such profile of biomarkers in our experience since 2001 with more than three million of FibroTest assessed in liver diseases (S1 File) [10, [12] [13] [14] [15] [16] [17] .

Although several mechanisms explaining the decrease in apolipoprotein-A1 in late severe Covid-19 pneumonia are known (S1 Table) , the reason for the early decrease before the acute phase, when haptoglobin remained normal is unclear. In patients with severe pneumonia, apolipoprotein-A1 decrease was associated with acute inflammation and the "cytokine storm" with an increase in IL6 and acute phase proteins such as CRP and haptoglobin. This dissociation suggests that different mechanisms play a role in the early influence of the SARS-CoV2 virus on the synthesis of apolipoprotein-A1, and the intestine could be more involved than the liver.

The SARS-CoV2 virus could impact several pathways leading to a decrease in the intestinal synthesis and absorption of apolipoprotein-A1 in the small intestine resulting in the decrease in serum [3] [4] [5] [6] [7] [8] [9] . These mechanisms are discussed in S1 File. The first could be the inhibition of lysophosphatidylcholine-acyltransferase-3 activity. There is evidence of direct SARS-CoV2 infection of the endothelial cell and diffuse endothelial inflammation in the intestine [7] [8] [9] . SARS-CoV2 uses angiotensin-converting enzyme-2 receptor (ACE-2) expressed on endothelial cells, to infect the host, widely expressed in the lung, intestine and liver [9, 20] . The second mechanism could be an impact of the virus through the intestinal mucus [21] . Apolipoprotein-A1 is released as a free apolipoprotein from the apical side of enterocytes into the lumen in the fasting state. Apolipoprotein-A1 had faster turnover in mucus, which could be a target for SARS-CoV2 [7, 8, 21] .

Despite the limitations of this study, these results suggest that apolipoprotein-A1 could be a component of multi-analyte Covid-19 population markers, but also diagnostic tests in individual patients at risk, and prognostic tests in individual Covid-19 patients. This ubiquitary protein seems useful in less symptomatic carriers of SARS-CoV2 as very early biomarker but also in patients with severe Covid-19 patients as an independent prognostic marker. It could also These results must be validated in independent cohorts, ideally with virologic and efficiency endpoints. The role of SARS-CoV2 in the possible "asymptomatic" decrease in apoliprotein-A1 could be related to intestinal infection without or before overt pulmonary disease. Finally, one hundred years after surrogate sentinel HDL-cholesterol for pneumonia [2] , apolipoprotein-A1 for the second time in a century, could be "one of the early warning systems that alert the world to potential outbreaks" [22] .

Supporting information S1 File. Mechanisms of the early decrease in apolipoprotein-A1 before recognition of the pandemic, liver or intestine. 

8;10.6-10100 patients with rheumatological diseases, prevalence = 58%. C. Diagnostic performance of apolipoprotein-A1 for the diagnosis of covid-19 in 136 covid-19 cases and

Funding acquisition: Thierry Poynard, Clemence Franc, Patrice Cacoub

Prediction models for diagnosis and prognosis of Covid-19 infection: systematic review and critical appraisal

Variation in the cholesterol content of the serum in pneumonia

Effect of low high-density lipoprotein levels on mortality of septic patients: A systematic review and meta-analysis of cohort studies

Haptoglobin polymorphism, iron metabolism and mortality in HIV infection

The haptoglobin beta subunit sequesters HMGB1 toxicity in sterile and infectious inflammation

Hypolipidemia is associated with the severity of COVID-19

Cholesterol: A new game player accelerating endothelial injuries caused by SARS-CoV-2?

SARS-CoV-2 productively infects human gut enterocytes

Is SARS-CoV-2 Also an enteric pathogen with potential fecal-oral Transmission? A COVID-19 virological and clinical review

Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study

Liver injury in COVID-19: management and challenges

Serum apolipoprotein A1 and haptoglobin, in patients with suspected drug-induced liver injury (DILI) as biomarkers of recovery

Validation of AshTest as a noninvasive alternative to transjugular liver biopsy in patients with suspected severe acute alcoholic hepatitis

Variability in definitions of transaminase upper limit of the normal impacts the APRI performance as a biomarker of fibrosis in patients with chronic hepatitis C: "APRI c'est fini?

Candidate biomarkers for the diagnosis and prognosis of drug-induced liver injury: An international collaborative effort

Prevalence of liver fibrosis and risk factors in a general population using non-invasive biomarkers (FibroTest)

Applicability and precautions of use of liver injury biomarker FibroTest. A reappraisal at 7 years of age

Estimating the burden of SARS-CoV-2 in France

Stability issues of RT-PCR testing of SARS-CoV-2 for hospitalized patients clinically diagnosed with COVID-19

High expression of ACE2 receptor of 2019-nCoV on the epithelial cells of oral mucosa

Apolipoprotein A-1 deposition in, and release from, the enterocyte brush border: a possible role in transintestinal cholesterol efflux (TICE)?

Responding to Covid-19. A Once-in-a-Century Pandemic?