Liver abnormalities following SARS-CoV-2 infection in children 1 to 10 years of age

Introduction

In early 2022, reports emerged of clusters of severe hepatitis not associated with any of the usual viruses in young children in Scotland and soon after, the USA and elsewhere in the UK.1–4 The coincidence in timing of this outbreak with the rise of childhood cases of COVID-19 as the Omicron variant spread made SARS-CoV-2 a tempting explanation for this new disease. For other viral infections, such as influenza, Epstein-Barr virus, Coxsackie B and SARS/MERS, damage to the liver is well documented and is often considered to result from T-cell-mediated attack without necessarily the presence of the virus or viral antigens in the liver.5 In addition, many viruses are known to leave patients vulnerable to secondary invaders. However, it soon became apparent that, although SARS-CoV-2 antibodies could be demonstrated in over half the patients, it was not possible to implicate the virus directly in severe liver damage.6 Subsequent studies have suggested that the culprit in the outbreak of severe hepatitis might be adenovirus type 41 species F (Ad-F41) or Ad-F41 plus adeno-associated virus 2.7 8 Some series indicate that as many as 80% of patients have evidence of infection with Ad-F41,9 so despite the absence of the usual pathologic signature of adenoviral hepatitis, inclusion bodies, this explanation seems the most likely. The reason for the sudden outbreak at widely separated geographical locations, however, is unclear. Concern persists that our lack of understanding of this outbreak may leave us less able to cope effectively with another such episode, should it occur.

The timing of the hepatitis outbreak coincident with the surge in COVID-19 in children suggested some role for SARS-CoV-2 in priming children for the devastating outbreak.10 In adults, SARS-CoV-2 is known to produce lingering alterations in the immune system that can also lead to autoimmune phenomena.11 12 In children, viral persistence has been detected in the gut with prolonged shedding in stool. This ongoing shedding has been postulated as a possible trigger for immune activation that may have affected the response to subsequent adenovirus infection.10 13–15

To investigate the possible role of SARS-CoV-2 in liver dysfunction in children, we examined the risk for protracted abnormalities in liver enzymes or bilirubin, or new diagnoses of hepatitis or liver disease, in children following diagnosis of COVID-19 compared with children who had other respiratory infections (ORIs).

Methods

We conducted a study using retrospective cohorts of aggregated and deidentified electronic health record (EHR) data from over 100 million patients on the TriNetX analytics platform contributed by 59 healthcare organisations (HCOs) in the USA representing a diverse population geographically, demographically and in insurance status. We employed the US Collaborative Network for the primary analyses. For additional privacy protection, some HCOs that contribute to the network date-shift individual EHRs from 1 to 365 days on the calendar while maintaining the relative timing of clinical data within each EHR. Analyses investigating secular trends during the pandemic or comparing periods predominated by SARS-CoV-2 variants (Delta, Omicron) were performed in the Research USA Minimal Date Shift Network including only HCOs that contributed minimally shifted EHR data (0–7 days). Built-in analytical functions allow for analysis of patient-level data in aggregate form only, without individual protected health information. See online supplemental eMethods for details. The MetroHealth System Institutional Review Board has designated the use of deindentified aggregated data on the TriNetX platform in ways such as described in this manuscript as not involving human participants. The study design and results reporting followed the guidelines of the Strengthening the Reporting of Observational Studies in Epidemiology.

Results

The overall study population was composed of children 1–10 years of age without pre-existing liver disease with EHR-documented COVID-19 (263 069) or a non-COVID-19 respiratory infection (1 111 362) between 1 January 2020 and 31 December 2022. The study population was drawn from 5 660 153 similarly aged children with at least one encounter during the study period. Prior to matching, patients in the COVID-19 cohort were significantly older than patients in the ORI cohort (table 1). After propensity matching on baseline characteristics (table 1), 2 60 132 patients were in each cohort. Tables summarising baseline characteristics for the subset of children aged 1–4 years, and the subsets of children in both age groups with a record of hospitalisation during the acute period (from 1 month prior to 1 month after the respiratory infection encounter diagnosis) can be found in online supplemental eTables 1–3.

Table 2 summarises the counts and hazard of hepatic complications after COVID-19 versus ORI for both age groups (1–4 years, 1–10 years) for all cases and for the subset with an associated hospitalisation. During the 1–6 months after the encounter diagnosis for the respiratory infection, patients with COVID-19 had a significantly greater hazard of elevated AST or ALT, and TBili with HR point estimates ranging from 2.13 to 4.40. HR point estimates were significantly increased for the composite outcome of diseases of the liver or viral hepatitis (K70–K77, B15–19: HR point estimate range, 1.48–4.50), and for diseases of the liver alone (K70–K77: HR point estimate range, 1.67–4.38). Few encounters were coded with the following ICD-10 encounter diagnoses: alcoholic liver disease (K70); toxic liver disease (K71); hepatic failure, not elsewhere classified (K72) (except for all patients aged 1–10 years with 12 in the COVID-19 cohort and 10 in the ORI cohort (HR (CI) 11.42 (1.49 to 87.83))); chronic hepatitis, not elsewhere classified (K73); fibrosis and cirrhosis of liver (K74). ICD-10 terms with sufficient counts to allow comparison had elevated HRs: other inflammatory diseases of the liver (K75: HR point estimate range, 2.85–9.85) which includes autoimmune hepatitis (K75.4) and other specified inflammatory liver diseases (K75.8); and other diseases of the liver (K76: HR point estimate range, 1.46–3.97) which includes fatty change of liver, not elsewhere classified (K76.0). The composite outcome of ICD-10 codes used for surveillance of hepatitis of unspecified aetiology surveillance also had significantly elevated HRs (HR point estimate range, 2.62–6.23). All HRs had 95% CIs that were greater than 1 except the outcome of other inflammatory diseases of the liver in the 1–4 years age group. The Cox proportionality assumption at p≥0.05 was met for all comparisons except elevated ALT/AST for 1–4 years age group with inpatient stay (p=0.049), and elevated TBil for 1–10 years age group with inpatient stay (p=0.01). A comparison of these outcomes without excluding patients with the encounter diagnosis for one of these outcomes during the acute period had increased RR estimates for all comparisons. See online supplemental eTable 4. Analyses conducted with cohorts of patients without excluding patients with pre-existing liver abnormalities showed similar results. Furthermore, additional ICD-10 terms had sufficient counts to report from this analysis, showing an elevated risk in the COVID-19 cohort for viral hepatitis (B15–B19) and hepatic failure, not otherwise classified (K72). See online supplemental eTable 5.

When comparison was restricted to patients with AST, ALT or TBil measured, but not shown to be elevated during the acute period, patients with COVID-19 had a greater hazard for elevated AST, ALT or TBil during the 1–6 months follow-up period in both age groups (table 3). Similarly, a comparison was made between patients with labs measured during the acute period, but without excluding those who may have had elevated labs during the acute period. The RR for persistent or newly incident lab elevations was greater for patients after COVID-19 (online supplemental eTable 6). Among patients aged 1–10 years with persistent lab elevations into the 1–6 months follow-up period after COVID-19, the most recent elevated lab value mean U/L (SD) were as follows: AST 236 (1760), ALT 133 (260) and TBil 2.16 (2.89).

The prevalence of pre-existing conditions (documented at least a month before the respiratory diagnosis) among patients who, during the 1–6 months follow-up period, had encounter diagnoses for diseases of the liver (K7–K77) or viral hepatitis (B15–B19) following COVID-19 versus ORI were compared. Cases with hepatic complications that followed COVID-19 versus ORI were older (mean, 5.51 vs 4.42 years) and had a higher prevalence of pre-existing conditions. However, when limiting the comparison to cases who were hospitalised during the acute phase of the respiratory infection, the prevalence of pre-existing conditions among COVID-19 or ORI-associated cases varied by condition. See online supplemental eTable 7 for details.

Analyses comparing shorter time periods yielded smaller total counts and did not uncover any temporal or variant-related trends. These analyses were conducted using the Research USA Minimally Shifted Network which has approximately 35% fewer records. When comparing COVID-19 to ORI during five serial 6-month time blocks from 1 March 2020 to 30 September 2022, there were no significant secular trends. Elevations in liver enzymes and bilirubin were evident in all time periods. Liver-related complications were seen in each time period, however, the relatively small counts when divided over five time periods made the comparison between COVID-19 and ORI underpowered. When restricting analyses to infections that occurred during COVID-19 waves in which either the Delta or the Omicron variant was predominant, results for liver enzymes and bilirubin were similar for the COVID-19 to ORI cohort comparison during each wave as when comparing the results over the entire study period. No differences in liver enzyme and bilirubin elevation were found when comparing COVID-19 cohorts between waves.

Discussion

In a large population sample from across the USA, children who had COVID-19 were at significantly higher risk for elevated liver enzymes and bilirubin or a new diagnosis of hepatitis of various types in the following 6 months. These elevations were observed for children aged 1–10 years and also for younger children aged 1–4 years when compared with patients with a non-COVID-19 respiratory infection, and it was true for patients with COVID-19 of unspecified severity, as well as for those with an associated hospitalisation. The risk was increased during the period 1–6 months after COVID-19 diagnosis whether or not patients with pre-existing liver disease were excluded or whether or not patients with liver abnormalities during the acute illness period were excluded from the sample. The risk was also noted when comparing only patients with normal labs measured during the acute phase of the illness. The increased risk for elevated transaminases or bilirubin was evident following a COVID-19 diagnosis during each time period of the pandemic. Because the surge in severe hepatitis among paediatric patients occurred coincident with the Omicron variant surge in the pandemic, we directly compared the time when the Delta variant predominated to the time when over 90% of sequenced viruses were the Omicron variant. No difference in the risk of subsequent liver enzyme or bilirubin elevations was found between variant waves.

Since several respiratory infections are known to be associated with subsequent liver abnormalities, the excess of new cases among children who contracted SARS-CoV-2 compared with those with prior respiratory infections may be especially telling. Indeed, the excess risk of elevated transaminases or bilirubin was much higher when children diagnosed with COVID-19 were compared with children who had a medical encounter for any reason, but this category could include well-child visits, fractures or allergies, diagnoses much less likely to display liver abnormalities than children who had viral respiratory infections. However, even compared with ORIs, COVID-19 posed more of a risk for subsequent liver diagnoses, suggesting that it does impose particular vulnerability. While elevated transaminases and bilirubin could reflect pathology of non-hepatic aetiology such as haemolysis or myositis and should prompt a clinical investigation that considers these other causes, the increased risk of these lab abnormalities in the COVID-19 population in comparison to non-COVID-19 respiratory infection population was mirrored by the elevated risk of encounter diagnoses for hepatic complications. The physician must be alert to ongoing vulnerability to other viral infections, or to lingering inflammatory damage following COVID-19.

We tested whether prior infection with SARS-CoV-2 is associated with subsequent diagnoses of hepatitis of known viral origin (hepatitis A–E), unspecified viral hepatitis, or cases of acute liver complications for which an infectious origin was not diagnosed, including toxic liver disease, hepatic failure, other inflammatory diseases of the liver (including autoimmune hepatitis) or other diseases of the liver (including fatty change). We reasoned that if COVID-19 conferred vulnerability by any of the mechanisms involving the immune system, this vulnerability may extend to many infectious or even non-infectious disorders of the liver. For all acute disease, ICD-10 codes examined for which there was a large enough number of cases for comparison, the risk was higher among children 1–6 months following COVID-19 than it was following ORIs. For the following codes: viral hepatitis (B15–B19); toxic liver disease (K71); hepatic failure, not elsewhere classified (K72); the number of cases was too small for the primary analysis. We speculate that ongoing low-grade liver inflammation following COVID-19, though rare (approximately 10 in 10 000) among children without a history of liver abnormalities, leaves children vulnerable to either develop liver disease or to acquire an additional infection. Nevertheless, if 10 million children were infected, 10 000 or more might be left at risk.

Children were more likely to have liver enzymes or bilirubin measured during the following 1–6 months if they were diagnosed with COVID-19 at the index encounter (RR, 1.8) even though the risk of a follow-up visit for any reason was only modestly increased for children with COVID-19 (RR, 1.07). This probably speaks to the index of suspicion in the physicians caring for these children. However, this excess of sampling might create some bias of ascertainment in detecting liver abnormalities among children following COVID-19. Therefore, we compared only those children who had blood drawn to measure liver enzymes or bilirubin around the time of the diagnosis of COVID-19 or ORI. For this comparison, laboratory measurements showing liver enzymes and bilirubin were also elevated among children following COVID-19.

How COVID-19 might predispose to later liver abnormalities is a matter for speculation. For adults, four mechanisms have been suggested: direct infection, toxicity from cytokine storm or hypoxic injury associated with severe COVID-19, toxicity from the drugs used to treat COVID-19 and immune phenomena.17 Direct infection of hepatocytes is unlikely because viral receptors are sparse on hepatocytes, although they do occur in abundance on cholangiocytes. However, there are case reports of acute hepatitis associated with COVID-19, some of them accompanying multisystem inflammatory syndrome in children (MIS-C), and others, without significant respiratory involvement.18 Though in adults, worse acute liver injury is associated with more severe COVID-19, we did not observe an association with severity of COVID-19 as assessed by need for hospitalisation. Moreover, children tend to have less severe disease than adults. Therefore, it is less likely that cytokine storm, vascular injury or hypoxia during the acute illness account for our findings. Few antiviral drugs are approved for children under 10 years of age (paxlovid has not been approved for children <12 years of age and remdesivir has been seldom used), so drug-induced hepatotoxicity is also less likely than in adults. Therefore, immune phenomena seem the most likely explanation. Many autoimmune phenomena have been documented following SARS-CoV-2 infection, so an autoimmune mechanism of liver damage is possible.19 MIS-C is thought to be a result of immune dysregulation following SARS-CoV-2 infection.20 Case reports of autoimmune hepatitis following COVID-19 in adults have appeared.12 21 22 In addition, for other viruses, T-cell-mediated damage to the liver has been observed even in the absence of persistence of the virus or viral antigens in the liver.5 Further work is needed to sort out these possibilities.