Persistent Cholestatic Injury and Secondary Sclerosing Cholangitis in COVID-19 Patients

A search of the surgical pathology files at Massachusetts General Hospital (Boston) for medical liver biopsies between May and July 2020 identified 8 patients with a documented history of a positive nasopharyngeal swab for COVID-19, who had a total of 9 liver specimens submitted for routine clinical evaluation. Among these patients, 6 patients had a total of 7 liver core biopsies (including 1 follow-up biopsy), and 1 patient had a liver wedge biopsy; these 7 combined patients comprise this study. The eighth patient had an explant liver specimen, which was excluded from this study because of the presence of cirrhosis and multiple liver comorbidities prior to infection.

Clinical information was obtained from the electronic medical record as available. Clinical details that were collected included demographic information; medical history; immunosuppression status; COVID-19–related length of stay, with length of intubation and tracheostomy; COVID-19–associated medications administered; and significant clinical complications during the patient's COVID-19 illness. Clinical evidence of hepatobiliary injury was also assessed, with a particular focus on the temporal aspects of liver biochemistries from admission to long-term follow-up.

Radiographic information was assessed in detail for all patients who underwent abdominal magnetic resonance imaging or magnetic resonance cholangiopancreatography (MRCP) during their clinical course. Ultrasound imaging of the right upper quadrant with Doppler imaging as well as any subsequent follow-up imaging studies were also assessed as available.

The liver specimens were reviewed in detail for semiquantitative assessment of histologic parameters, including inflammation, cholestasis, acidophil bodies, portal edema, ductular reaction, and bile duct injury, among others. Ancillary stains (including trichrome stain, iron stain, periodic acid–Schiff with diastase special stain, copper stain, and immunohistochemistry for C4d) were evaluated as available. Additionally, in situ hybridization for COVID-19 was performed on each specimen. Appropriate controls were performed for all additional stains. This study was approved by the Institutional Review Board at Massachusetts General Hospital in accordance with the ethical guidelines of the 1975 Declaration of Helsinki.

The 6 men and 1 woman in this cohort had a median age of 63 years (range, 60–76 years). They predominantly self-identified as minorities (1 Hispanic, 2 black, 1 Haitian, 1 Burmese/Chinese, and 1 Italian/white). The most common comorbidities for these patients were obesity (median body mass index, 29.9 kg/m²; range, 25.1–42.4 kg/m²), hyperlipidemia, diabetes mellitus (median A_1c, 6.9%; range, 5.7%–10.1%), and hypertension. All patients were confirmed to have normal liver function and normal liver enzymes prior to COVID-19 infection. It is noteworthy that 3 patients were chronically immunosuppressed because of 1 case each of a history of liver transplant (hepatitis C virus/hepatocellular carcinoma, 8 years prior), kidney transplant (immunoglobulin [Ig] A nephropathy, 3 years prior), and chronic hydroxychloroquine treatment for arthritis (Table 1).

Prior to diagnosis, all patients experienced a typical viral prodrome of cough, fever, and shortness of breath in the very early stages of the pandemic (March and April 2020). Within 2 weeks of symptom onset, all patients were confirmed to have COVID-19 infection by a positive polymerase chain reaction test on a nasopharyngeal swab and were admitted. At the time of admission, liver biochemistries for all patients were normal or near normal (Table 2).

All patients were critically ill with prolonged hospitalization, with a median length of stay of 43 days (range, 34–98 days). All patients were intubated, for a median duration of 19 days (range, 17–36 days), and 6 patients underwent tracheostomy, for a median duration of 23.5 days (range, 17–76 days). As this was early in the pandemic, patients were trialed with a variety of agents, most commonly inhaled nitric oxide, hydroxychloroquine, and atorvastatin. Only 1 patient was trialed on remdesivir (Table 1). All patients were also on a considerable but variable battery of broad-spectrum antibiotics during their hospital stays, including combinations of cephalosporins, macrolides, tetracyclines, and glycopeptide antibiotics (vancomycin). All patients were administered ketamine as a peri-intubation sedative for a median duration of 15 days (range, 3–40 days), with evidence of cholestatic injury occurring a median of 2 days after initiation of ketamine (range, 0–12 days).

During hospitalization, all patients showed acute elevation of liver biochemistries, from baseline normal or near-normal enzyme levels at admission (Table 2). Median peak ALT and AST of these patients were 529 IU/L (range, 137–1065 IU/L) and 311 IU/L (range, 150–1887 IU/L), respectively, at a median of 33 days (for ALT) and 20 days (for AST) from COVID-19 diagnosis (range, 5–70 days). Median peak alkaline phosphatase (ALP) for these patients was 1498 IU/L (range, 512 to >2200 IU/L) at a median of 84 days from diagnosis (range, 26–178 days). In several of the cases, liver biochemistries during the clinical course showed 2 peaks, with a delayed elevation of ALP by a median of 48 days from peak ALT/AST (range, 10–146 days) (Figure 1, A through G). Additionally, elevation of ALP appeared to persist late in the clinical course, even after discharge from the hospital. In several cases, ALP was elevated more than 10-fold above baseline. Most patients also had a roughly parallel increase in total bilirubin (Tbili; median, 3.3 mg/dL; range, 1.3–13.6 mg/dL) and direct bilirubin (median, 2.5 mg/dL; range, 0.8–12.3 mg/dL). The median R factor of the series was calculated to be 0.57 (range, 0.06–1.3), which supports cholestatic liver injury.

Other relevant laboratory values included elevated D-dimer (ranging from 2442 to >10 000 ng/mL fibrinogen equivalent units) and elevated fibrinogen (ranging from 136 to 1123 mg/dL), both of which generally indicate high levels of systemic inflammation and a hypercoagulable state. Serum autoantibody titers were also obtained for all patients, with 6 patients demonstrating positive anti-nuclear antibody (1:160–1:640) and 4 patients demonstrating elevated anti–smooth muscle antibody (1:160–1:320). Immunoglobulin levels were assessed in 5 patients, and all of them had elevated levels of IgG. However, none of these patients had features of autoimmune hepatitis on liver biopsy. Additionally, all patients were clinically evaluated for evidence of other superimposed viral infections (including cytomegalovirus and Epstein-Barr virus) and found to be negative.

All 7 patients had a complicated and extended clinical course with numerous complications (Table 1). All patients had acute kidney injury, with 3 patients requiring continuous veno-venous hemofiltration. More than half of the patients had ventilator-associated pneumonia, and nearly all patients experienced septic shock and required vasopressors. Four patients also showed cardiac abnormalities, ranging from prolonged QT interval on electrocardiogram to atrial flutter with rapid ventricular response. One patient's clinical course was also complicated by small-bowel obstruction, and resection showed intestinal ischemia with microthrombi typical of COVID-19–associated ischemia, which has been previously published.¹³  Only 1 patient was well enough to be discharged home after hospitalization; the remaining patients required extensive stays in skilled nursing or rehabilitation facilities.

After recovering from acute COVID-19 and evaluation of persistent biochemical abnormalities, only 2 patients underwent any intervention: patient 2 was trialed on prednisone and ursodiol, and patient 7 was trialed on ursodiol. None of the other patients underwent any specific interventions. At a median follow-up time from COVID-19 diagnosis of 365 days (range, 119–400 days), many patients showed marked improvement and near normalization of ALT, AST, Tbili, and direct bilirubin. Five patients, however, continued to show persistent elevation of ALP, although improvement was consistent and gradual over time. In the long term, because of persistent clinical and radiographic evidence of sclerosing cholangitis, 1 patient (patient 6) was treated with ursodiol and had a subsequent admission for suspected hepatic encephalopathy, with radiographic evidence of portal hypertension at that time. One patient (patient 7) showed persistent elevation in Tbili, which was direct predominant.

All patients were evaluated with at least one ultrasound of the right upper quadrant during their clinical course, and 4 patients had Doppler imaging, all of which showed patent hepatic vasculature. One patient (patient 1, the liver transplant patient) had an MRCP while hospitalized, and an additional 5 patients had MRCP after discharge for evaluation of persistent ALP elevation (median, 74 days from diagnosis; range, 24–224 days). Three patients (patients 2, 3, and 5) had a normal MRCP without biliary obstruction or filling defect.

Patient 1 was found to have unchanged mild common bile duct prominence without filling defects compared with pre–COVID-19 imaging. However, there was new periportal edema and mild irregularity of intrahepatic ducts (Figure 2, A and B).

The remaining 2 patients (patients 6 and 7) had more significant abnormal findings on later MRCPs that persisted on repeat examinations. MRCPs of patient 6 on days 79 and 141 after COVID-19 diagnosis showed diffuse stricturing and dilatation of intrahepatic ducts throughout the liver but sparing of the extrahepatic ducts (Figure 2, C). MRCPs of patient 7 on days 105 and 345 after COVID-19 diagnosis had more mild abnormalities, limited to beading in the right hepatic lobe but without ductal dilatation, notably with the second MRCP degraded by motion artifact (Figure 2, D).

Six patients had core liver biopsies, at a median time after COVID-19 diagnosis of 68 days (range, 53–170 days). The median length of the core biopsies was 24 mm (range, 20–42 mm), with a median of 18.5 portal tracts (range, 8–38 portal tracts). The histologic findings in the liver biopsies spanned the spectrum of biliary injury, from mild portal changes with cholestatic injury (4 cases) to more dramatic evidence of a biliary obstructive process (2 cases) (Table 3). The wedge biopsy was taken for evaluation of fatty liver disease during a routine cholecystectomy for biliary colic 271 days after COVID-19 diagnosis, a much longer interval than that for the core biopsies, and is discussed separately.

Four cases showed mild cholestatic injury (patients 1 and 3–5). In these cases, portal tracts were variably expanded with a mild mixed inflammatory infiltrate composed of mononuclear cells and scattered neutrophils. Native bile ducts were present but showed mild to moderate bile duct injury with cytoplasmic vacuolization, nuclear hyperchromasia, nuclear disarray, and occasional infiltration by inflammatory cells (Figure 3, A). Canalicular cholestasis was focally present in most cases (Figure 3, B). The lobules showed scattered acidophil bodies and a spotty lobular inflammatory infiltrate (Figure 3, C), and 1 case showed multifocal areas of lytic necrosis consistent with bile infarcts and associated cholestatic resetting (Figure 3, D).

Two cases demonstrated more severe biliary changes, with biliary obstructive features (patients 2 and 6). In these 2 cases, portal tracts showed marked expansion with portal edema, marked ductular reaction, and a moderate to marked mixed inflammatory infiltrate (including abundant neutrophils), consistent with bile duct obstruction (Figure 4, A). Both of these cases were profoundly cholestatic, with both cholangiolar and canalicular cholestasis as well as bile infarcts (Figure 4, B and C). Scattered acidophil bodies, areas of extensive hepatocyte dropout (with a plasma cell inflammatory infiltrate in 1 case), and areas of feathery degeneration and cholate stasis were present (Figure 4, D). There was mild periductal fibrosis and bile duct injury. Although occasional portal tracts suggested loss of native bile duct profiles, no definitive ductopenia was identified in these biopsies. A copper stain performed on the biopsy of patient 6 showed rare granules of copper in periportal hepatocytes, suggesting prolonged cholestasis.

The single case of nonalcoholic fatty liver disease (patient 7) was the wedge biopsy, which (as discussed previously) was procured for evaluation of fatty liver disease during a routine cholecystectomy. It showed mild centrilobular steatosis without definitive steatohepatitis. No significant portal or lobular changes were present, and there was no evidence of a cholangiopathic process. None of the cases showed microthrombi, portal vein endophlebitis, endothelial swelling, or evidence of veno-occlusive disease.

Trichrome stain on all core biopsies showed at least focally expanded portal tracts, with patchy sinusoidal fibrosis present in many of the cases, and focal periductal fibrosis. A copper stain was performed on all cases, to reveal no abnormal copper accumulation in any case except one (patient 6; described above). An iron stain was examined in 5 cases and showed scattered iron accumulation in Kupffer cells in the majority of cases. Periodic acid–Schiff with diastase stains in all cases highlighted scattered ceroid-laden macrophages and no intracytoplasmic globules. Immunohistochemistry for C4d was negative in the 5 cases examined. In situ hybridization for COVID-19 was negative in all cases.

We describe a series of patients with hepatic injury in the setting of COVID-19 infection who developed progressive cholestatic liver injury that persisted for weeks to months after recovery from COVID-19. Our cohort showed 2 peaks in liver biochemical tests during the course of hospitalization, with significant and prolonged increased ALP beyond ALT/AST for an extended duration after hospitalization and evidence of at least a transient secondary sclerosing cholangiopathy (SSC) on histology and radiology in a subset of patients.

Similar cases of prolonged cholangiopathy in critically ill COVID-19 patients have been reported, with a higher frequency than is reported in critically ill patients who do not have COVID-19.^10,11  The largest clinical cohorts were described by Bütikofer et al,¹¹  Faruqui et al,¹⁴  and Roth et al.¹⁰  Together, the published literature describes prolonged and severe cholestasis after critical COVID-19 (often with ALP >3 times the upper limit of normal), and a significant subset of patients show abnormal findings on MRCP.^10,11,14  Histologic data among these cohorts are limited, but examined liver biopsies showed features of large duct obstruction and varying degrees of portal fibrosis,^10,11,14  similar to the findings in this study. One series¹⁰  includes a description of vascular findings, with portal vein endophlebitis, endothelial swelling, and focal veno-occlusive disease. A small series¹¹  of autopsies in patients with COVID-19 also showed evidence of ischemic injury in perihilar ducts with centrilobular necrosis and occasional parenchymal infarcts. Although histologic findings of vasculopathy were not identified in this study, our evaluation for vasculopathy may have been limited by the absence of large portal tracts on biopsy.

Interestingly, among reported cases of patients with post–COVID-19 cholangiopathy, nearly half required consideration for liver transplant, and authors have suggested that liver transplantation may be necessary for treatment of post–COVID-19 cholangiopathy.^12,14,15  However, our cohort demonstrates that these patients commonly show gradual clinical improvement of biochemical abnormalities without drastic intervention. None of the patients described in this study required evaluation for transplant.

The etiology of the cholestatic pattern of liver biochemistries in these patients is incompletely understood. RNA sequencing data from the Human Protein Atlas database and subsequent single-cell sequencing studies have suggested that there is abundant angiotensin-converting enzyme 2 (ACE2) expression on cholangiocytes and endothelial cells but not on hepatocytes,^16,17  providing a potential biological basis for direct SARS-CoV-2 infection. However, the lack of biliary-specific clinical and histologic injury in the acute setting and the persistence of elevated biochemistry months after clearance of the infection suggest that the etiology of liver injury in COVID-19 may not be due to direct viral injury and infection, and is more likely to be multifactorial. Also, these biopsies were taken after resolution of COVID-19, and none showed evidence of viral infection in the bile ducts by in situ hybridization.

An alternative hypothesis for COVID-19–associated secondary sclerosing cholangitis is that immune and inflammatory mechanisms mediate or perpetuate liver injury in these cases, even if the initial trigger may have been viral infection.¹⁸  A number of other respiratory viruses cause mild liver abnormalities, including previous coronaviruses such as severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1) and the Middle East respiratory syndrome coronavirus,¹⁹  with evidence for both direct viral infection and collateral damage mediated by virus-specific T cells in causing liver damage.¹⁸  However, SARS-CoV-2 has not yet been detected in the liver directly, and early literature on small autopsy cohorts demonstrated only mild steatosis and injury. In this context, it would be reasonable to speculate that the role of SARS-CoV-2 infection in liver injury may largely be inflammatory in nature, although this would need further study.

The third and most likely possibility is that these findings represent changes associated with critical illness, which can cause a number of liver abnormalities, including secondary sclerosing cholangitis in critically ill patients (SSC-CIP).²⁰  COVID-19 patients are known to have a long clinical course requiring extended hospital stays and rehabilitation, raising the possibility that these long-term biliary abnormalities are best categorized as SSC-CIP. As such, the findings in this study may be attributable to critical illness, rather than directly caused by COVID-19 specifically. The pathophysiology of SSC-CIP, however, is not well understood and is likely multifactorial, with a number of possible etiologies, including ischemic injury, cholestatic injury due to altered bile metabolism, and lobular injury due to drug toxicity or systemic inflammation.²¹  All of these are potential contributors of COVID-19–associated SSC.

The concept of long-term ischemic injury in the setting of COVID-19 infection as a potential etiology is particularly intriguing. Patients in this cohort had elevated D-dimer and fibrinogen levels, suggesting generally high levels of systemic inflammation as well as an overall procoagulable state, and many of these patients required vasopressor support. Patients with severe COVID-19 also experience other ischemic complications, including ischemic enterocolitis, acroischemia and acute limb ischemia, and other acute thrombotic events.^13,22–24  Histologic examination of ischemic enterocolitis has shown fibrin thrombi.¹³  Therefore, the possibility of ischemic injury to the bile duct would explain the obstructive physiology as well as the MRCP changes. Taken together, it is likely that ischemia, whether from direct thrombotic complications or distributive/septic shock, may be contributing to a clinical secondary sclerosing cholangitis in these patients in the longer term.

Additionally, intrahepatic cholestasis in sepsis has been attributed to altered bile metabolism in critically ill patients.²⁵  Although the patients in this cohort were not septic at the time of biopsy, some of the biopsies showed cholangiolar cholestasis, a finding associated with, but not specific for, sepsis. Altered bile metabolism, therefore, may be a contributor to the cholestatic picture in these patients.

Lastly, the possibility of drug-induced liver injury in COVID-19–associated SSC should be considered given the numerous antiviral agents and ancillary therapies that were administered in these patients, many of which have hepatotoxic potential. In this cohort, there were several notable drug exposures, including cefepime (6 patients), acetaminophen (4 patients), hydroxychloroquine (5 patients), azithromycin (2 patients), atorvastatin (4 patients), remdesivir (1 patient), and peri-intubation sedatives such as ketamine (7 patients). In particular, ketamine has been associated with SSC,^26,27  but it is difficult to definitively implicate ketamine with the cholestatic injury in the patients in this study given the variable presentation and duration of administration. So far, no reports of liver toxicity from remdesivir have been published. However, a randomized controlled trial of lopinavir and ritonavir showed that elevated AST, ALT, and Tbili occurred in a small subset of patients.^28,29  The patients in our small cohort were treated with a variety of different regimens that were completed months prior to development of a cholangiopathy, and it would be somewhat unusual (although not impossible) to see such effects in the long term after cessation of therapy.

In summary, a subset of critically ill COVID-19 patients show a biochemical pattern of cholestatic injury that is marked and extensive in duration in the posthospitalization course. These patients may have obstructive changes on liver biopsy, and radiologic evidence of secondary sclerosing cholangitis that persist up to a year after COVID-19 infection. It should be emphasized that these findings are similar to those seen in other critically ill patients and may not be a direct manifestation of COVID-19 specifically. The mechanism of cholestatic injury remains unclear but is likely multifactorial, with potential contributors including ischemia, widespread systemic inflammation, altered bile metabolism, or drug injury. Our findings support that smoldering cholestatic liver disease and secondary sclerosing cholangitis may be considered long-term sequelae of COVID-19 acute illness, which likely is a form of SSC-CIP, and that this condition often resolves with conservative management.