Incidence and Sequelae of Liver Injury Among Children Treated for Solid Tumors: Analysis of a Large Single-Center Prospective Cohort

Abstract

Objectives: 

Pediatric oncology patients are at risk of adverse drug events. The incidence and etiologies of liver injury in this population are not well characterized. We utilized a large, single-center pediatric oncology registry to investigate the incidence, causes, and outcomes of liver injury during treatment for solid tumor malignancies.

Methods: 

We reviewed all young individuals (age <25 years) who received treatment for any solid tumor at the University of Michigan between January 2004 and July 2016. Subjects with liver injury meeting predetermined laboratory criteria were identified. Cases were independently reviewed by 2 expert hepatologists to assign a cause of liver injury. Clinical characteristics of drug-induced liver injury (DILI) and non-DILI cases were compared. Cases of liver injury occurring after bone marrow or liver transplant were excluded.

Results: 

Of 1136 solid tumor patients, 160 (14%) experienced liver injury, and the overall frequency of DILI was 4%. DILI was the leading identified cause of liver injury (31%), followed by infection (17%), metastatic/malignant biliary disease (13%), and perioperative liver injury (13%). Most DILI cases (>90%) were mild acute hepatocellular injury episodes that did not result in modification to the chemotherapy plan, and all DILI eventually resolved. Severe presentations involving jaundice and/or prolonged hospital course were significantly more common among non-DILI versus DILI cases (23% vs 2%, P < 0.001).

Conclusions: 

DILI is the leading cause of liver injury events among pediatric solid tumor patients. In our registry, DILI was of mild severity and did not result in an alteration of the treatment plan in most patients. In contrast, non-DILI-related liver injury events, including infection, were more likely to have a more severe presentation and a complicated course with a greater mortality during follow-up.

What Is Known/What Is New

What Is Known

• A high percentage of long-term survivors of childhood cancers have chronically elevated aminotransferases.
• Little data are available regarding the liver safety of chemotherapy, and the phenotypes and long-term outcomes of patients with drug-induced liver injury.

What Is New

• Drug-induced liver injury is the most common cause of liver injury among young individuals treated for solid tumors.
• Drug-induced liver injury is generally mild without jaundice, and most patients do not require modification of their cancer treatment because of drug-induced liver injury.
• High-dose methotrexate is the most commonly implicated agent to cause drug-induced liver injury particularly in sarcoma patients.
• Nondrug-induced liver injury liver injury is more severe and more likely to lead to a prolonged hospitalization with a higher observed mortality during follow-up.

Cancer patients are frequently prescribed single or combination regimen therapy in high doses. Among adult patients receiving cytotoxic chemotherapy, incident liver enzyme abnormalities to greater than laboratory upper limit of normal (ULN) are common (noted in 10%–40%) and more significant liver injury (alanine aminotransferase >3× ULN or total bilirubin >2× ULN) is noted in approximately 1% to 4% of treated patients ⁽¹⁾.

There is a relative paucity of literature regarding the liver safety of chemotherapy in children. This represents a critical knowledge gap, as children are generally known to receive higher dose-intensity chemotherapy as compared with adults, because of their more biologically aggressive tumors ^(2,3). Therefore, we set out to investigate the frequency and triggers of bonafide liver injury events among consecutive young individuals treated at a large pediatric referral center. Furthermore, in view of recent literature demonstrating that a high percentage of long-term survivors of childhood cancers have chronically elevated aminotransferases ^(4,5), we set out to determine the phenotypes and long-term outcomes of DILI.

We retrospectively reviewed a large registry of nearly 2000 pediatric oncology patients to identify cases of liver injury, and then applied the rigorous causality assessment methodology of the US Drug-Induced Liver Injury Network (DILIN) to characterize the overall frequency, triggers and clinical characteristics of drug-induced liver injury (DILI). Herein, we present the results of our analysis of patients treated for solid tumors.

METHODS

This retrospective cohort study was approved by the Institutional Review Board (IRB) of the University of Michigan. The need for patient or parent/guardian consent was waived by the IRB.

Since January 1, 2004, young individuals (here defined as children and young adults age <25 years) evaluated at the University of Michigan pediatric oncology clinic have been prospectively enrolled into a comprehensive database, hereafter referred to as the Pediatric Hematology and Oncology (PHO) database. Demographics and clinical characteristics are recorded in the PHO database. We retrospectively reviewed all patients with solid tumors enrolled in the PHO database between January 1, 2004 and July 31, 2016 (see Supplemental Table 1, Supplemental Digital Content, https://links.lww.com/MPG/B833, for list of included tumor types).

Patients who experienced liver injury during treatment were identified by screening the electronic medical record using the software package DataDirect. Liver injury was defined as liver enzyme abnormalities at any time following initiation of cancer treatment, as follows: serum alanine (ALT) or aspartate (AST) aminotransferase >5× upper limit of normal (ULN) or serum alkaline phosphatase (Alk P) >2× ULN on 2 consecutive lab checks within 2 months, or serum total bilirubin > 2.5 mg/dL accompanied by any elevation of AST, ALT, or Alk P > ULN on a single laboratory draw. We followed age-adjusted reference standards of the University of Michigan clinical laboratory. We analyzed the first episode of liver injury in patients less than 25 years of age at the time of liver injury. We excluded patients who did not receive any antineoplastic therapy at our center, and liver injury after bone marrow transplant or liver transplant.

A preliminary chart review of cases meeting laboratory criteria was conducted by 2 expert hepatologists (V.C., F.D.). For cases obviously unrelated to DILI, an appropriate alternative diagnosis was assigned.

The remaining cases were considered possible DILI cases, and each medical record was systematically reviewed to populate a standardized template. The template included a brief summary of all past cancer-related treatment, full details of all medication exposure for 2 months prior, and 1 month following liver injury onset. The names, doses, routes, and frequency of all medications were recorded. The type and date of any surgical procedure, the date, duration, field of treatment, and dose of any radiation therapy received, and the date and duration of any exposure to total parenteral nutrition (TPN) were recorded.

DILI cases were identified by following the methodology of the US DILIN ^(6,7). The 2 hepatologists (V.C., F.D.) independently reviewed the medical records of each possible DILI case, and assigned liver injury pattern (hepatocellular, cholestatic, mixed), causality scores, and severity scores. For cases in which more than 1 agent was felt to be a possible cause of DILI, an overall causality score and causality scores for each agent were assigned. When reviewer scores did not agree, the 2 reviewers conducted a second, joint chart review and discussion in order to assign a final causality score. If agreement could not be reached, the case and discussion were reviewed by a third expert hepatologist (R.J.F.) who then assigned the final score. Cases with a final causality score of definite, highly likely, or probable were considered confirmed DILI cases, and cases with a score of possible or unlikely were assigned an alternative cause or categorized as indeterminate. Peak laboratory values were used to determine severity scores.

The outcomes of the patients with confirmed DILI were described including development of chronic DILI (defined as persistently elevated ALT, AST, Alk P, or bilirubin above ULN or prior baseline for at least 6 months following DILI onset), development of complications including sinusoidal obstruction syndrome, liver imaging abnormalities including evidence of cirrhosis, nodular regenerative hyperplasia, intrahepatic mass lesions, need for liver transplant, or death. Final outcomes data were abstracted on July 2017.

Statistical Analysis

Analyses were performed using R version 3.4.4 and the package tableone version 0.10.0. Bivariate comparisons of demographic and clinical variables (age at diagnosis, sex, obesity, race, cancer type, pattern of injury, DILI severity index, cause of death, AST, ALT, Alk P and total bilirubin) by liver injury types (DILI vs non-DILI, liver injury vs nonliver injury, DILI vs nonliver injury, and non-DILI vs nonliver injury) were done using chi-squared tests or t-tests assuming equal variances for large sample sizes and approximately normally distributed variables. For categorical comparisons, where cell counts were too low to permit large sample approximation, Fisher exact test was used. For continuous variables that were highly skewed based on visual inspection of histograms, Wilcoxon rank sum test was used instead of the t-test. Where statistically significant associations of cancer type with liver injury type were identified, further comparisons were made of each cancer category (brain tumor, lymphoma, neuroblastoma, renal tumor, sarcoma, and other) by liver injury category to determine, which cancer types had the strongest associations, using chi-squared or Fisher exact tests.

RESULTS

During the study period, 1916 young individuals were enrolled in the PHO database. A total of 317 patients were excluded from further analysis. Of the remaining 1599 patients, 1136 patients had a solid tumor. A total of 160 solid tumor patients (14%) met lab criteria and were designated as liver injury cases, and the remaining 976 patients were designated as nonliver injury cases (Supplemental Figure 1, Supplemental Digital Content, https://links.lww.com/MPG/B833).

Expert review of the 160 liver injury cases yielded 49 confirmed cases of DILI, and 111 cases were assigned an alternative diagnosis. The 2 reviewers independently reached a score consistent with confirmed DILI (both reviewers assigned a score of either definite, highly likely, or probable DILI) in 45 of the 49 cases (Cohen K statistic = 0.67). There was irreconcilable disagreement between the 2 reviewers in only 1 case, which was resolved as non-DILI liver injury by the third expert reviewer (R.J.F.). DILI was the most common identifiable cause of liver injury (31%), followed by viral/bacterial infection (17%) and metastatic liver or biliary tract disease (13%) (Fig. 1). The overall frequency of DILI among solid tumor patients (after exclusions) was 4%.

FIGURE 1: Etiologies of liver injury. ^∗Myositis, total parenteral nutrition-induced cholestasis, hemophagocytic lymphohistiocytosis, malnutrition, neutropenic enteritis, necrotizing pancreatitis, abdominal compartment syndrome, radiation-induced injury (exclusion of 3 cases of myositis and 1 case of neutropenic enteritis from the group of diagnoses categorized as non-DILI liver injury resulted in no change in the statistical significance of any of the comparisons in the manuscript.). ^∗∗Chronic hepatitis, hepatoblastoma, congenital hepatic fibrosis, Alagille syndrome, Caroli syndrome. DILI = drug-induced liver injury.

There were neither significant differences between liver injury and nonliver injury cases with respect to age at cancer diagnosis, sex nor ethnicity. There were, however, significant differences between the 2 groups with respect to primary cancer diagnosis, specifically sarcoma was more common among the liver injury patients (32% vs 18%, P < 0.001), and brain tumors more common among the nonliver injury patients (20% vs 9%, P = 0.001) (Table 1).

TABLE 1: Presenting clinical features and outcomes of patients with and without liver injury during treatment

Clinical Characteristics of the Drug-induced Liver Injury and Nondrug-induced Liver Injury Cases

The mean age at cancer diagnosis among DILI cases was 10 years and 47% were boys. Eighty-four percentage of patients were Caucasian, 8% African American, and 4% Hispanic. Median BMI at DILI onset was 18.8 (range 13.1–39.9) and 73% of patients had a BMI below the 95th percentile. None of the DILI cases had a preexisting diagnosis of chronic liver disease (Table 2).

TABLE 2: Presenting features and outcomes of confirmed drug-induced liver injury cases to nondrug-induced liver injury cases

The majority (51%) of DILI patients were asymptomatic with the most common signs/symptoms present at DILI onset being nausea (31%), fever (6%), abdominal pain (4%), and rash (4%). None had peripheral eosinophilia (defined as an absolute eosinophil count >500 μL). The median latency between drug initiation and liver injury onset for the 29 single-agent cases was 7 days (range: 0–528).

Almost all DILI cases (96%) presented with a hepatocellular pattern of liver injury and a median ALT of 315 IU/L. The median follow-up time after DILI onset was 49 months (range 1--154 months). Almost all cases (96%) were scored as mild severity. There were 2 other cases scored as moderate and moderate-severe, of which 1 was associated with a prolonged hospitalization related to DILI and 1 case associated with jaundice and hospitalization. Both cases fully recovered from DILI within 6 months without long-term sequelae (Table 2).

Sarcoma was the primary cancer diagnosis in 51% of DILI cases, and 22 of the 25 sarcoma cases were osteosarcoma. The next most common tumor types were lymphoma (25%), neuroblastoma (8%), and CNS tumors (8%). Sarcoma was significantly more common among DILI cases as compared to non-DILI liver injury cases (51% vs 23%, P = 0.001), but not significantly different in frequency between non-DILI liver injury cases and nonliver injury cases (23% vs 18%, P = 0.17) (Table 2).

The mixed or cholestatic lab patterns at presentation were more common among the non-DILI liver injury patients versus DILI patients (41% vs 4%), and patients with non-DILI liver injury overall had a higher frequency of severe presentations of liver injury as compared with the DILI cases (Table 2). Typically, the higher severity scores in the non-DILI liver injury group were related to development of jaundice with or without need for hospitalization because of liver injury.

Agents Implicated in Drug-induced Liver Injury Cases

A single agent was implicated in 29 DILI cases, and 20 cases involved complex multiagent regimens, such that a single agent could not be implicated as the sole cause of DILI. The cancer diagnoses and implicated agents in the multi-agent DILI cases are summarized in Supplemental Table 2 (Supplemental Digital Content, https://links.lww.com/MPG/B833).

Among the single-agent cases, the most commonly implicated agent was high-dose intravenous methotrexate (HD IV MTX) (86%), and in 22 of these 25 cases, the MTX was being used to treat osteosarcoma. In these cases, the median latency was 9 days (range: 0–371) days. Twenty of the 22 cases presented with hepatocellular injury and 2 cases with mixed injury. Twenty-one of 22 cases were of mild severity. DILI led to a modification or delay in the dose of methotrexate in only 2 cases.

The other implicated single agents were single cases of doxorubicin, mitotane, crizotinib, and bevacizumab hepatotoxicity (see Supplemental Digital Content and Supplemental Table 3, Supplemental Digital Content, https://links.lww.com/MPG/B833, for case summaries).

Outcomes of Drug-induced Liver Injury

The overall median time to DILI resolution was 28 days (range: 6–1693). The majority of patients (69%) were able to continue their chemotherapy regimen without any modification in the dose or timing, and another 23% had chemotherapy modified following DILI onset because of other side effects. Therefore, the occurrence of DILI did not result in modification of cancer treatment plan in 92% of cases. The remaining 8% (4 cases) had their chemotherapy paused or modified because of DILI, these included 2 cases related to HD IV MTX, 1 related to crizotinib, and 1 related to mitotane.

The overall incidence of chronic liver injury 6 months after DILI onset was 14% (7/49). The majority of chronic DILI cases (58%) were being treated for lymphoma. Two of the cases had DILI attributed to a single agent (crizotinib and bevacizumab), and the other 5 cases involved multiple agents as possible triggers. All chronic DILI cases eventually resolved without further complications. The median time to lab normalization among the chronic DILI cases was 259 days (230–369).

There was a higher overall incidence of death among the liver injury cases versus nonliver injury cases (38% vs 14%, P < 0.001). There was also a trend towards a higher incidence of death between non-DILI and DILI liver injury cases (43% vs 27%, P = 0.067). In both the DILI and non-DILI liver injury groups, however, most died because of progression of their primary cancer and no patients died because of DILI. Septic shock was the second leading cause of death (15%) in the non-DILI liver injury group but was not observed as a cause of death among the DILI cases (Table 2).

Overall, a total of 13 DILI patients died with 12/13 deaths attributed to cancer progression, and in 1 case, the cause of death was not able to be determined because of inadequate follow-up. None of the deaths were chronic DILI cases and none had reported signs or symptoms of liver dysfunction or imaging evidence of advanced fibrosis at any time following DILI.

DISCUSSION

Accurate identification of DILI cases is very challenging particularly when the clinical context is complex. To address this issue in our cohort of pediatric oncology patients, we applied the expert adjudication methodology of US DILIN. The expert opinion adjudication process is a comprehensive approach that has been shown to outperform the widely utilized Roussel-Uclaf Causality Assessment Method (RUCAM), and is the best available means to confirm that cases labeled as DILI are indeed bonafide cases ⁽⁸⁾. The high level of independent agreement regarding confirmed DILI cases in this study suggests the high quality of the DILI cases.

The overall incidence of liver injury events meeting our lab inclusion criteria was 14%, and the incidence of DILI was 4%. Thus, our data clearly establish that young individuals treated for solid tumors are at special risk of liver injury overall, and of DILI relative to the general population ^(9,10). Further, DILI was the leading identifiable cause of liver injury and more common even than infection.

DILI in this population tends to have distinctive features. Almost all cases demonstrated the hepatocellular pattern of liver injury at DILI onset. It must be pointed out that the hepatocellular pattern of liver injury is generally more common in children as compared with adults, given that the ULN for Alk P is much higher among most children versus adults ^(11,12). Nonetheless, jaundice was an uncommon occurrence. Among those with symptoms, nausea was most common (31%). In our review of cases, nausea was most likely related to treatment-induced mucositis. In addition, immunoallergic features (fever, rash, eosinophilic) were rare.

We found a trend but not statistically significant difference in the frequency of obesity between DILI and non-DILI liver injury cases (27% vs 14%, P = 0.09). The frequency of obesity among both DILI and non-DILI cases was roughly similar to the overall prevalence of obesity among US children (18.5%) as reported by the CDC ⁽¹³⁾. Notably, obesity and nonalcoholic fatty liver disease (NAFLD) have been shown to act as risk factors for DILI related to specific agents, including MTX ⁽¹⁴⁾. The large number of cases of MTX-induced liver injury in our study cohort may have influenced the apparent trend toward an association between DILI and obesity. Further investigation is warranted to ascertain whether and how obesity acts as a risk factor for DILI among young individuals treated for cancer.

Indeed, the clinical features of DILI in our study cohort were impacted heavily by the phenotype of HD IV MTX-induced liver injury, as this was by far the most frequently implicated trigger of DILI in our study. MTX is a well-known hepatotoxin, and the profile of HD MTX-induced liver injury involves severe, acute rise in ALT and AST without jaundice, followed by normalization within 1 to 2 weeks ⁽¹⁵⁾. The very short median latency in our study (7 days) reflects the high number of cases related to MTX. By comparison, the median latency, which has been associated with DILI in the general US pediatric population is 55 days, but with a very wide range (1 day to 7.6 years) ⁽¹⁶⁾. This again is a reflection of the specific drug triggers reported by US DILIN, some which are linked to shorter (e.g. trimethoprim-sulfamethoxazole) or longer (e.g. minocycline) drug latencies ⁽¹⁷⁾.

The type of primary solid tumor strongly influenced the risk of developing DILI. Sarcoma was significantly more common among DILI cases as compared with both the non-DILI liver injury and nonliver injury groups. This is despite the fact that sarcomas represented less than 20% of all solid tumor diagnoses in the PHO database. The higher frequency of DILI among sarcoma patients was driven largely by cases of HD IV MTX in treatment of osteosarcoma, which was the cause of DILI in 22 of the 25 sarcoma DILI cases. We found no significant association between sex, ethnicity, or obesity, and frequency of DILI.

It is worth noting that antimicrobials, such as trimethoprim-sulfamethoxazole or fluconazole, which were commonly used among our patients and which are known potential hepatotoxins, were not identified as triggers of DILI in our study. This could be due in part to the fact that, while antimicrobials are the most commonly reported class of agent linked to DILI events in US general population (children and adults), the estimated overall incidence of idiosyncratic DILI related to these and other agents used in the general population is very low, on the order of 14–19 per 100,000 per year ^(16,18). The fact, however, remains that the incidence of antimicrobial-induced DILI specifically among young individuals treated for cancer is not well described, and it is important for studies of hepatotoxicity in this population to consider all drug exposures in the adjudication of liver injury cases.

It is also notable that the occurrence of DILI in our study was not associated with serious clinical consequences in the majority of our cases. Most cases were mild, and although there was a 14% frequency of chronic DILI, all chronic cases eventually resolved with no other complications. Only 8% of the DILI cases had their cancer treatment modified because of DILI.

These data should not be interpreted as a rationale to continue an agent(s) suspected of inducing DILI irrespective of the clinical circumstances. Each DILI case should be managed with recognition of what is known about the typical phenotype of DILI related to the specific implicated agent(s), and with consideration of the presence or absence of risk factors for severe outcomes (eg, death or need for liver transplant), such as DILI-induced hepatocellular injury with jaundice (ie, Hy's Law), DILI-induced hepatic synthetic dysfunction (liver failure), or DILI manifesting as a hypersensitivity syndrome ⁽¹⁹⁾. The importance of the treatment regimen, and whether there are any viable alternatives, should also be carefully considered in decisions regarding whether to modify or stop a treatment regimen.

In addition, there were other important causes of liver injury noted in the cohort, including infection. These etiologies must, therefore, be considered on the differential of liver injury in young individuals treated for solid tumors particularly in view of the significantly higher frequency of jaundice and more severe liver injury among the patients with alternative causes of liver dysfunction.

Death was more common among patients with history of any form of liver injury versus those without (Table 1). The majority of the deaths were, however, related to progression of cancer irrespective of the cause of liver injury. The highest frequency of death was observed among patients with non-DILI liver injury (though not statistically significantly different as compared with DILI patients), and a significant percentage of these deaths were related to infection. Therefore, while DILI is not a major cause of death, the occurrence of liver injury in general signals a higher risk of death and the possibility of infection should be thoroughly considered.

There are several important limitations of our study beyond its retrospective nature. The first is that DILI is clinical diagnosis of exclusion, and patients being treated for cancer often have complicated medical histories. Again, we believe that the impact of these limitations was lessened by the rigor of the expert adjudication process.

The second limitation is that the study design resulted in a likely underestimation of the overall frequency of DILI among young individuals who experienced liver injury, as we analyzed only the first instance of meeting lab criteria. It is expected that there were patients with liver injury who experienced multiple subsequent episodes of liver injury that we did not characterize. We accepted this limitation as it would have been an overwhelming task to analyze all episodes of liver injury. We can definitively report that 86% of young individuals in the PHO registry never experienced clinically significant liver injury during their course of treatment.

A final limitation is that we were not able to assess important aspects of the DILI cases attributed to combinations of agents (cases in which a single agent could not be implicated), such as the exact offending agent(s) and latency. In these cases, we decided to label all of the agents as possible triggers of DILI, as all of the cases involved complex, asynchronous exposure to multiple potentially hepatotoxic agents over time, making confident implication of one of the agents impossible. This problem is very commonly encountered when considering the possibility of DILI in a child with cancer, and further highlights the importance of reporting of all DILI cases to study registries, such as the US DILIN, particularly cases where a single agent can be confidently implicated. These single agent profiles provide important data regarding the liver injury phenotype secondary to a particular agent, that can then be applied to adjudication of future cases involving multiple suspect agents.

CONCLUSIONS

Fourteen percentage of solid tumor patients experience liver injury while receiving antineoplastic agents, and DILI is the most common cause. DILI occurs most commonly among osteosarcoma patients being treated with HD IV MTX. DILI is almost always mild acute hepatocellular injury, jaundice is rare and other serious outcomes, such as liver failure or death related to DILI are not observed. The occurrence of DILI does not impact treatment strategy for most of these patients. In contrast, liver injury attributed to other etiologies are more likely to require hospitalization and die because of tumor progression. Therefore, rapidly identifying the etiology of liver injury during cancer treatment is important so that the appropriate interventions can be undertaken. DILI should be considered on the differential of liver injury in a child receiving treatment for solid tumor malignancy. However, the occurrence of mild DILI may not warrant treatment modification.