Healthcare costs related to adverse events in hepatocellular carcinoma treatment: A retrospective observational claims study

Abstract Background Hepatocellular carcinoma (HCC) is an aggressive form of liver cancer with increasing incidence and mortality worldwide. For metastatic disease, systemic treatment is recommended. In addition to tumor characteristics, adverse events (AEs) may influence regimen choice. Aim To analyze healthcare burden among patients with advanced HCC, by treatment type and AEs observed. Methods Included were adult commercial and Medicare Advantage enrollees with ≥2 non‐diagnostic claims coded for HCC (the first setting the index date); ≥1 claim for systemic treatment of advanced/metastatic HCC; and continuous enrollment for a 6‐month pre‐index baseline period to ≥1 month post‐index (follow‐up). Patients were excluded by lack of systemic treatment; incomplete demographic information; pregnancy, liver transplant, other cancers during baseline or clinical trial participation. We describe patient characteristics, common AEs, overall survival, and healthcare burden in 2017 USD up to 12 months after initiation of tyrosine kinase inhibitor (TKI) monotherapy; immune checkpoint inhibitor (ICI) monotherapy; or FOLFOX combination therapy. Results The analytic sample consisted of 322 patients (median age 65.8 years, 76% male) who had 12 months' (unless death occurred prior) available follow‐up, with median follow‐up of 9 months. Among these, 241 (75%) had TKI monotherapy, 23 (7%) had ICI monotherapy, and 58 had FOLFOX (18%) first‐line treatment. Overall, patients had a high burden of AEs (mean 3.2), with the most prevalent being pain (75%), infection (39%), ascites (34%), and bleeding (29%). After adjusting for covariates, infection ($50 374), fever ($47 443), and diarrhea ($29 912) imposed the highest incremental annual costs versus patients without the AE. Up to 90% of costs were attributable to inpatient admissions, with 56% to 60% involving intensive care. Median 1‐year survival was 32%. Conclusions This real‐world study demonstrated AE burden in alignment with previous clinical studies. Regardless of regimen used, AEs are associated with substantial healthcare costs due to inpatient care.


| INTRODUCTION
Hepatocellular carcinoma (HCC) is an aggressive tumor that is typically diagnosed at a late stage and most often among patients with chronic liver disease. The presentation is heterogeneous in that patients in the United States (US) are mostly male and have cirrhosis due to alcohol abuse or hepatitis C virus. In contrast, among patients in the Asian-Pacific regions the underlying disease is more commonly hepatitis B virus infection. 1,2 HCC accounts for 75-85% of liver cancers and represents a significant cause of cancer-related mortality. 3 Despite the association of HCC with modifiable risk factors, incidence and mortality due to HCC continue to rise worldwide. 3,4 The American Cancer Society estimates that 42 810 new cases of liver cancer (including intrahepatic bile duct cancers) will be diagnosed in the US in 2020, and approximately three-quarters of those will be HCC. 5 In instances of early-stage disease, treatment may include surgical and non-surgical therapies. Systemic therapies are most often used in advanced-stage disease. [6][7][8][9] For metastatic disease, systemic treatments are recommended by the most recent National Comprehensive Cancer Network (NCCN) guidelines. 10 Unfortunately, underutilization of treatment is common. 2,11 Indeed, approximately 50% of patients with HCC of any stage do not receive treatment. In addition to performance status and comorbid conditions, demographic factors associated with nontreatment include older age, non-Caucasian race, and lack of insurance or low socioeconomic status. 11 Even with treatment for advanced stage disease, US 5-year survival is 12% or less. 12 Although tumor characteristics influence the treatment selected, in addition to survival, other important outcomes affecting the choice of regimen include the occurrence of adverse events (AEs) during treatment. Among clinical trials of systemic regimens, common adverse events include hypertension, diarrhea, decreased appetite, nausea/vomiting, myelosuppression, infections, rash, pruritus, and fatigue. [13][14][15][16][17] Such AEs may lead to reduced survival and quality of life, discontinuation of HCC treatment, and increased costs for healthcare related to treatment of these events. 2 Randomized clinical trial literature on efficacy and safety exist for systemic treatments of advanced and unresectable HCC. However, data from US realworld studies have been limited primarily to the study of sorafenib, until recent Food and Drug Administration (FDA) approvals of additional tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs) added newer options. In this study, NCCN-recommended treatment regimens for advanced/ metastatic and unresectable HCC were studied to describe patient characteristics, the most common AEs during treatment, overall survival, and healthcare costs and utilization for up to 12 months after treatment initiation.

| Study design
This was a retrospective database study of costs related to AEs, as observed in administrative healthcare claims, following initiation of treatment of advanced/metastatic HCC. Claims and enrollment information, and linked mortality data, were accessed for the period July 1, 2009 to August 31, 2019 ( Figure 1). Baseline patient characteristics (clinical and demographic) were observed during the 6-month period prior to the index date, which was set as the date of the first claim with an HCC diagnosis during the observation period. The primary outcomes included AEs and healthcare costs and utilization by study cohort assigned by treatment regimen, for the post-index period of at least 1 month. The sample of patients described herein had at least 12 months of follow-up available, except in the case of death as end to their study period.

| Patient selection/eligibility criteria
The study patients were commercial and Medicare Advantage enrollees with evidence of HCC (at least two non-diagnostic claims for HCC in any claim position on two separate days during the identification period). Additional inclusion criteria were age ≥ 18 years as of the index date; ≥1 claim for treatment of HCC on or after the index date with

| Baseline characteristics
Demographic variables included age as of the index date, gender, insurance type (commercial or Medicare Advantage), and US Censusdesignated geographic region. 18 Clinical comorbid status was indicated by the Charslon 19,20 Comorbidity Index (CCI) score. In addition, mean and standard deviation (SD) and median time from metastatic diagnosis to start of first line of treatment and follow-up time were measured in days.

| Treatment patterns
The NCCN-recommended agents for systemic treatment of HCC were identified per the Healthcare Common Procedure Coding System (HCPCS) and an algorithm classified the start of therapy as the date of first claim for a fill or infusion within the first 30 days after the index date. The end of a line of therapy (LOT) was identified by start of a new agent, death, or a 60-day period of no fills added to a 30-day run -out period, used to identify discontinuation. In addition, a censored (incomplete) LOT was indicated by end of the study period or disenrollment. The proportion of patients who had a second LOT, and the time to the start of LOT2 were recorded.

| Outcomes
The AEs of interest were selected through literature and clinical review and package insert reviews for TKIs, ICIs, and FOLFOX, the agents included in NCCN guidelines as of October 2019. [21][22][23][24][25] Twenty-five AEs were identified by ICD-9/10 codes (Appendix 2) appearing as of the first date in which a diagnosis code was observed in the first claim position during the available follow-up period. All immune-mediated AEs were reported, as well all other AEs with a prevalence of ≥5% in the in the total sample of patients. discharge and disenrollment reason indicating death; or death dates from NDI file. Death data were used to calculate overall survival for all patients who met the inclusion and exclusion criteria. The period in days between the index date and the end of study was determined for each patient. This allowed determination as to whether the patient died by the end date of the study or was censored at the end of the 12-month follow-up period or at disenrollment.

| Statistical analyses
All study variables were analyzed descriptively as numbers and proportions (%) for dichotomous and polychotomous variables and means, medians, and SD provided for continuous variables. Descriptive techniques that account for variable length of observation time (e.g., PPPM) were used where appropriate.
Results were stratified by treatment cohort. Bivariate comparisons of pre-index characteristics and outcome measures were F I G U R E 1 Study design describing a hypothetic index date and pre-and post-index data obtained. AE = adverse event provided, and appropriate tests (e.g., t-test, Wilcoxon rank-sum, chi-square test) were used based on the distribution of the measure.
Statistical significance was set at p < 0.05. In addition, multivariable analysis of total healthcare costs was conducted using a generalized linear model with log link and gamma distribution, adjusting for the presence of covariates, including the prevalent AEs of at least 5% in the study population. Predicted incremental healthcare costs for patients with versus without select AEs were presented as recycled predictions. Kaplan-Meier (KM) analysis was used to estimate patient overall survival and the fraction of patients surviving for a set period of time after treatment. All analyses were conducted using SAS software version 9.0.

| RESULTS
After meeting inclusion criteria, 1246 eligible patients were identified

| Sample characteristics
The total analytic sample consisted of 322 patients (median age 65.8 years, 76% male) with 12 months of follow-up available, except in the event of death before 12 months. The median follow-up period was 9 months (Table 1). By treatment, 241 (75%) were treated with TKI monotherapy, 23 (7%) with ICI monotherapy, and 58 with FOLFOX (18%), as LOT1. Among all 322 patients, 79 (25%) had ≥2 LOTs. At diagnosis, the mean CCI was 5.1. Geographic distribution aligned with the database overall, with the highest proportion of patients in the South region.

| Outcomes
All-cause utilization by regimen during follow-up among the analytic sample (n = 322) was high for ED and inpatient hospitalizations. In total, 84% of patients had ≥1 ED visit, 81% had ≥1 inpatient hospitalization, and among inpatients, 46% had ≥1 ICU stay. The values for ED, inpatient, and ICU all-cause utilization for those receiving TKI monotherapy (n = 241) were 82%, 81%, and 45%, respectively.
Payer-paid all-cause PPPM costs by category and first-line treatment are shown in Figure 3

| DISCUSSION
This study is one of few studies based on real-world samples; most studies present highly selected RCT populations, which have homogeneous characteristics by design. Recent systematic reviews, 2,26 including studies not necessarily designed to examine costs associated with adverse events, have supported the conclusion that HCC imposes a substantial burden upon patients, caregivers, and the healthcare system.
The current analytic sample included 322 patients with 12 months of follow-up available, except for those whose study period ended earlier due to death, but not due to discontinuation, end of enrollment, or end of the study. The majority of patients (75%) received TKI, with 7% The most prevalent AEs during follow-up were pain (75%), infection (39%), ascites (34%), bleeding (29%), and anemia (18%

| Limitations
Caution is required when interpreting results of comparative observational studies, given the lack of randomization and subsequent biases (e.g., channeling) introduced in an observational design. The use of claims for outcomes studies is limited in that such data are created for the purpose of billing, rather than research, and can only reflect costs paid for care in covered venues/providers and are subject to coding errors. Additionally, claims cannot provide all clinical data, such as grade and severity, which would be useful for comparing AEs in the context of this study relative to those described in randomized controlled trials. As such, it must also be acknowledged that while some AEs are referred to as "immune-mediated," there is no way to absolutely confirm this for many of them, as for example in hypothyroidism. In addition, some of the AEs-in particular, pain, ascites, and asthenia/fatigue-could have been due to the HCC itself and possibly exacerbated or not strictly associated with the treatment, but there is no way to determine this. The use of claims also limits the interpreta-

ETHICAL STATEMENT
Institutional Review Board review and approval was obtained. A HIPAA waiver was received specific to the use of identifiable health information.

DATA AVAILABILITY STATEMENT
Third party use of the proprietary data obtained from the Optum Research Database requires strict data security and privacy protocols and a restrictive license agreement. Thus, data used to generate the results presented cannot be disclosed publicly.