Exposure to Agent Orange and Hepatocellular Carcinoma Among US Military Personnel

Key Points Question Is Agent Orange (AO) exposure during the Vietnam War (1966 to 1975) associated with incident hepatocellular carcinoma (HCC)? Findings This cohort study of 296 505 eligible veterans, no association between AO and incident HCC was observed. However, among veterans, viral hepatitis, nonalcoholic fatty liver disease, and alcohol and tobacco use were the most important clinical risk factors for HCC, which were further modified by cirrhosis status. Meaning These findings suggest that AO exposure is not associated with incident HCC.


Introduction
Hepatocellular carcinoma (HCC) incidence is increasing globally with a relative 5-year survival of only 18%, with cirrhosis being the leading risk factor. 1 Identifying key clinical drivers for HCC can guide screening practices of at-risk populations, which are associated with improved survivals 2 ; can guide implementation of early interventions to modify risk factors; and has the potential to affect health care policy and delivery.For instance, as data emerged demonstrating the efficacy of mitigating the risk of liver complications with direct acting antiviral therapies for hepatitis C virus (HCV), 3 stakeholders including the US Centers for Disease Control and Prevention and the US Preventative Services Task Force recommended universal HCV screening for adults. 4Similarly, as the metabolic syndrome and its hepatic manifestation of nonalcoholic fatty liver disease (NAFLD) were recognized an independent risk factors for HCC, society guidelines including recommended screening for NAFLD in patients with diabetes. 5e Veterans Healthcare Administration (VHA) is the largest integrated health care system in the United States serving 9 million veterans yearly.The VHA has placed a large emphasis on population health and implementing quality improvement measures to expedite the delivery of care to veterans.For instance, HCV screening and treatment were implemented in the VHA in a more comprehensive manner than any other health care system in the US. 6 More recently, the national Hepatic Innovation Team Collaborative shifted their focus on improving the care of veterans living with cirrhosis by using automated and integrated clinical systems. 7,8Similarly, demonstrating how environmental exposures during combat affect health has also been emphasized within the VHA and is linked to a veterans' medical service connection and ability to access health care. 9This is reflected in current government policies, including the 2022 Sergeant First Class Health Robinson Honoring Our Promise to Address Comprehensive Toxics (PACT) Act, 10 signed into law by the Biden administration.The PACT Act is the largest health care benefit expansion for veteran exposed to toxins.This includes recognizing Agent Orange (AO)'s association with many malignant neoplasms. 11,12ring the Vietnam war, the US military sprayed herbicides to defoliate forests to clear land of military occupation and control enemy food supplies. 13Most herbicides contained 2,4,5trichlorophenoxyacetic acid (2,4,5-T), which was contaminated with 2,3,7,8-tetrachlorodibenzo-pdioxin (TCDD), including in AO. 13 It was recognized as early as 1970 that 2,4,5-T was teratogenic and was banned from US domestic use. 14An estimated 366 kg of TCDD was sprayed in South Vietnam, 15 with large and persistent clinical ramifications to this day.Over the years, AO has been studied extensively given its association with several cancers, including bladder, Hodgkin and non-Hodgkin lymphoma, prostate, and lung and soft tissue sarcomas, leading to its classification in 1997 as a carcinogen by the International Agency for Research on Cancer. 16Its effects on HCC remain controversial.While animal models link TCDD to HCC, 17 epidemiological studies in veterans have not been conclusive.9][20][21][22] Type 2 diabetes has also been linked to AO exposure and is recognized by the VHA as an AO-associated condition, which is also an independent risk factor for HCC. 23Given the new government incentives to expand eligibility of veterans and the lack of strong data linking AO to HCC, large, statistically rigorous studies addressing the limitations of previous work are needed to better evaluate if AO as a clinically relevant risk for HCC.In the current study, we investigate the association between AO and incident HCC using the largest cohort to-date of Vietnam veterans with a long-follow up in the VHA.

Methods
This cohort study was conducted according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 24This study was approved by the institutional review board of the Veterans Affairs Greater Los Angeles Healthcare System (GLAHCS) with a waiver of informed consent because of the retrospective design.

Data Source
The US VHA is comprised of 168 Veteran Affairs (VA) Medical Centers across the US and its territories and is the largest integrated, nationwide health care system in the US.The VHA contains an integrated network of electronic health data that can be accessed by approved researchers.For this study, we used the VA Informatics and Computer Infrastructure (VINCI) platform to extract demographics, clinical laboratory tests, medical comorbidities, alcohol use questionnaires and clinical outcomes data from the Corporate Data Warehouse (CDW), which is the national data repository for VA electronic health records.

Study Cohort
To capture veterans who primarily obtained their care within the VHA, we selected veterans who had at least 2 outpatient visits that were separated by at least 6 months within a window of 3 years.From this group, we identified Vietnam veterans using their combat service location with the combat start date between the years January 1, 1966, and December 31, 1975.We excluded women (more than 99% of veterans deployed to Vietnam were males) and those who were younger than age 18 years at the time of deployment.Veterans were then followed up with from 2000 to 2019.The year 2000 was selected as the starting window because October 1999 was the earliest time when CDW data became available (eFigure in Supplement 1).

Variable Definitions AO Exposure
Veterans who were flagged as having AO exposure during the Vietnam War were categorized as AO exposed, while veterans deployed to Vietnam without exposure to AO were categorized as AO nonexposed.AO exposure was defined as any veteran who was flagged in the disability data under AgentOrangeExposure.These data are validated in the disability database to determine a veteran's claim, which are defined clinically using the PACT Act. 10,25AO exposure is based on (1) having a health condition associated with AO per VHA (updated over time) and (2) having served in a location with exposure to AO.If a veteran believes they have a condition associated with AO but is not currently recognized by the VHA, the veteran must show evidence that the medical condition started during military service or scientific evidence (ie, publications) that show the condition is associated with AO.
Additionally, we confirmed these methods to identify veterans who were AO exposed with VINCI Services.We also conducted a manual record review of a random set of 19 patients categorized as AO exposed who were seen at the local VA and found that 95% were correctly identified to have been exposed to AO during the Vietnam War.

Baseline Patient Characteristics
Relevant clinical covariates and baseline characteristics were selected a priori based on their association with HCC.Age at deployment, sex, race and ethnicity (self-reported by patients in electronic health records of the VHA) and medical comorbidities were extracted (eTable 1 in Supplement 1). 26,27All medical comorbidities were identified the first time they occurred in the veteran's electronic health record.Diagnoses were identified using previously validated International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes. 26HIV and hepatitis B virus (HBV) 28 or HCV 29 were identified using laboratory data (HIV antibodies, HBV surface antigen, HCV RNA, respectively) (eTable 2 in Supplement 1).Alcohol-associated liver disease was defined as having an Alcohol Use Disorders Identification Test score of 4 or more for men 30 or previously validated ICD-9 and ICD-10 codes 27 (eTable 1 in Supplement 1).Obesity was defined by a body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) of more than 30, which was calculated from height (limited height by 48 to 84 inches) and weight (limited weight by 75 to 500 lbs) with the peak BMI recorded for each veteran. 31Smoking status was defined as nonsmoker (never or current nonsmoker in Health Factor table) or smoker (current smoker, former smoker, history of tobacco use).

Cirrhosis Ascertainment
A diagnosis of cirrhosis was defined using: (1) validated ICD-9 and ICD-10 codes for cirrhosis 26 ; (2) validated ICD-9 and ICD-10 for decompensated cirrhosis (as defined by ascites, hepatic encephalopathy, hepatopulmonary syndrome, hepatorenal syndrome, spontaneous bacterial peritonitis, and variceal hemorrhage) 32 ; or (3) having a Fibrosis-4 (FIB-4) score of more than 3.25 for at least 2 consecutive years (eTable 2 in Supplement 1). 33FIB-4 is calculated as follows using the aspartate aminotransferase (AST) level and alanine aminotransferase (ALT) value: All laboratory data in the FIB-4 calculations were within the last 3 months for each year.For each diagnosis, the earliest date was used to define the occurrence.While FIB-4 is a commonly used noninvasive tests clinically to assess the risk of cirrhosis, it can overestimate it given the large component of age, which can be a confounder given the aging population of Vietnam veterans. 34We next evaluated the number of patients with the validated ICD-9 and ICD-10 for cirrhosis and decompensated cirrhosis and an elevated AST to Platelet Ratio Index (APRI) score 35 (excludes age): (AST level/AST upper limit of normal) / (Platelet count × 100)

Outcomes Primary Outcomes
HCC cases were identified using 2 methods.First, we used the VA Central Cancer Registry (VACCR), which constitutes a national repository of all cancer cases within the VHA since 1995.In the VACCR, each case is validated by a manual record abstraction using the North American Association for Central Cancer Registries standards with more than 90% accuracy. 36Second, we used ICD-9 and ICD-10 32 codes if they occurred twice in the patient's outpatient medical records, which has been well validated with positive estimated values of 84% to 94%. 33To assess the accuracy of the VACCR, we identified 40 cases and confirmed the diagnosis by conducting a manual record abstraction at GLAHCS.Of the 40 patients selected in the VACCR at GLAHCS, 90% were correctly identified, as confirmed by manual record abstraction of hepatology, gastroenterology, or oncology clinical notes.
Using the ICD-9 and ICD-10 codes, we identified 3469 veterans with HCC.Of those, we explored 31 cases, and HCC was correctly identified in 74% of the cases.We found that most incorrectly identified cases were secondary malignant neoplasms to the liver from other primary cancers, mostly of gastrointestinal origin (67% of the cases).To enrich cases with HCC using the ICD-9 and ICD-10 algorithm, we identified all gastrointestinal cancer malignant neoplasms using ICD-9 and ICD-10 (eTable 3 in Supplement 1) and removed them from the HCCs.

Secondary Outcomes
To identify death outcomes, we linked the VA CDW cohort to the VA-Department of Defense Mortality Data Repository (MDR), 37 which provides reliable and quality-controlled data for death in veterans and a sensitivity of more than 90%. 38Next, we cross-referenced the death date from the VACCR, CDW, and MDR.We defined death outcomes as in MDR death file, in CDW death (if the date was between January 1, 2000, and December 31, 2019), or in VACCR death data.All death dates were between January 1, 2000, and December 31, 2019, with the latter date used as a censor date.

Statistical Analysis
Patient baseline characteristics are presented as the number and percentage of patients or mean (SD).Covariates were selected based on known clinical associations with HCC.HBV and HCV were combined into 1 covariate (viral hepatitis).NASH and NAFLD diagnoses were also combined into 1 covariate.Diabetes, hypertension, and HBV and HCV were used as time-dependent covariates.Given the link between diabetes and obesity, we assessed the interaction between the 2 diagnoses.
Interactions were also assessed between AO and viral hepatitis, age, alcohol, NAFLD or NASH, smoking, and diabetes.
We conducted a competing risk analysis to assess the risk of incident HCC as a primary outcome with death and LT as competing events.Type-specific adjusted hazard ratios (aHRs) are reported for both univariate and multivariate models.Covariate selection was determined a priori and included age (at deployment); race and ethnicity; smoking; portal hypertension; and etiologies of cirrhosis, including NAFLD or NASH, viral hepatitis, alcohol associated liver disease, hemochromatosis, alpha-1antitrypsin, secondary biliary or unspecified biliary cirrhosis, and autoimmune hepatitis.Given that cirrhosis is the most important risk factor for HCC and that cirrhosis could not be ascertained until after CDW were available, which was approximately in the year 2000, the results were stratified into cirrhosis and noncirrhosis status for each veteran prior to any statistical analyses.Furthermore, to adjust for the impact of cirrhosis on HCC, we conducted inverse probability weighting (IPW) model.
All statistical analyses were conducted using SAS version 9.4 (SAS Institute).Statistical significance was set at P <.05, and tests were 2-sided.Data were collected from January 1, 2019, to December 31, 2020, and analyzed from December 2020 to October 2023.

AO Exposure and HCC Risk
Incident HCC At a mean (SD) of follow-up time is 51.1 (1.9) years (51.2 [1.8] years and 51.0 [2.1] years in AO exposed and AO nonexposed group, respectively), we found 2654 veterans who developed HCC in our final cohort (1463 [0.9%] in the AO exposed group and 1191 [0.9%] in the AO nonexposed group).HCC incidence rate was 0.00018 per year (0.00018 per year and 0.00017 per year).In our multivariable model adjusting for other confounders of HCC, we found that AO was not associated with HCC both in veterans with cirrhosis (aHR, 1.00; 95% CI, 0.91-1.10;P = .98)and without cirrhosis (aHR, 1.05; 95% CI, 0.89-1.23;P = .57).The lack of association between AO and incident HCC based on cirrhosis status was also substantiated using the APRI score, which is another surrogate marker for advanced fibrosis and cirrhosis (eTable 4 in Supplement 1).Next, we conducted IPW to adjust for cirrhosis and did not find a significant association between AO exposure and incident HCC (eTable 5 in Supplement 1).
To further confirm our findings and explore the association of death data accuracy, we conducted a sensitivity analysis using (1) HCC as an event with death as the only competing risk, and (2) HCC as an event and death was the censored event if occurred before HCC (instead of censored at last follow up in 2019) and transplantation as the competing risk.After adjusting for all the HCC risk factors, neither analysis found an association between AO and HCC risk.Given the inverse association between age and incident HCC, we explored the association between AO and age further.There were no significant interactions between age and AO exposure which was further confirmed by assessing their association after stratification of age at deployment (eMethods in Supplement 1).

Other Risk Factors Associated With HCC
We found that self-reported race and ethnicity were associated with incident HCC (Table 2; eTable 4 in Supplement 1).Veterans with cirrhosis who self-identified as Hispanic or Latino individuals (aHR, 1.51; 95% CI, 1.30-1.75;P <.001) or Black individuals (aHR, 1.18; 95% CI, 1.05-1.32;P = .004)had a higher risk for HCC compared with non-Hispanic White individuals.
In For both alcohol and smoking, we found that early exposure during ages 20 to 30 years were significantly associated with incident HCC, independent of cirrhosis status (eTables 6 to 8 in

Discussion
In this large nationwide analysis of approximately 300 000 Vietnam veterans, we report on key, modifiable clinical risk factors associated with incident HCC.After adjusting for relevant clinical risk factors in a competing risk model, we found that AO exposure was not associated with incident HCC.
HCC is a complex disease that results from many clinical and environmental risk factors that can change over time. 40AO has been linked diabetes, 11 which is independently associated with incident HCC. 23,32We explored these associations in our analyses but did not find that AO exposure modified the HCC risk.While our study aimed to investigate the association of AO with HCC, we found other important associations.Dyslipidemia was inversely associated with the risk of HCC in veterans irrespective of cirrhosis and after adjusting for confounders.2][43]  To appropriately assess the risk of AO exposure and HCC development, we defined our control group as veterans who were deployed to Vietnam but were not flagged in the VHA as having been exposed to AO.This study design allows to control for other nonmeasurable factors, such as other potential geographical, diet, and environmental-related exposures that are not captured by the electronic health records.Using a non-Vietnam War group as a control may not have appropriately isolated these differences and could have potentially introduced bias.This is especially relevant given the consideration of the healthy soldier effect (HSE) that has previously been described, where veterans who are deployed are less likely to develop chronic diseases, including during the Vietnam War. 44The HSE has been attributed to early physical training and health care access and has been shown to occur as far out as 40 years from war. 45 found that self-reported race and ethnicity were significantly associated with incident HCC.
Specifically, patients who self-identified as Hispanic or non-Hispanic Black were more likely to develop HCC, compared with White individuals, even after adjusting for key clinical confounders.
These findings suggest that other nonmeasured variables could contribute to these disparities.
7][48][49][50][51] However, few studies have addressed this in the VHA.Our findings support that even within the VHA, where patients have access to health care, racial and ethnic disparities persist.Small studies have proposed that mental health and substance use may drive these disparities; however, larger studies are needed to understand these observations.
We found that a minority of veterans underwent LT.While LT is a life-saving measure for patients with advanced liver disease, 52 it remains limited resource due to the scarcity of organs and a complex evaluation process that affects candidates every level of the evaluation, including referral, eligibility and listing, especially in veterans who have been shown to disproportionally receive fewer LTs. 53Therefore, HCC risk modification of clinical factors is even more imperative in veterans to prevent HCC.

Table 2 .
Univariable and Multivariable Models for HCC Development With Liver Transplantation and Death as Competing Events 1).While we found an association between incident HCC and age at deployment in veterans without cirrhosis (aHR, 1.02; 95% CI, 1.00-1.04;P = .047),this was not reproduced using the age-independent APRI score (eTable 4 in Supplement 1).
a Cirrhosis defined using validated International Classification of Diseases, Ninth Revision and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision and consecutive Fibrosis-4 scores.bAdjusting for age at deployment and race and ethnicity.cCombined NAFLD and NASH in multivariable model.dCombined HCV and HBV. e No cases in the noncirrhosis group.fObesity and diabetes interactions assessed which was not significant in multivariable model.Supplement Early exposure to alcohol and tobacco use were also important risks for HCC development, especially in veterans with cirrhosis.Taken together, our work supports that HCC risk factors are dynamic and can change during the course of a veteran's life depending on the presence of cirrhosis and other contributing factors, such as alcohol and tobacco use.These findings, in conjunction with our novel findings that AO is not associated with HCC, help inform which risk factors (ie, viral hepatitis, NAFLD or NASH, and ALD) should be prioritized and modified to mitigate the rise in HCC among veterans.
Flow Diagram of Vietnam Veteran Cohort eTable 1. List of ICD-9 and ICD-10 Codes Used for Etiologies of HCC and Medical Comorbidities eTable 2. List of Laboratory LOINCs Including HIV, HCV, HBV, ALT, AST and Platelet Counts eTable 3. ICD-9 Codes of Gastrointestinal Malignant Neoplasms Other Than HCC eTable 4. Association Between AO and Other Risk Factors for HCC Using APRI Score to Identify HCC Cases eMethods.Additional Methods With eTables 5-8 eTable 5. Association Between AO and Other Risk Factors for HCC Using Inverse Probability Weighing eTable 6. Age (In Years) Stratification for AO Exposure eTable 7. Age (In Years) Stratification for Tobacco Use eTable 8. Age (In Years) Stratification for Alcohol Use