Incidence of non-Hodgkin's lymphoma among individuals with chronic hepatitis B virus infection
Hepatology July 2007
Marianne Ulcickas Yood 1 2 3 *, Charles P. Quesenberry Jr. 4, Dianlin Guo 5, Cary Caldwell 6, Karen Wells 1, Jun Shan 4, Lynn Sanders 5, Mary Lou Skovron 5, Uchenna Iloeje 5, M. Michele Manos 4
1Josephine Ford Cancer Center, Henry Ford Health System, Detroit, MI
2Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT
3EpiSource, Hamden, CT
4Division of Research, Kaiser Permanente Medical Care Program of Northern California, Oakland, CA
5Global Epidemiology and Outcomes Research, Bristol Myers Squibb Company, Wallingford CT/Plainsboro, NJ
6Department of Gastroenterology, Yale New Haven Hospital, New Haven, CT
Bristol-Myers Squibb Company
Although hepatitis C virus (HCV) infection has been shown to be associated with development of non-Hodgkin's lymphoma (NHL), few studies have investigated the association between chronic HBV infection and NHL. The purpose of this study was to compare the incidence of NHL between patients with and without chronic hepatitis B virus (HBV) infection. Using automated laboratory result and clinical data from two United States health systems, we identified individuals with chronic HBV infection from January 1, 1995 through December 31, 2001. Using each health system's population-based tumor registry, we identified all cases of NHL diagnosed through December 31, 2002. We excluded any individual with a history of NHL or human immunodeficiency virus (HIV). We fit Cox proportional hazards models to calculate hazard ratios comparing the incidence of NHL between chronic HBV-infected patients (N = 3,888) and patients without HBV (N = 205,203) drawn from the source populations. We identified 8 NHL cases in the chronic HBV infection cohort and 111 cases in the comparison cohort. Patients with chronic HBV infection were 2.8 times more likely to develop NHL than matched comparison patients (adjusted hazard ratio = 2.80, 95% confidence interval = 1.16-6.75), after controlling for age, race, sex, income, Charlson comorbidity index, study site, and HCV infection. Conclusion: chronic HBV-infected patients were nearly 3 times more likely to develop NHL than comparison patients.
The year 2005 saw 56,390 projected new cases of non-Hodgkin's lymphoma (NHL) and 19,200 deaths due to NHL in the United States. The overall incidence rate of NHL is increasing worldwide, especially in developed countries,[2-4] whereas the rate in the United States has been relatively stable during 1995 to 2003. NHL disproportionately affects males and is more common in white non-Hispanics than in other racial or ethnic groups. For the period of 1995 to 2002, the median age of NHL diagnosis in the United States was 67 years. Immune suppression and pesticide/insecticide exposure are thought to be related to NHL development, and there is strong evidence that infections with human T-cell leukemia virus, Epstein-Barr virus, helicobacter pylori, and hepatitis C virus (HCV) are involved in the etiology of NHL.
Infection with hepatitis B virus (HBV) is a major global public health problem with 50 million people infected per year and more than 350 million people chronically infected worldwide. In the United States, chronic HBV infection is more common in Asians than in other racial-ethnic groups. Among the common sequelae of chronic HBV infection are cirrhosis of the liver and hepatocellular carcinoma (HCC). Given the relationship between chronic HCV infection and NHL,[12-16] an association between HBV infection and NHL is quite plausible. HCV infection is thought to promote the development of NHL by inducing chronic lymphoproliferation through persistent antigenic stimulation. Despite the plausible relationship between chronic HBV infection and NHL, few studies have investigated this potential link.[18-20] The few studies, conducted in Eastern Europe and Asia, in which chronic HBV infection is highly prevalent, suggest that HBV infection is 2 to 5 times more prevalent among NHL cases compared with hospital controls without NHL or general population controls.
We investigated the potential role of chronic hepatitis B infection in the development of NHL. To estimate the effect of chronic HBV infection on the incidence of NHL, we conducted a cohort study at two United States health care delivery systems.
We found that patients with chronic HBV infection were almost 3 times more likely to develop NHL compared with comparison patients, after adjusting for demographics, comorbidity, and HCV co-infection. Although the literature on the association between HBV and NHL is limited, our effect estimate is consistent with previous reports.[18-20] For example, in a case-control study conducted in Korea, Kim et al. observed an odds ratio of 2 for HBV chronic infection when NHL cases were compared with non-hematological cancer controls.
One limitation of the few previous comparative studies is that, in contrast to our study, they collected HBsAg data on or after NHL diagnosis, or limited the definition of an HBV case to HBsAg positivity before NHL treatment initiation. Of note are literature reports of reactivation of HBV among NHL patients treated with chemotherapy.[22-24] Therefore, previously reported associations between HBV infection and NHL may reflect an effect of treatment on the presence of HBV. However, our study design minimizes the possibility of reverse causation. We first excluded patients with a history of NHL from both HBV and non-HBV cohorts, and then followed them for the development of NHL.
Our calculated NHL crude incidence rate of 14 cases per 100,000 person-years (95% CI 12-17) for the uninfected comparison group is fairly similar to the United States (SEER) age-adjusted rates of 19.1 per 100,000 people per year for 2000-2003. Because we excluded individuals with HIV, however, our incidence rate may not be directly comparable with the rate reported by SEER. Consistent with other studies, our results indicate that individuals with HCV infection were more than 2 times more likely to develop NHL, although small numbers resulted in imprecise estimates. We observed the predicted increasing risk of NHL with increasing age, the increased risk among males, and lowest risk among Asians relative to the other race/ethnic groups. Sex and race differences in risk were not statistically significant, nor did we observe risk differences between whites, blacks, and Hispanics. This is likely due to sample size limitations, and more importantly, to the age distribution of our cohort. At younger ages, racial and ethnic differences in NHL are not large.
Our results should be interpreted in light of their limitations. Although we used an automated data algorithm designed to strictly assess HBV chronicity, we recognize that our case assessment provides neither perfect sensitivity nor perfect specificity. Although we attempted to limit the study cohort to patients with chronic HBV infection by requiring other test results and clinical parameters in addition to a positive HBsAg result, some misclassification of HBV cases may still exist. For example, a liver test elevation (ALT or AST) at least 6 months after the index positive HBsAg test may have been due to reasons unrelated to actual chronic HBV infection. (Note that fewer than 8% of cases qualified as chronic HBV infection based solely on this criterion.) More importantly, our case definition algorithm excluded a large number (7,523) of patients who had only a single positive HBsAg test and no further testing, but who may have been true cases of chronic HBV. However, inclusion of these patients with the definite chronic HBV infection cases in a sensitivity analysis did not alter the results. An additional limitation is that the comparison cohort may well contain chronic HBV cases, since HBV screening is not universal and chronic cases may therefore remain undetected. We were unable to define the date of initial HBV diagnosis for our study population. Therefore, we did not conduct analyses that accounted for the interval between HBV diagnosis and NHL. However, even if available, initial diagnosis date has limited value in defining the true duration of chronic infection. In addition, this study did not evaluate treatment for HBV that might have been secondarily associated with the development of NHL. We also did not examine the effect of viral characteristics such as hepatitis B e antigen status and HBV viral load, nor did we investigate NHL tumor subtypes.
Another limitation of this study is the fact that race and ethnicity data were missing for 7% of chronic HBV cases and almost 23% of comparison patients. Therefore, our ability to accurately adjust for the effect of this important covariate is somewhat limited, particularly for Asian race.
Despite these limitations, this health-plan population-based study supports a relationship between chronic HBV infection and the development of NHL. After controlling for key risk factors, patients with chronic HBV infection were almost 3 times more likely to develop NHL as comparison patients. To our knowledge, this is the first cohort study of the association between chronic HBV infection and NHL incidence in a United States population. If confirmed through further studies, consideration of this risk may be warranted in chronic HBV patient assessment and care.
We identified 3,888 chronic HBV patients (15,481 person-years) and 205,203 comparison patients (787,029 person-years) for analysis. Because they were matched prospectively on age and sex, little difference was seen in the distribution of these variables between the chronic HBV cohort and the comparison cohort (Table 1). The mean age of both cohorts was 40 years, and slightly more than half were male. In contrast, we found large and expected differences in the distribution of racial/ethnic groups. In the chronic HBV cohort, Asian race was much more prevalent than in the comparison cohort (66% vs. 11%), whereas white race was much less prevalent (15% versus 44%). The median follow-up time (3+ years) was similar in both groups. Within the chronic HBV infection cohort, 3.6% had HCV infection; 0.9% of the comparison cohort had HCV infection. Overall, we identified 8 cases of NHL (0.52 cases per 1,000 person-years) in the chronic HBV group (95% CI 0.24-0.96) and 111 cases (0.14 cases per 1,000 person-years) in the comparison group (95% CI 0.12-0.17).
Using Cox proportional hazard models to obtain unadjusted and adjusted hazard ratios for NHL, we found that individuals with chronic HBV infection were more than 2 times more likely to develop NHL than matched comparison patients (crude hazard ratio = 2.30; 95% CI = 1.01-5.24). The Kaplan-Meier plot (Fig. 2) depicts the incidence of NHL over time. After adjusting for age, race, sex, income, comorbidity index, study site, and HCV infection, the rate of NHL was nearly 3 times higher in patients with chronic HBV (adjusted hazard ratio = 2.80, 95% CI = 1.16-6.75) (Table 2). A total of 3 NHL cases occurred in patients with HCV infection (1 NHL case in a patient with chronic HBV and 2 NHL cases in patients without chronic HBV infection). Adjusting for covariates (Table 2), we also found an increased risk of NHL in individuals with HCV (adjusted hazard ratio = 2.36, 95% CI = 0.74-7.52).
We conducted a sensitivity analysis, excluding from the analysis all patients with HCV infection (N = 1,929) and found that although the estimate is less precise due to decreased sample size, the relative effect of chronic HBV infection on NHL incidence remained (adjusted HR 2.43, 95% CI 0.94-6.29). Among the 3,888 patients in the chronic HBV infection cohort, 2,675 (68.8%) received at least one HCV test. We conducted an additional sensitivity analysis by including the 7,523 patients with a single positive HBsAg who represent possible additional chronic HBV cases that were excluded from the primary analysis because they did not meet the additional criteria for chronic HBV infection. We found no difference in results (adjusted HR = 2.79, 95% CI 1.48-5.28).
Materials and Methods
This study was conducted at two United States health care delivery systems: Henry Ford Health System (HFHS) (Detroit, MI) and the Kaiser Permanente Medical Care Program of Northern California (KPMCP) (Oakland, CA). The HFHS health maintenance organization/medical group is a large, integrated health system serving the primary and specialty health care needs of approximately 350,000 residents in southeastern Michigan, including Detroit and its surrounding metropolitan area. HFHS patients who are enrolled in the health maintenance organization, Health Alliance Plan, constitute a stable population, with a mean length of enrollment equal to 9.3 years and annual member retention of approximately 80%. The KPMCP is the largest and oldest integrated health care delivery system in the United States, currently providing health care to approximately 3.2 million members through 3,300 physicians, over 45 outpatient clinics, and 18 hospitals. The 8-year mortality-adjusted retention from 1988 to 1996 of KPMCP members aged 35 to 84 years was 72%.
Both KPMCP and HFHS maintain extensive electronic databases, including all outpatient pharmacy, inpatient and outpatient encounter and claims records (including dates of services and diagnoses), and laboratory results data for all tests performed. Both sites also maintain tumor registries with active, thorough case ascertainment processes.
This study was approved by the institutional review board at each site.
Patients with Chronic HBV Infection (Chronic HBV).
Cohort identification is summarized in Fig. 1. From laboratory result data, we identified all individuals 16 years of age or older who had at least one positive hepatitis B virus surface antigen test (HBsAg+) recorded during January 1, 1995 through December 31, 2001 (N = 11,643). The date of the first HBsAg+ test result was considered the index date (the reference point at which follow-up began and clinical history and age variables calculated). Remaining patients met criteria for the chronic HBV cohort if they met any of the following additional criteria:
1 Repeat HBsAg+ test recorded 6 months or more after the initial HBsAg+ test
2 Positive hepatitis B DNA or hepatitis B e antigen at least 6 months after index HBsAg+ test
3 At least one prescription for interferon or lamivudine (with no evidence of malignancy or multiple sclerosis among interferon-exposed patients)
4 A liver biopsy with evidence of chronic HBV
5 Elevated alanine aminotransferase or aspartate aminotransferase tests performed at least 6 months after the index HBsAg+ test
6 Medical record documentation of chronic HBV found after conducting record review of patients with at least two inpatient ICD-9-coded diagnoses for chronic HBV
From the 4,120 patients meeting the above criteria, we excluded patients with a known history of NHL before the index date (n = 4) (based on tumor registry data, see below) or a diagnosis of human immunodeficiency virus (HIV) before the index date (N = 228) (based on HIV registry and laboratory result data), leaving a final sample of 3,888 patients with chronic HBV infection.
Identification of Comparison Cohort.
For each of the 4,120 individuals meeting the initial chronic HBV cohort criteria, we randomly selected 50 comparison patients (N = 206,000) from the general population of health system members without an HBsAg+ test or an ICD-9-coded encounter indicative of HBV infection. The comparison patients were matched to the chronic HBV patients on health system (KPMCP vs. HFHS), age at index date (±5 years), and sex. In addition, we required that the comparison patient be a health system member or patient within 6 months of the date of the matched patient's index date. Like the chronic HBV cases, the comparison group excluded patients with a history of NHL recorded before the index date (N = 139) and those with a diagnosis of HIV infection before the index date (N = 658), leaving a final sample of 205,203 patients.
Identification of Non-Hodgkin's Lymphoma (NHL) Cases.
Cases of NHL newly diagnosed from January 1, 1995 through December 31, 2002 were identified from the tumor registries maintained by each health system. The registries at HFHS and KPMCP inform the National Cancer Institute Surveillance Epidemiology and End Results (SEER) Program and employ thorough case finding systems that include review of all pathology and cytology reports, as well as radiation and oncology consultations. The SEER program assures maintenance of a population-based case reporting standard of 97% or greater. Follow-up for newly diagnosed NHL cases began on the index date (date of chronic HBV diagnosis for cases and corresponding date for matched comparison patients). In both cohorts, NHL follow-up data were obtained through December 31, 2002.
Potential Confounding Variables
We collected information on diagnosis of HCV (based on laboratory result data), ecological estimates of median household income (derived from matching patient address to 2000 census block group data), age at index date, sex (obtained from membership files), race, and ethnicity. Race and ethnicity data were available directly from patient records at HFHS. At KPMCP, race and ethnicity were available only for a proportion of patients and was derived from existing data sources, including surname matching. We constructed a comorbidity score using automated health care encounter data according to the Deyo adaptation of the Charlson comorbidity index. This adaptation of Charlson comorbidity index scoring categorizes patients based on the presence/history of: myocardial infarction, congestive heart failure, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcerative disease, liver disease, diabetes, hemiplegia, renal disease, cancer, or HIV.
We calculated the incidence of NHL (number of events/person-years of follow-up) and associated 95% confidence intervals. Person-time was calculated from the date of chronic HBV diagnosis (or index date for comparison cohort) until the earliest of: NHL diagnosis, death, disenrollment from the health plan, or the end of the follow-up period (December 31, 2002). We fit Kaplan-Meier curves and quantified the association between chronic HBV infection and NHL by fitting Cox proportional hazards models and obtaining relative hazards and associated 95% confidence intervals. Analyses were adjusted for the following variables: race (white, black, Asian, Hispanic); age (in years; <30, 30-39, 40-49, 50-59, 60+); median residential neighborhood income based on geocoding to 2000 Census Data (<$44,000, $44,000-$59,999, $60,000-76,000, >$76,000); sex (male, female); study site (HFHS, KPMCP); Charlson comorbidity score (0, 1, >2); and HCV infection (yes/no). Patients with missing race data and patients with the category gotherh were excluded from multivariate analyses because no NHL was found in those groups (race is collected by the tumor registries).