Neighborhood Socioeconomic Status and Hodgkin's Lymphoma Incidence in California

Hodgkin's lymphoma, an uncommon malignancy of B lymphocytes, has long been noted for an unusual age-specific incidence pattern that varies by socioeconomic status (SES), with a pronounced peak in young adults (ages 15-44 years) and a second peak in persons ages >55 years in economically developed populations (1-6). MacMahon was the first to speculate from these and other features a multiple-disease model of Hodgkin's lymphoma etiology in which Hodgkin's lymphoma represented in young adults a rare response to delayed exposure to a common infection and in older adults a malignancy like non-Hodgkin's lymphoma (7, 8). Subsequent case-control studies seemed to bolster the notions of age-specific etiologies by demonstrating associations of young-adult Hodgkin's lymphoma with markers of higher childhood SES (9-12), by showing limited associations with these markers in older adults, and by finding associations with lower levels of parental education and other markers of lower SES in children (13, 14). For young adults, associations with higher maternal education, lower housing density, and fewer siblings (9, 10, 12) were thought to indicate later exposure to infections, giving rise to the “delayed infection” etiologic hypothesis. Later work singled out delayed infection with EBV as being of particular interest (15).

More recently, published studies have not replicated associations of Hodgkin's lymphoma with markers of childhood or adult SES. In 1995, Macfarlane et al. reported that the positive ecologic correlation in international data of young-adult Hodgkin's lymphoma incidence with national level of economic development first described in 1971 (2) had weakened over time and was sensitive to the inclusion of Asian data omitted from the original analysis (16). Several case-control studies conducted in the 1990s reported that previously implicated markers of SES did not associate with young-adult Hodgkin's lymphoma risk clearly (17-19), even in analyses taking into consideration variation by race (17). These observations have led us and others to hypothesize that societal changes, such as the increased use of daycare, may have reduced socioeconomic variation in etiologically relevant exposures for Hodgkin's lymphoma (17, 19, 20).

Before concluding that SES variation in Hodgkin's lymphoma is no longer a relevant correlate of Hodgkin's lymphoma incidence, it is important to reexamine the association. In order for such an analysis to be informative, it must be conducted in a contemporary population with adequate size to stratify analyses by factors like race/ethnicity and histologic subtype, both known to influence Hodgkin's lymphoma incidence (21) and the latter shown previously to affect socioeconomic gradients in Hodgkin's lymphoma (3-5). Therefore, taking advantage of ∼150 million person-years of observation from the population-based cancer registry covering the large, ethnically diverse population of California, we examined the association of a previously validated, neighborhood-level, multidimensional measure of SES with Hodgkin's lymphoma incidence overall and in subgroups defined by age group, sex, race/ethnicity, and histologic subtype.

Eligible for this study were all 3,906 patients newly diagnosed with Hodgkin's lymphoma [International Classification of Diseases for Oncology, Second Edition (22) morphology codes 9650-9667] in California and reported by state mandate to the California Cancer Registry (CCR) during the period January 1, 1988 to December 31, 1992. This pericensal period was chosen based on the availability of appropriate population denominators from the 1990 U.S. Census and the fact that adequate numerators (e.g., 1998-2002) had not yet been released by the CCR for an analysis around the 2000 census data. For this study, cancer registry data, routinely abstracted from the medical record, for this study included patient age at diagnosis, sex, race, Hispanic ethnicity, residential address at diagnosis, and tumor histologic subtype. Hispanic ethnicity was additionally classified by the CCR through matching of last (or maiden) names with lists of known Hispanic surnames because of prior evidence of medical record underreporting of Hispanic ethnicity (23). For the present analysis, race/ethnicity was grouped into four mutually exclusive categories: Hispanic, non-Hispanic White, non-Hispanic Black, and non-Hispanic Asian/Pacific Islander/other (hereafter called Hispanic, White, Black, and Asian); persons with unknown race/ethnicity (n = 5) were excluded. We also excluded patients (n = 107) with evidence of HIV infection based on the cancer registry HIV/AIDS code or AIDS-related cause of death because of the epidemiologic differences between HIV-associated and sporadic Hodgkin's lymphoma (24). These exclusions resulted in a final study population of 3,794 Hodgkin's lymphoma patients.

Individual-level patient SES characteristics (e.g., education, income, and occupation) are not collected routinely by most U.S. cancer registries, including the CCR. However, because the residential address at diagnosis is routinely geocoded for each patient, SES characteristics of the geographic region in which the patient resided at the time of diagnosis can be obtained from the U.S. Census Bureau. For the 1990 census, the smallest geographic unit with both SES information and the detailed population counts needed to calculate cancer incidence rates was the census block group, an area containing on average 1,500 residents. To determine block group SES, we used the method of Yost et al. (25), which employed principal components analysis to develop a single index from seven census indicator variables of SES (education index, median household income, percent living 200% below poverty level, percent blue-collar workers, percent ages >16 years in workforce without job, median rent, and median house value). This index thus incorporates information about three critical domains of SES—education, income, and occupation (26)—as well cost of living, which varies within California, and has been used previously to show substantial socioeconomic gradients in breast cancer incidence in California (25). Using this index, we assigned a standardized score to each of the 20,919 block groups in California in 1990 and then categorized these scores into tertiles (low, medium, and high); tertiles were selected over finer categorizations to provide adequate sample size for all planned analyses. Table 1 shows for each SES tertile the median block group value of selected indicator variables.

Each Hodgkin's lymphoma patient was then assigned a neighborhood SES tertile based on geocoded census block group of residence. Geocoded block group was not available for 292 (7.7%) of the eligible patients; these patients did not differ significantly (P < 0.05) from patients with known block group on age at diagnosis, race/ethnicity, vital status, year of diagnosis, or tumor characteristics but did differ on county of residence, because patients with unknown block group were more likely to live in counties with rural areas. Patients with unknown block group were randomly allocated to block groups within the same county, and all analyses were conducted in parallel data sets, one including and one excluding these patients. As only minor differences were observed in results of these analyses, the randomly allocated patients were included in final calculations to maximize rate stability.

Population data for Hodgkin's lymphoma incidence rate denominators were obtained from age-, sex-, and race-specific population counts for census block groups from the modified age, race, sex, and Hispanic origin (“MARS”) files from the 1990 U.S. Census. Because census block group population estimates were not available for intercensal years, we multiplied the 1990 population counts by 5 to estimate the population at risk for the 5-year period 1988 to 1992.

Case counts and population estimates were sorted by 5-year age group, sex, race/ethnicity, and neighborhood SES tertile. Hodgkin's lymphoma incidence rates were calculated per 100,000 persons and age adjusted to the 2000 U.S. population. SES tertile-specific incidence rate ratios (RR) and associated 95% confidence intervals (95% CI) were estimated and adjusted for age and, when appropriate, sex and race/ethnicity using Poisson regression, a probability model for counts of rare events like Hodgkin's lymphoma. We conducted tests for linear trend across neighborhood SES tertiles by using Poisson regression, exchanging a continuous SES variable for the categorical SES variables and assessing the significance of the continuous term. All calculations were carried out using SAS software (version 9.0). In accordance with CCR confidentiality guidelines at the time of the study, patient categories with fewer than five cases are shown in tables as “<5” and rates are not presented. Hodgkin's lymphoma incidence patterns were examined separately for persons ages <15, 15 to 44, and ≥45 years based on California incidence patterns for the period 1988 to 1992 and on the well-established epidemiologic differences in Hodgkin's lymphoma characteristics by age at diagnosis in this population (21).

The overall age-adjusted rate of Hodgkin's lymphoma in this population was 2.6 per 100,000, and the age-specific incidence patterns are shown in Fig. 1. The majority of the 3,794 study patients were young adults ages 15 to 44 years at diagnosis, children ages <15 years comprised 5% of the case series, and older adults ages ≥45 years comprised 32%. Across these age groups, patients varied in distributions of sex, race/ethnicity, histologic subtype, and neighborhood SES tertile (Table 2). A male excess was apparent in all age groups but was most pronounced in children. Equal proportions of childhood Hodgkin's lymphoma patients were of White and Hispanic race/ethnicity, whereas young-adult and older-adult patients were most likely to be White. The majority of childhood and young adult patients had the nodular sclerosis Hodgkin's lymphoma subtype; in older adults, this subtype comprised a smaller proportion of all cases, whereas unclassified disease was substantially more frequent than in younger Hodgkin's lymphoma patients.

In children, Hodgkin's lymphoma incidence did not vary markedly or significantly by neighborhood SES (incidence rates per 100,000 children for low, medium, and high tertiles: 0.61, 0.53, and 0.66, respectively; P_trend = 0.12). The small numbers of patients within neighborhood SES tertiles precluded further stratification by race/ethnicity, sex, or histologic subtype.

Table 3 shows that, for all races combined, Hodgkin's lymphoma incidence was 22% higher for males and 44% higher for females in the highest tertile of neighborhood SES than in the lowest. Socioeconomic gradients seemed to vary by race/ethnicity, although within these groups stronger associations persisted in females than males. Gradients were generally observed in all groups, except Asian males, but statistically significant associations were limited to White males and females, Hispanic females, and Asian females, with a borderline (P = 0.06) association observed in Hispanic males. The strongest associations were observed in Hispanic and Asian females for whom rates were 2.7-fold higher in the highest compared with the lowest neighborhood SES tertiles. Whites had higher sex-specific Hodgkin's lymphoma rates than other racial/ethnic groups; for White females, the rate in the highest SES tertile was 24% higher than the rate in the lowest, whereas for White males rates suggested a weak association with neighborhood SES and a trend of borderline significance. However, the rate of Hodgkin's lymphoma in White males of high SES, the highest risk group, was 11 times higher than that in Asian females of low SES, the lowest risk group.

Hodgkin's lymphoma histologic subtype modified associations with SES among young adults, with significant associations observed only for the nodular sclerosis form. Table 4 shows that rates of nodular sclerosis Hodgkin's lymphoma for all races combined were 39% higher for males and 53% higher for females in the highest SES tertiles compared with the lowest. In Whites, rates increased significantly with increasing SES and were 32% to 34% higher in the highest tertile than the lowest, with similar elevations and trends for males and females. Similar patterns, but no significant associations, were observed in Blacks. Stronger associations with SES were observed in Hispanics of both sexes and in Asian females. For the mixed cellularity subtype, rates of young-adult Hodgkin's lymphoma did not associate consistently with neighborhood SES in males (counts and incidence rates per 100,000 males for low, medium, and high tertiles: n = 75, rate = 0.57; n = 93, rate = 0.68; and n = 77, rate = 0.62, respectively; P_trend from Poisson regression = 0.49) or females (counts and incidence rates per 100,000 females for low, medium, and high tertiles: n = 21, rate = 0.17; n = 21, rate = 0.16; and n = 38, rate = 0.30, respectively; P_trend from Poisson regression = 0.19) for all races combined. No socioeconomic gradients in the incidence of mixed cellularity Hodgkin's lymphoma were evident for White males or females, Hispanic males or females, or Blacks or Asians for whom sex-specific analyses were based on small numbers (data not shown). Rates of the uncommon histologic subtypes (lymphocyte predominant, lymphocyte depleted) and of unspecified subtype did not differ significantly by SES tertile across all racial/ethnic groups combined (data not shown).

Among older adults, associations between neighborhood SES and Hodgkin's lymphoma incidence were inconsistent across racial/ethnic and sex categories (Table 5). No associations were observed in most groups. However, elevated risks occurred for Black females and Hispanic males in the highest compared with lowest SES tertile, and a lower risk was suggested for Black males in the highest tertile. Stratification of Hodgkin's lymphoma rates by the nodular sclerosis and mixed cellularity histologic subtypes did not resolve all of these inconsistencies for older adults. For the nodular sclerosis subtype in older adults, rates did not associate consistently with neighborhood SES in males (counts and incidence rates per 100,000 males for low, medium, and high tertiles: n = 58, rate = 1.15; n = 77, rate = 1.20; and n = 112, rate = 1.50, respectively; P_trend from Poisson regression = 0.30) or females (counts and incidence rates per 100,000 females for low, medium, and high tertiles: n = 50, rate = 0.80; n = 67, rate = 0.87; and n = 85, rate = 1.00, respectively; P_trend from Poisson regression = 0.31) for all races combined. Within racial/ethnic groups, an association with higher SES was suggested only for Black females for whom the rate of nodular sclerosis Hodgkin's lymphoma in the highest SES tertile was significantly higher than the rate in the lowest (RR, 6.3; 95% CI, 1.97-21.54) despite small numbers of cases (n = 7 and 5, respectively). Table 6 shows that for mixed cellularity disease a significantly increasing association of neighborhood SES was observed in White males for whom rates of mixed cellularity Hodgkin's lymphoma were 56% higher in the highest SES tertile compared with the lowest. Among Hispanic males, rates were 2.1-fold higher in the highest compared with the lowest tertile with a significant trend (P = 0.05). Rates in Blacks and Asians were based on very small numbers. Rates of lymphocyte predominant, lymphocyte depleted, and unspecified subtypes did not vary significantly by neighborhood SES (data not shown).

This analysis of 150 million person-years of observation in the California population allowed us to examine the association of neighborhood SES to Hodgkin's lymphoma incidence in detailed patient subgroups relevant to Hodgkin's lymphoma epidemiology but difficult to study due to sample size constraints. In this large and diverse population, we did not detect uniform associations of our neighborhood SES measure with Hodgkin's lymphoma incidence across groups defined by age, race/ethnicity, sex, and Hodgkin's lymphoma histologic subtype. Rather, we found positive associations primarily in two subgroups: young adults (especially Hispanic and Asian females) with the nodular sclerosis histologic subtype and older adults with the mixed cellularity subtype. Within each neighborhood SES tertile, we observed racial/ethnic variation in Hodgkin's lymphoma rates remarkably similar to that reported previously (5, 21), with the highest rates in Whites, intermediate rates in Blacks and Hispanics, and very low rates in Asians.

To our knowledge, socioeconomic gradients in Hodgkin's lymphoma incidence have not been described systematically across racial/ethnic groups. Our finding of a positive socioeconomic gradient in nodular sclerosis Hodgkin's lymphoma among White young adults confirms the results of prior cancer registry-based studies using area-level composite SES measures (3-5). A lesser influence of neighborhood SES is consistent with the Macfarlane et al. observation of a weakening socioeconomic gradient over time in international cancer registry data (16) and with contemporary studies conducted in mostly White populations that report weak or null associations of young-adult Hodgkin's lymphoma risk with individual-level markers of SES (17, 18, 27). Our findings of a positive association of neighborhood SES among some older adults with mixed cellularity Hodgkin's lymphoma have not been reported previously, to our knowledge. Rather, mixed cellularity Hodgkin's lymphoma incidence has been negatively associated with county- or census-tract level SES indicators across all ages using 1973 to 1980 U.S. data (4) and for older adults in 1972 to 1985 Los Angeles data (5). The discrepancy of our findings with these earlier observations may be explained in part by improved accuracy of diagnostic and histologic classifications in cancer registry data over time for older adults (28). However, as the mixed cellularity classification has been shown to have moderate to poor interobserver reliability even in more contemporary data (29), it is difficult to know the effect of subtype misclassification, especially the misdiagnosis of Hodgkin's lymphoma as non-Hodgkin's lymphoma in older adults, on the mixed cellularity results of this and prior studies.

Although area-level associations are not necessarily predictive of individual-level associations (the ecologic fallacy), neighborhood socioeconomic conditions have been assumed to affect Hodgkin's lymphoma development largely through their correlation with individual-level SES (30), which has been interpreted previously to proxy for factors reducing or delaying exposure to infectious agents. After Cohen et al.'s observation of higher Hodgkin's lymphoma incidence in soldiers with more years of schooling (31), multiple studies showed that links of Hodgkin's lymphoma with educational level and other socioeconomic characteristics measured at diagnosis were generally explained by socioeconomic characteristics of childhood like smaller family size, lower birth order, living in a single family home, home ownership, and fewer playmates (10, 11, 32-34) in support of the “delayed infection” hypothesis. If the timing of infections is important to Hodgkin's lymphoma causation, this factor could be additionally modulated by neighborhood-level socioeconomic characteristics like low population density or population composition (e.g., fewer children or immigrants), a possibility that has been addressed by only one prior study, to our knowledge. Among persons ages 0 to 24 years in Great Britain during the period 1984 to 1988, Alexander et al. found independent and positive associations of Hodgkin's lymphoma risk with neighborhood characteristics, including higher SES and close proximity to “built-up areas,” but no relation with rural status or average length of commute to work (6). The relative contributions of individual- and community-level SES factors to Hodgkin's lymphoma incidence could be addressed by contextual analysis, which allows the effects of both to be measured separately (35).

Regardless, the incidence patterns observed in this analysis raise interesting questions for the etiology of the most common form of Hodgkin's lymphoma, young-adult nodular sclerosis. First, the marked socioeconomic gradients in Hispanic males and females and in Asian females suggest that, at least in contemporary U.S. populations, etiologically important conditions align more closely with foreign birthplace than with SES, as 74% of Hispanics and 65% of Asians living in the United States in 1990 were born in other countries compared with <10% in Whites and Blacks (36). In our case-control study of Hodgkin's lymphoma in northern California women, U.S. birthplace was associated with higher Hodgkin's lymphoma risk (17), and childhood social class risk factors had stronger associations with risk in non-Whites (a group with more foreign-born persons) than Whites (17, 37, 38). These data would support the idea that childhood environments protecting against Hodgkin's lymphoma may have become less common over time in all U.S. social strata, while persisting outside the United States, particularly in the birthplaces of contemporary Mexican and Asian immigrants to California. The weakly positive associations of SES we observed among Whites are compatible with the hypothesis that previously noted socioeconomic gradients in Hodgkin's lymphoma have been mitigated by the increased use of daycare and preschool, recently associated with reduced risk of Hodgkin's lymphoma (19), especially by higher SES families (20). Second, we observed that stratification by neighborhood SES level only weakly attenuated racial/ethnic gradients in incidence. Regardless of neighborhood SES, Whites had substantially higher rates than other groups, and Asians had uniformly low rates, even those living in high SES areas. Given that the Asian population had higher median household income and education than White, Black, and Hispanic populations in the United States in 2000 (36), these findings support the notion that Asians as a group have less genetic susceptibility to Hodgkin's lymphoma (37). It is interesting to note that the racial/ethnic variation observed in young-adult Hodgkin's lymphoma is similar to that observed for breast cancer, for which adjustment for SES only slightly reduced a comparable racial/ethnic gradient in breast cancer, with rates also highest in Whites, intermediate in Blacks and Hispanics, and lowest in Asians (25). These observations implicate cultural or lifestyle differences among racial/ethnic groups (especially immigrant groups) as important and relatively untapped sources of new hypotheses for Hodgkin's lymphoma causal factors. Third, in most groups, stronger socioeconomic gradients were observed in females than males. This could reflect socially determined differences in exposure opportunities, biologically determined differences in immune response to exposures, or modification of disease pathogenesis by hormonal factors as proposed previously (39). Hormonal exposures through pregnancy and breast-feeding may interact with childhood exposures to affect Hodgkin's lymphoma development in women (38). Thus, the higher Hodgkin's lymphoma incidence in Hispanic women living in higher SES neighborhoods may reflect the protection to lower SES women afforded by both early exposure to infection and higher parity or lactation.

Our findings have implications for the interpretation of past and future etiologic studies of Hodgkin's lymphoma. Because effect estimates for risk factors associated with SES should be sensitive to the racial/ethnic or immigrant composition of the population studied, comparisons may be misleading among studies conducted in homogeneous populations or that have not accounted for the racial/ethnic or immigrant composition of the study population. The complexity of incidence patterns we report here also underscores the critical importance to Hodgkin's lymphoma etiologic research of considering variation by age at diagnosis, histologic subtype, sex, and race/ethnicity. At the very least, future etiologic studies of Hodgkin's lymphoma must describe these features of their study population in detail and attempt to stratify analyses accordingly. Although our cancer registry-based data provided the benefits of representativeness, large numbers of patients, and uniform means of SES assignment, they were limited in their scope, particularly their lack of individual-level SES information. Future cancer registry-based analyses should use more recent cancer incidence and 2000 census data when they become available. To provide better-defined exposures and mechanisms for further etiologic study, future work should use more comprehensive data resources with the capacity to assess joint contributions of individual- and community-level factors, including characteristics more specific than the composite measure of SES used here. Interesting factors for more detailed study of the young-adult, nodular sclerosis and older-adult, mixed cellularity forms of Hodgkin's lymphoma might include immigration status and timing, characteristics of birthplace (especially factors common to Hispanic or Asian immigrants to California), household composition, housing type and density, and childcare practices. As part of this comprehensive assessment, specific aspects of the neighborhood SES index should be separated to try to isolate aspects of importance.

We thank the staff of the CCR, including Amy Laurent, Mark Allen, and William Wright, Ph.D., for providing the data used in this analysis and Lyndsey Darrow, Erin Eberle, Dawn Beahm, Scarlett Gomez, and Kathleen Torres for their contributions to this study.