Residential History, Family Characteristics and Non–Hodgkin Lymphoma, A Population-Based Case-Control Study in the San Francisco Bay Area Free

Incidence rates of non–Hodgkin lymphoma (NHL) have nearly doubled since the early 1970s (1) with 58,870 new cases and 18,840 deaths estimated in the United States in 2006 (2). The few known risk factors include impaired immune function associated with chronic autoimmune conditions and immunosuppressive therapies, increasing age, viral infections such as EBV and human T-cell leukemia/lymphoma virus-I (1), and family history of NHL or other hematopoietic cancer (3). Molecular, genetic, and epidemiologic data also indicate that NHL etiology is likely to vary by histologic subtype (4-6).

The hygiene hypothesis was developed to explain the increased incidence of allergy, asthma, and atopy in Westernized countries (7) but also has been tested to explain other immune-related conditions, including the incidence of leukemia (8), Hodgkin's disease (9), and NHL (10-12) over the past several decades. In general, it states that better sanitation and small family size have led to fewer and later onset of childhood infections and, thus, altered immune function (7). The proposed immunologic mechanism involves T-helper cell 1 (Th1) and Th2 immunity. Th1 typically responds against infectious agents and includes phagocytic activity and cell-mediated immunity (13). Th2 responses are associated with humoral immunity and development of antibodies and also involve inflammation and suppression of phagocytosis (13). It has been suggested that either in utero or at birth, the T-cell immunologic environment is skewed toward a Th2 response pathway (14-16). The hypothesis also purports that a Th1/Th2 imbalance is countered postnatally by Th1 stimulation via infections; otherwise, an allergic or Th2 response develops (16, 17). Based on these assumptions, childhood infections are important in immune system development and may decrease the incidence of immune-related disease.

Factors integral to the hygiene hypothesis, such as family size, residential history, childhood exposure to infectious agents, and social class, have been studied in association with risk of childhood leukemia and Hodgkin's disease (8, 9, 18-22); however, few studies have published detailed analyses of the associations between these factors and NHL (10-12, 22). In the present analyses, we examined the association between NHL, residential history, family characteristics, history of common childhood infections, and allergic rhinitis to determine whether these key hygiene hypothesis factors also play a role in the development of NHL.

Details of the study design and methods have been published (6, 23-29) and therefore are presented in brief here. The Northern California Cancer Center rapid case ascertainment was used to identify NHL patients ages 21 to 74 years in six Bay Area counties usually within 1 month of diagnosis. One thousand five hundred ninety-three eligible NHL patients completed in-person interviews (72%) and 1,304 HIV-negative patients were included in these analyses. Control participants were identified by random-digit dial (30-32) that was supplemented by random sampling of the Health Care Financing Administration (now Centers for Medicare and Medicaid Services) files for participants ages >65 years. Controls were frequency-matched to patients by 5-year age groups, sex, and county of residence (23). Two thousand five hundred fifteen eligible control participants completed in-person interviews (78%) and 2,402 HIV-negative individuals were included in these analyses. No proxy interviews were conducted. Study protocols and procedures were approved by the University of California San Francisco Committee on Human Research and all participants provided written informed consent before interview.

Diagnostic pathology materials were re-reviewed and classified using the Working Formulation (33) by the expert study pathologist for 97% of cases. To approximate the WHO classification for NHL, all follicular subtypes were combined into one follicular lymphoma (FL) category and diffuse large-cell and immunoblastic large-cell subtypes were combined into diffuse large-cell lymphoma (DLCL; refs. 34-36). Due to sample size constraints, the subtype results are presented only for FL, DLCL, and working formulation small lymphocytic (SL) lymphoma.

Data were collected about residential characteristics at ages 8, 15, 25, 35, 45, and 55 years, including the number of persons in the household (hereafter household size), number of rooms, dwelling type, and type of residential area. Demographic characteristics, foreign born, birth order, number of full or half-siblings, ever lived or worked on a farm, number of different childhood infections, and history of allergic rhinitis also were included in these analyses.

Most continuous variables were categorized based on the quartiles or tertiles of the distribution among controls. Categories were collapsed for final analyses if the odds ratio (OR) estimates for the initial categorical groupings differed by <10%. Residential crowding was computed as the number of people divided by the number of rooms in the household and was categorized based on quartiles of the distribution among the control participants. Crowding in childhood (age ≤15 years) was determined as the maximum crowding at ages 8 and 15 years. Dwelling type was analyzed as an ordinal categorical variable based on increasing potential crowding in the dwelling environment with single family home the a priori reference group. Areas of residence were described as urban/suburban, rural nonfarm, and rural farm. Ever having lived or worked on a farm was integrated with residential area to create the final categories urban/suburban without farm exposure (reference), urban/suburban with farm exposure, rural nonfarm, and rural farm. Living on a farm as a child was analyzed as a categorical variable grouped as never having lived on a farm (reference), first having lived on a farm when ≤1-year old, >1 to 8 years old, or ≥8 years old.

Variables for the number of different infectious diseases or conditions before age 8 years and before age 15 years were computed as the sum of self-reported history of measles, mumps, rubella, scarlet fever, rheumatic fever, chicken pox, whooping cough, mononucleosis, tuberculosis, polio, and pneumonia, and analyzed as an ordinal variable [0 (reference), 1, 2-3, ≥4]. History of allergic rhinitis (ever/never) was used as a measure of atopy and defined as self-reported history of allergies to plants, seasonal allergies, or hayfever. Total number of siblings, birth order, and number of younger siblings were analyzed as ordinal variables with no siblings defined as the reference group.

Age- and sex-adjusted unconditional logistic regression models were used to obtain OR and 95% confidence intervals (95% CI) as estimates of the relative risk (hereafter called risk). Statistical interaction was determined using the likelihood ratio test for nested models. Statistical tests for trend for categorical variables were based on the effect estimate when the factor was modeled as an ordinal variable. Potential confounding effects of number of infections before age 8 years, foreign born, White non-Hispanic (yes/no), and education were evaluated and remained in the final models if they altered the OR for the main effect by ≥10%. Investigation of an association between NHL and the hygiene hypothesis was a focus of these analyses. Given that the hygiene hypothesis emphasizes exposures during early childhood, a statistical model to test the hygiene hypothesis was restricted to direct and indirect effects (mediated through allergic rhinitis) of common infectious diseases when ≤8 years old. All statistical tests were two-sided and considered statistically significant for P ≤ 0.05 and somewhat significant for 0.05 < P ≤ 0.10.

Overall, risk increased with the number of people in the household. The OR for NHL was increased 30% among those who at 8 years old lived in households with more than six people compared with those who lived in households with three people or less. ORs for NHL were increased for those living in homes with more than one person at ages 25, 35, 45, and 55 years compared with those living alone at each of these ages although the risk estimates did not consistently follow a linear pattern (Table 1).

At age 8 years, compared with those who lived in homes with four rooms or less, the ORs peaked for five- to six-room households and declined with further increases in number of rooms. Decreasing risk estimates for NHL were associated with increasing number of household rooms at age 15 years with the OR reduced nearly 40% for the highest room category. Risk estimates for NHL associated with number of household rooms were increased for participants in the highest room categories at ages 35, 45, and 55 years. Education was a confounder for ages 45 and 55 years and tabled values are adjusted accordingly. There was no statistical interaction between the number of rooms and number of individuals living in the home (Table 2).

Median crowding density (these data not shown in table) was 0.67 persons per room at age 8 years and 0.62 persons per room at age 15 years, whereas the median of the maximum level of crowding at age 8 or 15 years was 0.71 persons per room. Median levels of crowding decreased with increasing age; 0.50 at 25 or 35 years old, 0.44 at 45 years old, and 0.33 at 55 years old. The third quartile of crowding never exceeded 1.0 person per room at any age. Participants who lived in more crowded homes (third and fourth quartile groups) at 8 or at 15 years old were 40% more likely to have been diagnosed with NHL compared with participants in the lowest quartile of crowding (P_trend < 0.0001, P = 0.0004, respectively). A similar increasing trend for NHL with increasing crowding was observed for the maximum level of crowding by age 15 years (P_trend < 0.0001; Table 3).

Compared with those who lived in a single family home, ORs for NHL increased with increasing residential units in the dwelling at age 8 or 15 years (moderate: ≥2 family house or <10-unit apartment building, OR, 1.2; 95% CI, 0.99-1.5; large: ≥10-unit apartment building or dormitory, OR, 1.3; 95% CI, 0.98-1.6; P_trend = 0.02). In contrast, risk estimates for NHL were reduced among those who resided in dwellings with a moderate number of units at ages 25, 35, and 45 years (OR, 0.76; 95% CI, 0.64-0.90; OR, 0.68; 95% CI, 0.55-0.85; OR, 0.65; 95% CI, 0.48-0.87, respectively) or with a large number of units at age 55 years (OR, 0.72; 95% CI, 0.65-1.0). No other housing type was associated with NHL although some estimates were unstable because of the small number of exposed study participants.

NHL risk was somewhat reduced among participants who ever had lived on a farm with decreased ORs observed among those who lived on a farm for <2.5 years (OR, 0.71; 95% CI, 0.58-0.87). ORs for NHL decreased with increasing age first lived on a farm (P_trend = 0.03) with a nearly 20% reduced risk among those who first lived on a farm when >8 years old (OR, 0.81; 95% CI, 0.66-0.99).

Those who lived in an urban/suburban area and had no history of living or working on a farm were the reference group for analyses of residential regions. At all ages, ORs were reduced ∼10% to 20% for those who lived in urban/suburban areas and also had a farm history. However, confidence intervals included unity except for estimates at ages 15 and 25 years (OR, 0.82; 95% CI, 0.68-0.99; OR, 0.82; 95% CI, 0.70-0.95; respectively). History of habitation in other residential areas was not associated with NHL at any age.

Compared with only children, increased ORs for NHL were observed for participants who had five or more siblings and increased with increasing birth order (i.e., increasing number of older siblings, P_trend = 0.01) although estimates were only somewhat increased for individual birth-order categories (Table 4). ORs were increased 50% among participants who had five or more younger siblings, whereas ORs for those with fewer younger siblings were elevated only slightly and were not different from chance. Compared with U.S.-born study participants, those who were foreign born were more likely to have been diagnosed with NHL.

An increasing number of different childhood infectious diseases by ages 8 and 15 years was associated with a decreasing trend in ORs for NHL (P_trend < 0.0001, 0.0003, respectively; Table 4). ORs for NHL were decreased among those with a history of allergic rhinitis. Race/ethnicity (White non-Hispanic yes/no), foreign born, residential crowding, and number of siblings were neither effect modifiers nor confounding factors for these associations.

In analyses stratified by allergic rhinitis status, ORs for NHL decreased with increasing number of different childhood infections by age 8 years with a possible greater reduction observed among those with a history of allergic rhinitis although confidence intervals were wide (data not shown). ORs were not confounded by residential crowding at age 8 years, number of siblings (<5, ≥5), race/ethnicity (White non-Hispanic yes/no), or foreign born (yes/no).

In analyses by NHL subtype, risk estimates at nearly all ages were increased for participants who resided in the largest households although some CIs overlapped unity (data not shown in table). In general, ORs were greatest for participants who lived with at least one other person at age ≥35 years, regardless of histologic subtype. Subtype ORs for those who reported living in homes with more rooms at age 15 years were decreased for DLCL and SL and generally were increased at age ≥35 years (data not shown in table).

Increased crowding at ages 8, 15, and 8 and 15 years combined was associated with increased ORs for FL and DLCL although estimates generally were similar for the third and fourth quartiles of crowding and race/ethnicity attenuated the ORs for DLCL (Table 5). The elevated ORs for SL did not follow a linear pattern with fourth quartile estimates that were ∼20% lower than estimates for the third quartile of crowding.

Consistent with results for all NHL, a decreasing trend in ORs by NHL subtype was observed with increasing number of different childhood infections by age 8 and 15 years most notably for FL (Table 5). A decreasing trend in ORs for DLCL with increasing number of infections was attenuated after adjustment for race/ethnicity (P_trend = 0.02 for age 8 and 15 years). Although ORs for SL also decreased with increasing number of infections at age 8 years, all estimates included unity (P_trend = 0.25). ORs for DLCL and SL were reduced among participants who reported a history of allergic rhinitis. Participants who had five or more siblings were more likely to have been diagnosed with FL or DLCL. There was a suggestion of increasing trends in ORs for DLCL with increasing birth order and for FL with increasing number of younger siblings, although confidence intervals included unity. Sibling variables did not confound the relationships between NHL subtypes and number of infections or history of allergic rhinitis.

There was no association between DLCL or SL and history of living or working on a farm (data not shown in table). OR estimates hovered around unity and all confidence intervals included unity. Results showed that those who ever had lived on a farm were less likely to be diagnosed with FL (P = 0.03), although there was no linear trend in the ORs with increasing number of years lived on a farm (P_trend = 0.23). There was no association between NHL subtypes and age first lived on a farm.

Results from our analyses of residential history indicate that there was some evidence of increased risk of NHL with increased crowding during childhood. Adjusted ORs were increased for those who at 8 years old had resided in households with more people or had resided in more crowded homes. Results for residential history factors at age 15 years were consistent with those at age 8 years although the pattern of decreasing ORs with increasing number of household rooms was observed only at age 15 years. Having more siblings, more older siblings (higher birth order), or five or more younger siblings also was associated with increased ORs for NHL, whereas an increased number of different childhood infections and a history of allergic rhinitis were associated with decreased ORs for NHL. The direct effect between NHL and increasing number of different childhood infections was similar to the indirect effect estimated when analyses were stratified by history of allergic rhinitis and adjusted for crowding or number of siblings. At adult ages, most ORs for NHL were elevated for any number of people in the household greater than the reference group and for those people who lived in the largest homes after adjustment for education.

To explain the increased incidence of atopy and asthma in developed countries, Strachan (7) proposed that less crowding and improved sanitation resulted in fewer childhood infections and a switch from an infection-induced Th1-dominant immune response to an atopy/allergy–related Th2-dominant immune response. The hypothesized Th1/Th2 switch has been investigated in many epidemiologic studies of immune-related conditions mainly through measures of crowding, childhood infections, and atopy. Several epidemiologic studies have investigated the association between NHL and number of siblings, number of childhood infections, history of allergies, and some residential characteristics although most studies evaluated only the main effects of the factors of interest.

The reported associations between NHL and number of siblings have been inconsistent. Published results have shown no association between NHL and family/sibship size (11, 12, 22, 37), an increased risk among only children (38), decreasing risk with decreasing number of siblings with only children having the lowest risk estimates (10), and increasing risk with increasing number of siblings among men who have sex with men (23) and among heterosexual men and women (26). Our results that showed associations between NHL and birth order/number of siblings are consistent with the recently published results from Australia (10). However, to have been consistent with the premises of the hygiene hypothesis, we had expected to find a decreased risk for adult NHL among participants with more and/or older siblings as these individuals would have been more likely to have been exposed to infectious agents at a younger age compared with children with fewer siblings (39). If the association between infectious agents and NHL is not due to chance, then our data indicate that number of siblings may not be a good proxy for childhood exposure to infectious agents, may need to be further qualified with regard to the number of siblings and other children in the home as a young child, or that the hypothesis that number of siblings is associated with childhood exposure to infectious agents is invalid.

An infectious etiology for NHL has been investigated in many epidemiologic studies. However, viruses known to cause NHL (e.g., EBV and human T-cell lymphotrophic virus, type I) account for few NHL cases in the United States. Risk of NHL has been positively associated with exposure to other infectious agents that include childhood infections, varicella virus (22), and scarlet fever (38, 40), as well as history of herpes zoster (38, 40), tuberculosis (12, 41), infectious mononucleosis (12, 42), hepatitis C virus (43), respiratory flu in men (44), eye infections in women (44), and history of polio among men (26). Although studies often considered the lag time between onset of infectious conditions and diagnosis, few studies have collected detailed information about infectious disease onset during childhood, which is relevant to the hygiene hypothesis. Furthermore, results have been somewhat inconsistent from the few studies that specifically evaluated childhood infectious disease onset. In one early study (22), age at onset of infectious diseases as recorded on college physical exams of 50,000 male alumni did not differ between patients and controls. In comparison, a recent German study showed a decreased risk of NHL among men who reported childhood measles (but not other childhood infections) and among participants who reported that they were more often sick or absent from school as child when <10 years old (11). An Italian study showed an increased NHL risk among participants whose average age at onset of their first infectious disease was ≥4 years compared with children <4 years old if they also were from a small family (one sibling or none; ref. 12). The disparate results may be related to the variation in study measurements used to test the association between early childhood infections and NHL and/or the underlying differences in childhood vaccine regimens. The timing, number, and frequency of other infections that commonly occur during childhood (colds, diarrheal conditions) may be a better marker for early childhood exposure to infectious agents because they usually are prevented by sanitation and hygiene rather than through immunizations. However, the potential for recall bias associated with specific occurrences of common infections as a young child could have been substantial in a case-control study.

Residential characteristics of childhood also may be a proxy for childhood exposure to infectious agents but seldom have been evaluated in NHL studies. An early study of NHL found no association between urban versus rural childhood origin (22), a result that is consistent with a recent study that evaluated early childhood rural living (11). Two studies that analyzed sleeping accommodations reported no associations between NHL and sharing a bed or bedroom as child (10, 11). Similar to our results for residential crowding at age 8 years, one study also reported that living in a household with six or more people when young (≤3 years old) was associated with a somewhat increased risk for NHL (11). To be consistent with the hygiene hypothesis, we would have expected risk of NHL to decrease with increased crowding. When combined, the variable results from these few studies provide little evidence that residential crowding during childhood is related to adult-onset NHL. Given that residential history effects were not consistent with infectious disease results in our study or in the German study (the only studies that have published data for both factors), it seems that these residential factors may not be a good proxy for childhood infectious disease exposures.

Our results showed an inverse association between NHL and allergic rhinitis that is consistent with our earlier results (6, 23, 26) and with the majority of studies that have investigated the association between NHL and allergies or atopy (10, 11, 38, 41, 44, 45). Although the results have varied by specific allergen and some studies have reported null effects (40, 46), the overall evidence supports an inverse association between NHL and allergic conditions. It has been suggested that allergies/atopy may result in more efficient immune function because of allergy-related immune conditioning. Additionally, histamine has antitumor effects and its release during an allergic response may partly explain the observed associations between allergies and NHL (47).

Studies that have examined farm exposure and NHL have yielded inconsistent results (24, 48-51) with exposures to pesticides, herbicides, solvents, and animal pathogens all suggested as possible risk factors. We reported reduced risk estimates for those who had lived on a farm for <2.5 years and with older age when first lived on a farm. Consistent with the hygiene hypothesis, control participants who first lived on a farm when ≤1-year old were less likely to have reported a history of allergic rhinitis. Results summarized in a recent review showed that children on farms, especially those with farm-animal contact, seemed to have a decreased risk for allergies possibly due to endotoxin exposures, whereas adult farmers' allergies were exacerbated by endotoxin exposures (as reviewed in ref. 52). Exposures on a farm are complex and often not mutually exclusive. It is possible that the farm-related early childhood exposures associated with reduced risk for immune-related conditions are negated in adult farmers who may be at increased risk of NHL after having accumulated years of adverse agricultural-related exposures (51, 53). Studies that incorporate methodology to accurately assess the various exposures and conditions that occur among farmers and farm residents over time are needed to clarify the role that farm-related exposures play in risk of NHL.

The present study has examined factors related to the hygiene hypothesis that have been examined infrequently in relation to NHL. The large study size and extensive information collected on residential history and infections provided adequate statistical power for most analyses. However, potential for recall bias can be a problem in case-control studies and we relied on self-reported information about conditions and exposures that occurred decades before diagnosis or interview. Because childhood infections are difficult to recall, particularly age of occurrence, we included only well-known childhood infections and used broad age categories (8 and 15 years) in our analyses. Differential recall of factors not known to be related to NHL, such as childhood infectious diseases, allergies, and residential characteristics, would be unlikely. To moderate potential selection and participation biases, all incident cases of NHL were identified by the cancer registry using rapid case ascertainment with random-digit dial used to identify controls from the same target population as patients. All interviews used a standardized questionnaire, were in-person, and were conducted by experienced survey workers to decrease response bias. Limitations include the potential for confounding and results due to chance. Sibling data that included stepbrothers and stepsisters, data from other sources of exposure such as playmates, baby sitters, day-care environment, public transportation, and information about other childhood infections may have provided a more complete assessment of childhood exposures to infectious agents. However, our results are somewhat consistent with other studies and provide insight into the potential for exposure to infectious agents in the household setting.

The Th1/Th2 paradigm proposed in the hygiene hypothesis is likely an overly simplistic model for development of immune-related diseases (17). The associations between infections, allergies, and NHL in this study population provide some support for a role of immune mechanisms as outlined by the hygiene hypothesis and by the alternative T-cell regulation hypotheses. Further research to test alternate hypotheses that involve T-cell regulation and consider specific types of infectious exposures are needed to clarify the effect of childhood infections and allergy on immune conditions (17, 54). The interplay between these exposures provides a basis for further research into the biology of immune function as it relates to infections, atopy, and NHL using larger pooled data sets, such as in the InterLymph Consortium.

We thank Dr. Ronald F. Dorfman at Stanford University for the pathology review.