Our objective in this study was to evaluate whether the useof hormone replacement therapy (HRT) is associated with non-Hodgkin lymphoma (NHL) or chronic lymphocytic leukemia (CLL). A cohort of 37,220 Iowa women ages 55 to 69 years in 1986 with no history of prior cancer was linked annually to a population-based cancer registry. Through 1998 (13 years of follow-up), 258 incident cases of NHL were identified, including 135 cases of diffuse NHL, 58 cases of follicular NHL, and 31 cases of small lymphocytic NHL. In addition, 63 cases of CLL were identified. Current and former use of HRT (primarily estrogen) and other cancer risk factors were self-reported on the baseline (1986) questionnaire. Compared with never users of HRT at study baseline, current [multivariate relative risk (RR), 1.4; 95% confidence intervals (CIs), 0.9–2.0) but not former (RR, 1.1; 95% CI, 0.8–1.4) users were at increased risk of NHL after adjustment for age and other confounding factors. This association was seen only in nodal NHL [RRcurrent, 1.5 (95% CI, 1.0–2.4); RRformer, 1.1 (95% CI, 0.8–1.6)] and was not apparent for extra-nodal sites. Of the common subtypes, there was a strong positive association with follicular NHL [RRcurrent, 3.3 (95% CI, 1.6–6.9); RRformer, 2.6 (95% CI, 1.4–4.7)], and women who were current users for more than 5 years had the highest risk (RR, 3.9; 95% CI, 1.8–8.6). There was no association with diffuse or small lymphocytic NHL, or with CLL. Most of the follicular NHLs were nodal (88%), and exclusion of extra-nodal sites slightly strengthened the association with HRT. For diffuse NHL, 64% of the cases were nodal, and there was no association of HRT with either nodal or extra-nodal sites. These data suggest that HRT is a risk factor for follicular NHL but not for diffuse or small lymphocyte NHL or CLL.
HRT3 is classified as a Group 1 carcinogen by the IARC based on both animal and human data (1) . Estrogen is also known to have immunomodulatory effects (2) , and NHL is a tumor the development of which is sensitive to immunological function (3) . Animal studies have clearly shown that estrogen administration induces endometrial, breast, and ovarian tumors, and there are data showing that s.c. estrogen can also induce lymphoma (4) . In humans, HRT is associated with a substantially increased risk of endometrial cancer (when estrogen is unopposed by progestins) and a moderately elevated risk of breast cancer (5) . HRT also appears to be associated with a reduced risk of colorectal cancer, whereas data for ovarian, liver, and thyroid cancer as well as cutaneous melanoma are equivocal or show no association (5) .
There are few data regarding HRT and NHL. A case-control study reported a positive association of estrogen replacement therapy and NHL (6) , whereas another case-control study of intermediate and high grade NHL found a weak inverse association that was not statistically significant (7) . The early results (7-year follow-up) from the Iowa Women’s Health Study cohort found no association with HRT use (8) . We now report more extended follow-up of the Iowa cohort (13 years and nearly double the number of cases). In addition, we report results for common NHL subtypes as well as for CLL, which is currently considered to be part of the spectrum of NHL (9) .
Subjects and Methods
The Iowa Women’s Health Study Cohort.
This study was reviewed and approved by the institutional review boards at the University of Iowa and the University of Minnesota. Details on the Iowa Women’s Health Study design specific to NHL and CLL have been published previously (8 , 10) . Briefly, in January 1986, a mailed questionnaire was returned by 41,836 randomly selected women who were ages 55–69 years and had a valid Iowa driver’s license in 1985 (42.7% response rate). There were only minor demographic differences between respondents and nonrespondents to the baseline survey (11) and, compared with nonrespondents, respondents have had somewhat lower cancer incidence and mortality rates for smoking-related cancers (12) .
The baseline questionnaire inquired about current and past HRT, reproductive history, anthropometrics, and dietary and other lifestyle factors. HRT was determined by the following questions: “Have you ever used pills other than birth control pills which contain estrogen or other female hormones (e.g., at the change of life or menopause, after surgery, or at any other time)?” Response categories were “yes, currently,” “yes, but not currently,” and “never.” Subjects responding yes were then asked “How long did you take estrogens or other female hormone pills (other than birth control pills)?” Response categories for duration of use were “one month or less,” “2–6 months,” “7–12 months,” “13 months-2 years,” “3–5 years,” or “more than 5 years.” Names and doses of pills were not obtained.
Vital status and cancer incidence in the cohort was ascertained through 13 years of follow-up (1986–1998). Follow-up questionnaires were mailed in 1987, 1989, 1992, and 1997 to ascertain vital status and address changes. Deaths were also ascertained by annual linkage to Iowa death-certificate data, supplemented by linkage to the National Death Index for survey nonrespondents and emigrants from Iowa. Out migration has been estimated at 1%/year.
Cancer incidence, except for nonmelanoma skin cancer, was ascertained by annual linkage to the Iowa Cancer Registry, part of the National Cancer Institute’s SEER program (13) . All of the participants were linked by a combination of social security number; first, last, and maiden names, birth date, and postal (ZIP) code. The Iowa Cancer Registry collects cancer data, including identifying information, tumor site, morphology, histological grade, and extent of disease, on all persons who were Iowa residents at the time of their diagnosis. All of the tumor site and morphology data were derived from pathology reports of the diagnosing pathologist, and there was no centralized review of the tumor material.
Topographic and morphological data were coded using the ICD-O, second edition (14) , which is based on the Working Formulation (15) . The histological subtypes of NHL were grouped according to Groves et al. (16) into the following subtypes: small lymphocytic, follicular, diffuse, high grade, and peripheral T-cell (ICD-O codes given in Table 1⇓ ). This system is a slight modification of the Working Formulation for use in epidemiological studies. In addition, the primary site was classified as nodal (lymph nodes, tonsil, and spleen) or extranodal (all other sites that are not primary lymphoid organs) according to the ICD-O definition (14) . CLL (ICD-O morphology code 9823) was also analyzed, but as a distinct end point (i.e., it was not aggregated with small lymphocytic NHL).
Before data analysis, we excluded women with a self-reported history of cancer or cancer chemotherapy on the baseline (1986) questionnaire (n = 3,902) to provide a cancer-free at-risk cohort. We further excluded women who were premenopausal (n = 545), had missing data on the HRT question (n = 167), or who had their diagnosis at or before the date of return of their questionnaire (n = 2). This left a final cohort of 37,220 available for analysis. Each woman accumulated person-years of follow-up from the date of receipt of the 1986 baseline questionnaire until the date of NHL diagnosis, date of emigration from Iowa, or date of death; if none of these events occurred, person-years were accumulated through December 31, 1998.
Age-adjusted and multivariate RRs, along with 95% CIs, were calculated as measures of association between HRT and NHL incidence, and were estimated using Cox proportional hazards regression (17) . Other risk factors for NHL in this dataset, including age, marital status, farm residence (18) , transfusion history (8) , adult-onset diabetes (8) , alcohol use (19) , cigarette smoking (20) , and red meat and fruit intake (21) were included in the multivariate models as potential confounders. Although not associated with the risk of NHL in this cohort, in additional models, we adjusted for education and age at menopause (22) .
The mean age of the 37,220 women in the at-risk cohort was 61.7 years, and more than 99% were white. During 441,409 person-years of follow-up (1986–1998), 258 women developed NHL. The mean age at diagnosis was 69.7 years (range, 57–81 years). Approximately 70% of the NHL cases were nodal, and of the extra-nodal sites, the stomach was the most common (Table 1)⇓ . The most common NHL subtype was diffuse NHL (52%), followed by follicular (23%) and small lymphocytic (12%) NHL; only 9% of the cases could not be subclassified.
Characteristics of women who used HRT are provided in Table 2⇓ . Compared with never users of HRT, current users, on average, were slightly younger, had a lower body mass index, and consumed fewer servings of red meat. They were also more likely to have higher educational attainment, to be currently married, to have ever had a blood transfusion, to have used oral contraceptives, and to drink alcohol, whereas they were less likely to have a history of adult-onset diabetes. However, the differences across groups were generally small. There were few notable differences in smoking history or the proportion living on a farm.
As shown in Table 3⇓ , compared with never users of HRT at baseline, the incidence of NHL was slightly elevated in former users (RR, 1.1) and somewhat more elevated in current users (RR, 1.4), although neither estimate was statistically significant. When evaluated by duration of use, the greatest risk of NHL was for current users at baseline with ≤5 years of use (RR, 1.6); RRs in the other categories were ≤1.2. None of these estimates was statistically significant, and adjustment for major NHL risk factors in these data did not alter these estimates.
With respect to anatomical site, there was a statistically significant positive association with current HRT use and the risk of nodal NHL (RR, 1.6), which changed little after multivariate adjustment (RR, 1.5). Again, elevated risk was mainly among current users with ≤5 years of use (RRmultivariate, 2.1). There was essentially no association of HRT with extra-nodal NHL.
We next evaluated common subtypes of NHL. For follicular NHL, risk was elevated in former (RR, 2.7) and current users (RR, 3.3), and this was not attenuated in multivariate adjustment. When evaluated by the duration of use, risk was generally higher with longer use, and the highest risk was seen for current users of >5 years (RR, 3.9). Most (88%) of the follicular NHLs were nodal. When we analyzed only nodal follicular NHL, the multivariate association with former (RR, 2.7; 95% CI; 1.4–5.1) and current (RR, 3.7; 95% CI; 1.7–8.0) HRT use strengthened slightly.
Of the other common subtypes of NHL, there was no consistent association of HRT use with the risk of diffuse NHL, small lymphocytic NHL, or CLL. An analyses of duration of use revealed no consistent trends. Of the diffuse NHLs, 64% were nodal. There was no association between HRT use and the risk of either diffuse nodal [(RRcurrent, 1.0 (95% CI 0.5–2.1); RRformer, 0.9 (95% CI, 0.5–1.4)]] or diffuse extra-nodal [RRcurrent, 1.0 (95% CI, 0.4–2.6); RRformer, 1.0 (95% CI, 0.4–2.6] NHL.
In all of the multivariate models, additional adjustment for education and age at menopause did not alter the risk estimates (data not shown).
We found a weak positive association between current use of HRT and the risk of NHL overall. Among the common subtypes, HRT use was associated with follicular NHL, particularly nodal follicular NHL, but not with diffuse or small lymphocytic NHL nor with CLL. Longer duration of use was associated with a greater risk of follicular NHL among both former and current HRT users, and the association with follicular NHL was not confounded by several key demographic, medical history, lifestyle, or dietary risk factors.
Strengths of this study include the prospective cohort study design in a well-defined, community-based population, case ascertainment using a SEER cancer registry, inclusion of the full spectrum of NHL cases (i.e., aggressive cases that often do not get enrolled into case-control studies), and virtually complete follow-up of the cohort. Limitations include an abbreviated assessment of HRT based on self-report, including no data on the type of preparation used, nor precise length of use. However, even with the limited HRT assessment used in the study, HRT use was inversely associated with total and coronary heart disease mortality, colon cancer incidence, and hip fracture, and was positively associated with endometrial and breast cancer incidence (23) . On the basis of other data collected around the time of the baseline survey, Folsom et al. (23) have estimated that no more than 20% of the current users, and a smaller proportion of former users, were also using a progestin in combination with estrogen. Thus, the issue of combination therapy and the risk of NHL remains open.
Other limitations of this study include relatively small sample sizes for some subtypes, and the lack of data on other subtypes of potential interest (e.g., mantle cell lymphoma; MALT lymphoma; high-grade lymphoma; peripheral T-cell lymphomas). Subtype classification relied on reports from the local pathologist, and there was no central review of the cases. In addition, our classification of subtypes relied on the SEER ICD-O second-edition codes, which are based on the Working Formulation; however, this does not necessarily translate into the more current WHO Classification system (9) . In particular, the diffuse lymphomas are likely to include some lymphomas of T-cell origin, as well as other more uniquely defined subtypes. However, follicular NHL appears to be reliably classified by pathologists (24) , and this subtype has seen little modification with adoption of the revised European-American lymphoma (REAL) classification (25) and WHO systems (9) . In addition, we based our subtype categorization on the recent summary data, which suggests that these are reasonable groupings of the SEER data for etiological epidemiology studies (16) .
Besides the early results form this cohort (8) , which showed no association with the risk of NHL, only two other studies have addressed the association of HRT and NHL. Bernstein and Ross (6) published preliminary results from a population-based case-control study (n = 337 female cases and matched controls) conducted from 1979 to 1982 in Los Angeles. Compared with women who had never used estrogen replacement therapy, women who had ever used it were at a slightly increased risk of NHL (OR, 1.29; 95% CI, 0.94–1.79), mainly attributable to women who had used it 13 or more months (OR, 1.58; 95% CI, 1.09–2.29). These estimates were not adjusted for other risk factors, and the distribution of cases by subtypes (or a subtype analysis) was not reported. In a case-control study of intermediate- and high-grade B-cell NHL, conducted in Los Angeles from 1989 to 1992 (7) , ever use of HRT was inversely associated with risk (OR, 0.60) among HIV-negative women, although this estimate was not statistically significant (95% CI, 0.31–1.18). Additional adjustment for age at menopause did not alter this association. The proportion of intermediate to high-grade cases was not reported; therefore, we cannot directly compare these results with our data. In our study, the use of HRT for >5 years among both former and current users was associated with RRs of <1 for diffuse NHL (all intermediate grade), although the estimates were based on small numbers and lacked precision. Thus, the association of HRT with intermediate- and high-grade NHL remains unresolved at this time. To our knowledge, no prior study has evaluated HRT use and the risk of follicular or small lymphocytic NHL or CLL.
Correlates of endogenous estrogen, including reproductive (22 , 26, 27, 28, 29) and menstrual history (22) factors, have not shown a consistent association with NHL in the few studies to evaluate these associations. No data are currently published evaluating these factors with specific subtypes of NHL.
Although we cannot rule out the possibility that the specific association of HRT use with follicular NHL may be a chance finding, this association is supported by several lines of evidence. Overall, the incidence rate of NHL is about 50% higher in men than women (in both Iowa and in the United States), and this is true for nearly all subtypes except follicular NHL, for which the rates are approximately equal (13) . In Iowa from 1978 to 1997, the age-incidence curve for follicular NHL among women is similar to that seen for female breast cancer (Fig. 1)⇓ , in contrast to the age-incidence curves for diffuse, small lymphocytic leukemia and CLL (which continue to increase rapidly with age) or high-grade NHL (which shows neither pattern; Fig. 2⇓ ). These patterns are similar for all SEER data for white females for 1993–1997. The age-incidence pattern for female breast cancer (which follicular NHL follows) has been widely interpreted as suggesting a role for ovarian hormones in this malignancy (30) .
Evaluation of ovarian hormones (particularly estrogen) in the etiology of NHL generally, or follicular NHL specifically, is limited. However, estrogen clearly plays a role in immunological function (31, 32, 33, 34, 35) , and immunological function is thought to be important in lymphomagenesis (3) . For example, estrogen is known to influence lymphopoiesis, Th1/Th2 balance, antigen presentation, lymphocyte homing, and cytokine production (2 , 33) . Estrogen at physiological levels is considered to be immunostimulatory, in part by increasing Th1 cytokine production (e.g., secretion of IL-2, IFN-γ, and tumor necrosis factor β), which is associated with an augmentation of the cellular immune system and the suppression of the humoral immune system. At higher concentrations (e.g., in pregnancy), there is a shift toward Th2 cytokine production (e.g., secretion of IL-4, IL-5, IL-6, and IL-10), which is associated with the suppression of the cellular immune system and stimulation of humoral immunity (33 , 36 , 37) . However, the exact roles and mechanisms of estrogen in the immune system are not completely understood (2) , in part because estrogen effects vary between sexes and across species (32) , and even across strains of mice (38) . Even less is known about the role of estrogen in lymphomagenesis. Both estrogen-receptor-mediated and non-receptor-mediated mechanisms may be relevant (39 , 40) .
In conclusion, we found a weak positive association of HRT with NHL overall, and a strong positive association for follicular NHL, particularly for nodal follicular NHL. The association with follicular NHL is novel, and requires replication. Future studies should assess HRT formulation and duration of use and should classify cases by the current WHO system (9) . Further research on the role of estrogen in the development and progression of NHL, and follicular NHL in particular, may lead to new insights into the etiology and treatment of this disease.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 Supported in part by the National Cancer Institute Grant R01 CA39741 (to The Iowa Women’s Health Study). J. R. C. was supported in part by a National Cancer Institute Preventive Oncology Award (K07 CA64220).
↵2 To whom requests for reprints should be addressed, at Department of Health Sciences Research, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. Phone: (507) 538-0499; Fax: (507)266-2478; E-mail:
↵3 The abbreviations used are: HRT, hormone replacement therapy; NHL, non-Hodgkin lymphoma; CLL, chronic lymphocytic leukemia; SEER, Surveillance, Epidemiology and End Results; ICD-O, International Classification of Diseases for Oncology; RR, relative risk; CI, confidence interval; OR, odds ratio; IL, interleukin.
- Received February 4, 2002.
- Revision received July 26, 2002.
- Accepted August 29, 2002.