Association of aspirin and other non-steroidal anti-inflammatory drug use with incidence of non-hodgkin lymphoma
Non-steroidal anti-inflammatory drugs (NSAIDs), including aspirin, seem to have chemopreventive properties against several types of cancer, particularly colon cancer. Persons with rheumatoid arthritis, an autoimmune disease for which NSAIDs are used commonly, have been reported to be at lower risk of colon cancer but at elevated risk of non-Hodgkin lymphoma (NHL), raising the possibility that NSAIDs may be a risk factor for NHL. We evaluated the association of use of NSAIDs, arthritis history, and risk of NHL in a prospective cohort of 27,290 postmenopausal women from the state of Iowa. The frequency of use of aspirin and of other NSAIDs excluding aspirin (e.g., ibuprofen), as well as a physician diagnosis of rheumatoid arthritis (RA) or osteoarthritis (OA), were self-reported on a questionnaire mailed in 1992. The incidence of NHL was ascertained through annual linkages to the Iowa SEER Cancer Registry. Relative risks (RR) and 95% confidence intervals (CI) were estimated using Cox proportional hazards regression. Through 7 years of follow-up, 131 cases of NHL were identified. Compared to women who did not use either aspirin or other non-aspirin NSAIDs, women using aspirin exclusively (RR = 1.71; 95% CI = 0.94–3.13), non-aspirin NSAIDs exclusively (RR = 2.39; 95% CI = 1.18–4.83), or both types of drugs (RR = 1.97; 95% CI = 1.06–3.68) were at increased risk of NHL. A diagnosis of RA (RR = 1.75; 95% CI = 1.09–2.79), but not OA (RR = 1.06; 95% CI = 0.67–1.68), was associated with risk of NHL, but the positive association of use of aspirin and other NSAIDs with NHL was independent of RA history. Multivariate adjustment for other NHL risk factors only attenuated slightly these associations, whereas exclusion of cases occurring during the first 2 years of follow-up strengthened the associations. These data suggest that use of NSAIDs, either aspirin or other non-aspirin NSAIDs, are associated positively with risk of NHL, and that this association is independent of RA history. © 2003 Wiley-Liss, Inc.
Nonsteroidal anti-inflammatory drugs (NSAIDs), both aspirin and non-aspirin drugs including over-the-counter (e.g., ibuprofen) and prescription (e.g., Naprosyn, Clinoril) agents, have gained attention as chemopreventive agents against the development of colorectal cancer, and possibly against other cancers including stomach, breast, lung, pancreas and ovary.1 Although the exact mechanistic basis underlying these observations remains unknown, it is thought that the putative chemopreventive properties of these agents are mainly due to their ability to inhibit the cyclooxygenase (COX) activity of prostaglandin G/H synthase isoforms 1 and 2 (commonly referred to as COX-1 and COX-2), although COX independent mechanisms have been proposed.2 NSAIDs vary in their ability to inhibit COX-1 and COX-2. Aspirin is relatively selective for COX-1, whereas older, commonly used NSAIDs, including ibuprofen, sulindac and indomethacin, inhibit both COX isoforms to the same extent. In contrast, newer selective NSAIDs (e.g., celecoxib) specifically inhibit COX-2. COX is the rate-limiting step in the conversion of arachidonic acid to prostaglandin H2, which can then be converted to a variety of tissue specific prostanoids. Elevated prostaglandins may promote carcinogenesis through stimulation of cell proliferation and suppression of immune response.1, 2 COX may also act as initiators of carcinogenesis through direct carcinogen activation and through the production of carcinogenic metabolites such as malondialdehyde, which is produced from the metabolism of prostaglandin H2.1
Rheumatoid arthritis patients, who frequently use both aspirin and other non-aspirin NSAIDs in high doses and for a long duration, have a decreased risk of colorectal cancer, and possibly other digestive tract cancer as well as female breast cancer.1, 3 These patients, however, also have an increased risk of several lymphatic or hematopoietic cancers, with the most consistent finding for an increased risk of non-Hodgkin lymphoma (NHL).1, 3 The increased risk of NHL in rheumatoid arthritis patients is thought to be due to either treatment with cytotoxic or immunosuppressive agents or from immune effects of the disease.3, 4 Independent of rheumatoid arthritis status, only 4 epidemiologic studies have assessed aspirin or other NSAIDs with risk of NHL, and results have been mixed, with null,5, 6 positive7 and inverse8 associations reported.
Because of the limited data available, we evaluated the association of aspirin and non-aspirin NSAID use with risk of NHL in a large, prospective cohort of older women. In addition, we evaluated the association of rheumatoid arthritis and osteoarthritis with risk of NHL.
MATERIAL AND METHODS
Iowa Women's Health Study cohort
Our study was reviewed and approved by the Human Subjects Protection committees at the University of Iowa and the University of Minnesota. Details on the Iowa Women's Health Study design specific to NHL have been previously published.9, 10 In 1986, a questionnaire was returned by 41,836 randomly selected women who were 55–69 years of age and had a valid Iowa driver's license in 1985 (42.7% response rate). There were only minor demographic differences between respondents and non-respondents to the baseline survey11 and, compared to non-respondents, respondents have had somewhat lower incidence and mortality rates for smoking related cancers.12
Follow-up questionnaires were mailed in 1987 (91% response rate), 1989 (89% response rate), 1992 (83% response rate) and 1997 (79% response rate) to ascertain vital status and to update residence and risk factor information. Residence in Iowa was also update by periodic linkages to the Iowa driver's license database. Vital status and NHL incidence in the cohort was ascertained through annual linkages to the Iowa death certificate files and the Iowa Cancer Registry files. Registry staff collect cancer data including identifying information, tumor site, morphology, histologic grade, and extent of disease on all persons who were Iowa residents at the time of their diagnosis. The Iowa Cancer Registry is part of the National Cancer Institute Surveillance's Epidemiology and End Results (SEER) program.13 NHL was defined based on International Classification of Diseases-Oncology (ICD-O) second edition14 codes 9590-9595 and 9670-9717, which is based on the Working Formulation15 and does not include chronic lymphocytic leukemia (9803) as part of the definition of NHL. Deaths among survey non-respondents and emigrants from Iowa were identified by linkage to the National Death Index. Participants were linked by a combination of social security number, first, last, and maiden name, birth date and zip code.
Risk factor assessment
NSAID use was first ascertained in this cohort on the 1992 survey. The respondents were first asked how often they took aspirin, and then they were given examples (Bufferin™, Anacin™, enteric-coated aspirin, Ecotrin™ and Excedrin™). Respondents were specifically asked to exclude use of acetaminophen and Tylenol™. Respondents were next asked how often they took other non-steroidal anti-inflammatory drugs or arthritis medicines, and they were also given examples (ibuprofen, Advil™, Nuprin™, Motrin™, Naprosyn™, Feldene™ and Clinoril™). On this question, respondents were specifically asked to exclude use of aspirin, acetaminophen, Tylenol™, prednisone, cortisone and Deltasone™. The categories for frequency of use for both questions were the following: never, <once per week, once per week, 2–5 times per week, and 6 or more times per week. Doses were not ascertained.
Marital status, height, weight, adult onset-diabetes, history of blood transfusion, smoking status, use of alcohol, use of estrogen replacement therapy and history of rheumatoid arthritis, osteoarthritis, migraine headaches or cardiovascular disease was ascertained on the 1992 questionnaire, supplemented with data from the baseline (1986) questionnaire for some of the variables. Farm vs. non-farm residence and consumption of red meat and fruit were determined solely from the baseline questionnaire.
Before data analysis, we excluded women with a self-reported history of cancer or cancer chemotherapy on the baseline (1986) questionnaire (n = 3,902) and women who developed any cancer except non-melanoma skin cancer (n = 2,837). We further excluded women who died between baseline and the 1992 survey or women who were alive but did not respond to the 1992 survey (n = 7,728), and women with missing data for both of the NSAIDs questions (n = 79). This left 27,290 women available for analysis.
The following age time-scales were constructed for each woman: 1) age at receipt of the 1992 survey; and 2) age at NHL diagnosis, emigration from Iowa, death or December 31, 1999 (whichever came first). Age at emigration was based on the date moved from Iowa (if known) or the midpoint of the interval between the last known date in Iowa and the date identified outside of Iowa. Out migration is approximately 1%/year.
Aspirin and other non-aspirin NSAID use was categorized into the response categories provided on the survey, with the exception that the <once per week and once per week categories were combined to provide more stable risk estimates. Relative risks (RR), along with 95% confidence intervals (CI), were estimated as measures of association for the exposures of interest with NHL incidence using Cox proportional hazards regression16 with age as the time scale.17 Other risk factors for NHL in this cohort, including age, marital status, farm residence,18 transfusion history,9 adult-onset diabetes,9 alcohol use,19 cigarette smoking,20 red meat and fruit intake21 and use of hormone replacement therapy22 were included in multivariate models as potential confounders. All analyses were carried out using SAS (SAS Institute, Cary, NC) and SPLUS (Mathsoft, Seattle, WA).
The mean age in 1992 of the 27,290 women in the at-risk cohort was 67.5 years, and over 99% were white. Twenty-eight percent of the cohort reported never using aspirin, 33% used aspirin once a week or less, 18% used aspirin 2–5 times per week, and 21% used aspirin 6 or more times per week. Sixty percent reported never using other non-aspirin NSAIDs, whereas 18% used these agents once a week or less, 8% used them 2–5 times per week, and 13% used them daily. With respect to combined use of these agents, 17% of the women reported taking neither, whereas 29% used both. Aspirin users were just as likely to use other NSAIDs (40%) as non-aspirin users (38%). Use of both aspirin and other NSAIDs was strongly and positively associated with history of rheumatoid arthritis and osteoarthritis (Table I). There were weaker but positive associations of both aspirin and other NSAIDs with history of migraine headache, history of cardiovascular disease, and ever having had a blood transfusion. Body mass index was positively associated with the use of other NSAIDs but not with use of aspirin. Other NHL risk factors in this cohort, including consumption of red meat and fruit, marital status, adult-onset diabetes, cigarette smoking, use of alcohol and use of estrogen replacement therapy were not associated with either aspirin or other NSAID use.
Table I. Mean or Percent Distribution of Selected Characteristics of the Cohort by Frequency of Use of Aspirin and Non-Aspirin NSAIDs
|Mean age in 1992||67.6||67.2||67.4||67.8||67.7||67.0||67.3||67.5|
|Red meat (mean servings/week)2||7.9||8.1||7.9||7.8||8.0||7.8||8.4||8.0|
|Fruit (mean servings/week)2||19.1||18.7||18.7||19.0||18.9||18.6||18.8||19.0|
|Mean body mass index (kg/m2)||27.2||26.8||26.9||27.5||26.6||27.2||27.7||28.9|
|Lives on farm2||20||22||22||19||21||20||21||19|
|History of adult onset diabetes3||9||6||6||11||8||7||8||11|
|Ever received a blood transfusion||29||24||26||33||26||27||31||36|
|Any use of alcohol||8||10||11||10||9||11||10||9|
|Current user of HRT||15||17||18||19||15||19||19||23|
|History of rheumatoid arthritis||12||7||9||15||7||8||16||28|
|History of osteoarthritis||20||12||15||24||11||14||24||45|
|History of migraine headaches||9||7||11||13||8||10||13||14|
|History of cardiovascular disease3||15||9||12||31||16||13||16||20|
From 1992–99 (195,573 person-years), 131 cases of NHL were identified in the cohort. The mean age at diagnosis was 72.9 years (range = 63–82 years). We first considered aspirin and other NSAID use independently. There was no association between aspirin use and risk of NHL, either in age or multivariate models, or after exclusion of the first two years of follow-up (Table II). In contrast, there was a suggestive positive association of other NSAID use with risk of NHL (RR = 1.40; 95% CI = 0.99–1.97), but there was no evidence for dose-response with increasing frequency of use. Multivariate adjustment only slightly attenuated the association. Because many NHL patients experience B-symptoms (i.e., fever, night sweats and weight loss) before their diagnosis, and this could lead to increased use of aspirin or NSAIDs, we ran additional analyses excluding cases diagnosed during the first 2 years of follow-up in the cohort (Table II); these exclusions slightly strengthened the associations.
Table II. Association of Aspirin and Other Nsaids With Risk of NHL, Iowa Women's Health Study, 1992–1999
|Aspirin use|| || || || || || || || |
| Frequency of use|| || || || || || || || |
|Non-aspirin NSAID use|| || || || || || || || |
| Frequency of use|| || || || || || || || |
|Combined use|| || || || || || || || |
| No use||13||33258||1||Reference||1||Reference||1||Reference|
| Aspirin only||56||84354||1.71||0.94–3.13||1.68||0.92–3.07||2.31||1.04–5.15|
| Non-aspirin NSAIDs only||19||20800||2.39||1.18–4.83||2.21||1.08–4.52||3.39||1.38–8.32|
We next assessed the joint association of aspirin and other NSAIDs on the risk of NHL, so that the reference group would have no use of either of these agents. Compared to non-users of both drugs, women using aspirin only (RR = 1.71; 95% CI = 0.94–3.13), NSAIDs only (RR = 2.39; 95% CI = 1.18–4.83), or both agents (RR = 1.97; 95% CI = 1.06–3.68) were all at increased risk of NHL (Table II). These associations were only slightly attenuated after multivariate adjustment. Exclusion of cases diagnosed during early follow-up strengthened these associations.
Our study sample was too small to formally evaluate associations with individual NHL subtypes. Among the cases, extra-nodal disease was equally common among non-users of both aspirin and other NSAIDs (30%) as it was users of any type of NSAIDs (32%). Among the major subtypes, distributions for no NSAIDs vs. any NSAIDs were similar for small lymphocytic (7 vs. 8%), follicular (14 vs. 21%), diffuse (61 vs. 58%), high-grade (3 vs. 6%) and unclassified (14 vs. 6%).
We next evaluated the association of several chronic diseases for which aspirin or other NSAIDs are often used with risk of NHL. A history of rheumatoid arthritis (RR = 1.75; 95% CI = 1.09–2.79) was positively associated with NHL, but a history of osteoarthritis (RR = 1.06; 95% CI = 0.67–1.68) or heart disease (RR = 0.98; 95% CI = 0.61–1.58) were not. When NSAID use and rheumatoid arthritis were included in the same model, both predicted NHL incidence (Table III). The association of class of NSAIDs used with NHL was similar in women with and without a history of rheumatoid arthritis (Table III), although many of the risk estimates were not significant statistically, perhaps due to small numbers.
Table III. Association of Joint Use of Aspirin and Other NSAIDs With Risk of NHL, After Adjustment or Stratification by Rheumatoid Arthritis History, Iowa Women's Health Study, 1992–1999.
|Non-aspirin NSAIDs only||17||2.24||1.06–4.71||13||2.34||1.05–5.22||4||2.53||0.28–22.7|
We found a suggestive positive association of non-aspirin NSAID use with risk of NHL, but there was no dose-response with frequency of use. This association strengthened after the exclusion of cases diagnosed during the first 2 years of follow-up. In contrast, there was no association for aspirin use when evaluated alone, although when use of aspirin and non-aspirin NSAIDs were jointly evaluated, aspirin users had an increased risk of NHL, particularly after the exclusion of early follow-up. Although rheumatoid arthritis was positively associated with NHL risk, the association of class of NSAID (aspirin vs. non-aspirin) with NHL seemed to be independent of rheumatoid arthritis.
Strengths of our study include the prospective design, use of a SEER cancer registry to identify cases, and the ability to assess confounding by a variety of NHL risk factors. One of the major limitations of these data is our limited exposure assessment. In particular, we did not have data on type, dose or duration of NSAID use. These details are often difficult to recall, however, given the multiple number of products on the market and the sporadic use of these agents. Nevertheless, these simple questions were sufficient to detect inverse associations between use of aspirin with risk of both breast and pancreatic cancer in this cohort.23, 24 We also did not have data on the reason for using these agents, and confounding by indication remains a potential explanation. We did have data on history of rheumatoid arthritis, osteoarthritis, migraine headache and heart disease, however, and these conditions did not confound the associations we reported. This cohort is unlikely to have had significant exposure to selective COX-2 agents, and the risk of NHL from these agents, if any, is undetermined. Another limitation is that we did not have a sufficient number of cases to evaluate NHL by subtypes, which may have distinct etiologies.25 We saw no striking differences for nodal vs. extranodal or NHL subtype distribution by NSAIDs history.
There are few data on the association of these agents with risk of NHL. Bernstein et al.7 found a positive association between use of aspirin and other pain relievers (including Tylenol™) and risk of NHL (RR = 1.96 for 13+ months of use compared to no regular use; 95% CI = 1.25–3.08) in a population-based case-control study conducted from 1979–82. In contrast, a more recent population-based case-control study conducted from 1988–95 by Holly et al.8 found an inverse association of use of aspirin or other NSAIDs (combined) for at least 4 consecutive weeks with risk of NHL (OR = 0.72; 95% CI = 0.56–0.91). Two hospital-based case-control studies reported no association of aspirin or other analgesics with NHL5 or lymphoma and leukemia combined.26 Of 2 cohort studies, both reported no association between aspirin use with lymphoma incidence (based on 48 cases)6 or lymphoma and hematopoietic cancer mortality combined (number of cases not reported).27
Rheumatoid arthritis patients often use a large amount of aspirin and non-aspirin NSAIDs, and one highly consistent finding from the follow-up of large cohorts of these patients (>1,000) is that they have a reduced risk of colorectal cancer (RR = 0.5–0.8) and an increased risk of NHL (RR = 1.9–4.1).4, 28, 29, 30 Although the inverse association reported for colorectal cancer is thought to be due to use of NSAIDs,1, 3 the positive association with NHL is often ascribed to the use of cytotoxic or immunosuppressive therapies in rheumatoid arthritis or to the disease process itself.3, 4 Of the therapies used in the management of rheumatoid arthritis patients, however, the data are strongest for the involvement of azathioprine and cyclophosphamide in NHL development,3 treatments for severe disease that are not used in most rheumatoid arthritis patients. Methotrexate, a commonly used treatment, does not seem to increase risk of lymphoma,3, 31, 32 although high doses of methotrexate may be associated with EBV-associated lymphoproliferative disease and Hodgkin lymphoma (reviewed in Mariette et al.32). In a small case-control study nested in a cohort of rheumatoid arthritis patients, inflammatory activity was strongly and positively associated with risk of lymphoma, but there was no association of drug treatment.33 No data on the latter results were provided, however, and it is unclear to what extent use of aspirin or other NSAIDs was collected. Another study found that NHLs that develop in RA patients do not share characteristics with lymphomas generally seen in immunosuppressed patients.34 Our data raise the hypothesis that use of NSAIDs might explain some of the association of rheumatoid arthritis with NHL risk.
Hypothesized mechanisms for the chemopreventive properties of NSAIDs, at least for colon cancer, include the inhibition of cell proliferation, induction of apoptosis, blocking of angiogenesis, decreased activation of carcinogens by COX, and enhancement of immune surveillance.1 The latter mechanism is the most poorly understood, but potentially of most relevance to NHL, a cancer of the immune system. Prostaglandins, particularly prostaglandin E2, have long been know to have important immunoregulatory properties, and it is generally thought that drugs that decrease prostaglandin E2 synthesis enhance cellular immune responses.35 Suppression of cellular immunity, not enhancement, however, would be predicted to be associated with NHL. Nevertheless, the in vivo effects of NSAIDs on the cellular and humoral immune system are currently incomplete and often contradictory, in part due to the experimental systems utilized and the complexity of the human immune system.35 Furthermore, the immunomodulatory effects of NSAIDs may act through COX-independent mechanisms, including direct effects on cytokines36 and transcription factors.37 Of potential interest, the constitutive activation of NF-κB has been associated with inflammatory diseases such as rheumatoid arthritis, and aspirin and other NSAIDs are thought to elicit some of their anti-inflammatory effects by blocking activation NF-κB.38 NF-κB is also essential, however, in the development of Th1-type (cellular) immune response,38 and this raises the hypothesis that the interference of NF-κB function by NSAIDs may predispose to B cell malignancy.
In conclusion, our data raise the concern that NSAIDs, either aspirin or non-aspirin agents, increase the risk of NHL. This association is independent of a history of rheumatoid arthritis. If confirmed, the mechanistic basis to this observation warrants closer evaluation.
We thank M.J. Eversman for assistance in preparing this manuscript. The Iowa Women's Health Study was supported the National Cancer Institute Grant R01 CA39741. Dr. Cerhan was supported in part by a National Cancer Institute Preventive Oncology Award (K07 CA64220).