Aspirin and Other Nonsteroidal Anti-Inflammatory Drugs and Risk of Non-Hodgkin Lymphoma

There is accumulating evidence that aspirin reduces the incidence and mortality of several cancers (1, 2), beyond its established preventive effect on colorectal cancer (3). However, results of epidemiologic studies of nonsteroidal anti-inflammatory drugs (NSAIDs) in relation to non-Hodgkin lymphoma (NHL) are inconsistent. Positive, negative, and null associations have been reported, with some studies reporting an association with all NSAIDs, whereas others with just aspirin or nonaspirin NSAIDs (4–20). Most studies had limited statistical power, particularly for detailed analyses by frequency, duration, and type of NSAID use, and/or NHL subtype. Most previous studies were case control (13) and only 3 prospective studies had more than 135 cases (7, 19, 20), including one which examined only selective COX-2 inhibitors (7), and one which examined only aspirin (19). Summary measures from meta-analyses have been null (12–14), but, again, were based primarily on results from case-control studies, and were therefore subject to potential recall and selection bias. In addition, the meta-analyses were limited by the need to summarize across studies with widely varying definitions of NSAID use.

In this study, we examine the associations of 2 measures of NSAID use (frequency of current use, and long duration, regular use) with NHL incidence in the Cancer Prevention Study-II (CPS-II) Nutrition Cohort. This cohort is well suited to examine these associations because of its size, the availability of detailed, regularly updated information on aspirin and nonaspirin NSAID use, and the availability of NHL subtypes.

The CPS-II Nutrition Cohort (n = 184,188) is a prospective study conducted in 21 U.S. states and described in detail elsewhere (21). A mailed questionnaire in 1992/1993 collected sociodemographic, lifestyle, and medical history information. Biennial follow-up surveys beginning in 1997 collected updated exposure information and ascertained new cancer diagnoses. Participants were excluded from this analysis for nonresponse to follow-up surveys (n = 6,276), previous cancer diagnoses (n = 21,083), unknown NSAID information at enrollment (n = 7,134), and unverified hematopoietic cancer reports on the first follow-up survey (n = 125). Among the 149,570 participants in the analytic cohort, 1,709 incident NHLs were identified during follow-up. The majority of these cancers were self-reported and then verified by medical record (n = 989) or state cancer registry linkage (n = 333). An additional 186 NHL cases were identified through linkage with the National Death Index and 133 were subsequently verified (22). Finally, 201 cases were identified through the process of verifying another cancer reported by the participant. NHL subtypes (diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), and multiple myeloma) were defined using guidelines published by the International Lymphoma Epidemiology Consortium Pathology Working Group (23).

Information on use of aspirin and other NSAIDs was reported on mailed questionnaires in 1992/1993, 1997, 1999, 2001, 2003, and 2005. Historical information on use of aspirin (but not other NSAIDs) was also reported in 1982, when subjects were enrolled into the parent mortality cohort (24). The 1982 questionnaire asked about “times/month” of aspirin use. The 1992 and later questionnaires asked separately about aspirin and other NSAIDs. For each type, participants were asked about average days/month used, average number of pills taken on days used, and number of years used. We calculated a time-dependent variable for frequency in pills/month by multiplying days used per month by pills/day for each type of NSAID. For the primary analysis we considered a pill of low-dose aspirin the same as a pill of regular aspirin, but in a sensitivity analysis each low-dose aspirin pill (typically 81 mg) was counted as 1 quarter of a regular-dose aspirin pill (typically 325 mg). Participants who reported days/month they used a particular NSAID but did not report pills/day were assigned a value of 1 pill/day. The current number of pills/month of NSAIDs in each follow-up interval was categorized as 1 to <15, 15 to <30, 30 to <60, and 60+ pills/month and was updated at the time of each follow-up questionnaire (in 1997, 1999, 2001, 2003, 2005). In addition, because NSAID use for subclinical symptoms of NHL is of particular concern and the induction/latency period is unknown, we also calculated a lagged frequency variable defined by use reported on the next-to-most-recent questionnaire, rather than the most recent questionnaire. For example, use reported in 1992 was used to predict risk during the interval between the 1997 and 1999 questionnaires and use reported in 1997 was used to predict risk during the 1999–2001 interval. The first interval (1992–1997) was lagged by excluding the first 2 years of follow-up.

Analyses of duration of regular (≥30 NSAID pills/month) use of NSAIDs used the following categories: (a) never reported use (never users), (b) past or less than regular NSAID use only, (c) current regular NSAID use of <5 years, and (d) current regular NSAID use of ≥5 years. During the 1992–1997 follow-up interval, participants were categorized as having ≥5 years of regular NSAID use if they reported at least 5 years of use on their 1992 questionnaire and also reported regular NSAID use on both the 1982 and 1992 questionnaires. During the 1997 to 1999 follow-up interval, participants were categorized as having ≥5 years of regular NSAID use if they reported regular NSAID use on both the 1992 and 1997 questionnaires. During later follow-up intervals, participants were categorized as having ≥5 years of regular NSAID use if they reported regular NSAID use on the 3 most recent questionnaires (e.g. on the 1992, 1997, and 1999 questionnaires for the 1999–2001 interval). In each follow-up interval, participants who had not reported NSAID use on any previous questionnaire were categorized as never users, whereas participants who were neither never users nor current regular users were categorized as “past or less than regular use only.”

Both frequency and duration of regular NSAID use were calculated by summing pills per month for each type of NSAID to create a total NSAID variable. In addition, we also examined aspirin and nonaspirin NSAID use separately. Based on results from other studies, we could not rule out different associations of these 2 NSAID types with NHL risk. To ensure we were studying the association of current use of each type of NSAID without the influence of the other, participants were classified into mutually exclusive categories: never NSAID use, former NSAID use, current aspirin-only, current nonaspirin NSAIDs only.

Person-time was calculated as days between enrollment date and cancer diagnosis date, death date, date of last returned survey, or end of follow-up (June 30, 2007). Participants with missing NSAID information on a follow-up survey were censored during the subsequent follow-up interval but were reentered into the analysis if valid NSAID information was reported on the next survey. However, participants were permanently censored if NSAID information was missing on 2 consecutive surveys. In analyses of specific NSAID types (aspirin/nonaspirin), participants were censored on the date they reported use of the other type of NSAID. In addition, participants who self-reported a hematopoietic cancer that could not be verified were censored on the date of their last cancer-free questionnaire.

Extended Cox regression (25) was used to calculate time-dependent hazard ratios (HR) and 95% confidence intervals (CI) for the associations between NSAID use and NHL incidence overall, DLBCL, follicular lymphoma, CLL/SLL, and multiple myeloma. All models were stratified on enrollment age and used follow-up time as the time scale. Multivariable models included sex, family history of hematopoietic cancers, race, alcohol intake, education, smoking status, body mass index (BMI), sitting time, diabetes status, rheumatoid arthritis status (starting in 2001), cholesterol-lowering drug use, acetaminophen use, and postmenopausal hormone use.

On average, participants were aged 63 years at the start of follow-up and most identified as white (97%). In 1992, 43% of the participants reported never use of NSAIDs and 31% were regular users. Those who reported regular NSAID use in 1992/1993 were more likely to be slightly older, have a BMI ≥ 30 kg/m², be former smokers, drink at least 2 drinks daily, currently use cholesterol-lowering drugs, and to report diabetes than participants who never or only occasionally used NSAIDs (Table 1). And in 2001, regular NSAID users were more likely to report rheumatoid arthritis than other participants.

Total NSAID use was not associated with NHL risk when defined as frequency of current use or duration of regular use (Table 2). The hazard ratio (HR) for participants who reported currently taking ≥60 pills/month compared to never users was statistically significant (HR = 1.26, 95% CI, 1.04–1.53) but was attenuated (HR = 1.08, 95% CI, 0.88–1.32) in the lagged model (Table 2). For NHL subtypes, 60+ pills/month of any NSAID use compared to never use was associated with a higher risk of follicular lymphoma (HR = 1.87, 95% CI, 1.08–3.24). This association remained when the lagged NSAID variable was used (HR = 1.76, 95% CI, 1.04–2.98). The hazard ratio for 30-<60 pills/month was elevated but not statistically significant, and there was no association with any other frequency of use. The elevated risk of follicular lymphoma was strongest for long-term (5+ years), regular NSAID users.

In analyses by NSAID type (Table 3), use of 60+ pills/month of aspirin-only or nonaspirin-NSAIDs-only, was associated with increased NHL risk. However, the association with aspirin was attenuated when the lagged variable was used, whereas the association with nonaspirin strengthened slightly. Both aspirin and nonaspirin NSAID use were associated with a statistically significant 2- to 3-fold higher risk of follicular lymphoma which remained when the lagged variables were used. Results were not meaningfully changed in sensitivity analyses counting low-dose aspirin as one-fourth of a regular aspirin (data not shown).

Results from this prospective study do not support a strong association between NSAID use and overall NHL incidence. A statistically significant positive association was observed between current use of 60+ pills/month of NSAIDs and NHL risk. However, this association was no longer apparent when NSAID exposure was lagged, suggesting that the association with current NSAID use might be attributed to NSAID use for symptoms of undiagnosed NHL.

These results for all NSAIDs extend our previous null findings for long-term regular aspirin use and NHL overall (not including multiple myeloma or CLL; ref. 19) and are consistent with 3 meta-analyses that found no relationship between NHL incidence and ever/never or a heterogeneous “maximum exposure” to aspirin (12, 14) or between all NSAIDs and all hematopoietic malignancies (13). In the Vitamins and Lifestyle cohort (20), the largest (n = 577 cases) study to examine associations with aspirin and other NSAIDs, no associations of aspirin or other NSAID use with risk of hematologic malignancies, NHL, or NHL subtypes were found. However, NSAID use was not updated throughout follow-up in that study and the resulting potential misclassification is unknown. Among 2 earlier cohort studies, one reported a nonsignificant reduced risk of lymphoma with aspirin used 1 month before interview (n = 94 cases; ref. 18), and the other a statistically significant higher risk of NHL with aspirin and other NSAIDs, but no dose response was observed (n = 131 cases; ref. 5). In general, previous cohort studies reported positive (4, 5, 7) or null (19, 20) associations between NSAID use and NHL whereas most case-control studies reported inverse (8–11, 15), associations. Potential recall bias in the case-controls studies may have influenced the study results and contributed to these inconsistent findings.In analyses by NHL subtype, we observed an association between current use of 60+ pills/month of NSAIDs use and increased risk of follicular lymphoma that persisted when NSAID exposure was lagged and was strongest for long-term, regular NSAID users. Few other studies examined NSAID use and NHL subtypes. One case-control study reported a positive association between long-term, regular NSAID use and DLBCL only (17), whereas another reported a lower risk of several NHL subtypes with 4 or more consecutive weeks of NSAID use (8).

Our study has some important strengths, including its size and the availability of prospectively collected, regularly updated NSAID use data. With over 1,700 cases this is one of the largest studies to date on this topic. Second, the availability of regularly updated information on type and frequency of NSAID use allowed us to examine mutually exclusive groups of never users, current aspirin-only users, and current nonaspirin only users. In addition, it allowed us to examine a highly exposed group of NSAID users (regular users of 5+ years). This is one of the first studies to comprehensively examine NSAID use and the most common NHL subtypes, but sample sizes for these analyses were still modest and results require replication. In addition, the impact of unknown confounders (e.g., autoimmune conditions other than rheumatoid arthritis) on our results is unknown.

In summary, results of this study support an association between regular NSAID use and follicular lymphoma incidence but no other NHL subtype. This association with follicular lymphoma should be further investigated in large (or consortial) studies. This is particularly important for aspirin as the risks and benefits of prophylactic, daily use are weighed.

No potential conflicts of interest were disclosed.

Conception and design: L.R. Teras, S.M. Gapstur, A.V. Patel

Development of methodology: L.R. Teras, E.J. Jacobs

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.M. Gapstur, A.V. Patel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.R. Teras, W.R. Diver, Y. Zhai, E.J. Jacobs

Writing, review, and/or revision of the manuscript: L.R. Teras, S.M. Gapstur, A.V. Patel, W.R. Diver, E.J. Jacobs

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.R. Teras

Study supervision: S.M. Gapstur

The American Cancer Society (ACS) has funded the creation, maintenance, and follow-up of the Cancer Prevention Study II Nutrition cohort. All of the authors were employed by the Society during the course of this analysis.