Abstract
Background
Results from prospective studies suggest that non-steroidal anti-inflammatory drugs (NSAID) may decrease lung cancer risk; however, any protective effect appears to be most evident in men.
Methods
We evaluated the associations between NSAID use and lung cancer incidence in postmenopausal women in the Women’s Health Initiative (WHI) adjusting for female specific potential confounders such as hormone therapy in addition to smoking histories and other potential confounders. We identified 143,841 women from ages 50 to 79 and 1,902 centrally confirmed lung cancer cases were included in the analysis. We used Cox regression models to estimate hazard ratios and their 95% confidence intervals.
Results
Compared to non-use, regular NSAID use was not associated with overall lung cancer incidence (NSAID use >10 years HR 0.87, 95% CI 0.71–1.08, p-trend =0.13). No statistically significant associations were found when examined by histological subtypes and although there was a trend of decreased risk with longer duration of NSAID use in the adenocarcinoma subtype, this was not statistically significant (NSAID use >10 years HR 0.80, 95% CI 0.58–1.10, p trend = 0.07).
Conclusion
Our study did not show that NSAID use is associated with lung cancer risk in women even after adjusting for female-specific confounders. There was a trend of decreased risk in the adenocarcinoma subtype; however, this was not statistically significant.
Impact
Future studies will need to take in account the various molecular subtypes of non-small cell lung cancer to further elucidate the role of NSAIDS in lung cancer, especially for the adenocarcinoma subtype.
BACKGROUND
The potential chemopreventive effect of aspirin and other non-steroidal anti-inflammatory drugs (NSAIDs) on cancer has been reported since the 1980s with various observational studies suggesting beneficial effect on prevalent cancers such as colorectal, breast and lung cancer (1, 2). Potential anti-carcinogenic mechanisms of NSAIDs include inhibition of the cyclooxygenase (COX) pathway, which is involved in inflammation, apoptosis, and angiogenesis (3), and overexpression of COX2 has shown to increase the survival of lung adenocarcinoma in vitro (4). Also, NSAID induced tumor regression has been observed in lung cancer mouse models (5). Additionally, it has been reported that up to 70–90% of non-small cell lung cancers, particularly adenocarcinoma subtype, overexpress COX2 enzyme (6, 7).
Several prospective cohort studies have investigated the relationship between NSAID use and lung cancer risk. Published cohort studies include the Vitamin and Lifestyle Study (VITAL), Iowa Women’s Health Study (IWHS), Health Professional Follow-Up Study (HPFS), National Health and Nutrition Examination Survey (NHANES), Cancer Prevention Study II (CPS) and the Nurses’ Health Study (NHS), which have reported hazard ratios ranging from 0.69 to 1.10 (8–13). The inconsistent results may be due to heterogeneity in exposure definition such as dosage, frequency, duration of use and adjusted covariates.
Inconsistencies between studies are further complicated by possible effect modification by gender with decreased risks primarily reported among men but not in women, suggesting either a real biological difference or residual confounding by female-specific factors, such as postmenopausal hormone therapy. Preclinical studies show that female hormones may play a significant role in lung carcinogenesis (14, 15) and postmenopausal estrogen and progestin combination hormone use has been associated with increased lung cancer mortality in the Women’s Health Initiative (WHI) hormone therapy (HT) trial (16). In the CPS cohort, long term daily aspirin users were more likely to be HT users compared to non-aspirin users (41% vs. 30%) (12), and similarly in the NHS, current aspirin users were more likely to be HT users (36% vs. 28%) (13). This suggests that HT use may confound the association between NSAID use and lung cancer in studies which include postmenopausal women; thus HT may be an important covariate which needs to be adjusted when evaluating aspirin and NSAID use in relation to lung cancer risk in women.
We report here on our investigation into the association between use of aspirin and non-aspirin NSAIDs and lung cancer risk in the WHI, a large multi-center prospective study of postmenopausal women in the US. We recently reported the results of our investigation on the association between NSAID use and overall cancer risk in women within the WHI and this study included results in overall lung cancer (17). In the present study, we will expand on the reported results by further examining the associations by histologic subtypes and smoking status.
METHODS
Study population
The WHI is a large study of postmenopausal women designed to investigate the determinants of major chronic diseases in women. Approximately 161,000 postmenopausal women, ages 50–79, were recruited at 40 clinical centers across the US between September 1993 and December 1998 (18, 19). The study consisted of an observational study (OS) and clinical trials (CT) with aims to identify risk factors and develop prevention strategies for major causes of morbidity and mortality in postmenopausal women including cancer, cardiovascular disease and osteoporotic fractures. The details of study design have been previously described (20, 21).
The WHI-CT (n=68,133; Trial registration: clinicaltrials.gov identifier, NCT00000611) included three overlapping components: two placebo-controlled hormone therapy trials [estrogen-alone (n=10,739) and estrogen plus progestin (n=16,608)]; a diet modification compared to usual diet trial (n=48,836); and a calcium/vitamin D supplementation placebo controlled trial (n=36,282). Participants in the WHI-OS were 93,676 women who were screened for participation in the CT but were ineligible or unwilling to participate, or who were directly recruited (21). Participants provided written informed consent and the WHI protocol was approved by institutional review boards at each participating institutions (19,20).
At baseline, women with history of any cancer except non-melanoma skin cancer (n=16,255), women with missing baseline medication collection (n=2) and missing baseline smoking status (n=2,126) were excluded. After exclusions there were 143,841 women available for inclusion in the analysis.
Data collection
WHI participants attended baseline screening visits, during which they completed self-administered questionnaires that collected detailed information on demographics, medical and reproductive history, family history of cancer, physical activity, and other risk factors. Women in the CT were followed regularly at annual clinic visits and exposure data were updated at 3, 6 and 9 years from randomization. Participants in the HT trial also had 6-month follow ups in their first two years. The participants in the OS were initially assessed at a baseline screening visit in which demographic, baseline biometric and exposure data were collected. Participants were mailed annual forms for updated exposure data and medical history and they returned for a follow up visit at 3 years after entry. Medication and supplement inventory was repeated at the year 3 visit (20).
Case ascertainment
Incident lung cancers were identified through self-reports semi-annually for CT participants and annually for OS participants, or by death certificates. Self-reported cases of lung cancer were confirmed with a pathology report. Reports of death were confirmed by medical record and death certificate review at the clinical coordinating center. National Death Index was used for participants who could not be contacted. The initial assignment of an outcome was assigned by the local clinical center physician adjudicator and this was further assessed by central review (22). All outcomes including lung cancers were centrally adjudicated by reviewers blinded to randomization assignments. Cancer stage, histology and grade were coded per Surveillance, Epidemiology and End Results guidelines (16). At the end of September 2010, 1,902 incident lung cancer cases were identified after 11.4 years of follow up.
Exposure variable (NSAIDS)
The primary exposure variables were non-steroidal anti-inflammatory drug (NSAID) use which included aspirin and non-aspirin NSAIDs. Non-aspirin NSAIDs included non-aspirin salicylates, ibuprofen, indomethacin, naproxen, piroxicam, celecoxib and others. Details on medication usage were collected from baseline questionnaires and were updated at year 3 clinic visit for the OS and at years 1, 3, 6 and 9 for CT. Participants were asked to bring all prescription and over-the-counter medications that they used regularly. A regular NSAID user was defined as use of at least twice a week in each of the two weeks preceding the interview. Regular users were further asked for the type of medication, dosage and duration of use (21).
Smoking exposure and other covariates
The adjusted smoking covariates were assessed at WHI enrollment and include smoking status (never, former, current), pack-years smoking, age at smoking initiation, years since quitting for former smokers and environmental smoking exposure. Smoking status was obtained based on self-report on standard questionnaires. Participants were asked on their initial questionnaire whether they were current or former smokers. Former smokers were asked the age at which they discontinued smoking and both current and former smokers were asked to report their average number of cigarettes smoked per day, years of smoking and the age at initiation. Never smokers were defined as smoking less than 100 cigarettes in their entire life. Participants with missing smoking status were excluded at baseline. Environment smoking (passive smoking) exposure data was collected only in the OS. Participants were asked if they had ever lived with someone who smoked cigarettes inside their homes, both when they were less than 18 years old and when they were 18 years or older. If so, the number of years lived with a smoker was assessed (21). Inclusion of this covariate in multivariate analysis did not change the risk estimates thus was not included in the final multivariate model.
Covariates other than smoking included ethnicity, body mass index, alcohol intake, use of multivitamins, postmenopausal hormone use, reproductive history, fruit and vegetable intake, family history of lung cancer, personal history of heart and respiratory diseases. Data on covariates were obtained by self-report on questionnaires and dietary intake was assessed by a semi-quantitative food frequency questionnaire (21). Reproductive history including parity and age at menopause were initially included as covariates, however, these were removed from the final model as they did not significantly alter the point estimates in the multivariate models.
Special care was taken for HT adjustment as this was associated with increased lung cancer mortality and a nonstatistical increase in lung cancer incidence in a previously published analysis of WHI HT trial (16). Prior use of HT was assessed at baseline via personal interviews in both the CT and OS. In the HT trial, the participants in the placebo arm were considered to be never/former users based on the baseline data. Participants in the intervention arm were considered to be current users.
Statistical analysis
Cox proportional hazards models were used, with time from enrollment as the basic time variable with stratification on baseline age, smoking status (never/former/current) and cohort (CT versus OS), to estimate the hazard ratios of lung cancer incidence in each exposure category compared with a reference category in order to evaluate the association between NSAID use and lung cancer incidence. For continuous variables, p-values for linear trend were calculated using the Cochran-Mantel-Haenszel test. The multivariate models adjusted for the aforementioned covariates and also were adjusted for randomization in the trials. In regards to smoking adjustment, age at started smoking was modeled as a linear term using the median of the reported age category and years since quitting smoking was also modeled as a linear term subtracting the median of the age category at quitting from age at baseline. Pack-years smoking was modeled as a categorical term. Updated NSAID data were used in time-dependent models taking in account the updated medication information in follow up questionnaires. In the time-dependent models of various NSAID types (e.g. ASA only, non-ASA NSAID only), participants were censored if they started a different type of NSAIDs during follow up. Participants who initially reported to be a NSAID user who reported no use at a later time remained within the user category and their duration of use remained as reported at the time of last use. In the absence of updated information, participants remained in the analysis and their NSAID user status remained as last reported if their last medication status was a user. However, if their last status was non-user and they had any prior use, they were retained in the models as a user. In regards to the duration of NSAID use, participants who reported regular NSAID use were asked about the duration of use at each follow up and the reported duration of use at each update was considered to represent cumulative use. Separate analyses were performed stratifying by smoking status (never, former, current) and histologic subtypes (adenocarcinoma, squamous cell carcinoma and small cell carcinoma). Participants were right-censored from analysis when any other cancer, with the exception of non-melanoma skin cancer, was reported or at the time of non-lung cancer death, withdrawal from the study or loss of contact. Statistical analyses were performed using SAS 9.2 and a two-sided p<0.05 was considered statistically significant.
RESULTS
The characteristics of participants are shown in Table 1 by duration of NSAID use at baseline. Among the 161,808 participants in the CT and OS, 143,841 participants were included in the analysis after exclusions were made. Among the 143,841 participants, 1,902 centrally confirmed lung cancer cases were identified as of September 2010. The most frequent histologic subtypes were adenocarcinoma (46%), squamous cell carcinoma (14%) and small cell carcinoma (10%). Less frequent subtypes included large cell carcinoma, carcinoid tumors, spindle cell carcinoma and others.
Table 1.
Baseline characteristics and lung cancer risk factors by duration of NSAID exposure in the Women’s Health Initiative (WHI) observational study and clinical trial (n=143, 841)
| Non User (N=94,051) | < 5 year (N=30,199) | 5 – 10 years (N=12,171) | > 10 years (N=7,420) | |||||
|---|---|---|---|---|---|---|---|---|
| N | (%) | N | (%) | N | (%) | N | (%) | |
| Age group at screening, mean (SD) | 62.6 | (7.2) | 64.0 | (7.1) | 64.1 | (7.2) | 63.5 | (7.2) |
| 50–59 | 34590 | (36.78) | 8517 | (28.2) | 3433 | (28.21) | 2342 | (31.56) |
| 60–69 | 41573 | (44.20) | 14238 | (47.15) | 5535 | (45.48) | 3360 | (45.28) |
| 70–79 | 17888 | (19.02) | 7444 | (24.65) | 3203 | (26.32) | 1718 | (23.15) |
|
| ||||||||
| Race/ethnicity | ||||||||
| White | 75066 | (79.81) | 25746 | (85.25) | 10858 | (89.21) | 6759 | (91.09) |
| Black | 9496 | (10.10) | 2490 | (8.25) | 708 | (5.82) | 332 | (4.47) |
| Hispanic | 4406 | (4.68) | 1014 | (3.36) | 297 | (2.44) | 151 | (2.04) |
| American Indian | 415 | (0.44) | 122 | (0.4) | 46 | (0.38) | 33 | (0.44) |
| Asian/Pacific Islander | 3249 | (3.45) | 454 | (1.5) | 126 | (1.04) | 61 | (0.82) |
| Unknown | 1419 | (1.51) | 373 | (1.24) | 136 | (1.12) | 84 | (1.13) |
|
| ||||||||
| Smoking status | ||||||||
| Never | 48983 | (52.08) | 15300 | (50.66) | 5872 | (48.25) | 3689 | (49.72) |
| Past | 38403 | (40.83) | 12988 | (43.01) | 5472 | (44.96) | 3131 | (42.2) |
| Current | 6665 | (7.09) | 1911 | (6.33) | 827 | (6.79) | 600 | (8.09) |
|
| ||||||||
| History of emphysema | ||||||||
| No | 85451 | (96.58) | 27495 | (96.34) | 11129 | (96.11) | 6767 | (96.31) |
| Yes | 3026 | (3.42) | 1045 | (3.66) | 451 | (3.89) | 259 | (3.69) |
|
| ||||||||
| History of CVD | ||||||||
| No | 86858 | (92.35) | 25707 | (85.13) | 10071 | (82.75) | 6567 | (88.5) |
| Yes | 7193 | (7.65) | 4492 | (14.87) | 2100 | (17.25) | 853 | (11.5) |
|
| ||||||||
| Family history of cancer | ||||||||
| No | 31126 | (34.54) | 9565 | (33.15) | 3764 | (32.35) | 2361 | (33.01) |
| Yes | 58985 | (65.46) | 19290 | (66.85) | 7873 | (67.65) | 4792 | (66.99) |
|
| ||||||||
| BMI (kg/m2) | ||||||||
| <25 | 34280 | (36.78) | 9091 | (30.34) | 3950 | (32.73) | 2606 | (35.42) |
| ≥25 | 58931 | (63.22) | 20875 | (69.66) | 8120 | (67.27) | 4752 | (64.58) |
|
| ||||||||
| Multivitamin use | ||||||||
| No | 60327 | (64.14) | 17039 | (56.42) | 6566 | (53.95) | 4137 | (55.75) |
| Yes | 33722 | (35.86) | 13160 | (43.58) | 5605 | (46.05) | 3283 | (44.25) |
|
| ||||||||
| Fruit/vegetable servings per day | ||||||||
| 0–2 | 14255 | (15.65) | 4350 | (14.8) | 1614 | (13.63) | 1086 | (14.94) |
| >2–4 | 35893 | (39.42) | 11637 | (39.6) | 4607 | (38.90) | 2869 | (39.46) |
| >4–6 | 25274 | (27.76) | 8298 | (28.24) | 3443 | (29.07) | 2158 | (29.68) |
| >6 | 15638 | (17.17) | 5098 | (17.35) | 2178 | (18.39) | 1158 | (15.93) |
|
| ||||||||
| Post-menopausal HT use1 | ||||||||
| Never | 36750 | (39.11) | 10627 | (35.22) | 3928 | (32.31) | 2473 | (33.34) |
| Past E-alone | 7419 | (7.89) | 2720 | (9.01) | 1055 | (8.68) | 645 | (8.7) |
| Past E+P | 5139 | (5.47) | 1677 | (5.56) | 740 | (6.09) | 454 | (6.12) |
| Current E-alone | 22533 | (23.98) | 8197 | (27.16) | 3498 | (28.77) | 2035 | (27.44) |
| Current E+P | 22131 | (23.55) | 6955 | (23.05) | 2938 | (24.16) | 1810 | (24.4) |
Current use includes participants randomized to the active arms of the hormone trial, or current use reported at baseline for participants not in the hormone trial. Past use includes participants randomized to the placebo arms of the hormone trial who reported either past or current use at baseline, and past use reported at baseline for participants not in the hormone trial.
Abbreviations: NSAID, non-steroidal anti-inflammatory drug; CVD, cardiovascular disease; BMI, body mass index; HT, hormone therapy; E, unopposed estrogen; E+P, estrogen and progestin combination
Among the participants, 51% had no smoking history while 42% were former and 7% were current smokers. Any regular NSAID use was reported by 34.6% of the participants while 18.6% reported use of aspirin only and 12.3% reported non-aspirin NSAID use only.
After adjusting for age, women with greater than 10 years of NSAID use were more likely to have a history of cardiovascular disease, more likely to use multivitamins and have a history of HT use compared to non-NSAID users. They were also slightly more likely to be former or current smokers (Table 1).
Table 2 shows the association between lung cancer and NSAID use by type and duration. The initial model adjusted for various smoking variables including smoking status, pack-years of smoking, age started and years since quitting smoking. The second model further adjusted for hormone therapy use and the final multivariate model included other covariates including body mass index, multivitamin use, history of cardiovascular disease, fruit and vegetable intake and alcohol use (Table 2). After adjustment for the above covariates, no significant associations between overall lung cancer and NSAID use by type and duration were observed (NSAID use >10 years HR 0.87, 95% CI 0.71–1.08, p-trend=0.13). The risk estimates were changed somewhat with smoking adjustment but they were little altered by adjustment for HT or other covariates in the multivariate models. For example, the risk estimate for NSAID use >10 years without any smoking adjustments was HR 0.96 95% CI 0.78–1.19, p-trend=0.45.
Table 2.
Hazard Ratios of Lung Cancer by Type and Duration of NSAID use in the WHI Clinical Trial and Observational Study
| # participants | # cases1 | Smoking-adjusted2 | Smoking and HT-adjusted3 | Multivariate-adjusted4 | |
|---|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | HR (95% CI) | |||
| All NSAID | |||||
| Non-user | 94051 | 1176 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 years | 30199 | 439 | 1.01 (0.90, 1.12) | 1.00 (0.90, 1.11) | 0.98 (0.87, 1.11) |
| 5 – 10 years | 12171 | 188 | 1.07 (0.93, 1.23) | 1.06 (0.92, 1.22) | 1.03 (0.89, 1.20) |
| > 10 years | 7420 | 99 | 0.90 (0.74, 1.11) | 0.90 (0.73, 1.10) | 0.87 (0.71, 1.08) |
| p-value for trend5 | 0.10 | 0.10 | 0.13 | ||
| ASA only: < 100 mg | |||||
| Non-user | 94051 | 848 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 year | 4901 | 61 | 1.06 (0.86, 1.31) | 1.06 (0.85, 1.31) | 1.02 (0.82, 1.28) |
| Non-user | 94051 | 848 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 year | 4901 | 61 | 1.06 (0.86, 1.31) | 1.06 (0.85, 1.31) | 1.02 (0.82, 1.28) |
| ≥ 5 years | 1239 | 14 | 1.19 (0.86, 1.63) | 1.18 (0.86, 1.62) | 1.16 (0.83, 1.62) |
| p-value for trend5 | 0.61 | 0.63 | 0.50 | ||
| ASA only: > 100 mg | |||||
| Non-user | 94051 | 870 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 years | 10974 | 116 | 1.05 (0.88, 1.26) | 1.04 (0.87, 1.24) | 1.00 (0.83, 1.22) |
| ≥ 5 years | 9704 | 111 | 1.05 (0.87, 1.28) | 1.05 (0.86, 1.28) | 1.02 (0.83, 1.26) |
| p-value for trend5 | 0.09 | 0.10 | 0.15 | ||
| ASA only (all) | |||||
| Non-user | 94051 | 955 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 years | 15875 | 203 | 1.01 (0.88, 1.17) | 1.01 (0.87, 1.16) | 0.97 (0.84, 1.13) |
| 5 – 10 years | 6002 | 81 | 1.12 (0.93, 1.35) | 1.12 (0.93, 1.34) | 1.03 (0.84, 1.26) |
| > 10 years | 4941 | 58 | 0.85 (0.65, 1.11) | 0.85 (0.65, 1.11) | 0.82 (0.62, 1.08) |
| p-value for trend5 | 0.12 | 0.12 | 0.15 | ||
| Non-ASA NSAID only | |||||
| Non-user | 94051 | 880 | 1 (reference) | 1 (reference) | 1 (reference) |
| < 5 years | 12257 | 124 | 1.02 (0.86, 1.20) | 1.01 (0.85, 1.19) | 1.02 (0.85, 1.22) |
| 5 – 10 years | 4506 | 41 | 0.92 (0.68, 1.25) | 0.91 (0.67, 1.23) | 0.90 (0.66, 1.24) |
| > 10 years | 958 | 9 | 0.91 (0.51, 1.61) | 0.89 (0.50, 1.58) | 0.78 (0.42, 1.47) |
| p-value for trend5 | 0.81 | 0.77 | 0.46 | ||
Case numbers for non-users differ in the various categories due to time dependent censoring. Follow up time was censored if participants started a NSAID that is different from the defined NSAID in the category. For example, in the ASA only analysis, if non-users started taking a non-ASA during follow up, they were censored from analysis.
From a Cox proportional hazards regression model stratified by 5-year age intervals, CT vs OS, Extension study enrollment and smoking status (never/former/current); adjusted for linear age, pack years of smoking, age started and years since quitting smoking.
Stratified and adjusted as in Model 1, with additional adjustment for postmenopausal hormone use (never, past/E-alone, past/E+P, current/E-alone, current/E+P).
Stratified and adjusted as in Model 2, with additional adjustment for BMI, race/ethnicity, hx of emphysema, hx CVD, family hx cancer, alcohol intake, multivitamin use, fruit/vegetable intake, and randomization arm of the DM trial.
Tested using a linear form of years of use.
Abbreviations: ASA, aspirin; NSAID, non-steroidal anti-inflammatory drug; HT, hormone therapy; HR, hazard ratio; CI, confidence interval; WHI, Women’s Health Initiative
We also examined associations between NSAID use and lung cancer risk defined by histologic subtype. Again, there were no statistically significant associations; however, there was a statistically non-significant trend of decreased hazard ratio with longer duration of all NSAID use (p-trend = 0.07) and aspirin use (p-trend=0.08) for adenocarcinoma subtype (Table 3). No significant associations were observed for squamous cell and small cell lung cancers although case numbers were small. Finally, in analyses stratified on smoking status, no significant interactions were observed (Table 4).
Table 3.
Hazard Ratios of Lung Cancer Histology Subtypes by Type and Duration of NSAID use in the WHI Clinical Trial and Observational Study
| Adenocarcinoma | Squamous cell | Small cell | ||||
|---|---|---|---|---|---|---|
| # Cases | HR (95% CI)1 | # Cases | HR (95% CI)1 | # Cases | HR (95% CI)1 | |
| All NSAID | ||||||
| Non-user | 538 | 1 (reference) | 162 | 1 (reference) | 119 | 1 (reference) |
| < 5 years | 212 | 0.98 (0.83, 1.16) | 63 | 1.03 (0.77, 1.39) | 36 | 0.80 (0.54, 1.18) |
| 5 – 10 years | 85 | 0.86 (0.68, 1.09) | 22 | 0.85 (0.55, 1.30) | 23 | 1.29 (0.83, 2.00) |
| > 10 years | 39 | 0.80 (0.58, 1.10) | 17 | 0.80 (0.45, 1.42) | 13 | 0.89 (0.46, 2.73) |
| p-value for trend2 | 0.07 | 0.26 | 0.51 | |||
| ASA only (all) | ||||||
| Non-user | 443 | 1 (reference) | 130 | 1 (reference) | 93 | 1 (reference) |
| < 5 years | 102 | 0.96 (0.77, 1.19) | 27 | 1.03 (0.70, 1.51) | 15 | 0.86 (0.51, 1.44) |
| ≥ 5 years | 61 | 0.80 (0.61, 1.05) | 17 | 0.59 (0.34, 1.01) | 17 | 1.25 (0.75, 2.07) |
| p-value for trend2 | 0.08 | 0.22 | 0.63 | |||
| Non-ASA NSAID only | ||||||
| Non-user | 414 | 1 (reference) | 125 | 1 (reference) | 89 | 1 (reference) |
| < 5 years | 62 | 1.16 (0.91, 1.49) | 19 | 1.31 (0.72, 1.80) | 8 | 0.69 (0.34, 1.35) |
| ≥ 5 years | 23 | 0.78 (0.50, 1.21) | 11 | 1.68 (0.91, 3.08) | 4 | 0.44 (0.14, 1.40) |
| p-value for trend2 | 0.12 | 0.65 | 0.26 | |||
From a Cox proportional hazards regression model stratified by 5-year age intervals, CT vs OS, Extension study enrollment, and smoking status (never/former/current); adjusted for linear age, pack years of smoking, age started and years since quitting smoking, postmenopausal hormone use (never, past/E-alone, past/E+P, current/E-alone, current/E+P), BMI, race/ethnicity, hx of emphysema, hx CVD, family hx cancer, alcohol intake, multivitamin use, fruit/vegetable intake, and randomization arm of the DM trial.
Tested using a linear form of years of use.
Table 4.
Hazard Ratios of Lung Cancer by Baseline Smoking Status and Type and Duration of NSAID use in the WHI Clinical Trial and Observational Study
| Never smokers | Former smokers | Current smokers | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| # Cases | HR (95% CI)1 | # Cases | HR (95% CI) 1 | # Cases | HR (95% CI)1 | |
| All NSAID | ||||||
| Non-user | 203 | 1 (reference) | 612 | 1 (reference) | 361 | 1 (reference) |
| < 5 years | 70 | 1.02 (0.78, 1.35) | 259 | 1.00 (0.86, 1.17) | 110 | 0.91 (0.74, 1.13) |
| 5 – 10 years | 20 | 0.93 (0.63, 1.39) | 109 | 0.99 (0.80, 1.22) | 59 | 1.19 (0.91, 1.56) |
| > 10 years | 11 | 0.60 (0.32, 1.14) | 53 | 1.02 (0.77, 1.34) | 35 | 0.76 (0.50, 1.14) |
| p-value for trend2 | 0.16 | 0.64 | 0.24 | |||
| Test of heterogeneity3 | p = 0.80 | |||||
| ASA only (all) | ||||||
| Non-user | 178 | 1 (reference) | 458 | 1 (reference) | 319 | 1 (reference) |
| < 5 years | 41 | 1.08 (0.76, 1.51) | 106 | 0.92 (0.74, 1.13) | 56 | 1.02 (0.78, 1.33) |
| ≥ 5 years | 14 | 0.80 (0.50, 1.26) | 77 | 1.00 (0.79, 1.26) | 48 | 0.92 (0.67, 1.25) |
| p-value for trend2 | 0.20 | 0.93 | 0.08 | |||
| Test of heterogeneity3 | p = 0.54 | |||||
| Non-ASA NSAID only | ||||||
| Non-user | 161 | 1 (reference) | 441 | 1 (reference) | 278 | 1 (reference) |
| < 5 years | 18 | 1.02 (0.66, 1.57) | 73 | 1.31 (0.90, 1.43) | 33 | 0.78 (0.54, 1.14) |
| ≥ 5 years | 7 | 0.77 (0.36, 1.64) | 29 | 0.84 (0.57, 1.24) | 14 | 1.04 (0.63, 1.71) |
| p-value for trend2 | 0.81 | 0.33 | 0.77 | |||
| Test of heterogeneity3 | p = 0.72 | |||||
From a Cox proportional hazards regression model stratified by 5-year age intervals, CT vs OS, and Extension study enrollment; adjusted for linear age, for smokers: pack years of smoking, age started and years since quitting smoking if former, postmenopausal hormone use (never, past/E-alone, past/E+P, current/E-alone, current/E+P), BMI, race/ethnicity, hx of emphysema, hx CVD, family hx cancer, alcohol intake, multivitamin use, fruit/vegetable intake, and randomization arm of the DM trial.
Tested using a linear from of years of use.
Tested for an interaction of duration of use (linear) and smoking status (never/former/current).
We performed several sensitivity analyses in order to ensure the robustness of our findings. As the participants in OS did not provide updated information on NSAID use at years 6 and 9, we performed an analysis without updates at years 6 and 9 and results were not significantly different from the primary analysis (NSAID use >10 years HR 0.91, 95% CI 0.73–1.13, p-trend = 0.20). In another analysis with all updates but which only included participants from CT, results were also not significantly altered from the primary analysis (NSAID use >10 years HR 0.86, 95% CI 0.42–1.18, p-trend=0.33).
DISCUSSION
In this analysis of lung cancer incidence, no significant associations between overall NSAID use and lung cancer were observed. However, there was a statistically non-significant trend of decreased risk of adenocarcinoma with longer duration of NSAID use.
The effect of NSAID use in lung adenocarcinoma has been inconsistent in the literature with reports of both protective and null findings. The VITAL cohort, a prospective study of approximately 77,000 participants, reported a decreased lung cancer incidence with NSAID use (>4.2 days/week use for >10 years) and the association was strongest for adenocarcinoma (HR 0.59, 95% CI 0.37–0.94) while no significant association was seen in squamous cell carcinoma (HR 0.97, 95% CI 0.57–1.64)8. In contrast, in the Iowa Women’s Health Study in which 27,000 women were assessed, there were no significant associations in overall lung cancer or in the adenocarcinoma subtype (9). Of note, this study assessed the role of current NSAID use and did not report the association by duration of use. Other prospective observational studies including the Cancer Prevention Study, National Health and Nutrition Examination Study, the Nurses’ Health Study and Health Professional Follow-Up study did not report risk estimates by histologic subtypes (10–13).
There have been a few randomized studies that aimed to assess the effect of aspirin in cancer. One of these studies is the Women’s Health Study in which approximately 40,000 women were randomized to receive either aspirin 100mg every other day or placebo and the primary outcome was newly diagnosed invasive cancers. After 10-years of active intervention, the study reported that there was no reduction in risk of incident cancers (RR 1.01, 95% CI 0.94–1.08) except for lung cancer in which there was a trend of reduced risk (RR 0.78, 95% CI 0.59–1.03) (23). There was no reduction in cancer mortality overall or by site, except for lung cancer mortality (RR 0.70, 95% CI 0.50–0.99). However, this was not confirmed in a subsequent follow up analysis after 18-years of follow up (lung cancer incidence HR 1.04, 95% CI 0.86 – 1.26) and no significant outcomes were observed in adenocarcinoma versus non-adenocarcinoma metastases in all cancer (24).
Although we did not examine the associations in lung cancer mortality, studies indicate that aspirin may decrease the risk of lung cancer mortality, particularly in adenocarcinoma. An inverse association for adenocarcinoma was observed in the study by Rothwell et al in which the associations between daily aspirin for five years or longer and mortality from various cancer sites were evaluated in a pooled analysis of individual patient data from randomized trials (25). This study showed that for lung cancer, the lower risk was confined to adenocarcinoma. For instance, the hazard ratio at 20-year follow up was 0.55 (95% CI 0.33–0.94) for adenocarcinoma while the hazard ratio in squamous cell was 1.26 (95% CI 0.73–2.18).
A potential reason for the inconsistent results across various studies of NSAID use in lung cancer is that the effect of aspirin on lung cancer may vary by the driver molecular pathway of each tumor. The past decade has been marked by significant advances in our understanding of lung cancer biology and it is now known that lung adenocarcinoma is not a homogeneous disease but is comprised of various molecular subtypes with distinct oncogenic drivers such as an epidermal growth factor receptor (EGFR) mutation and anaplastic lymphoma kinase (ALK) rearrangement (26). There is evidence from the literature that the role of aspirin in colorectal cancer risk differs by BRAF (V-Raf murine sarcoma viral oncogene homolog B) mutational status in which the protective effect is observed in BRAF wildtype but not in BRAF mutated cancers (27). This is thought to be due to the upregulation of COX2 by Raf kinases whose activity is upregulated in BRAF mutation positive colorectal cancers (28, 29). Similarly, the effect of aspirin on lung cancer may differ in the various molecular subtypes and the inconsistent risk estimates may be due to differing prevalence of the various molecular subtypes. Therefore, further molecular epidemiological studies are needed to clarify the role of aspirin and other NSAIDS in the various molecular subtypes of lung cancer.
In this study, one of the objectives was to evaluate whether the associations of NSAID use is confounded by HT use and other reproductive factors. Literature has reported that there appears to be a gender specific chemopreventive effect of NSAID use. For instance, the VITAL study reported a decreased risk in men (HR 0.66, 95% CI 0.47–0.92) but not in women (HR 1.07, 95% CI 0.75–1.51) although there was no significant effect modification by gender when the analysis was limited to adenocarcinoma subtype (8). A gender difference was also observed in the National Health and Nutrition Examination Study where there was a decrease in risk of lung cancer mortality in men (RR 0.69, 95% CI 0.49–0.96) but not in women (RR 1.10, 95% CI 0.67–1.81) (11). Given the previous report from the WHI HT trial which reported increased lung cancer mortality with HT use (16), we hypothesized that the observed null effect in the above studies may have been due to inadequate adjustment for HT use and reproductive factors. However, the adjustment for these covariates in our analysis did not alter the hazard ratio estimates in overall lung cancer. The apparent null association in women may be due to an interaction between the cyclooxygenase (COX) and estrogen pathways. Studies indicate that there may be a cross-talk between COX and estrogen pathways; for example, an analysis of Nurses’ Health Study participants showed that women who were regular NSAID users had lower levels of estradiol (30) and increased pro-inflammatory prostaglandin E2 production was associated with increased aromatase activity (31).
The strengths of this study are that this is a large study with a long follow up period and well annotated exposure data including updated exposure information on dose and duration, and detailed data on covariates such as smoking history, reproductive history and HT use, and centrally confirmed cancer diagnosis. Despite this, this analysis was still limited by small numbers when stratified by histologic subgroups. Also, the analysis may have been affected by misclassification of NSAID use, particularly in the OS, where NSAID use was updated twice (baseline and year 3) while in the CT, the use was updated at multiple time points. Additionally, the role of frequency of use could not be fully addressed in this study as participants were asked whether they used NSAIDs at least twice a week but no additional data on frequency (e.g. daily versus alternate day use) were collected. The exposure data were based on self-reported use thus studies incorporating biomarker measurements of anti-inflammatory effect would strengthen future studies investigating the role of NSAIDs.
Lung cancer continues to be the leading cause of cancer mortality globally and there is a need to continue to develop prevention strategies to reduce its disease burden. Smoking cessation is imperative in a successful lung cancer prevention program but there is a need for additional prevention strategies for the growing number of former smokers who have successfully quit smoking. Therefore, ongoing research in identifying potential chemopreventive agents is warranted. Future studies of NSAIDS in lung cancer will need to take in account the various molecular subtypes in addition to the histologic subtypes.
Acknowledgments
Financial Information
The Woman’s Health Initiative (WHI) program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, U.S. Department of Health and Human Services through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C.
Footnotes
Conflict of Interest Statement:
The authors of this paper have no conflicts of interest to report.
References
- 1.Bosetti C, VR, Gallus S, Cuzick J, LaVecchia C. Aspirin and cancer risk: a quantitative review to 2011. Ann Oncol. 2012;23:1403–15. doi: 10.1093/annonc/mds113. [DOI] [PubMed] [Google Scholar]
- 2.Cuzick J, Otto F, Baron J, Brown P, Burn J, Greenwald P, et al. Aspirin and non-steroidal anti-inflammatory drugs for cancer prevention: an international consensus statement. Lancet Oncol. 2009;10(5):501–7. doi: 10.1016/S1470-2045(09)70035-X. [DOI] [PubMed] [Google Scholar]
- 3.Horn L, Backlund M, Johnson D. Targeting the eicosanoid pathway in non-small-cell lung cancer. Expert Opin Ther Targets. 2009;13:675–88. doi: 10.1517/14728220902915567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lin M, Lee R, Yang P, Ho F, Kuo M. Cyclooxygenase-2 inducing Mcl-1 dependent survival mechanism in human lung adenocarcinoma CL1.0 cells. Involvement of phosphatidylinositol 3-kinase/Akt pathway. J Biol Chem. 2001;276:48997–9002. doi: 10.1074/jbc.M107829200. [DOI] [PubMed] [Google Scholar]
- 5.Saini R, Sanyal S. Chemopreventive effect of nonsteroidal anti-inflammatory drugs on 9,10-dimethylbenz(a)anthracene-induced lung carcinogenesis in mice. Oncology Research. 2009;17:505–18. doi: 10.3727/096504009789745520. [DOI] [PubMed] [Google Scholar]
- 6.Ermert L, Dierkes C, Ermert M. Immunohistochemical expression of cyclooxygenase isoenzymes and downstream enzymes in human lung tumors. Clin Cancer Res. 2003;9:1604–10. [PubMed] [Google Scholar]
- 7.Hida T, Yatabe Y, Achiwa H, Muramatsu H, Kozaki K, Nakamura S, et al. Increased expression of cyclooxygenase 2 occurs frequently in human lung cancers, specifically in adenocarcinomas. Cancer Res. 1998;58:3761–4. [PubMed] [Google Scholar]
- 8.Slatore C, Au D, Littman A, Satia J, White E. Association of nonsteroidal anti-inflammatory drugs with lung cancer: results from a large cohort study. Cancer Epidemiol Biomarkers Prev. 2009;18:1203–7. doi: 10.1158/1055-9965.EPI-08-1110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hayes J, Anderson K, Folsom A. Association between nonsteroidal anti-inflammatory drug use and the incidence of lung cancer in the Iowa Women’s Health Study. Cancer Epidemiol Biomarkers Prev. 2006;15:2226–31. doi: 10.1158/1055-9965.EPI-06-0172. [DOI] [PubMed] [Google Scholar]
- 10.Holick C, Michaud D, Leitzmann M, Willett W, Giovannucci E. Aspirin and lung cancer in men. British J Cancer. 2003;89:1705–8. doi: 10.1038/sj.bjc.6601343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ratnasinghe L, Graubard B, Kahle L, Tangrea J, Taylor P, Hawk E. Aspirin use and mortality from cancer in a prospective study. Anticancer Research. 2004;24:3177–84. [PubMed] [Google Scholar]
- 12.Jacobs E, Thun M, Bain E, Rodriguez C, Henley S, Calle E. A large cohort study of long term daily use of adult-strength aspirin and cancer incidence. J Natl Cancer Inst. 2007;99:608–15. doi: 10.1093/jnci/djk132. [DOI] [PubMed] [Google Scholar]
- 13.Feskanich D, Bain C, Chan A, Pandeya N, Speizer F, Colditz G. Aspirin and lung cancer risk in a cohort study of women: dosage, duration and latency. British J Cancer. 2007;97:1295–99. doi: 10.1038/sj.bjc.6603996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Stabile L, Gaither A, Gubish C, Hopkins T, Luketich J, et al. Human non-small cell lung tumors and cells derived from normal lung express both estrogen receptor α and β and show biological responses to estrogen. Cancer Res. 2002;62:2141–50. [PubMed] [Google Scholar]
- 15.Ishibashi H, Suzuki T, Suzuki S, Niikawa H, Lu L, Miki Y, et al. Progesterone receptor in non-small cell lung cancer - A potent prognostic factor and possible target for endocrine therapy. Cancer Res. 2005;65:6450–8. doi: 10.1158/0008-5472.CAN-04-3087. [DOI] [PubMed] [Google Scholar]
- 16.Chlebowski R, Schwartz A, Wakelee H, Anderson G, Stefanick M, Manson J, et al. Oestrogen plus progestin in lung cancer in postmenopausal women (Women’s Health Initiative trial): a post-hoc analysis of a randomised controlled trial. Lancet. 2009;374:1243–51. doi: 10.1016/S0140-6736(09)61526-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Brasky T, Liu J, White E, Peters U, Potter J, Walter R, et al. Non-steroidal anti-inflammatory drugs and cancer risk in women: Results from the Women’s Health Initiative. International J Cancer. 2014;135:1869–83. doi: 10.1002/ijc.28823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Chlebowski R, Manson J, Anderson G, Cauley J, Aragaki A, Stefanick M, et al. Estrogen plus progestin and breast cancer incidence and mortality in the Women’s Health Initiative Observational Study. J Natl Cancer Inst. 2013;105:526–35. doi: 10.1093/jnci/djt043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Langer R, White E, Lewis C, Kotchen J, Hendrix S, Trevisan M. The Women’s Health Initiative Observational Study: Baseline characteristics of participants and reliability of baseline measures. Ann Epidemiol. 2003;13:S107–S21. doi: 10.1016/s1047-2797(03)00047-4. [DOI] [PubMed] [Google Scholar]
- 20.The Women’s Health Initiative Study Group. Design of the Women’s Health Initiative Clinical Trial and Observational Study. Controlled Clinical Trials. 1998;19:61–109. doi: 10.1016/s0197-2456(97)00078-0. [DOI] [PubMed] [Google Scholar]
- 21.Anderson G, Manson J, Wallace R, Lund B, Hall D, Davis S, et al. Implementation of the Women’s Health Initiative study design. Ann Epidemiol. 2003;13:S5–S17. doi: 10.1016/s1047-2797(03)00043-7. [DOI] [PubMed] [Google Scholar]
- 22.Curb J, McTiernan A, Heckbert S, Kooperberg C, Stanford J, Nevitt M, et al. Outcomes ascertainment and adjudication methods in the Women’s Health Initiative. Ann Epidemiol. 2003;13:S122–S8. doi: 10.1016/s1047-2797(03)00048-6. [DOI] [PubMed] [Google Scholar]
- 23.Cook N, Lee I-M, Gaziano J, Gordon D, Ridker P, Manson J, et al. Low-dose aspirin in the primary prevention of cancer. JAMA. 2005;294:47–55. doi: 10.1001/jama.294.1.47. [DOI] [PubMed] [Google Scholar]
- 24.Cook N, Lee I-M, Zhang S, Moorthy V, Buring J. Alternate-day, low-dose aspirin and cancer risk: long-term observational follow up of a randomized trial. Ann Intern Med. 2013;159:77–85. doi: 10.7326/0003-4819-159-2-201307160-00002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rothwell P, Fowkes F, Belch J, Ogawa H, Warlow C, Meade T. Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials. Lancet. 2011;377:31–41. doi: 10.1016/S0140-6736(10)62110-1. [DOI] [PubMed] [Google Scholar]
- 26.Li T, Kung H, Mack PG, Gandara DR. Genotyping and genomic profiling of non-small cell lung cancer: Implications for current and future therapies. J Clin Oncol. 2013;31:1039–49. doi: 10.1200/JCO.2012.45.3753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Nishihara R, Lochhead P, Kuchiba A, Jung S, Yamauchi M, Liao X, et al. Aspirin use and risk of colorectal cancer according to BRAF mutation status. JAMA. 2013;309:2563–71. doi: 10.1001/jama.2013.6599. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wagner E, Nebreda A. Signal integration by JNK and p38 MAPK pathways in cancer development. Nat Rev Cancer. 2009;9:537–49. doi: 10.1038/nrc2694. [DOI] [PubMed] [Google Scholar]
- 29.Sumimoto H, Imabayashi F, Iwata T, Kawakami Y. The BRAF-MAPK signaling pathway is essential for cancer-immune evasion in human melanoma cells. J Exp Med. 2006;203:1651–6. doi: 10.1084/jem.20051848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gates M, Tworoger S, Eliassen H, Missmer S, Hankinson S. Analgesic use and sex steroid hormone concentrations in postmenopausal women. Cancer Epidemiol Biomarkers Prev. 2010;19:1033–41. doi: 10.1158/1055-9965.EPI-09-0975. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Subbaramaiah K, Morris P, Zhou X, Morrow M, Du B, Giri D, et al. Increased levels of COX-2 and prostaglandin E2 contribute to elevated aromatase expression in inflamed breast tissue of obese women. Cancer Discov. 2012;2:356–65. doi: 10.1158/2159-8290.CD-11-0241. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]