Abstract
Background: Prostate cancer is considered a major health problem in western countries. Promising results from observational studies on cancer at other sites fuelled the publication of several studies assessing the association between nonsteroidal anti-inflammatory drug (NSAID) use and prostate cancer. However, these studies show conflicting results. Methods: We conducted a cohort study with a nested case-control analysis to further study the association between NSAIDs and prostate cancer. We used data from the General Practice Research Database in United Kingdom. Results: Aspirin use was associated with a reduced risk of prostate cancer [odds ratio (OR) = 0.70, 95% confidence interval (95% CI) = 0.61–0.79]. We also found that paracetamol use with a treatment duration longer than 1 year was associated with a decreased risk (OR = 0.65, 95% CI = 0.54–0.78). Non-aspirin-NSAID (NA-NSAID) and paracetamol short-term use was associated with a small increased risk whereas long-term users of NA-NSAIDs presented an OR of 0.89 (95% CI = 0.73–1.08). Discussion: Our findings support a protective effect of aspirin and paracetamol against prostate cancer. The transient elevated risk observed among newly started users of NA-NSAIDs and paracetamol is most likely explained by prothopathic bias. We found some suggestion of a reduced risk with long-term use of NA-NSAID.
Introduction
Prostate cancer is considered a major health problem in western countries. The incidence has substantially increased and age-standardized incidence rates have doubled between 1971 and 1993 in England and Wales. The largest increase occurred in men in their sixth decade (1). Because there is no curative therapy for advanced prostate cancer, a growing interest has focused on prostate cancer prevention.
The potential role of aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) in cancer chemoprevention has been extensively studied in animal and human studies. One of the mechanisms advanced for this beneficial effect would be through inhibition of cyclooxygenase-2 (COX-2; Ref. 2), the enzyme responsible for the production of various inflammatory prostaglandins. Prostaglandins play a key role on the accelerated proliferation of tumor tissue (3). In addition, there is evidence showing that COX-2 is overexpressed in prostate cancer and that prostate tumor grade is positively correlated to COX-2 levels (4).
Several observational studies have studied the association between NSAID use and prostate cancer with conflicting results. While some of these reports suggest an inverse association between NSAID use and prostate cancer (5–10), others found no protective effect (11–16). This inconsistency in the literature spurred us to conduct a cohort study with a nested case-control analysis to assess the association between NSAID use and prostate cancer.
Materials and Methods
We used data from the General Practice Research Database (GPRD). The GPRD contains computerized information entered by general practitioners in the United Kingdom (17). Data on over 2 million patients are systematically recorded and sent anonymously to the Medicines and Healthcare Products Regulatory Agency, which collects and organizes this information to be used for research projects. The computerized information includes demographics, details from general practitioner visits, diagnoses from specialist referrals and hospital admissions, results of laboratory tests, and a free text section. Prescriptions issued by the general practitioner are directly generated from the computer. Several studies with the GPRD have documented the validity and completeness of this database (18). An additional requirement for participating practices is recording of the indication for new courses of therapy.
Study Population
We identified all males 50–79 years old between January 1995 and December 2001. Patients became members of the study population on the first day of the study period when they met the criteria of at least 1-year enrollment with the general practitioner and 1 year since the first computerized prescription. That day was their study starting date. Patients with a code for any cancer before starting date were excluded. We also excluded men 70 years and older at start date who had a follow-up greater than 1 year and no data recorded during their total follow-up time. This study is an extension of a previous study, the primary goal of which was to assess the effect of antihypertensive medication on prostate cancer (19). For the present study, the final study cohort comprised 339,462 patients.
Follow-up
All members from the study cohort were followed from starting date until the earliest occurrence of one of the following end points: recorded diagnosis of prostate cancer (International Classification of Diseases, Eighth Revision: 1850), any cancer other than prostate, aged 80 years, death or end of study period (December 2001).
Case Ascertainment and Validation
We identified 2364 patients with a code of prostate cancer and manually reviewed their computerized patient profiles. Information included demographic data and all clinical information with no personal identifiers. After the review of the computerized information, 2185 patients were considered incident cases of prostate cancer. We performed a validation study in a random sample of 100 patients requesting the general practitioner to confirm the diagnosis of prostate cancer. Because 98% of cases for whom valid information was received were confirmed as cases, we decided not to request records for the remaining cases ascertained with the computerized review. At the end, 2183 patients were considered cases of prostate cancer.
Cohort and Nested Case-Control Analysis
Incidence rates of prostate cancer in our study population and risk factors were described in detail in our previous report (19). A nested case-control analysis was performed to assess the association between NSAIDs and prostate cancer. All cases of prostate cancer (n = 2183) identified in the study cohort were used in the case-control analysis and we considered the date of initial diagnosis as index date. A date during the study period was generated at random for every member of the study cohort. If the random date of a study member was included in his eligible person-time, we used his random date as the index date and marked that person as an eligible control. This selection mechanism allows that the likelihood of being selected as a control is proportional to the person-time at risk. The same exclusion criteria were applied to controls as to cases. Ten thousand controls were frequency matched by age (interval of 1 year) and calendar year from the list of all eligible controls.
Estimates of odds ratio (OR), assumed to be valid estimates of the relative risk, and 95% confidence interval (CI) associated with use of NSAIDs compared with nonuse were computed using unconditional logistic regression. We ascertained patients with antecedents of prostatism, defined as any mention of benign prostatic hyperplasia (BPH) and/or prostatism (with or without surgical intervention) recorded at least more than 1 year before the index date. We also elicited subjects' use of health services (visits to the general practitioner, specialist referrals, and hospital admissions) in the 2 years prior to the index date. All estimates of OR were adjusted for age, calendar year, BPH prior history, and health services utilization.
Exposure Definition
We studied use of aspirin, non-aspirin-NSAIDs (NA-NSAIDs), and paracetamol. We defined three time windows of exposure for each class of drugs: current use, past use, and no use. Current use was categorized as use that lasted until the index date or ended in the year prior to the index date based on the supply of drug therapy as prescribed by the general practitioner. Past use was use that ended more than 1 year before the index date. Finally, the time window of no use was defined as no use of each respective drug group at any time before the index date. Current users were subdivided according to treatment duration and trend tests were performed. Among those with duration of treatment longer than 1 year, we also explored the effect of dose.
An additional analysis using 1-year lag time (advancing the index date by 1 year in cases and controls) was performed.
Results
We found that having a prior history of BPH was associated with a small increased risk of prostate cancer (OR = 1.38, 95% CI = 1.23–1.56). We found that greater use of health services (visits to the general practitioner, referrals to specialist, or hospitalizations) within 2 years prior to the index date was associated with a greater risk of being diagnosed with prostate cancer (Table 1).
. | Cases (n = 2183) . | Controls (n = 10,000) . | OR (95% CI)a . | |||
---|---|---|---|---|---|---|
. | n (%) . | n (%) . | . | |||
BPH | ||||||
No prior historyb,c | 1622 (74.3) | 8391 (83.1) | ||||
Between 1 and 2 yr before | 140 (6.4) | 211 (2.1) | 2.11 (1.67–2.66) | |||
More than 2 yr before | 421 (19.3) | 1398 (14.0) | 1.25 (1.10–1.42) | |||
General practitioner visitsd | ||||||
0–2b | 159 (7.3) | 1946 (19.5) | ||||
3–5 | 334 (15.3) | 2031 (20.3) | 1.63 (1.33–2.00) | |||
6–12 | 802 (36.7) | 3301 (33.0) | 1.84 (1.52–2.22) | |||
≥13 | 888 (40.7) | 2722 (27.2) | 1.78 (1.45–2.18) | |||
Referralsd | ||||||
0–1b | 545 (25.0) | 5808 (58.1) | ||||
2–4 | 855 (39.2) | 2533 (25.3) | 2.90 (2.56–3.29) | |||
≥5 | 783 (35.9) | 1659 (16.6) | 3.49 (3.03–4.03) | |||
Hospital admissionsd | ||||||
0b | 1232 (56.4) | 7774 (77.7) | ||||
1 | 481 (22.0) | 1170 (11.7) | 1.75 (1.54–2.00) | |||
≥2 | 470 (21.5) | 1056 (10.6) | 1.76 (1.53–2.02) |
. | Cases (n = 2183) . | Controls (n = 10,000) . | OR (95% CI)a . | |||
---|---|---|---|---|---|---|
. | n (%) . | n (%) . | . | |||
BPH | ||||||
No prior historyb,c | 1622 (74.3) | 8391 (83.1) | ||||
Between 1 and 2 yr before | 140 (6.4) | 211 (2.1) | 2.11 (1.67–2.66) | |||
More than 2 yr before | 421 (19.3) | 1398 (14.0) | 1.25 (1.10–1.42) | |||
General practitioner visitsd | ||||||
0–2b | 159 (7.3) | 1946 (19.5) | ||||
3–5 | 334 (15.3) | 2031 (20.3) | 1.63 (1.33–2.00) | |||
6–12 | 802 (36.7) | 3301 (33.0) | 1.84 (1.52–2.22) | |||
≥13 | 888 (40.7) | 2722 (27.2) | 1.78 (1.45–2.18) | |||
Referralsd | ||||||
0–1b | 545 (25.0) | 5808 (58.1) | ||||
2–4 | 855 (39.2) | 2533 (25.3) | 2.90 (2.56–3.29) | |||
≥5 | 783 (35.9) | 1659 (16.6) | 3.49 (3.03–4.03) | |||
Hospital admissionsd | ||||||
0b | 1232 (56.4) | 7774 (77.7) | ||||
1 | 481 (22.0) | 1170 (11.7) | 1.75 (1.54–2.00) | |||
≥2 | 470 (21.5) | 1056 (10.6) | 1.76 (1.53–2.02) |
Estimates of risk are adjusted for age, calendar year, use of aspirin, NA-NSAIDs, and paracetamol, and all the variables included in the table using logistic regression.
Reference category.
BPH diagnosed less than 1 year before the index date is included in the reference category.
In the 2 years prior to the index date.
Use of aspirin was associated with a decreased risk of prostate cancer (Table 2). Current users of aspirin presented an OR of 0.70 (95% CI = 0.61–0.79) compared with nonusers. The reduced risk was observed irrespective of treatment duration. Among those using aspirin for 1 year or more, we assessed the effect of dose and a similar protective effect was seen across all dose groups (Table 3).
. | Cases (n = 2183) . | Controls (n = 10,000) . | Model 1a . | Model 2b . | ||||
---|---|---|---|---|---|---|---|---|
. | n (%) . | n (%) . | OR (95% CI) . | OR (95% CI) . | ||||
Aspirin | ||||||||
No usec | 1675 (76.7) | 7675 (76.8) | ||||||
Current use (yr) | 421 (19.3) | 1904 (19.0) | 0.95 (0.84–1.07) | 0.70 (0.61–0.79) | ||||
0–1 | 111 (5.1) | 519 (5.2) | 0.91 (0.74–1.13) | 0.62 (0.50–0.77) | ||||
1–2 | 76 (3.5) | 331 (3.3) | 1.01 (0.78–1.31) | 0.67 (0.51–0.88) | ||||
2–4 | 102 (4.7) | 490 (4.9) | 0.92 (0.73–1.15) | 0.69 (0.55–0.87) | ||||
≥4 | 132 (6.1) | 564 (5.6) | 1.03 (0.84–1.26) | 0.83 (0.67–1.02) | ||||
Test for trend: P = 0.22 | ||||||||
NA-NSAIDs | ||||||||
No usec | 879 (40.3) | 4582 (45.8) | ||||||
Current use (yr) | 622 (28.5) | 2100 (21.0) | 1.40 (1.24–1.59) | 1.14 (1.00–1.29) | ||||
0–1 | 453 (20.8) | 1360 (13.6) | 1.55 (1.36–1.77) | 1.26 (1.09–1.45) | ||||
1–2 | 38 (1.7) | 188 (1.9) | 0.94 (0.66–1.35) | 0.70 (0.49–1.02) | ||||
2–4 | 47 (2.2) | 203 (2.0) | 1.09 (0.78–1.52) | 0.99 (0.70–1.40) | ||||
≥4 | 84 (3.9) | 349 (3.5) | 1.16 (0.90–1.49) | 0.93 (0.71–1.21) | ||||
Test for trend: P = 0.12 | ||||||||
Paracetamol | ||||||||
No usec | 972 (44.5) | 5009 (50.1) | ||||||
Current use (yr) | 721 (33.0) | 2510 (25.1) | 1.33 (1.19–1.49) | 0.95 (0.84–1.07) | ||||
0–1 | 516 (23.6) | 1468 (14.7) | 1.63 (1.43–1.85) | 1.14 (1.00–1.31) | ||||
1–2 | 58 (2.7) | 255 (2.6) | 1.04 (0.77–1.40) | 0.67 (0.49–0.92) | ||||
2–4 | 76 (3.5) | 291 (2.9) | 1.25 (0.96–1.63) | 0.87 (0.66–1.15) | ||||
≥4 | 71 (3.3) | 496 (5.0) | 0.69 (0.53–0.89) | 0.50 (0.38–0.65) | ||||
Test for trend: P = 0.02 |
. | Cases (n = 2183) . | Controls (n = 10,000) . | Model 1a . | Model 2b . | ||||
---|---|---|---|---|---|---|---|---|
. | n (%) . | n (%) . | OR (95% CI) . | OR (95% CI) . | ||||
Aspirin | ||||||||
No usec | 1675 (76.7) | 7675 (76.8) | ||||||
Current use (yr) | 421 (19.3) | 1904 (19.0) | 0.95 (0.84–1.07) | 0.70 (0.61–0.79) | ||||
0–1 | 111 (5.1) | 519 (5.2) | 0.91 (0.74–1.13) | 0.62 (0.50–0.77) | ||||
1–2 | 76 (3.5) | 331 (3.3) | 1.01 (0.78–1.31) | 0.67 (0.51–0.88) | ||||
2–4 | 102 (4.7) | 490 (4.9) | 0.92 (0.73–1.15) | 0.69 (0.55–0.87) | ||||
≥4 | 132 (6.1) | 564 (5.6) | 1.03 (0.84–1.26) | 0.83 (0.67–1.02) | ||||
Test for trend: P = 0.22 | ||||||||
NA-NSAIDs | ||||||||
No usec | 879 (40.3) | 4582 (45.8) | ||||||
Current use (yr) | 622 (28.5) | 2100 (21.0) | 1.40 (1.24–1.59) | 1.14 (1.00–1.29) | ||||
0–1 | 453 (20.8) | 1360 (13.6) | 1.55 (1.36–1.77) | 1.26 (1.09–1.45) | ||||
1–2 | 38 (1.7) | 188 (1.9) | 0.94 (0.66–1.35) | 0.70 (0.49–1.02) | ||||
2–4 | 47 (2.2) | 203 (2.0) | 1.09 (0.78–1.52) | 0.99 (0.70–1.40) | ||||
≥4 | 84 (3.9) | 349 (3.5) | 1.16 (0.90–1.49) | 0.93 (0.71–1.21) | ||||
Test for trend: P = 0.12 | ||||||||
Paracetamol | ||||||||
No usec | 972 (44.5) | 5009 (50.1) | ||||||
Current use (yr) | 721 (33.0) | 2510 (25.1) | 1.33 (1.19–1.49) | 0.95 (0.84–1.07) | ||||
0–1 | 516 (23.6) | 1468 (14.7) | 1.63 (1.43–1.85) | 1.14 (1.00–1.31) | ||||
1–2 | 58 (2.7) | 255 (2.6) | 1.04 (0.77–1.40) | 0.67 (0.49–0.92) | ||||
2–4 | 76 (3.5) | 291 (2.9) | 1.25 (0.96–1.63) | 0.87 (0.66–1.15) | ||||
≥4 | 71 (3.3) | 496 (5.0) | 0.69 (0.53–0.89) | 0.50 (0.38–0.65) | ||||
Test for trend: P = 0.02 |
Model 1: Estimates of risk are adjusted for age, calendar year, prior BPH history, and all the variables included in the table using logistic regression.
Model 2 (fully adjusted): Estimates of risk are adjusted for age, calendar year, prior BPH history, number of visits to general practitioners, referrals, hospitalizations, and all the variables included in the table using logistic regression.
Reference category.
. | Cases (n = 2183) . | Controls (n = 10,000) . | OR (95% CI)a . | |||
---|---|---|---|---|---|---|
. | n (%) . | n (%) . | . | |||
Aspirin | ||||||
No use | 1675 (76.7) | 7675 (76.8) | ||||
75 mg | 163 (7.5) | 751 (7.5) | 0.70 (0.58–0.85) | |||
150 mg | 94 (4.3) | 379 (3.8) | 0.82 (0.64–1.05) | |||
300 mg | 53 (2.4) | 255 (2.6) | 0.71 (0.52–0.98) | |||
NA-NSAIDs | ||||||
No use | 879 (40.3) | 4582 (45.8) | ||||
Low-mediumb | 92 (4.2) | 397 (4.0) | 0.88 (0.68–1.13) | |||
High | 77 (3.5) | 343 (3.4) | 0.90 (0.68–1.18) | |||
Paracetamol | ||||||
No use | 972 (44.5) | 5009 (50.1) | ||||
Low-mediumc | 180 (8.3) | 909 (9.1) | 0.66 (0.54–0.80) | |||
High | 25 (1.2) | 133 (1.3) | 0.59 (0.38–0.93) |
. | Cases (n = 2183) . | Controls (n = 10,000) . | OR (95% CI)a . | |||
---|---|---|---|---|---|---|
. | n (%) . | n (%) . | . | |||
Aspirin | ||||||
No use | 1675 (76.7) | 7675 (76.8) | ||||
75 mg | 163 (7.5) | 751 (7.5) | 0.70 (0.58–0.85) | |||
150 mg | 94 (4.3) | 379 (3.8) | 0.82 (0.64–1.05) | |||
300 mg | 53 (2.4) | 255 (2.6) | 0.71 (0.52–0.98) | |||
NA-NSAIDs | ||||||
No use | 879 (40.3) | 4582 (45.8) | ||||
Low-mediumb | 92 (4.2) | 397 (4.0) | 0.88 (0.68–1.13) | |||
High | 77 (3.5) | 343 (3.4) | 0.90 (0.68–1.18) | |||
Paracetamol | ||||||
No use | 972 (44.5) | 5009 (50.1) | ||||
Low-mediumc | 180 (8.3) | 909 (9.1) | 0.66 (0.54–0.80) | |||
High | 25 (1.2) | 133 (1.3) | 0.59 (0.38–0.93) |
Estimates of risk are adjusted for age, calendar year, BPH prior history, number of visits to general practitioners, referrals, hospitalizations, and all the variables included in the table using logistic regression.
Specific cutoff values for dose (in mg) by NA-NSAID were as follows: aceclofenac 100, acemetacin 120, apazone 600, diclofenac 100, etodolac 400, fenbufen 900, fenoprofen 1200, flurbiprofen 150, ibuprofen 1200, indomethacin 75, ketoprofen 150, mefenamic 1000, meloxicam 7.5, nabumetone 1000, naproxen 750, piroxicam 10, sulindac 200, tenoxicam 10, and tiaprofenic acid 450.
Low-medium ≤2000 mg/day, high >2000 mg/day.
NA-NSAID use was not associated with the risk of developing prostate cancer (Table 2). Current users of NA-NSAIDs presented a small elevated risk of prostate cancer (OR = 1.14, 95% CI = 1.00–1.29). When only duration of use greater than 1 year was considered, the estimate dropped to 0.89 (95% CI = 0.73–1.08). Among this group, users of both low-medium and high dose shared similar results (Table 3).
Current paracetamol users presented an OR of 0.95 (95% CI = 0.84–1.07) compared with nonusers (Table 2). Within current users, those with treatment duration shorter than 1 year had a slightly elevated risk (OR = 1.14, 95% CI = 1.00–1.31) whereas those treated for 1 year or longer experienced a risk reduction (OR = 0.65, 95% CI = 0.54–0.78). We found a statistically significant trend toward greater risk reduction with longer duration of use (P = 0.02).
We further explored the risk of prostate cancer among the group of long-term paracetamol users. We subdivided this group into two groups according to daily dose (low-medium ≤2000 mg, high >2000 mg) and found a similar effect in both groups (Table 3). Paracetamol is available in three different main preparations in the United Kingdom: paracetamol alone, paracetamol with codeine/hydrocodeine, and paracetamol with propoxyphene. We found that the observed inverse association was unrelated to the preparation being used (OR = 0.53, 95% CI = 0.37–0.75; OR = 0.75, 95% CI = 0.58–0.97; and OR = 0.57, 95% CI = 0.44–0.75, respectively). The vast majority of long-term paracetamol treatment was prescribed for relief of osteoarthritis-related pain (90%). When we restricted the analysis to subjects using NA-NSAID, the results for paracetamol remained unchanged (data not shown).
In a secondary analysis, we advanced the index date by 1 year in cases and controls. The previously observed elevated risk among short-term users of NA-NSAIDs or paracetamol was not observed anymore in this analysis. All other results replicated the ones from the standard analysis (data not shown). Further adjustment for body mass index, smoking, and alcohol use did not materially change any of our results.
Discussion
The results of our study suggest that aspirin is associated with a decreased risk of prostate cancer. This finding, in agreement with some previous studies, supports a protective effect of aspirin for this type of cancer. The pooled estimate of relative risk from seven studies that had previously assessed the effect of aspirin is 0.92 (95% CI = 0.81–1.05; Ref. 20).
This reduced risk was present with daily doses as low as 75 mg. When we analyzed the effect of aspirin in the subset of patients with a history of angina or myocardial infarction, the two most frequent indications for aspirin use in our study, we saw that the reduced risk among aspirin users was approximately of the same magnitude than overall (data not shown), further supporting that the reduced risk among aspirin users was not due to the reason why it was prescribed.
We observed a small elevated risk among short-term users of NA-NSAIDs. We hypothesize this finding could be due to the existence of prothopathic bias whereas early symptoms of prostate cancer would induce a greater prescription of this type of drug treatment. Several circumstances support this hypothesis. Firstly, the elevated risk was exclusively confined to short-term users. Indeed, the 1-year lag time analysis designed to avoid this type of bias showed no increased risk among short-term users (data not shown). Secondly, NSAIDs and paracetamol are prescribed to relieve pain but not aspirin where the vast majority is for cardioprotection. This would explain why aspirin was relatively insensitive to this bias whereas short-term users of paracetamol together with NA-NSAIDs had a modestly increased risk. Our results suggest that chronic use of NA-NSAIDs could be associated with a small reduced risk of prostate cancer. The evidence for long-term use of paracetamol was stronger, compatible with a 40% reduction. As paracetamol and NA-NSAIDs are sometimes used as substitute treatment for each other (21), we analyzed the effect of paracetamol in the subset of NA-NSAID users and vice versa. We found very similar results than the overall ones [i.e., a patent reduced risk among users of paracetamol and a less stronger reduced risk among users of NA-NSAIDs (data not shown)].
Caution must be taken in interpreting the inverse association between paracetamol use and prostate cancer. Because to our knowledge this is only the second study addressing the effect of paracetamol (Nelson et al. reported an OR = 0.81 with 95% CI = 0.25–3.23; Ref. 5), we believe it might be premature to speculate about a potential mechanism. Further observational studies are needed to either confirm or refute this association before embarking on mechanistic studies.
One would expect that those subjects who have regular contacts with the health care system might be more likely to be screened for prostate cancer and consequently more likely to be diagnosed, hence introducing some degree of detection bias. Furthermore, this group of patients would tend to be different in other ways (higher prevalence of drug use, higher prevalence of other diseases, etc.). To control for this bias, we included use of health services (general practitioner visits, specialist referrals, and hospitalizations) in our multivariate model. As expected, controlling for resources utilization resulted in lower estimates for the vast majority of analyzed variables. For instance, the estimate for history of BPH decreased from 1.79 to 1.38 after adjustment for use of health services. Additionally, we performed a subanalysis in which we included only subjects with five or more visits to the general practitioner and/or one or more specialist referrals in the 2 years prior to the index date. This subanalysis yielded identical results than the adjusted overall analysis.
Previous studies suggest that aspirin might be effective in reducing the risk in advance prostate cancer only (8, 11). We performed an additional analysis in which only cases who died within 1 year of the index date as a consequence of the disease were included. Both aspirin and NA-NSAIDs had similar estimates although slightly more protective than those obtained considering all cases (OR = 0.59, 95% CI = 0.38–0.92 and OR = 0.72, 95% CI = 0.41–1.26, respectively, for use of aspirin and NSAIDs for 1 year or longer).
In our study, information on drug exposure came from prescriptions written by general practitioners. Over-the-counter use of NSAIDs, paracetamol, or aspirin is not recorded on computer files. However, this potential misclassification would tend to be nondifferential among cases and controls, which at worse would lead us to underestimate any protective effect of these drugs. However, the impact of over-the-counter use (mainly short-term treatment) on estimates of chronic use will be of no importance (22). We could indirectly assess the magnitude of the misclassification using data from a study in the United Kingdom in a similar population in which subjects were interviewed to obtain information on use of prescribed and self-medication in the last 3 months (23). Prevalence of NA-NSAID use in this study was 18% in the control series compared with 12% in our study using a similar time window. In their study, 25% of the controls were over 80 years old whereas our study included only males and the upper age limit was 79. Because the use of NSAIDs is greater among elderly and women, no major underrecording of long-term NSAID use likely was present in our data after allowing for the different age and sex distributions. In the same study (23), the authors reported a prevalence of 20% for acetaminophen; the corresponding figure in our study was 16%.
In summary, we documented a protective effect of aspirin and paracetamol on the risk of developing prostate cancer and found some suggestion of a reduced risk with long-term use of NA-NSAIDs.
Grant support: AstraZeneca.
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.
Acknowledgments
We thank the staff at GPRD and the participating general practitioners for their collaboration and the Boston Collaborative Drug Surveillance Program for providing access to the database.