Portal de Periódicos da CAPES

Non‐steroidal anti‐inflammatory drug use and risk of venous thromboembolism

2011; Elsevier BV; Volume: 9; Issue: 7 Linguagem: Inglês

10.1111/j.1538-7836.2011.04354.x

1538-7933

Morten Schmidt, Christian Fynbo Christiansen, Erzsébet Horváth‐Puhó, Robert J. Glynn, Kenneth J. Rothman, Henrik Toft Sørensen,

Atrial Fibrillation Management and Outcomes

Journal of Thrombosis and HaemostasisVolume 9, Issue 7 p. 1326-1333 ORIGINAL ARTICLEFree Access Non-steroidal anti-inflammatory drug use and risk of venous thromboembolism M. SCHMIDT, M. SCHMIDT Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus Department of Cardiology, Aarhus University Hospital, Skejby, DenmarkSearch for more papers by this authorC. F. CHRISTIANSEN, C. F. CHRISTIANSEN Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this authorE. HORVÁTH-PUHÓ, E. HORVÁTH-PUHÓ Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this authorR. J. GLYNN, R. J. GLYNN Divisions of Pharmacoepidemiology and Pharmacoeconomics and Preventive Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MASearch for more papers by this authorK. J. ROTHMAN, K. J. ROTHMAN RTI Health Solutions, Research Triangle Institute, Research Triangle Park, NC, USASearch for more papers by this authorH. T. SØRENSEN, H. T. SØRENSEN Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this author M. SCHMIDT, M. SCHMIDT Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus Department of Cardiology, Aarhus University Hospital, Skejby, DenmarkSearch for more papers by this authorC. F. CHRISTIANSEN, C. F. CHRISTIANSEN Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this authorE. HORVÁTH-PUHÓ, E. HORVÁTH-PUHÓ Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this authorR. J. GLYNN, R. J. GLYNN Divisions of Pharmacoepidemiology and Pharmacoeconomics and Preventive Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MASearch for more papers by this authorK. J. ROTHMAN, K. J. ROTHMAN RTI Health Solutions, Research Triangle Institute, Research Triangle Park, NC, USASearch for more papers by this authorH. T. SØRENSEN, H. T. SØRENSEN Department of Clinical Epidemiology, Aarhus University Hospital, AarhusSearch for more papers by this author First published: 19 May 2011 https://doi.org/10.1111/j.1538-7836.2011.04354.xCitations: 40 Morten Schmidt, Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Allé 43-45, DK-8200 Aarhus N, Denmark.Tel.: +45 8942 4800; fax: +45 8942 4801.E-mail: msc@dce.au.dk AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract Summary. Background: The association between the use of non-selective non-steroidal anti-inflammatory drugs (NSAIDs) or cyclooxygenase-2-selective inhibitors (COX2Is) and the risk of venous thromboembolism (VTE) remains unclear. Objectives: To examine this association. Patients/Methods: We conducted a population-based case–control study in northern Denmark (population of 1.7 million). Using the National Patient Registry, we identified patients with a first hospital VTE diagnosis during 1999–2006 (n = 8368) and their comorbidities. For each case, we selected 10 controls (n = 82 218) matched by age and sex. From the prescription database, we ascertained the use of NSAIDs at the time of diagnosis (current use) or before (recent use), and comedications. Current use was further classified as new use (first-ever prescription redemption within 60 days before diagnosis date) or long-term use. We used odds ratios from a logistic regression model to estimate incidence rate ratios (IRRs) with 95% confidence intervals (CIs). Results: As compared with no use, the adjusted IRR associating current non-selective NSAID use with VTE was 2.51 (95% CI 2.29–2.76), and that for current COX2I use was 2.19 (95% CI 1.99–2.41). Recent users had substantially smaller increases than current users. The adjusted IRRs among long-term users were 2.06 for non-selective NSAIDs (95% CI 1.85–2.29) and 1.92 for COX2Is (95% CI 1.72–2.15). Similarly increased risks were found for unprovoked VTE (occurrence in the absence of pregnancy, cancer, major trauma, fracture or surgery within 3 months preceding the VTE), deep vein thrombosis, pulmonary embolism, and individual NSAIDs. Conclusions: The use of non-selective NSAIDs or COX2Is was associated with a two-fold or more increased risk of VTE. Introduction Non-steroidal anti-inflammatory drugs (NSAIDs) are widely employed to treat inflammatory conditions and pain [1]. By inhibiting cyclooxygenase (COX)-1-mediated production of prostaglandins [1], non-selective NSAIDs are known to cause gastrointestinal toxicity [1]. An alternative is provided by COX-2-selective inhibitors (COX2Is), which are available in the form of older or newer agents [2]. The newer COX2Is (coxibs), introduced into clinical practice in 1998, were developed as NSAIDs with an improved gastrointestinal side effect profile [1]. The safety of both traditional NSAIDs (i.e. older COX2Is and non-selective NSAIDs) and coxibs is controversial, because several of these drugs increase the risk of arterial thromboembolic events [3]. Whether the use of NSAIDs is related to the risk of venous thrombosis remains unclear [4, 5]. Venous thrombosis occurs predominantly in the deep vessels of the lower limbs (deep vein thrombosis [DVT]) and is a common disease process affecting more than one per 1000 persons each year in Western populations [6-8]. It is associated with serious complications such as pulmonary embolism (PE) and post-thrombotic syndrome [6, 8]. DVT and PE are collectively referred to as venous thromboembolism (VTE) [6]. VTE incidence increases exponentially with age for both men and women [6], with a recurrence rate as high as 30% within 10 years [6]. The classic risk factors for VTE include immobilization, cancer, fractures, pregnancy, and recent surgery [7, 8]. We hypothesized that prothrombotic drugs such as non-aspirin NSAIDs increase the risk of VTE [3]. Whereas conflicting results exist for traditional NSAIDs [4, 5], no data exist on the clinical association between coxibs and VTE. Any increased VTE risk associated with NSAID use would have major clinical and public health implications, especially in the elderly, where the prevalence of NSAID use and the incidence of VTE are high. We conducted a large population-based case–control study examining the association between the use of non-selective NSAIDs or COX2Is and the risk of VTE. Methods Setting We conducted this study in northern Denmark, which has 1.7 million inhabitants (approximately 30% of the Danish population). Since 1998, complete computerized prescription records have been available for this population. Our study period began on 1 January 1999, thus providing at least 1 year of prescription history for all study participants. We included subjects to 31 December 2006. The Danish National Health Service provides universal tax-supported healthcare, guaranteeing unfettered access to general practitioners and hospitals, and partial reimbursement for prescribed medications, including NSAIDs [9]. Linkage among national registries is possible in Denmark by use of the unique central personal registry number assigned to each Danish citizen at birth and to residents upon immigration [10]. VTE We used the Danish National Patient Registry [11], covering all Danish hospitals, to identify all VTE patients defined by an incident inpatient or outpatient diagnosis of lower limb DVT or PE during the study period. This registry contains data on dates of admission and discharge, all discharge diagnoses from non-psychiatric hospitals after 1977, and emergency room and outpatient clinic visits after 1995 [11]. Each discharge is associated with one primary diagnosis and one or more secondary diagnoses classified according to the International Classification of Diseases, 8th revision (ICD-8) until the end of 1993, and the 10th revision (ICD-10) thereafter [11]. We identified both primary and secondary diagnoses of DVT (ICD-8, 451.00; ICD-10, I80.1–3) and PE (ICD-8, 450.99; ICD-10, I26). To reduce potential coding errors, we excluded patients who had an outpatient PE diagnosis with no subsequent inpatient VTE diagnosis. In a secondary analysis, we excluded VTE cases with the following classic risk factors: pregnancy, major trauma, fracture, surgery within 3 months preceding VTE, pre-existing cancer, or a new cancer diagnosis within 3 months after VTE [12]. The date of the first VTE diagnosis was taken as the index date for cases. Controls We used the Danish Civil Registration System to select 10 population controls for each case, matched on age and sex [10]. This registry has maintained data on all vital statistics – including date of birth, change of address, date of emigration, and exact date of death – for the Danish population since 1968, with daily updates [10]. We selected controls using risk-set sampling: controls had to be alive and at risk for a first VTE hospitalization on the index date of the case to whom each was matched. Controls were assigned an index date identical to that of corresponding cases. NSAID use We used the regional prescription database [13] to identify prospectively all NSAID prescriptions filled by cases and controls before their index date. Pharmacies in Denmark are equipped with electronic accounting systems, which are primarily used to secure reimbursement from the National Health Service. For each filled prescription, the patient’s personal registry number, the type and amount of drug prescribed according to the Anatomical Therapeutic Chemical (ATC) classification system and the date on which the drug was dispensed are transferred electronically from the pharmacies to the prescription database [13]. Except for ibuprofen in the 200 mg per tablet dose, all non-aspirin NSAIDs are available by prescription only [9]. Regular users of ibuprofen are typically registered in the database, because the cost is partly refunded when the drug is prescribed by a physician. We identified prescriptions for non-selective non-aspirin NSAIDs (ibuprofen, naproxen, ketoprofen, dexibuprofen, piroxicam, tolfenamic acid, and indomethacin), older COX2Is (diclofenac, etodolac, nabumeton, and meloxicam), and newer COX2Is (celecoxib, rofecoxib, valdecoxib, parecoxib, and etoricoxib) [2]. Because of overlapping COX-2 selectivity, we collapsed the groups of older and newer COX2Is into one group named COX2Is [2]. In primary analyses, the exposures consisted of the NSAID subclasses of non-selective NSAIDs and COX2Is. In addition, preplanned analyses were conducted for the six individual NSAIDs most frequently prescribed in the study population. The ATC codes are provided in Data S1. We defined current NSAID users as persons who filled their most recent NSAID prescription within 60 days before their index date. We chose an exposure window of 60 days to capture most current users, as NSAID prescriptions are seldom provided for more than 60 days at a time in Denmark [14]. Because some side effects may arise shortly after therapy initiation and inclusion of long-term users may lead to underestimation of these complications [15], we further categorized current users into two groups: new users, defined by having filled their first-ever prescription within 60 days before their index date; and long-term users, defined by having filled additional prescriptions 61–365 days before their index date. The long-term user group was of interest because the longer period of use should eliminate any protopathic bias, i.e. the association between new NSAID use and prodromal symptoms related to an incipient occurrence of VTE [16]. We defined persons who had filled their most recent prescription between 61 and 365 days before their index dates as recent users. We defined persons with no filled NSAID prescriptions 365 days before their index date as non-users (reference group). Other patient characteristics We obtained information from 1977 from the Danish National Patient Registry [11] on inpatient and outpatient diagnoses of the following conditions that may be associated with NSAID use: cardiovascular disease, chronic obstructive pulmonary disease (COPD) or asthma, diabetes mellitus, liver disease, obesity, osteoarthritis, osteoporosis, renal failure, rheumatoid arthritis, and systemic connective tissue disease. To account further for potential unmeasured confounding from frailty and immobility, we included recent hospital admission as a dichotomous variable defined by any inpatient diagnosis of other diseases within 3 months before the index date. To increase the sensitivity of the diagnoses for diabetes mellitus, pulmonary disease, or cardiovascular disease, we used the prescription database to obtain data on any use since 1998 of the following drugs: antidiabetic drugs (oral antidiabetics and insulin), respiratory drugs, and cardiovascular drugs (angiotensin-converting enzyme inhibitors or angiotensin II receptor inhibitors, aspirin, β-blockers, calcium channel blockers, clopidogrel, diuretics, nitrates, statins, and other antihypertensives). We also obtained data on concurrent use of antipsychotics, hormone replacement therapy, oral glucocorticoids, and vitamin K antagonists, because these drugs affect the VTE risk [5, 7, 8]. The ICD and ATC codes are provided in Data S1. Statistical analysis Initially, we created contingency tables for the main study variables, from which we calculated the frequency of cases and controls in categories of exposures, and medical and demographic variables. We then stratified the contingency tables according to each of the potential confounding factors listed in Table 1. Table 1. Characteristics of cases with composite or unprovoked venous thromboembolism (VTE) and population controls from northern Denmark, 1999–2006 Composite VTE Unprovoked VTE Cases (%)n = 8368 Controls (%)n = 82 218 Cases (%)n = 4691 Controls (%)n = 40 152 Female sex 4493 (53.7) 44 143 (53.7) 2446 (52.1) 20 627 (51.4) Age (years) < 55 1922 (23.0) 19 115 (23.2) 1227 (26.2) 11 430 (28.5) 55-70 2621 (31.3) 25 889 (31.5) 1423 (30.3) 12 597 (31.4) ≥ 71 3825 (45.7) 37 214 (45.3) 2041 (43.5) 16 125 (40.2) Median age (IQR) 69 (56-78) 68 (56-78) 67 (54-78) 66 (52-77) Classic risk factors Cancer* 1788 (21.4) 7099 (8.6) – – Pregnancy† 47 (0.6) 151 (0.2) – – Surgery† 2431 (29.1) 4027 (4.9) – – Trauma or fracture† 722 (8.6) 1548 (1.9) – – Other comorbidities‡ Cardiovascular disease§ 5138 (61.4) 39 868 (48.5) 2746 (58.5) 17 765 (44.2) COPD or asthma§ 1994 (23.8) 12 531 (15.2) 1090 (23.2) 5696 (14.2) Diabetes mellitus§ 649 (7.8) 4857 (5.9) 345 (7.4) 2194 (5.5) Liver disease 103 (1.2) 413 (0.5) 54 (1.2) 180 (0.4) Obesity 383 (4.6) 1533 (1.9) 196 (4.2) 663 (1.7) Osteoarthritis 1270 (15.2) 8136 (9.9) 598 (12.7) 3435 (8.6) Osteoporosis 259 (3.1) 1870 (2.3) 113 (2.4) 800 (2.0) Renal failure 159 (1.9) 556 (0.7) 64 (1.4) 225 (0.6) Rheumatoid arthritis 201 (2.4) 1031 (1.3) 106 (2.3) 408 (1.0) Systemic connective tissue disease 277 (3.3) 1419 (1.7) 139 (3.0) 583 (1.5) Recent hospital admission¶ 2075 (24.8) 3563 (4.3) 582 (12.4) 779 (1.9) NSAID use** Ibuprofen 684 (8.2) 2323 (2.8) 380 (8.1) 1074 (2.7) Naproxen 37 (0.4) 224 (0.3) 16 (0.3) 116 (0.3) Diclofenac 385 (4.6) 1413 (1.7) 191 (4.1) 662 (1.6) Etodolac 105 (1.3) 475 (0.6) 54 (1.2) 210 (0.5) Celecoxib 115 (1.4) 431 (0.5) 47 (1.0) 183 (0.5) Rofecoxib 98 (1.2) 352 (0.4) 46 (1.0) 151 (0.4) Comedication use** Antipsychotics 370 (4.4) 1906 (2.3) 216 (4.6) 825 (2.1) Hormone replacement therapy 488 (5.8) 4213 (5.1) 265 (5.6) 1848 (4.6) Oral glucocorticoids 832 (9.9) 2092 (2.5) 384 (8.2) 872 (2.2) Vitamin K antagonists 221 (2.6) 1599 (1.9) 97 (2.1) 676 (1.7) COPD, chronic obstructive pulmonary disease; IQR, interquartile range; NSAID, non-steroidal anti-inflammatory drug. *Pre-existing cancer or a cancer diagnosis within 3 months after the index date. †Any inpatient or outpatient diagnosis within 3 months before the index date. ‡Any inpatient or outpatient diagnosis since 1977. §Any inpatient or outpatient diagnosis since 1977 or any filled prescription since 1998. ¶Any inpatient diagnosis, within 3 months before the index date, other than the diseases listed in Table 1. **Prescription redemption within 60 days before the index date (except for vitamin K antagonists [90 days] and hormone replacement therapy [120 days]). Next, we used unconditional logistic regression with adjustment for the matching factors of age and sex to estimate odds ratios with 95% confidence intervals (CIs) for VTE among current, new, long-term and recent users of non-selective NSAIDs or COX2Is as compared with non-users. Subjects with current use of both non-selective NSAIDs and COX2Is (51 cases and 86 controls) were included in each subclass analysis. Because we used risk-set sampling of controls, the odds ratios estimate the incidence rate ratios (IRRs) [17]. Afterwards, we fitted models with adjustments for the potential confounding factors listed in Table 1. To examine the effects of different exposure definitions, we repeated the analyses for exposure windows of 15, 30, 90 and 120 days. Stratified analysis was performed on subgroups of sex, age, cancer, cardiovascular disease, diabetes mellitus, musculoskeletal or connective tissue disease (osteoarthritis, rheumatoid arthritis, or systemic connective tissue disease), obesity, trauma or fracture, and recent hospital admission. To determine whether IRRs differed between all (composite) VTEs and unprovoked VTEs, between VTE subtypes, or between individual NSAIDs, the analyses were repeated for unprovoked VTE, DVT, PE, and the six individual NSAIDs most frequently prescribed. To evaluate clinically relevant heterogeneity across drugs in VTE risk, we added a direct comparison of VTE risk among the individual NSAIDs, using ibuprofen as a referent exposure. Patients with concomitant use of ibuprofen and another NSAID were excluded from this analysis. Because all patients had a need for pain relief, this comparison probably reduced confounding by indication. We identified the tablet dose from the last filled prescription, and examined the impact associated with low and high tablet dose. We quantified the influence of potential unmeasured confounding on the observed association by means of a rule-out approach [18]. We estimated how strongly a single unmeasured binary confounder would need to be associated with NSAID use and VTE to fully explain our findings. We illustrated this association graphically. We assumed, as a worst case scenario, that the prevalence of such a confounder was 30% in the population and that 10% of the population used NSAIDs. Analyses were performed with sas version 9.1 (SAS Institute, Cary, NC, USA). Results Patient characteristics Characteristics are provided in Table 1 for the 8368 patients with VTE and the 82 218 population controls. Slightly less than half of the cases were male and half were 70 years or older; 48.5% of controls and 61.4% of cases had been diagnosed previously with cardiovascular disease or had used cardiovascular drugs. COPD or asthma, diabetes mellitus, obesity and musculoskeletal and connective tissue diseases were also more common among cases than controls. Among all VTE patients, 4691 had unprovoked VTE. The distribution of characteristics among unprovoked VTE patients was similar to that for the overall group. Risk of VTE The age-adjusted and sex-adjusted IRRs for VTE among current users were 3.24 (95% CI 2.98–3.52) for non-selective NSAIDs and 3.10 (95% CI 2.84–3.38) for COX2Is as compared with no use (Table 2). The crude IRRs were similar to the age- and sex-adjusted IRRs. The matching factors were thus not strongly associated with the exposure. Table 2. Incidence rate ratios for venous thromboembolism (VTE) associated with non-steroidal anti-inflammatory drug (NSAID) use Incidence rate ratio (95% confidence interval) Composite VTE Unprovoked VTE No. of cases/controls Unadjusted* Adjusted† No. of cases/controls Unadjusted* Adjusted† No use 5483/66 311 1 (reference) 1 (reference) 3202/32 677 1 (reference) 1 (reference) Non-selective NSAIDs Current use‡ 794/2971 3.24 (2.98–3.52) 2.51 (2.29–2.76) 438/1365 3.28 (2.92–3.67) 2.71 (2.40–3.05) New use§ 257/543 5.78 (4.97–6.72) 4.56 (3.85–5.40) 152/257 6.19 (5.05–7.59) 5.43 (4.37–6.74) Long-term use¶ 537/2428 2.68 (2.43–2.95) 2.06 (1.85–2.29) 286/1108 2.62 (2.29–3.00) 2.13 (1.84–2.45) Recent use** 904/6282 1.75 (1.63–1.89) 1.44 (1.33–1.56) 456/3085 1.54 (1.38–1.71) 1.38 (1.24–1.54) COX2Is Current use‡ 709/2760 3.10 (2.84–3.38) 2.19 (1.99–2.41) 341/1240 2.76 (2.43–3.13) 2.15 (1.88–2.46) New use§ 198/546 4.40 (3.73–5.19) 3.23 (2.69–3.89) 109/242 4.63 (3.68–5.82) 4.18 (3.29–5.32) Long-term use¶ 511/2214 2.77 (2.50–3.06) 1.92 (1.72–2.15) 232/998 2.31 (1.99–2.67) 1.71 (1.46–2.00) Recent use** 806/5092 1.91 (1.76–2.07) 1.41 (1.30–1.54) 403/2340 1.75 (1.56–1.95) 1.46 (1.30–1.64) COX2I, cyclooxygenase-2-selective inhibitor. *Adjusted for the matching factors of age and sex. †Additional adjustments for the potential confounders listed in Table 1 (i.e. cancer, pregnancy, surgery, trauma, fracture, cardiovascular disease, chronic obstructive pulmonary disease, asthma, diabetes mellitus, liver disease, obesity, osteoarthritis, osteoporosis, renal failure, rheumatoid arthritis, systemic connective tissue disease, other inpatient hospital admission within 3 months before VTE, and current use of antipsychotics, hormone replacement therapy, oral glucocorticoids, and vitamin K antagonists). The classic VTE risk factors (cancer, pregnancy, surgery, trauma, and fracture) were not included, per definition, in the model for unprovoked VTE. ‡Prescription redemption within 60 days before the index date. §Current users who filled their first-ever prescription within 60 days before their index date. ¶Current users who filled their first prescription between 61 and 365 days before their index date. ** Most recent prescription redemption within 61-365 days before the index date. Adjusting for the potential confounders in Table 1 reduced the IRRs to 2.51 (95% CI 2.29–2.76) for non-selective NSAIDs and 2.19 (95% CI 1.99–2.41) for COX2Is. Among new users, confounder adjustment reduced the IRRs for VTE from 5.78 (95% CI 4.97–6.72) to 4.56 for non-selective NSAIDs (95% CI 3.85–5.40) and from 4.40 (95% CI 3.73–5.19) to 3.23 for COX2Is (95% CI 2.69–3.89). Among long-term users, the adjusted IRRs for VTE were 2.06 for non-selective NSAIDs (95% CI 1.85–2.29) and 1.92 for COX2Is (95% CI 1.72–2.15). Although the effect estimates were substantially smaller than for current use, recent use of non-selective NSAIDs (adjusted IRR 1.44, 95% CI 1.33–1.56) and COX2Is (adjusted IRR 1.41, 95% CI 1.30–1.54) was also moderately associated with an increased VTE risk. For all user definitions, the corresponding effect estimates were similarly increased for unprovoked VTE (Table 2) and VTE subtypes (Table 3). The IRRs were higher for DVT than for PE (Table 3). Table 3. Incidence rate ratios for deep vein thrombosis or pulmonary embolism associated with non-steroidal anti-inflammatory drug (NSAID) use Incidence rate ratio (95% confidence interval) Deep vein thrombosis Pulmonary embolism No. of cases/controls Unadjusted Adjusted No. of cases/controls Unadjusted Adjusted No use 3486/43 304 1 (reference) 1 (reference) 1997/23 007 1 (reference) 1 (reference) Non-selective NSAIDs Current use 568/1907 3.71 (3.36–4.10) 2.98 (2.67–3.32) 226/1064 2.45 (2.11–2.85) 1.74 (1.47–2.06) New use 194/354 6.87 (5.75–8.22) 5.72 (4.70–6.96) 63/189 3.87 (2.90–5.17) 2.58 (1.85–3.59) Long-term use 374/1553 2.99 (2.66–3.37) 2.36 (2.07–2.69) 163/875 2.15 (1.81–2.55) 1.55 (1.28–1.88) Recent use 596/4085 1.83 (1.66–2.00) 1.53 (1.38–1.69) 308/2197 1.63 (1.43–1.85) 1.30 (1.13–1.50) COX2Is Current use 473/1724 3.40 (3.05–3.78) 2.46 (2.19–2.77) 236/1036 2.63 (2.26–3.05) 1.76 (1.48–2.09) New use 139/340 5.10 (4.17–6.23) 3.93 (3.14–4.90) 59/206 3.32 (2.47–4.44) 2.19 (1.56–3.06) Long-term use 334/1384 2.97 (2.62–3.37) 2.10 (1.83–2.41) 177/830 2.45 (2.07–2.90) 1.64 (1.35–2.00) Recent use 539/3214 2.08 (1.88–2.29) 1.55 (1.39–1.73) 267/1878 1.64 (1.43–1.87) 1.20 (1.03–1.40) COX2I, cyclooxygenase-2-selective inhibitor. See user definitions and description of the unadjusted and adjusted model in the text or in Table 2. Current use of individual NSAIDs was associated with composite and unprovoked VTE (Table 4), for both high-dose and low-dose tablets (data not shown), as well as DVT and PE (Table S1), with a magnitude of the association similar to the results for the overall NSAID subclasses. In the direct drug comparison (Table 5), naproxen use was associated with a substantially lower risk of composite VTE (adjusted IRR 0.54, 95% CI 0.36–0.80) and unprovoked VTE (adjusted IRR 0.39, 95% CI 0.23–0.68) than ibuprofen. Table 4. Incidence rate ratios for composite or unprovoked venous thromboembolism (VTE) associated with individual non-steroidal anti-inflammatory drug use Incidence rate ratio (95% confidence interval) Composite VTE Unprovoked VTE Unadjusted Adjusted Unadjusted Adjusted No use 1 (reference) 1 (reference) 1 (reference) 1 (reference) Ibuprofen Current use 3.57 (3.26–3.90) 2.79 (2.52–3.08) 3.62 (3.20–4.09) 2.98 (2.62–3.39) Recent use 1.81 (1.67–1.96) 1.50 (1.37–1.64) 1.59 (1.42–1.79) 1.43 (1.27–1.61) Naproxen Current use 2.01 (1.42–2.86) 1.52 (1.03–2.25) 1.43 (0.85–2.42) 1.23 (0.71–2.13) Recent use 1.59 (1.24–2.04) 1.28 (0.97–1.68) 1.45 (1.03–2.04) 1.16 (0.81–1.68) Etodolac Current use 2.63 (2.12–3.25) 1.96 (1.55–2.47) 2.52 (1.86–3.40) 1.87 (1.36–2.57) Recent use 2.35 (1.97–2.80) 1.74 (1.43–2.12) 1.69 (1.29–2.23) 1.38 (1.03–1.84) Diclofenac Current use 3.30 (2.94–3.71) 2.38 (2.09–2.71) 2.95 (2.50–3.48) 2.41 (2.03–2.87) Recent use 1.93 (1.75–2.13) 1.47 (1.32–1.63) 1.83 (1.59–2.10) 1.58 (1.37–1.82) Celecoxib Current use 3.14 (2.55–3.87) 1.89 (1.49–2.39) 2.44 (1.77–3.38) 1.79 (1.27–2.52) Recent use 2.21 (1.84–2.65) 1.54 (1.26–1.89) 2.20 (1.69–2.86) 1.58 (1.20–2.08) Rofecoxib Current use 3.27 (2.61–4.10) 2.26 (1.75–2.91) 2.93 (2.10–4.08) 2.12 (1.49–3.04) Recent use 1.98 (1.63–2.40) 1.32 (1.06–1.64) 1.51 (1.12–2.02) 1.15 (0.84–1.56) See user definitions and description of the unadjusted and adjusted model in the text or in Table 2. Table 5. Incidence rate ratios for composite or unprovoked venous thromboembolism (VTE) comparing current use of individual non-steroidal anti-inflammatory drugs with ibuprofen as referent exposure Incidence rate ratio (95% confidence interval) Composite VTE Unprovoked VTE Unadjusted Adjusted Unadjusted Adjusted Ibuprofen 1 (reference) 1 (reference) 1 (reference) 1 (reference) Naproxen 0.56 (0.39–0.80) 0.54 (0.36–0.80) 0.39 (0.23–0.66) 0.39 (0.23–0.68) Etodolac 0.83 (0.66–1.05) 0.84 (0.65–1.08) 0.80 (0.57–1.11) 0.76 (0.54–1.07) Diclofenac 0.91 (0.79–1.06) 0.86 (0.74–1.01) 0.81 (0.66–0.99) 0.83 (0.67–1.03) Celecoxib 1.01 (0.80–1.28) 0.84 (0.65–1.09) 0.77 (0.54–1.10) 0.76 (0.53–1.10) Rofecoxib 1.13 (0.88–1.45) 1.01 (0.77–1.33) 0.95 (0.66–1.36) 0.91 (0.62–1.33) See current user definition and description of the unadjusted and adjusted model in the text or in Table 2. From the stratified analysis (Table S2), sex and age seemed to modify the rate ratio estimates for VTE associated with the use of non-selective NSAIDs and COX2Is, with the highest effect among males and persons younger than 55 years. Consistent with the principle that the effect estimates were lower among those at higher baseline risk, the estimates were slightly lower in strata of patients with cardiovascular disease, diabetes mellitus, obesity, osteoarthritis, rheumatoid arthritis, systemic connective tissue disease, and trauma or fracture. We estimated that an unmeasured confounder that is four times more frequent among NSAID users than non-users would need to increase the risk of VTE by a factor of 17 or more to explain our findings fully, if no increased risk actually existed. Figure 1 illustrates this association for current use of COX2Is. Even stronger confounders would be needed to explain the findings for current use of non-selective NSAIDs or new use of either subclass. The adjusted IRR for current use of non-selective NSAIDs or COX2Is decreased with increasing exposure windows (Table S3). Figure 1Open in figure viewerPowerPoint Required strength of an unmeasured confounder. Sensitivity analysis illustrating how strongly an unmeasured confounder would need to be associated with NSAID use (prevalence ratio for exposure–confounder association [PREC]) and venous thromboembolism (VTE) (relative risk of the disease in patients with the confounder [RRCD]) to fully explain our