Advertisement

The Risk of Major NSAID Toxicity with Celecoxib, Ibuprofen, or Naproxen: A Secondary Analysis of the PRECISION Trial

      Abstract

      Background

      The relative safety of long-term use of nonsteroidal anti-inflammatory drugs is unclear. Patients and providers are interested in an integrated view of risk . We examined the risk of major nonsteroidal anti-inflammatory drug toxicity in the PRECISION trial.

      Methods

      We conducted a post hoc analysis of a double-blind, randomized, controlled, multicenter trial enrolling 24,081 patients with osteoarthritis or rheumatoid arthritis at moderate or high cardiovascular risk. Patients were randomized to receive celecoxib 100 to 200 mg twice daily, ibuprofen 600 to 800 mg thrice daily, or naproxen 375 to 500 mg twice daily. All patients were provided with a proton pump inhibitor. The outcome was major nonsteroidal anti-inflammatory drug toxicity, including time to first occurrence of major adverse cardiovascular events, important gastrointestinal events, renal events, and all-cause mortality.

      Results

      During follow-up, 4.1% of subjects sustained any major toxicity in the celecoxib arm, 4.8% in the naproxen arm, and 5.3% in the ibuprofen arm. Analyses adjusted for aspirin use and geographic region found that subjects in the naproxen arm had a 20% (95% CI 4-39) higher risk of major toxicity than celecoxib users and that 38% (95% CI 19-59) higher risk. These risks translate into numbers needed to harm of 135 (95% CI, 72-971) for naproxen and 82 (95% CI, 53-173) for ibuprofen, both compared with celecoxib.

      Conclusions

      Among patients with symptomatic arthritis who had moderate to high risk of cardiovascular events, approximately 1 in 20 experienced a major toxicity over 1 to 2 years. Patients using naproxen or ibuprofen experienced significantly higher risk of major toxicity than those using celecoxib.

      Keywords

      Clinical Significance
      • Naproxen and ibuprofen at moderate dosages were associated with significantly more major nonsteroidal anti-inflammatory drug toxicity events than celecoxib among older adults with arthritis and established coronary artery disease or risk factors.
      • These risks translate into numbers needed to harm of 135 (95% CI, 72-971) for naproxen and 82 (95% CI, 53-173) for ibuprofen, both compared with celecoxib.

      Introduction

      Ischemic cardiovascular adverse events associated with nonsteroidal anti-inflammatory drugs have received considerable attention.
      • Ray W.A.
      • Varas-Lorenzo C.
      • Chung C.P.
      • et al.
      Cardiovascular risks of nonsteroidal antiinflammatory drugs in patients after hospitalization for serious coronary heart disease.
      • McGettigan P.
      • Henry D.
      Cardiovascular risk with non-steroidal anti-inflammatory drugs: systematic review of population-based controlled observational studies.
      • Trelle S.
      • Reichenbach S.
      • Wandel S.
      • et al.
      Cardiovascular safety of non-steroidal anti-inflammatory drugs: network meta-analysis.
      However, nonsteroidal anti-inflammatory drug (NSAID) use also can be tied to a spectrum of clinically important toxicities. These include gastrointestinal bleeding, acute and chronic kidney injury, heart failure, hypertension, and death.
      • Wolfe M.M.
      • Lichtenstein D.R.
      • Singh G.
      Gastrointestinal toxicity of nonsteroidal antiinflammatory drugs.
      • Solomon D.H.
      • Rassen J.A.
      • Glynn R.J.
      • Lee J.
      • Levin R.
      • Schneeweiss S.
      The comparative safety of analgesics in older adults with arthritis.
      • Ruschitzka F.
      Painful lessons: COX-2 inhibitors, NSAIDs, and hypertension.
      • Gislason G.H.
      • Jacobsen S.
      • Rasmussen J.N.
      • et al.
      Risk of death or reinfarction associated with the use of selective cyclooxygenase-2 inhibitors and nonselective nonsteroidal antiinflammatory drugs after acute myocardial infarction.
      • Arfe A.
      • Scotti L.
      • Varas-Lorenzo C.
      • et al.
      Non-steroidal anti-inflammatory drugs and risk of heart failure in four European countries: nested case-control study.
      Patients' and their care teams' concerns about analgesic toxicities reach beyond a single organ system; rather, they want to know the overall safety of a given drug. Thus, an integrated examination of the relative safety of these commonly used analgesics would provide useful information for healthcare providers and patients alike. Moreover, many patients who might receive chronic NSAID have advanced age and multiple comorbidities—both risk factors for common toxicities. These considerations highlight the importance of precise information regarding the relative safety of NSAID among common subgroups, such as those defined by age, sex, underlying arthritis, NSAID dosage, known cardiovascular disease, diabetes, aspirin use, or tobacco use.
      The PRECISION trial examined the relative cardiovascular safety of celecoxib, naproxen, and ibuprofen among patients with osteoarthritis or rheumatoid arthritis who required chronic NSAID.
      • Becker M.C.
      • Wang T.H.
      • Wisniewski L.
      • et al.
      Rationale, design, and governance of Prospective Randomized Evaluation of Celecoxib Integrated Safety versus Ibuprofen Or Naproxen (PRECISION), a cardiovascular end point trial of nonsteroidal antiinflammatory agents in patients with arthritis.
      The primary trial results demonstrated that celecoxib was not inferior to the nonselective NSAID with respect to cardiovascular safety.
      • Nissen S.E.
      • Yeomans N.D.
      • Solomon D.H.
      • et al.
      Cardiovascular Safety of Celecoxib, Naproxen, or Ibuprofen for Arthritis.
      Other end points also were adjudicated in a blinded fashion, allowing for examination of the risk of major NSAID toxicities across the treatment arms.
      The current analyses were secondary post hoc assessments of data acquired prospectively from the PRECISION trial. The goal was to examine the comparative risk of major NSAID toxicity related to celecoxib, naproxen, or ibuprofen. We hypothesized that the selective COX-2 inhibitor celecoxib would have a lower risk compared with the nonselective NSAID. This analysis also examined whether various patient subgroups had differential risk of any major NSAID toxicity based on treatment assignment.

      Methods

      Study Design and Population

      The PRECISION trial was a large phase IV study, testing whether celecoxib was noninferior to naproxen and ibuprofen with respect to cardiovascular toxicity. The design of PRECISION and the main results have been published.
      • Becker M.C.
      • Wang T.H.
      • Wisniewski L.
      • et al.
      Rationale, design, and governance of Prospective Randomized Evaluation of Celecoxib Integrated Safety versus Ibuprofen Or Naproxen (PRECISION), a cardiovascular end point trial of nonsteroidal antiinflammatory agents in patients with arthritis.
      • Nissen S.E.
      • Yeomans N.D.
      • Solomon D.H.
      • et al.
      Cardiovascular Safety of Celecoxib, Naproxen, or Ibuprofen for Arthritis.
      In brief, PRECISION was an event-driven randomized controlled trial conducted at 923 centers in North America, Central America, South America, Asia, and Eastern Europe between October 2006 and April 2016. The trial could not be performed in Western Europe because of restrictions placed on prescribing of coxibs by the European Medicines Agency.
      Eligible patients were those ≥18 years of age with a clinical diagnosis of osteoarthritis or rheumatoid arthritis
      • Altman R.
      • Alarcon G.
      • Appelrouth D.
      • et al.
      The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hip.
      • Altman R.
      • Alarcon G.
      • Appelrouth D.
      • et al.
      The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hand.
      • Altman R.
      • Asch E.
      • Bloch D.
      • et al.
      Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association.
      for at least 6 months who required chronic daily NSAID therapy. Participants could be included if they had known cardiovascular disease or cardiovascular risk factors. These entry criteria included any of the following: a known history of major adverse cardiovascular events; occlusive disease of coronary and noncoronary arteries; a clinical diagnosis of diabetes; or evidence of cardiovascular risk based on concomitant risk factors, including age ≥65 years in women and >55 years in men, hypertension, dyslipidemia, left ventricular hypertrophy, microalbuminuria, urine protein/creatinine ratio >2, Ankle Brachial Index <0.9, cigarette smoking, waist-hip ratio ≥0.90, and family history of premature cardiovascular disease. Patients with any of the following were excluded: any cardiovascular event within 3 months, such as major adverse cardiovascular event, unstable angina, evidence of cardiac rhythm instability, or any major cardiovascular surgery; a planned coronary, cerebrovascular, or peripheral revascularization; New York Heart Association (NYHA) Class III or IV heart failure or known left ventricular dysfunction with ejection fraction ≤35%; active, important gastrointestinal, hepatic, renovascular or coagulation disorders; history of acute joint trauma; allergy or hypersensitivity to celecoxib, ibuprofen, naproxen, or aspirin; poor responders to disease-modifying antirheumatic drugs or oral corticosteroid treatments; and required treatment with medications excluded during the course of the study. Women were excluded if they were pregnant, might have become pregnant, or were lactating. Supplemental Table 1 (available online) shows additional selection criteria.

      Patient Involvement

      Patients were not involved in designing the PRECISION trial or the current set of analyses. The patient burden from the trial was not formally assessed.

      Study Protocol

      Patients meeting selection criteria who signed informed consent were randomized 1:1:1 to celecoxib 100 mg twice per day for osteoarthritis and up to 200 mg twice per day for rheumatoid arthritis, ibuprofen 600 to 800 mg 3 times per day, or naproxen 375 to 500 mg twice per day with administration of double dummy tablets so that all subjects received the study drug thrice daily. Treatment was double-blinded with matching placebos. Randomization was stratified by geographic region, low-dose aspirin use (yes or no), and arthritis type (osteoarthritis or rheumatoid arthritis) and implemented using an interactive voice response system to ensure masking of allocation. All enrolled patients were provided open-label esomeprazole at 20 to 40 mg per day and allowed low-dose aspirin (≤325 mg/d) for cardiovascular event prevention. Patients had visits at baseline; at months 1, 2, 4, 8, and 12; and then every 6 months through month 42. Participants had a minimum follow-up of 18 months. Follow-up visits included clinical assessments and laboratory testing, as well as identification of new adverse events or changes in cardiovascular, renal, and gastrointestinal status.

      Outcomes

      The primary end point of the PRECISION trial was the first occurrence of a composite end point consisting of cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke, identical with the primary composite end point of the Antiplatelet Trialists' Collaboration.
      • Collaboration A.T.
      Collaborative overview of randomised trials of antiplatelet therapy—I: prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients.
      The current analyses report findings for composite outcomes, all of which were centrally adjudicated in a blinded fashion. The primary outcome was major NSAID toxicity composed of major adverse cardiovascular event (which includes Antiplatelet Trialists' Collaboration events plus revascularization or hospitalization for transient ischemic attack or unstable angina), plus clinically important gastrointestinal, renal, and all-cause mortality events. Clinically important gastrointestinal events were defined as gastroduodenal hemorrhage; gastric outlet obstruction; perforation of the gastroduodenum, small bowel, or large bowel; hemorrhage of the large bowel, small bowel, or acute gastrointestinal hemorrhage of unknown origin; or symptomatic gastric or duodenal ulcer. Clinically important renal events included development of renal insufficiency or renal failure, defined on the basis of development of any of the following: serum creatinine ≥2.0 mg/dL and increase of ≥0.7 mg/dL from baseline; hospitalization for acute renal failure with a doubling of the baseline serum creatinine or hyperkalemia with ≥50% elevation in serum creatinine; or initiation of dialysis.
      The secondary outcome defined major NSAID toxicity to also include iron-deficiency anemia believed to be of gastrointestinal origin, heart failure exacerbation (ie, incident heart failure or heart failure event requiring hospitalization), and worsening hypertension requiring hospitalization.
      A Clinical Events Committee blindly adjudicated all of these end points using prespecified definitions, except incident heart failure not requiring admission, which was recorded on adverse event forms.

      Statistical Analyses

      The current analyses focus on a composite safety outcome; thus, we used the modified intention-to-treat (or “on-treatment”) population, in which events were recorded only while patients were taking their randomly allocated treatment or during the following 30 days. Analysis of this population, and not the intention-to-treat population, was preplanned after consultation with the Food and Drug Administration (FDA), based on the expectation that treatment discontinuation could dilute the signal of safety events. Superiority hypotheses were tested, and no adjustments were made for multiple comparisons. Statistical significance, P < .05 for comparisons between treatment groups or P < .10 for treatment subgroup interaction, was based on nominal P values.
      Cumulative event curves were constructed for each of the 3 treatment arms for the primary and secondary outcomes. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) comparing treatment groups for the primary and secondary outcomes of interest were calculated using Cox proportional hazards regression models, adjusting for stratification factors (geographic region), arthritis diagnosis (osteoarthritis vs rheumatoid arthritis), and low-dose aspirin use. Interactions between treatment group and potential risk factors were tested in the Cox regression models for each drug-to-drug comparison by adding the interaction term to the model. The protocol specified the censoring of analyses after 43 months. Analyses were performed using SAS by a study statistician (DMB) (version 9.4; SAS Institute Inc, Cary, NC).

      Results

      The trial screened 31,857 patients, and 24,222 underwent randomization at 926 sites in 13 countries between 2006 and 2014. Fraudulent or repeated enrollment affected 141 study patients, who were therefore excluded. Thus, 24,081 patients were analyzed; 8072 patients were assigned to celecoxib, 7969 patients were assigned to naproxen, and 8040 patients were assigned to ibuprofen (Figure 1). The baseline characteristics of patients in each treatment were similar (Table 1). Mean age of patients was 63 years, 64% were female, 90% had osteoarthritis, 10% had rheumatoid arthritis, 23% had a prior cardiovascular event, 46% used low-dose aspirin, and 99% reported using a proton pump inhibitor. Median follow-up on treatment was 20 months and was similar across the 3 treatment arms.
      Figure 1
      Figure 1CONSORT diagram illustrating the participant allocation. The final population noted contributed to the intention-to-treat analysis shown in . However, in each treatment group, some patients never took a dosage of study drug and thus are not considered in the modified intention-to-treat analyses noted in .
      Table 1Baseline Characteristics of PRECISION Trial Participants
      Total

      (n = 24,081)
      Celecoxib

      (n = 8072)
      Naproxen

      (n = 7969)
      Ibuprofen

      (n = 8040)
      Age, mean (SD), y63.2 (9.4)63.0 (9.5)63.3 (9.4)63.2 (9.4)
      Female gender, n (%)15,445 (64.1)5175 (64.1)5096 (63.9)5174 (64.4)
      BMI, mean (SD), kg/m232.6 (7.3)32.7 (7.3)32.6 (7.3)32.5 (7.4)
      Aspirin use, n (%)11,065 (45.9)3701 (45.8)3652 (45.8)3712 (46.2)
      Indication for NSAID, n (%)
      Osteoarthritis21,645 (89.9)7259 (89.9)7178 (90.1)7208 (89.7)
      Rheumatoid arthritis2436 (10.1)813 (10.1)791 (9.9)832 (10.3)
      Cardiovascular risk category, n (%)
       Primary prevention18,601 (77.2)6209 (76.9)6186 (77.6)6206 (77.2)
       Secondary prevention5480 (22.8)1863 (23.1)1783 (22.4)1834 (22.8)
      History of diabetes, n (%)8496 (35.3)2843 (35.2)2768 (34.7)2885 (35.9)
      History of hypertension, n (%)18,744 (77.8)6296 (78.0)6145 (77.1)6303 (78.4)
      History of dyslipidemia, n (%)15,048 (62.5)5080 (62.9)4966 (62.3)5002 (62.2)
      Current smoker, n (%)5000 (20.8)1689 (20.9)1631 (20.5)1680 (20.9)
      Current statin use, n (%)12,978 (53.9)4367 (54.1)4304 (54.0)4307 (53.6)
      Current DMARD use, n (%)1758 (7.3)572 (7.1)602 (7.6)584 (7.3)
      Current corticosteroid use, n (%)3101 (12.9)1014 (12.6)1066 (13.4)1021 (12.7)
      Systolic BP, mean (SD), mm Hg125.2 (10.5)125.3 (10.5)125.0 (10.7)125.4 (10.4)
      Diastolic BP, mean (SD), mm Hg75.5 (8.0)75.5 (8.0)75.4 (8.0)75.5 (7.9)
      Prior ulcer disease, n (%)1336 (5.6)460 (5.7)436 (5.5)440 (5.5)
      Serum creatinine, mg/dL, mean (SD)0.90 (0.22)0.90 (0.22)0.90 (0.22)0.90 (0.23)
      BP = blood pressure; DMARD = disease-modifying antirheumatic drug; NSAID = nonsteroidal anti-inflammatory drug; SD = standard deviation.
      N (%) unless otherwise noted.
      The frequency of any major NSAID toxicity across all 3 treatment arms was 4.7% and 6.2% for the extended outcome (Table 2). The rates of outcomes differed across the 3 treatment arms: 4.1% of celecoxib users experienced a major NSAID toxicity, compared with 4.8% of naproxen users (P = .02) and 5.3% of ibuprofen users (P < .001). Examination of the time until any major NSAID toxicity shows that events began early and increased at consistent rates across the follow-up (Figure 2). For all 3 treatment arms, the adjusted HRs for the primary outcome show significantly higher risks for both naproxen users compared with the celecoxib users (HR, 1.20; 95% CI, 1.04-1.39, P = .02) and for ibuprofen users (HR, 1.38; 95% CI, 1.19-1.59, P < .001) (Figure 2a). Trends for the secondary outcome were similar (Figure 2b). These HRs translate into numbers needed to harm for the primary major NSAID toxicity as follows: naproxen compared with celecoxib 135 (95% CI, 72-971) and ibuprofen compared with celecoxib 82 (95% CI, 53-173). Naproxen users experienced a reduced risk of major NSAID toxicities compared with ibuprofen users for the primary outcome (HR, 0.84; 95% CI, 0.73-0.98; P = .048) and for the extended secondary outcome (HR, 0.87; 95% CI, 0.76-0.98; P = .048).
      Table 2Frequency of Major NSAID Toxicity in the PRECISION Trial Modified Intent-to-Treat Population
      Total

      N = 23,953
      Celecoxib

      N = 8030
      Naproxen

      n = 7933
      Ibuprofen

      N = 7990
      Major NSAID toxicity
      Major toxicity includes major adverse cardiovascular events, serious gastrointestinal events (gastroduodenal hemorrhage, gastric outlet obstruction, gastroduodenal small or large bowel perforation, large or small bowel hemorrhage, acute gastrointestinal hemorrhage, symptomatic gastric or duodenal ulcer or anemia defined as a decrease in hemoglobin ≥2 g/dL or hematocrit ≥10% with no clinical evidence of acute gastrointestinal bleed), renal events (development of renal insufficiency or renal failure, defined on the basis of development of any of the following: serum creatinine ≥2.0 mg/dL and increase of ≥0.7 mg/dL from baseline; hospitalization for acute renal failure with a doubling of the baseline serum creatinine or hyperkalemia with ≥50% elevation in serum creatinine; or initiation of dialysis), and all-cause mortality.
      1136 (4.7)328 (4.1)383 (4.8)425 (5.3)
       Major adverse CV events784 (3.3)247 (3.1)253 (3.2)284 (3.6)
       Renal events177 (0.7)42 (0.5)62 (0.8)73 (0.9)
       Serious gastrointestinal events138 (0.6)27 (0.3)52 (0.7)59 (0.7)
       All-cause mortality205 (0.9)53 (0.7)79 (1.0)73 (0.9)
      Expanded major NSAID toxicity
      Expanded major toxicity includes major toxicity; plus heart failure exacerbations, such as incident heart failure or heart failure hospitalizations; hypertension hospitalizations; or iron-deficiency anemia of gastrointestinal origin.
      1496 (6.2)418 (5.2)516 (6.5)562 (7.0)
       Heart failure exacerbations234 (1.0)63 (0.8)79 (1.0)92 (1.2)
       Hypertension admissions94 (0.4)25 (0.3)32 (0.4)37 (0.5)
       Iron-deficiency anemia151 (0.6)27 (0.3)66 (0.8)58 (0.7)
      CV = cardiovascular; NSAID = nonsteroidal anti-inflammatory drug.
      The totals are based on patients who took at least 1 dosage of study drug and thus are included in the modified intention-to-treat population. The cells in the table represent n (%). Major adverse cardiovascular events are defined as the first occurrence of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, revascularization, or hospital for transient ischemic attack.
      * Major toxicity includes major adverse cardiovascular events, serious gastrointestinal events (gastroduodenal hemorrhage, gastric outlet obstruction, gastroduodenal small or large bowel perforation, large or small bowel hemorrhage, acute gastrointestinal hemorrhage, symptomatic gastric or duodenal ulcer or anemia defined as a decrease in hemoglobin ≥2 g/dL or hematocrit ≥10% with no clinical evidence of acute gastrointestinal bleed), renal events (development of renal insufficiency or renal failure, defined on the basis of development of any of the following: serum creatinine ≥2.0 mg/dL and increase of ≥0.7 mg/dL from baseline; hospitalization for acute renal failure with a doubling of the baseline serum creatinine or hyperkalemia with ≥50% elevation in serum creatinine; or initiation of dialysis), and all-cause mortality.
      Expanded major toxicity includes major toxicity; plus heart failure exacerbations, such as incident heart failure or heart failure hospitalizations; hypertension hospitalizations; or iron-deficiency anemia of gastrointestinal origin.
      Figure 2
      Figure 2Cumulative incidence of major toxicity in PRECISION trial. The HRs are estimated in Cox proportional hazards regression models that include aspirin use and type of arthritis. A, The primary outcome of major NSAID toxicity. B, The expanded major NSAID toxicity outcome. CI = confidence interval; BID = 2 times per day; TID = 3 times per day.
      Subgroup analyses stratified the population across several a priori–determined patient characteristics that might be expected to associate with the risk of major NSAID toxicity (Figure 3). The interaction terms for several subgroups by NSAID stratum had P values < .10, including cardiovascular risk enrollment strata (primary with higher relative risks than secondary), low-dose aspirin use (no with higher relative risks than yes; only for naproxen), and diabetes (yes with higher relative risks than no; only for naproxen). The P values between other subgroups (ie, age, gender, arthritis diagnosis, tobacco use, NSAID dose escalation, and history of gastrointestinal ulceration) and treatment did not suggest interaction effects. These results were similar for the secondary outcome (Supplemental Table 2, available online).
      Figure 3
      Figure 3Forest plot demonstrates the risk of major NSAID toxicity by patient subgroups. CI = confidence interval; HR = hazard ratio.

      Discussion

      Pain represents the most common reason for visits to physicians, and healthcare providers in the United States write more than 100 million NSAID prescriptions annually.
      • Health I.
      National Prescription Audit, Oral NSAID Market.
      • Hsiao C.J.
      • Cherry D.K.
      • Beatty P.C.
      • Rechtsteiner E.A.
      National Ambulatory Medical Care Survey: 2007 summary.
      Although potential NSAID toxicities are numerous, the recent attention on opioid abuse and its consequences will ensure that NSAID remain a mainstay of treatment. Many patients who require chronic analgesics have multiple comorbidities, placing them at relatively high risk of drug-related adverse events. Therefore, we conducted secondary post hoc analyses of the PRECISION trial, a large randomized controlled trial that compared the overall safety of celecoxib, naproxen, and ibuprofen. The main trial found that celecoxib was not inferior to the 2 nonselective NSAID with respect to cardiovascular safety. We extended these analyses to examine a composite safety end point of major NSAID toxicity, including cardiovascular, gastrointestinal, renal, and all-cause mortality. The patients randomized to naproxen or ibuprofen had significantly higher risk of major NSAID toxicity than those receiving celecoxib; results for all of the components of the composite were directionally similar. In the overall study population, the numbers needed to harm were 135 for naproxen and 82 for ibuprofen compared with celecoxib.
      Beyond these overall results, the stratified analyses identified potentially important subgroups of patients with differential risk. Subgroups that had higher relative risks during treatment with nonselective NSAID include those without prior cardiovascular event (vs a prior cardiovascular event), those not receiving low-dose aspirin (vs receiving low-dose aspirin) for naproxen use, and those with diabetes (vs without diabetes) for naproxen use. It is not clear why these subgroups had differential relative risks for the nonselective NSAID. There was no subgroup interaction observed by age, gender, arthritis diagnosis, tobacco use, NSAID dose escalation, and history of gastrointestinal ulceration. These subgroup analyses were post hoc and should be considered hypothesis-generating.
      • Sun X.
      • Briel M.
      • Walter S.D.
      • Guyatt G.H.
      Is a subgroup effect believable? Updating criteria to evaluate the credibility of subgroup analyses.
      Several further observations should be highlighted. First, we observed an approximately 10% risk of any major NSAID toxicity associated with use of these agents in patients with prior cardiovascular events. This is not a new finding but one worth keeping in mind when considering prescribing.
      • Gislason G.H.
      • Jacobsen S.
      • Rasmussen J.N.
      • et al.
      Risk of death or reinfarction associated with the use of selective cyclooxygenase-2 inhibitors and nonselective nonsteroidal antiinflammatory drugs after acute myocardial infarction.
      Second, although many previous studies have focused on the gastrointestinal risks of NSAID, in this trial (in which 99% of patients used a proton pump inhibitor) serious gastrointestinal events made up <15% of the composite NSAID toxicity. It appears that the epidemiology of “NSAID gastropathy” may be changing with more frequent use of proton pump inhibitors. Third, all of the relative risks for the components of the major NSAID toxicity outcome were directionally similar. Finally, the absolute risks associated with the 3 interventions appear different by subgroup. These subgroup analyses were post hoc and should be considered hypothesis-generating. Further work will attempt to better risk stratify patients on the basis of easily accessible clinical characteristics.
      In striving to achieve precision medicine, providers aim to determine which treatments suit individual patients best. Safety weighs heavily in the choice of pharmacologic agents. The analyses presented in this article, as well as others from the PRECISION trial, provide new information about the relative safety of nonsteroidal anti-inflammatory drugs because the drugs compared had different safety profiles. As well, subtle differences exist between patients who experience toxicities. This article focused on a composite of major NSAID toxicities because patients and providers want to know which drug is “safest” overall, not just liable to provoke particular adverse events.

      Study Strengths and Limitations

      A major strength of the current analysis is the derivation of the data from a large randomized controlled trial with many individually adjudicated end points. Each of the component end points examined were directionally similar with respect to relative risks compared with celecoxib. As well, the population enrolled includes typical patients who require chronic analgesics and have common comorbidities. This study also has limitations. The PRECISION trial was not designed specifically to answer the question asked in the current post hoc analyses. The trial also had a higher than anticipated nonretention rate. Although the rates of noncompletion were similar across treatment arms, this could influence the validity of the current analyses. Disproportionate dose up-titration across the treatment arms, more for naproxen and ibuprofen than celecoxib, could have affected the presented results. This disparity was anticipated because celecoxib doses could not be increased in the osteoarthritis subgroup because of regulatory restrictions in the United States.

      Conclusions

      This secondary analysis of the PRECISION trial found an increased risk of major NSAID toxicity in patients randomized to naproxen or ibuprofen compared with celecoxib, with relative numbers needed to harm ranging from 82 to 135. These results have considerable public health implications, because large numbers of patients use these drugs. Results from this analysis and others from PRECISION demonstrate that the toxicities of NSAID vary across drugs and patient subgroups. The mechanistic reasons for these differences require further study but likely involve differential selectivity of the agents used on cyclooxygenase isoforms. Warnings from drug regulators, recommendations from professional societies, and prescribing by providers should take these differences into account and tailor NSAID selection to optimize patient outcomes.

      Appendix

      Supplemental A CONSORT 2010 Checklist of Information to Include When Reporting a Randomized Trial
      Tabled 1
      Section/TopicItem NoChecklist itemReported on page No
      Title and abstract
      1aIdentification as a randomized trial in the titleP1
      1bStructured summary of trial design, methods, results, and conclusions (for specific guidance see CONSORT for abstracts)P3-4
      Introduction
      Background and objectives2aScientific background and explanation of rationaleP6-7
      2bSpecific objectives or hypothesesP6-7
      Methods
      Trial design3aDescription of trial design (eg, parallel, factorial) including allocation ratioP7-8
      3bImportant changes to methods after trial commencement (eg, eligibility criteria), with reasonsNone
      Participants4aEligibility criteria for participantsP7; Supplemental Table 1
      4bSettings and locations where the data were collectedP7
      Interventions5The interventions for each group with sufficient details to allow replication, including how and when they were actually administeredP9
      Outcomes6aCompletely defined prespecified primary and secondary outcome measures, including how and when they were assessedP9-10
      6bAny changes to trial outcomes after the trial commenced, with reasonsNone
      Sample size7aHow sample size was determinedP7
      7bWhen applicable, explanation of any interim analyses and stopping guidelinesNone
      Randomization:
       Sequence generation8aMethod used to generate the random allocation sequenceP9
      8bType of randomization; details of any restriction (eg, blocking and block size)P9
       Allocation concealment mechanism9Mechanism used to implement the random allocation sequence (eg, sequentially numbered containers), describing any steps taken to conceal the sequence until interventions were assignedP9
       Implementation10Who generated the random allocation sequence, who enrolled participants, and who assigned participants to interventionsP9
      Blinding11aIf done, who was blinded after assignment to interventions (eg, participants, care providers, those assessing outcomes) and howP9
      11bIf relevant, description of the similarity of interventionsP9
      Statistical methods12aStatistical methods used to compare groups for primary and secondary outcomesP10-11
      12bMethods for additional analyses, such as subgroup analyses and adjusted analysesP10-11
      Results
      Participant flow (a diagram is strongly recommended)13aFor each group, the numbers of participants who were randomly assigned, received intended treatment, and were analyzed for the primary outcomeFigure 1
      13bFor each group, losses and exclusions after randomization, together with reasonsFigure 1
      Recruitment14aDates defining the periods of recruitment and follow-upP12
      14bWhy the trial ended or was stoppedP7
      Baseline data15A table showing baseline demographic and clinical characteristics for each groupTable 1
      Numbers analyzed16For each group, number of participants (denominator) included in each analysis and whether the analysis was by original assigned groupsTable 2
      Outcomes and estimation17aFor each primary and secondary outcome, results for each group, and the estimated effect size and its precision (eg, 95% CI)Figure 1
      17bFor binary outcomes, presentation of both absolute and relative effect sizes is recommendedFigure 2, P12-13
      Ancillary analyses18Results of any other analyses performed, including subgroup analyses and adjusted analyses, distinguishing prespecified from exploratoryP13, Figure 3
      Harms19All important harms or unintended effects in each group (for specific guidance see CONSORT for harms)Table 2
      Discussion
      Limitations20Trial limitations, addressing sources of potential bias, imprecision, and, if relevant, multiplicity of analysesP15-16
      Generalizability21Generalizability (external validity, applicability) of the trial findingsP15-16
      Interpretation22Interpretation consistent with results, balancing benefits and harms, and considering other relevant evidenceP16
      Other information
      Registration23Registration number and name of trial registryP1
      Protocol24Where the full trial protocol can be accessed, if availableSupplement
      Funding25Sources of funding and other support (eg, supply of drugs), role of fundersP17-18
      CI = confidence interval.
      Supplemental Table 1Full Inclusion and Exclusion Criteria
      Inclusion Criteria
      1Men and women aged ≥18 y at the time of consent.
      2If the subject is female and of childbearing potential, or <2 y postmenopause, she must have been using adequate contraception since her last menses and will use adequate contraception during the study. She should not be lactating and must have a negative pregnancy test within 24 h before receiving the first dose of study medication. Women <2 y postmenopause are considered of childbearing potential for the purpose of this study, and adequate contraception being considered medically acceptable is contraception such as hormonal contraception, intrauterine device, or barrier method plus spermicide.
      3Clinical diagnosis of OA or RA with a duration of at least 6 mo.
      4All subjects must have required a chronic analgesic regimen for at least 6 mo and taken chronic analgesic therapy ≥50% of the time.

      • Subjects receiving adequate pain management of their arthritis signs and symptoms with acetaminophen therapy are not eligible.
      • Subjects with RA who are receiving DMARD or oral corticosteroid therapy (≤20 mg prednisone or equivalent) in addition to their chronic analgesic regimen should have been on the same DMARD or corticosteroid for 3 mo and on a stable dosing regimen for 1 mo (defined as doses and frequency of administration are unchanged). DMARD or corticosteroid therapy alone does not qualify as chronic analgesic regimen; subjects receiving DMARD or corticosteroid therapy must also require a chronic analgesic regimen;
      5In the investigator's opinion, the subject requires and is eligible for chronic, daily therapy with an NSAID to control arthritis signs and symptoms;
      6Subject with established or at high risk for CVD defined as 1 of the following (a-d):
      • a.
        Coronary disease
        • History of stable angina; or
        • History of MI, unstable angina, PCI, or CABG surgery at least 3 mo before randomization; or
        • An angiographic stenosis >50% by visual estimation at catheterization.
      • b.
        Occlusive disease of noncoronary arteries
        • History of TIA or ischemic stroke at least 3 mo before randomization, or
        • Angiographic or ultrasound diagnosis of carotid artery stenosis ≥50%, or
        • History of carotid endarterectomy at least 3 mo before randomization, or
        • Symptomatic peripheral arterial disease (eg, intermittent claudication), or
        • Other arterial surgery or angioplasty for atherosclerotic vascular disease at least 3 mo before randomization
      • c.
        Diabetes mellitus: clinical diagnosis of type I or II diabetes.
        • Women must require current insulin treatment.
        • Men can have type I or II with or without insulin therapy.
      • d.
        High risk of atherosclerotic vascular disease
        • Women require at least 2 of the following 3:
          • Age ≥65 y
          • History of hypertension
          • Current smoking (defined as any cigarette smoking within the past 30 d)
        • Men require at least 3 of the following:
          • Age >55 y
          • History of hypertension;F
          • History of dyslipidemia (defined as LDL >160 mg/dL [4144 mmol/L] or high-density lipoprotein <35 mg/dL [0.906 mmol/L]). Subjects currently undergoing lipid-lowering therapy with a statin drug, fibrate, prescription omega-3-acid ethyl esters (eg, Omacor [Reliant Pharmaceuticals, New Jersey], Lovaza [GlazoSmithKline, North Carolina]), or prescription niacin (≥1000 mg/d) will automatically meet this criterion.
          • Family history of premature CVD (defined as history of myocardial infarction, angina pectoris, heart failure, cardiac death, or coronary revascularization [including PCI and CABG]; does NOT include history of hypertension), stroke, carotid endarterectomy, or other arterial surgery or angioplasty for atherosclerotic vascular disease in a parent, grandparent, or sibling with first symptom onset or diagnosis before age 55 y for men and 65 y for women;
          • Current smoking (defined as any cigarette smoking within the past 30 d);
          • History of microalbuminuria, urine protein/creatinine ratio >2;
          • Left ventricular hypertrophy as evidenced by electrocardiogram or echocardiography;
          • Documented ABI <0.9;
          • Waist–hip ratio ≥0.90.
      7In the opinion of the investigator, the subject is willing and able to comply with scheduled visits, treatment plan, laboratory tests, and other trial procedures for the duration of the study; evidence of a personally signed and dated informed consent document indicating that the subject (or legal representative) has been informed of all pertinent aspects of the trial.
      Exclusion Criteria
      1The subject has had acute joint trauma with active symptoms that may interfere with the assessment of arthritis;
      2The subject has a planned surgical or other invasive procedure to be performed during the course of the study;
      3The subject has:

      • Known allergy or hypersensitivity to celecoxib, ibuprofen or naproxen, or
      • Has experienced asthma, urticaria or allergic–type reactions after taking sulfonamides, *PPIs, lactose, or NSAIDs, or
      • Known true allergy to aspirin (acetylsalicylic acid, ie, has experienced urticaria, rash, severe difficulty breathing, asthma, nasal polyps, sinusitis/rhinitis)
        • *Subjects with hypersensitivity to or who cannot tolerate esomeprazole or PPIs may be treated with other gastroprotective agents permitted per protocol (ie, histamine-2 blockers) in place of esomeprazole provided by the study at the discretion of the investigator.
      • In addition, because of data suggesting that:
        • The concomitant use of PPIs in patients treated with clopidogrel may attenuate the benefits of antiplatelet therapy, and
        • Patients who are aged ≥50 y, have been treated with high doses of PPIs or have had long term (≥1 y) treatment with PPIs may be at increased risk for fractures of the hip, wrist, and spine
        • Subjects who are taking clopidogrel or believed to be at risk for fracture also may be treated with other gastroprotective agents permitted per protocol (ie, histamine-2 blockers) instead of esomeprazole or other PPIs at the discretion of the investigator. However, subjects taking clopidogrel should not be treated with cimetidine.
      • Otherwise, subjects should be treated with esomeprazole provided by the study;
      4The subject has received treatment with rheumatologic disease-modifying agents or oral corticosteroids and who has not been stable (defined as on the same medications for 3 mo and on the same dosing regimen for 1 mo before randomization);
      5The subject is receiving treatment with oral corticosteroids at a daily dose >20 mg prednisone or equivalent;
      6The subject requires and is receiving treatment with >325 mg aspirin/d;
      7The subject is currently taking or has a significant likelihood of requiring treatment during the study period with medication not permitted by the study protocol, including warfarin or other vitamin K antagonist anticoagulant or lithium;
      8The subject has a documented MI or stroke within 3 mo before randomization;
      9The subject has undergone CABG surgery or any major surgery (cardiac or noncardiac) within 3 mo before randomization;
      10The subject has planned coronary (PCI or CABG), cerebrovascular (ie, carotid endarterectomy/PTA with or without stent), or peripheral (ie, peripheral arterial bypass, PTA, or aortic aneurysm repair) revascularization at the time of study screening; in case of planned revascularization, the subject can be rescreened no sooner than 3 mo after revascularization;
      11The subject has an unstable condition defined as any of the following:

      • a.
        Unstable angina (AHA/ACC definition) within 3 mo before randomization
      • b.
        Uncontrolled hypertension (defined as systolic BP >140 mm Hg or diastolic BP >90 mm Hg at the baseline visit
      12The subject has evidence of cardiac electrophysiologic instability, including uncontrolled complex ventricular arrhythmia, uncontrolled atrial fibrillation or flutter, or uncontrolled supraventricular tachycardias within 3 mo before randomization. The presence of an implantable defibrillator is not a contraindication to enrollment;
      13The subject has NYHA Class III or IV CHF or known left ventricular dysfunction with ejection fraction ≤35%;
      14The subject has been diagnosed with or has been treated for esophageal, gastric, pyloric channel, or duodenal ulceration within 60 d before randomization;
      15The subject has a history of GI perforation, obstruction, or bleed within 6 mo before randomization;
      16The subject has:

      • Inflammatory bowel disease (eg, Crohn's disease or ulcerative colitis; subjects with a history of ulcerative colitis who have had a total colectomy and are considered cured are eligible; subjects with partial colectomy are excluded), or
      • Diverticulitis active at the time of randomization or within the 6 mo before randomization, or
      • Diverticulosis with any known history of bleeding, or
      • History or likelihood of bariatric procedure (eg, restrictive, malabsorptive, or combination procedures such as gastric banding, bypass, and Roux–en–Y), or
      • Other known, active, significant GI, hepatic, renal, or coagulation disorders.
      17The subject has an aspartate aminotransferase (AST [SGOT]), alanine aminotransferase (ALT [SGPT]), or blood urea nitrogen exceeding 2.0 times the upper limit of normal; creatinine exceeding 1.7 mg/dL (150 µmol/L) in men or 1.5 mg/dL (133 µmol/L) in women;
      18The subject has an active malignancy of any type. Subjects who have a history of basal cell or squamous cell carcinoma of the skin that has been successfully treated are eligible. Subjects with a history of other malignancies that have been successfully treated and who have no evidence of recurrence for at least 5 y before randomization are also eligible;
      19The subject has any medical (including known history of major hematologic, renal, vascular, or hepatic abnormalities) or psychologic/psychiatric condition (including documented unstable or uncontrolled alcoholism or drug abuse) or laboratory abnormality that may increase the risk associated with trial participation or investigational product administration or may interfere with the interpretation of trial results and in the judgment of the investigator would make the subject inappropriate for entry into this trial;
      20The subject:

      • was previously randomized in this study, or
      • participated in other studies within 30 d before screening
      • may not participate in other studies during participation in this study;
      21The subject has ongoing litigation or compensation related to his/her overall arthritic condition including past or current long-term treatment.
      ABI = Ankle–Brachial Index; ACC = American College of Cardiology; AHA = American Heart Association; AST = aspartate aminotransferase; ALT = alanine aminotransferase; BP = blood pressure; CABG = coronary artery bypass grafting; CHF = congestive heart failure; CVD = cardiovascular disease; DMARD = disease-modifying antirheumatic drug; GI = gastrointestinal; MI = myocardial infarction; NSAID = nonsteroidal anti-inflammatory drug; NYHA = New York Heart Association; OA = osteoarthritis; OR = odds ratio; PCI = percutaneous coronary intervention; PPI = proton pump inhibitor; PTA = percutaneous transluminal angioplasty; RA = rheumatoid arthritis; SGOT = serum glutamic-oxaloacetic transaminase; SGPT = serum glutamic pyruvic transaminase; TIA = transient ischemic attack.
      Supplemental Table 2Secondary Outcome in Modified Intention-to-Treat Population Stratified by Patient Characteristics
      StratumCelecoxibNaproxenIbuprofen
      Events/nEvents/nHR (95% CI)
      HRs comparing naproxen with celecoxib and ibuprofen with celecoxib.
      Interaction PEvents/nHR (95% CI)
      HRs comparing naproxen with celecoxib and ibuprofen with celecoxib.
      Interaction P
      Gender
       Male197/2880246/28651.26 (1.04-1.52).92262/28481.44 (1.20-1.74).99
       Female221/5150270/50681.28 (1.07-1.53)300/51421.44 (1.21-1.71)
      Age, y
       65+227/3553304/36291.31 (1.11-1.56).46314/36791.46 (1.23-1.73).79
       <65191/4477212/43041.19 (0.98-1.45)248/43111.41 (1.16-1.70)
      Arthritis
       OA367/7220446/71461.25 (1.09-1.43).38492/71621.44 (1.26-1.65).99
       RA51/81070/7871.43 (1.00-2.06)70/8281.45 (1.01-2.08)
      Aspirin
       Yes262/3683286/36401.12 (0.95-1.33).027311/36951.26 (1.07-1.49).017
       No156/4347230/42931.52 (1.24-1.86)251/42951.73 (1.42-2.12)
      CV risk group
       Primary222/6175321/61561.48 (1.25-1.76).013344/61651.66 (1.40-1.96).019
       Secondary196/1855195/17771.05 (0.86-1.28)218/18251.21 (1.00-1.47)
      Ulcer history
       Yes33/45850/4341.69 (1.09-2.64).1846/4371.70 (1.08-2.66).48
       No385/7572466/74991.24 (1.09-1.42)516/75531.42 (1.25-1.63)
      Smoking
       Yes93/168399/16271.13 (0.85-1.50).36112/16661.26 (0.96-1.66).31
       No325/6340416/62991.31 (1.13-1.52)449/63161.48 (1.29-1.71)
      Diabetes
       Yes168/2830229/27521.47 (1.20-1.79).09244/28641.57 (1.29-1.91).32
       No247/5114286/50961.17 (0.98-1.38)318/50471.37 (1.16-1.62)
      Dose escalation
       Yes31/477305/43981.24 (0.78-1.98).90324/44511.32 (0.83-2.10).28
       No387/7551211/35331.43 (1.20-1.70)238/35371.75 (1.48-2.06)
      CI = confidence interval; CV = cardiovascular; HR = hazard ratio; OA = osteoarthritis; RA = rheumatoid arthritis.
      * HRs comparing naproxen with celecoxib and ibuprofen with celecoxib.

      References

        • Ray W.A.
        • Varas-Lorenzo C.
        • Chung C.P.
        • et al.
        Cardiovascular risks of nonsteroidal antiinflammatory drugs in patients after hospitalization for serious coronary heart disease.
        Circ Cardiovasc Qual Outcomes. 2009; 2: 155-163
        • McGettigan P.
        • Henry D.
        Cardiovascular risk with non-steroidal anti-inflammatory drugs: systematic review of population-based controlled observational studies.
        PLoS Med. 2011; 8: e1001098
        • Trelle S.
        • Reichenbach S.
        • Wandel S.
        • et al.
        Cardiovascular safety of non-steroidal anti-inflammatory drugs: network meta-analysis.
        BMJ. 2011; 342: c7086
        • Wolfe M.M.
        • Lichtenstein D.R.
        • Singh G.
        Gastrointestinal toxicity of nonsteroidal antiinflammatory drugs.
        N Engl J Med. 1999; 340: 1888-1899
        • Solomon D.H.
        • Rassen J.A.
        • Glynn R.J.
        • Lee J.
        • Levin R.
        • Schneeweiss S.
        The comparative safety of analgesics in older adults with arthritis.
        Arch Intern Med. 2011; 170: 1968-1976
        • Ruschitzka F.
        Painful lessons: COX-2 inhibitors, NSAIDs, and hypertension.
        Curr Hypertens Rep. 2007; 9: 41-44
        • Gislason G.H.
        • Jacobsen S.
        • Rasmussen J.N.
        • et al.
        Risk of death or reinfarction associated with the use of selective cyclooxygenase-2 inhibitors and nonselective nonsteroidal antiinflammatory drugs after acute myocardial infarction.
        Circulation. 2006; 113: 2906-2913
        • Arfe A.
        • Scotti L.
        • Varas-Lorenzo C.
        • et al.
        Non-steroidal anti-inflammatory drugs and risk of heart failure in four European countries: nested case-control study.
        BMJ. 2016; 354: i4857
        • Becker M.C.
        • Wang T.H.
        • Wisniewski L.
        • et al.
        Rationale, design, and governance of Prospective Randomized Evaluation of Celecoxib Integrated Safety versus Ibuprofen Or Naproxen (PRECISION), a cardiovascular end point trial of nonsteroidal antiinflammatory agents in patients with arthritis.
        Am Heart J. 2009; 157: 606-612
        • Nissen S.E.
        • Yeomans N.D.
        • Solomon D.H.
        • et al.
        Cardiovascular Safety of Celecoxib, Naproxen, or Ibuprofen for Arthritis.
        N Engl J Med. 2016; 375: 2519-2529
        • Altman R.
        • Alarcon G.
        • Appelrouth D.
        • et al.
        The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hip.
        Arthritis Rheum. 1991; 34: 505-514
        • Altman R.
        • Alarcon G.
        • Appelrouth D.
        • et al.
        The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hand.
        Arthritis Rheum. 1990; 33: 1601-1610
        • Altman R.
        • Asch E.
        • Bloch D.
        • et al.
        Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association.
        Arthritis Rheum. 1986; 29: 1039-1049
        • Collaboration A.T.
        Collaborative overview of randomised trials of antiplatelet therapy—I: prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients.
        BMJ. 1994; 308: 81-106
        • Health I.
        National Prescription Audit, Oral NSAID Market.
        2013
        • Hsiao C.J.
        • Cherry D.K.
        • Beatty P.C.
        • Rechtsteiner E.A.
        National Ambulatory Medical Care Survey: 2007 summary.
        Natl Health Stat Report. 2010; 27: 1-32
        • Sun X.
        • Briel M.
        • Walter S.D.
        • Guyatt G.H.
        Is a subgroup effect believable? Updating criteria to evaluate the credibility of subgroup analyses.
        BMJ. 2010; 340: c117