Cardiovascular Risk of Celecoxib in 6 Randomized Placebo-controlled Trials: The Cross Trial Safety Analysis Scott D. Solomon, MD, Janet Wittes, PhD, Ernest Hawk, MD, MPH for the Celecoxib Cross Trials Safety Analysis Investigators On behalf of the individual trial investigators, I’m pleased to present the results of the Celecoxib Cross Trials Safety analysis. A manuscript based on these results will be available immediately following this presentation in Circulation. Manuscript available simultaneously online in Circulation

This research was funded entirely by the National Cancer Institute DISCLOSURES No Disclosures This research was funded entirely by the National Cancer Institute I have no conflicts of interest to disclose. This analysis was funded entirely by the National Cancer Institute.

Background Observational studies and randomized controlled trials have reported increased cardiovascular risk associated with cyclooxygenase-2 (cox-2) inhibitors (coxibs) 1,2,3,4 Strong biologic basis for this risk supported by abundant basic research5,6,7 Most clinical studies compared coxibs with active comparators in short-term arthritis trials Observational studies and randomized controlled trials have reported increased cardiovascular risk associated with cyclooxygenase-2 inhibitors, also known as coxibs. Moreover, numerous experimental studies have supported a strong biologic basis for this risk. While most clinical studies with these agents have compared coxibs with active comparators in relatively short-term arthritis trials, initial evidence of increased cardiovascular risk associated with rofecoxib, valdecoxib and celecoxib emerged from longer duration, placebo-controlled trials designed to study the role of coxibs in other therapeutic areas. 1McGettigan JAMA 2006; 2Graham et al. Lancet 2005; 3Bresalier et al. NEJM 2005; 4Solomon et al. NEJM 2005 5McAddam et al. PNAS 1999; 6Fitzgerald NEJM 2001; 7Fitgerald et al. NEJM 2004

Background In December 2004, Interim results from the Adenoma Prevention with Celecoxib (APC) trial results led to stopping drug in that trial and in 5 other long-term trials comparing celecoxib to placebo: The Prevention of Sporadic Adenomatous Polyps (PreSAP) trial1 The Alzheimer’s Disease Anti-inflammatory Prevention Trial (ADAPT)2 The MA-27 Breast Cancer Trial, The Celecoxib Diabetic Macular Edema (CDME) trial The Celecoxib/Selenium Trial. FDA hearing resulted in Black Box Warning. Celecoxib is the only available cox-2 inhibitor in US. In December 2004, two months after the withdrawal of rofecoxib because of increased cardiovascular risk observed in a polyp-prevention trial, similar findings in the NCI- and Pfizer-sponsored Adenoma Prevention with Celecoxib (APC) trial led to stopping drug administration in that and five other long-term trials comparing celecoxib to placebo: These trials, which are listed here, were the the Prevention of Sporadic Adenomatous Polyps (PreSAP) trial, the Alzheimer’s Disease Anti-inflammatory Prevention Trial (ADAPT), the MA-27 breast cancer recurrence trial, the Celecoxib Diabetic Macular Edema (CDME) trial, and The Celecoxib/Selenium Trial. Subsequent to the FDA hearing in february 2005, a black box warning was placed on celecoxib, which remains the only cox-2 inhibitor available currently in the US. 1 Arber et al. NEJM 2006; 2 ADAPT Invest. PLOS 2006

Low event rates Lead to Challenges in Risk Assessment with Coxibs Low Precision of the estimates Inability to test observational and RCT data suggesting coxib-associated CV risk may be dose related dose and interval may be important in CV risk.1 Inability to assess whether CV risk associated with celecoxib varies by baseline CV risk The relatively low cardiovascular event rate in all coxib cardiovascular risk analyses to date has reduced the certaintly of the estimates of risk and has thwarted our ability to answer some important yet vexing questions: While Observational data and our previous analyses of the APC and PreSAP studies have suggested that coxib-associated cardiovascular risk may be dose related, and that both dose and dosing interval may be important factors in cardiovascular risk, these relationships have been obscured by the low event rate. Moreover, Prior trials have had too few cardiovascular events to assess whether the cardiovascular risk associated with celecoxib use risk varies according to a patient’s baseline cardiovascular risk. 1Solomon et al Circulation 2006

Objective To understand more fully the cardiovascular risk profile associated with long-term use of celecoxib NCI commissioned and funded analysis of long-term placebo controlled trials with a planned follow-up  3 years To understand more fully the cardiovascular risk profile associated with long-term use of celecoxib, the National Cancer Institute asked investigators in four long-term, placebo-controlled trials with a planned follow-up of 3 years or greater to submit their data for central adjudication and combined analysis by the same process used to analyze the APC and PreSAP studies.

Selection of Studies Randomized, double-blind, placebo-controlled trials Planned follow-up of at least 3 years Source documentation available for adjudication 4 trials + APC and PreSAP fulfilled these criteria: ADAPT MA-27 CDME Celecoxib/Selenium Trial Prior to collecting data we specified that trials included in the analysis would have the following characterisitics: (1) they would be randomized, double blind, and placebo-controlled 2) the planned follow-up for each participant would be at least three years, and 3) source documentation would be available for central adjudication. All trials fulfilling these two criteria studied the therapeutic potential of celecoxib for a condition other than arthritis. In addition to the APC and PreSAP trials that we had previously adjudicated and analyzed we identified 4 trials that fulfilled these criteria.

Methodology Each study submitted patient-level data: Baseline data Outcomes Adverse events A blinded adjudication team identified all potential cardiovascular events from broad list of SAEs and AEs Requested source documentation for all relevant events All potential cardiovascular events were adjudicated by two reviewers masked to treatment allocation Categorized all deaths Adjudicated all non-fatal events Each study submitted patient-level data, including baseline data, outcomes data and adverse events. A blinded adjudication team identified all potential cardiovascular events from a broad list of SAEs and AEs, and requested source documentation for all relevant events All potential cardiovascular events were adjudicated by two reviewers masked to treatment allocation; we categorized all deaths and adjudicated all non-fatal events Endpoint Definitions: Solomon et al. NEJM 2005

Endpoints The following endpoints were adjudicated: Primary endpoint: Death (cardiovascular or non-cardiovascular) Myocardial Infarction Stroke Hospitalization for heart Failure Thromboembolic event Other cardiovascular Primary endpoint: CV death, MI, stroke, heart failure or thromboembolic event We adjudicated the following endpoints: Death, which was classified as cardiovascular or noncardiovascular, myocardial infarction, stroke, hosp for heart failure, thromboembolic event, and other cardiovascular. For the purposes of this analysis the primary endpoint was a composite of cardiovascular death, MI, stroke, Heart failure, or thromboembolic event.

Statistical Analysis Intention-to-treat Time-to-event analyses for each study Calculated incidence of each outcome, rate (per 1000 pt-yrs) by Rx group Cox models and KM curves Pooled (meta) analysis: Estimated hazard ratios calculated from the average of the log-hazard ratio for each individual trial weighted by the inverse of its variance Sensitivity of method assessed by standard Mantel-Haenszel pooled odds ratios and Cox models stratified by study. Analyses adjusted for baseline cardiovascular risk Pooled analyses assessed overall risk and dosing regimen-related risk We performed an intention to treat, time to event analysis for each study, and calculated the incidence of each outcome, and generate Cox models and Kaplan Meier curves. For the pooled analysis, we estimated hazard ratios calculated from the average of the log-hazard ratio for each individual trial weighted by the inverse of its variance, and the sensitivity of this method was assessed by both standard mantel-haenzel pooled odds ratios and Cox models stratified by study. We adjusted our analyses for baseline cardiovascular risk, and assessed both overall and dose-regimen related risk.

Dose and Baseline Risk Studies were grouped according to dose regimen: 400mg once daily (2 studies) 200mg twice daily (2 studies, 3 groups) 400mg twice daily (2 studies) We tested for interaction between dose regimen and celecoxib risk We created a 3-category risk score using a modified Framingham Risk model conforming to the availability of data from these studies: Low: No known risk factor Moderate: One of following , age > 75, hypertension, hyperlipidemia, current smoker, low-dose ASA High: Diabetes, prior CV disease, or ≥ 2 risk factors in “moderate” category Studies were grouped according to dose regimen We tested for interaction between dose regimen and celecoxib risk We created a 3-category risk score using a modified Framingham Risk model conforming to the availability of data from these studies, according to the following scheme: Patients were consider Low risk if they had No known risk factor Moderate risk if they had One of following , age > 75, hypertension, hyperlipidemia, current smoker, low-dose ASA High risk if they had Diabetes, prior history of CV disease, or two or more of the risk factors in the “moderate” category

Placebo-Controlled Trials Study N Sponsor Disease being Studied Celecoxib Dose Planned follow-up time APC 2035 NCI and Pfizer Colorectal polyps Celecoxib 200mg BID, celecoxib 400mg BID, or placebo 3+ Years PreSAP 1561 Pfizer Colorectal Polyps Celecoxib 400mg QD or placebo MA27 1635 NCI, NCI Canada, & Pfizer Breast Cancer Recurrence celecoxib 400 mg BID or placebo ADAPT 1809 NIA Alzheimer’s disease and cognitive decline Celecoxib 200mg BID or Naproxen sodium 220 mg BID, or placebo Up to 7 years CDME 86 NEI Diabetic Retinopathy Celecoxib 200mg BID or placebo Cel/Sel 824 NCI Celecoxib 400 mg QD or placebo 3-5 Years The six trials studied are listed here and with the exception of the PreSAP trial were all sponsored or jointly sponsored by the National Institutes of Health. The number of patients enrolled in each trial ranged from 86 in the CDME trial to 2035 in the APC trial. Each of the trials were designed to study the role of celecoxib for a condition other than arthritis, and these included prevention of colorectal adenoma recurrence, breast cancer recurrence, alzheimer’s disease, and diabetic retinopathy. The doses of celecoxib tested ranged from 400mg once daily in two studies, 200mg Twice Daily in 2 studies and 400 mg twice daily in two studies. Combined there were 16070 patient-years of follow-up.

Baseline Characteristics (%) ADAPT APC CDME MA27 PreSAP Cel/Sel Total # enrolled Pt-Years 1809 3530 2035 6234 86 101 1635 695 1561 4141 824 1369 7950 16070 Age, mean (SD) 75 ± 4 59 ± 10 59 ± 9 64 ± 9 60 ± 10 63 ± 9 64 ± 10 Male 54 68 62 66 50 White race 97 92 67 94 89 96 93 Diabetes 7.4 9.5 100 6.1 10 7.5 9.2 HTN or med 40 41 34 37 36 38 Hyperlipidemia or med 33 55 17 28 Current smoker 3 ? 24 16 14 Low-dose ASA use 31 45 Prior CV event 13 1.2 7 12 Low CV risk 32 19 Moderate CV risk 26 29 23 27 High CV risk 59 47 51 Baseline characteristics are shown here. Overall the mean age was 64, and ranged from 59 in the APC and CDME trials to 75 in the ADAPT study. Approximately half of the overall patients were male, the majority were caucasian. 9% of patients were diabetic; of note all of the patients in the retinopathy study were diabetic. 38% were hypertensive or on hypertensive medications; 28% were hyperlipidemic or on lipid lowering drugs;14% were current smokers. 31% were low-dose aspirin users. Based on our modified risk score, we categorized 28% of patients as low-risk, 27% as moderate risk, and 45% as high risk for CV events. There was some heterogeneity between the trials – most importantly, the ADAPT patients were older, by design, and the retinopathy study patients were all diabetic, also by design.

Event Numbers, Rates and Hazard Ratios Events (Event Rate per 1000/pt-yrs) Hazard Ratio 400mg QD placebo celecoxib PreSAP 12/628 (7.2) 23/933 (9.4) 1.3 (0.6, 2.5) Cel/Sel 8/410 (11.8) 7/414 (10.3) 0.9 (0.3, 2.4) 400mg QD Pooled 20/1038 (8.6) 30/1347 (9.6) 1.1 (0.6, 2.0) 200mg BID ADAPT 18/1083 (8.6) 18/725 (12.8) 1.5 (0.8, 2.9) APC 8/679 (3.9) 20/685 (9.7) 2.5 (1.1, 5.7) 200mg BID Pooled 29/1809 (6.9) 38/1450 (10.8) 1.8 (1.1, 3.1) 400mg BID 27/671 (13.4) 3.6 (1.6, 8.0) MA-27 3/817 (8.7) 6/818 (17.2) 1.8 (0.4, 7.3) 400mg BID pooled 11/1496 (4.6) 33/1489 (13.9) 3.1 (1.5, 6.1) Shown here are the event rates and hazard ratios for the individual trials, grouped by dose regimen. Of note, the CDME trial had a total of 3 events and very short follow-up time, so we performed all analyses with and without the CDME data included; the overall results were no different whether these data were included or not. The two trials of the 400mg once daily dose, PreSAP and CEL/SEL had very similar hazard ratios, with a combined hazard ratio of 1.1, and 95% confidence interval ranging from 0.6 to 2.0; The combined hazard ratio for the 200mg twice daily doses, ADAPT and APC, was 1.8, with a 95% confidence interval of 1.1 to 3.1 And the trials of the 400mg twice daily dose, APC and MA-27 had a combined hazard ratio of 3.1, with a 95% CI ranging from 1.5 to 6.1 *CDME Not included in this table because of extremely low event rates Solomon et al. Circulation 2008

CV Death, MI, Stroke, HF or thrombo-embolic event Hazard Associated with Celecoxib at Various Doses Stratified by Study and low-dose ASA use and Adjusted for Baseline CV Risk 400 mg bid 3.1 (1.5, 6.1) Dose-regimen effect P = 0.0005 Regimen 1.8 (1.1, 3.1) 200 mg bid Celecoxib 1.1 (0.6, 2.0) 400 mg qd We tested for a dose-regimen response in a linear trend model which showed a significant trend for increased risk progressing from 400mg qd to 400 mg bid, with a p-value of 0.0005. The overall risk associated with pooling all doses of celecoxib was 1.6, with 95% CI ranging from 1.1 to 2.3 1.6 (1.1, 2.3) Overall 0.5 0.5 0.7 0.7 0.9 0.9 1 1 2 2 3 3 4 4 5 5 6 6 Hazard Ratio CV Death, MI, Stroke, HF or thrombo-embolic event Solomon et al. Circulation 2008

Composite Outcomes (Hazard ratio and 95% CI) 400mg QD 200mg BID 400mg BID CV Death 0.5 (0.2, 1.7) 1.8 (0.5, 6.2) 6.5 (0.8, 54) + MI 1.0 (0.5, 2.1) 2.1 (1.0, 4.1) 3.4 (1.2, 9.6) +Stroke 1.0 (0.5, 1.9) 1.6 (0.9, 3.0) 2.9 (1.3, 6.6) + HF 1.1 (0.6, 2.1) 1.7 (1.0, 3.1) 2.7 (1.3, 5.6) + Embolic event 1.1 (0.6, 2.0) 1.8 (1.1, 3.1) 3.1 (1.5, 6.1) Any CV Event 1.3 (0.9, 1.9) 1.4 (1.0, 1.8) 1.6 (1.1, 2.3) We tested for the robustness of our findings by examining other composite endpoints. We obtained qualitatively similar results whether we used a traditional ATPC endpoint of CV death, MI and stroke or by adding failure or thromboembolic events to the primary composite. Stratified by study and baseline aspirin use and adjusted for baseline risk Solomon et al. Circulation 2008

Celecoxib Regimen and Baseline Cardiovascular Risk Baseline Risk – Dose Regimen Interaction p = 0.034 High Risk High Risk 400 bid 400 bid 200 bid 200 bid 400 400 qd qd Moderate Risk 400 bid Celecoxib Regimen and pre-treatment Cardiovascular Risk 200 bid 400 400 qd qd The overall event rate increased across the three baseline risk categories regardless of the use of celecoxib, with a doubling of risk between the low and moderate risk groups and a further doubling of risk between the moderate and high risk groups. We observed an interaction between baseline risk and dose with respect to outcomes. Patients in the low risk category showed the least effect of dose, and patients in the high risk category showed the greatest effect of dose, with an interaction p-value of 0.034. Low Risk 400 bid 200 bid 400qd 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 0.8 0.8 1 1 2 2 3 3 4 4 5 5 6 6 7 7 Hazard Ratio CV Death, MI, Stroke, HF or Thromboembolic Event Solomon et al. Circulation 2008

Prespecified Subgroups P-Interaction p = 0.37 p = 0.64 p = 0.54 p = 0.89 p = 0.17 p = 0.09 p = 0.40 p = 0.57 Non-Smoker Current Smoker No Diabetes Diabetes No Hyperlipidemia Hyperlipidemia No Hypertension Hypertension No CV Event CV Event History No Low Dose ASA Low Dose ASA Non-White White We observed no differential effect based on any of the prespecified subgroup comparisons, including smoking, diabetes, hypertension, hyperlipidemia, CV event history, race or gender. In particular, we did not observe an interaction with respect to low-dose aspirin use, although higher hazard ratios in the low-dose aspirin patients suggest that this characteristic is simply behaving as a surrogate of baseline risk. Female Male 0.3 0.4 0.6 0.8 1 2 3 4 5 6 Hazard Ratio Solomon et al. Circulation 2008

Limitations and Caveats None of the trials included in this analysis was designed or powered with the intent of assessing cardiovascular risk. Doses tested higher than those typically used in osteoarthritis patients. recommended doses in rheumatoid arthritis, acute pain and dysmenorrhea, FAP. These data provide the strongest evidence of a dose-related risk Data do not address the cardiovascular risk of doses lower than those tested. Some limitations and caveats of this analysis should be noted. None of the trials included in this analysis was designed or powered with the intent of assessing cardiovascular risk. As a result, we used data collected for other purposes to assess the effect of celecoxib on cardiovascular outcomes. The increased risk observed in this analysis needs to be considered in light of the high doses of celecoxib tested. All the tested doses are higher than the doses of celecoxib typically used in osteoarthritis patients (recommended daily dose 200mg); however, our data are directly relevant to doses recommended in the current celecoxib prescribing guidelines for patients with rheumatoid arthritis, acute pain and dysmennorhea, familial adenomatous polyposis, as well as doses currently being tested for non-arthritic conditions. Moreover, these data provide strong evidence of a dose and dose-regimen related risk. Our data do not, however, address whether doses lower than those tested in any of these trials would lead to lower cardiovascular risk, or whether non-selective non-steroidal anti-inflammatory agents would be associated with similar risk. We must therefore be cautious in extrapolating these results to doses not tested.

Conclusions (1) A pooled analysis of six randomized trials comparing celecoxib to placebo, with over 16,000 patient-years of follow-up, shows an overall increase in cardiovascular risk, with evidence for differences in risk based on the dose and dose-regimen of celecoxib. The data showed evidence of an interaction between baseline cardiovascular risk and the effect of celecoxib, suggesting that patients at highest baseline risk had an increased relative risk for celecoxib-related adverse cardiovascular events. In conclusion, A pooled analysis of six randomized trials comparing celecoxib to placebo, with over 16,000 patient-years of follow-up, shows an overall increase in cardiovascular risk, with evidence for differences in risk based on the dose and dose-regimen of celecoxib. The data showed evidence of an interaction between baseline cardiovascular risk and the effect of celecoxib, suggesting that patients at highest baseline risk had an increased relative – and not just absolute - risk for celecoxib-related adverse cardiovascular events.

Conclusions (2) That both the relative and absolute risks of cardiovascular events increases with baseline cardiovascular risk may provide more comfort in prescribing the drug in patients with very low baseline risk, and would argue for more caution in prescribing the drug in patients with higher baseline risk. Since celecoxib, which currently carries an FDA-mandated black-box warning, remains the only coxib available in the United States, and is the most commonly used coxib worldwide, these data should help guide rational clinical decisions regarding celecoxib use. That both the relative and absolute risks of cardiovascular events increases with baseline cardiovascular risk have important clinical implications; these data may provide a measure of comfort in prescribing celecoxib in patients with very low baseline risk, but would argue for caution in prescribing celecoxib in patients with higher baseline risk. Since celecoxib, which currently carries an FDA-mandated black-box warning, remains the only coxib available in the United States, and is the most commonly used coxib worldwide, these data should help guide rational clinical decisions regarding celecoxib use.

Composite KM Curves Combined studies (excluding CDME) 0.050 Combined studies (excluding CDME) 0.045 Log-rank p-value (3 df) = 0.026 0.040 0.035 0.030 Probability of event 0.025 0.020 0.015 0.010 0.005 0.000 6 12 18 24 30 36 Sample size Months of follow-up C 400 bid 1489 990 714 654 648 645 641 C 200 bid 1411 1342 1258 1157 1033 894 768 C 400 QD 1347 1231 1134 1042 939 841 514 Placebo 3617 2943 2453 2168 1894 1598 1196