The management of arthritis and chronic pain syndromes

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2006;4:1337 1345 Impact of Adherence to Concomitant Gastroprotective Therapy on Nonsteroidal-Related Gastroduodenal Ulcer Complications JAY L. GOLDSTEIN,* KIMBERLY B. HOWARD, SURREY M. WALTON,* TRENT P. MCLAUGHLIN, AND DENISE T. KRUZIKAS *University of Illinois at Chicago, Chicago, Illinois; Pfizer Inc, New York, New York; NDCHealth, Phoenix, Arizona; and NDCHealth, Yardley, Pennsylvania Background & Aims: The clinical impact of nonadherence to gastroprotective agents (GPAs) coprescribed with antiinflammatory therapies has not been evaluated. In a large, commercial, managed-care database, we retrospectively characterized the use of GPAs among patients receiving nonselective nonsteroidal anti-inflammatory drugs (ns-nsaids) or cyclooxygenase-2 selective inhibitors (coxibs) and determined the impact of nonadherence on the likelihood of gastroduodenal ulcer complications. Methods: Analyses identified the populations of patients with concomitant histamine-2 receptor antagonist or proton pump inhibitor (PPI) therapy and determined adherence with the prescribed therapy with respect to the duration of anti-inflammatory treatment. Multivariate regression analyses modeled the association between adherence with concomitant protective therapy and the likelihood of upper gastrointestinal (GI) complications including peptic ulcer disease, ulcer, and/or upper-gi bleed. Results: Among 144,203 patients newly prescribed anti-inflammatory therapies, 1.8% received concomitant GPA treatment (ns-nsaids, 1.4% vs coxibs, 2.6%; P.0001). The likelihood of GPA use increased with the presence of risk factors: age older than 65 years (odds ratio [OR], 1.40; 95% confidence interval [CI], 1.3 1.5) and prior history of peptic ulcer disease (OR, 2.5; 95% CI, 1.8 3.3), esophagitis/gastroesophageal reflux (OR, 3.8; 95% CI, 3.5 4.1), ulcer/upper-gi bleed (OR, 1.4; 95% CI, 1.2 1.5), or gastritis (OR, 2.5; 95% CI, 2.2 2.8). Of patients receiving concomitant PPI therapy, 68% had adherence rates of 80% or more. A significantly higher risk of upper-gi ulcers/complications was observed in ns-nsaid patients with adherence rates of less than 80% compared with adherence rates of 80% or more (OR, 2.4; 95% CI, 1.0 5.6), but no such relationship was observed among patients who took coxibs. Conclusions: Few patients receive concomitant GPA therapy when prescribed anti-inflammatory treatment, although use increased with the presence of risk factors. Adherence to concomitant therapy is paramount to reducing GI events among ns-nsaid users and educational efforts should be undertaken to promote use of and adherence to GPA therapy among these patients. The management of arthritis and chronic pain syndromes often involves continued use of analgesic medications. 1,2 Because of their efficacy and relatively inexpensive cost, nonselective nonsteroidal anti-inflammatory drugs (ns-nsaids) continue to be the mainstay of arthritis and pain management despite their associated risk of gastrointestinal (GI) toxicity. 3 6 With the aim of circumventing the upper-gi toxicity associated with use of ns-nsaids, multiple studies have shown that coadministration of so-called gastroprotective agents (GPAs) such as misoprostol or proton pump inhibitors (PPIs), reduces the rate of endoscopic gastric and/or duodenal ulcers compared with ns-nsaids alone. 7 10 In the case of misoprostol, there is also evidence of a reduction in the rate of upper-gi complications. 11 Although a single prospective endoscopic clinical trial suggested high-dose famotidine (40 mg twice a day) reduces the rate of endoscopic gastroduodenal ulcers compared with ns-nsaids alone, 12 there is a paucity of evidence that histamine-2 receptor antagonists (H 2 RAs) are effective in reducing ns-nsaid related upper-gi ulcer complications. 3,13,14 As an alternative to the use of coprescribed GPAs, cyclooxygenase-2 selective inhibitors (coxibs) are less likely to be associated with the development of endoscopic gastric and duodenal ulcers and upper-gi complications. 15 19 Recent studies also have suggested that PPIs co-administered with ns-nsaids are comparable with coxibs with respect to the rate of recurrent upper-gi ulcer bleeding in high-risk patients. 20 22 Based on these data, clinical guidelines have been forwarded by expert panels and developed by several national professional societies addressing the appropriate use of preventive strategies for patients at high risk. These guidelines generally recommend the concomitant use of GPAs such as a PPI or misoprostol, or the use of a coxib alone in place of an ns-nsaid among patients at high risk for GI complications. 2,23 27 Well-recognized risk factors for upper-gi ulcer complications include advanced age, history of upper-gi ulcers or bleeding, and concomitant use of corticosteroids or anticoagulants. 3,11,28 33 Despite the available data and the integrated guidelines, evidence suggests that significant proportions of high-risk patients are not receiving any protective strategies and, of those who do receive GPAs, many are treated inadequately with ineffective therapies. 34,35 For example, and despite the wealth of evidence supporting greater efficacy of PPIs compared with H 2 RAs, it is unfortunately still relatively common for physicians in clinical practice to prescribe standard doses of H 2 RAs (eg, ranitidine 150 mg twice a day) for prevention of ns-nsaid induced GI adverse events. 35 The fact that various national Abbreviations used in this paper: CI, confidence interval; GERD, gastroesophageal reflux disorder; GI, gastrointestinal; GPA, gastroprotective agents; H 2 RA, histamine-2 receptor antagonist; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; ns-nsaid, nonselective nonsteroidal anti-inflammatory drug; OR, odds ratio; PPI, proton pump inhibitor; PUD, peptic ulcer disease. 2006 by the AGA Institute 1542-3565/06/$32.00 doi:10.1016/j.cgh.2006.08.016

1338 GOLDSTEIN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 11 preventive guidelines for patients at high risk for NSAID-associated upper-gi toxicity are not applied uniformly in the clinical setting has been highlighted and quantified further. In a recent evaluation by Abraham et al 27 based on the use of a national Department of Veterans Affairs database, less than 30% of veterans considered to be at high risk for NSAIDassociated upper-gi toxicity were found to receive appropriate therapies. Patient adherence remains one of the important challenges of day-to-day clinical practice, and even when at-risk patients are identified and prescribed appropriate preventive strategies, nonadherence to the use of these medications may impact greatly on both short-term and long-term clinical outcomes. 36 38 Specific to anti-inflammatory treatment, Sturkenboom et al 35 determined that only 37% of patients newly receiving ns-nsaids had a greater than 75% adherence to their concomitant GPA therapy regimen. However, this study did not evaluate the clinical impact of this high level of nonadherence and, as such, leaves the issue of long-term GI safety and effectiveness of coprescription open to question. Therefore, this retrospective database study was undertaken to characterize the use of GPAs among patients receiving coxibs or ns-nsaids and to determine the impact of adherence to concomitant GPA therapy on the likelihood of coxib- and ns-nsaid related gastroduodenal toxicity. Patients and Methods This retrospective study was based on the patient-level clinical, longitudinal PharMetrics Integrated Outcomes database (PharMetrics, Watertown, MA), which offers administrative claims information collected from approximately 75 commercial managed-care plans covering more than 43 million enrollees across the United States. The database includes inpatient and outpatient diagnoses, procedures, and prescriptions filled within the plans. All medical and pharmaceutical claims include dates of service, and prescription data include date filled/administered, days supplied, and quantity dispensed. Additional data elements include demographic variables (age, sex, geographic region), health plan type (eg, health maintenance organization, preferred provider organization), payer type (eg, commercial, self-pay), provider specialty, and start and stop dates for plan enrollment. For the purposes of this study, we accessed a subset of 35 commercial managed-care plans from the PharMetrics database in which access to coxibs and GPA therapies were known to be available. We restricted our analysis to commercial managed-care plans in which claims for the agents of interest were recorded during the time frame of this study as an indicator showing the ability of physicians to prescribe these medications. Patient Sample The study time frame spanned a 3-year period from January 1, 2000, to December 31, 2002. Patients were eligible for inclusion in the study if they had an index prescription claim for an ns-nsaid or coxib and at least 1 refill for the same medication during this time frame. Because this study intended to examine the effects of long-term therapy, patients were excluded if they had less than a 10-day supply for their index medication or gaps in therapy of 120 days or more. Inclusion criteria also required no prescription claims for ns-nsaids, coxibs, or GPAs (misoprostol, PPIs, or H 2 RAs) during the 12 months before the index prescription date and 12 continuous months of enrollment in the plan both before and after the index prescription date. Patient data were analyzed during the 12-month preperiod to determine baseline demographic characteristics and the patients were followed-up for up to 12 months after the index prescription date to evaluate subsequent upper-gi outcomes related to ns-nsaid or coxib therapy. Treatment cohorts were defined by the index prescription claim during the study period. Ns-NSAIDs included ibuprofen, naproxen, nabumetone, diclofenac sodium, diclofenac potassium, etodolac, piroxicam, oxaprozen, sulindac, meloxicam, ketoprofen, flurbiprofen, and fenoprofen calcium. In these plans, aspirin use could not be measured objectively. Coxib products included celecoxib, rofecoxib, and valdecoxib. Patients were permitted to switch medications within their index cohort. For example, if a patient was initiated on celecoxib and had a subsequent prescription for a different coxib drug, they remained a coxib patient and were retained in the study. Similarly, a patient with an index claim for ibuprofen who switched to a different ns-nsaid treatment still was considered an ns-nsaid patient in the analyses. However, switching between treatment cohorts was not permitted; patients with any subsequent claims within 12 months after their index date for a medication listed in the alternative treatment group (ie, a coxib patient who had a subsequent claim for an ns-nsaid, or vice versa) were excluded from the analyses. In the case of patients switching between cohorts, the index time to the switch was not included in the analyses. Demographic data were collected to describe treatment cohorts with respect to age, sex, and health status. Health status was determined by comorbid illness, measured by the most common 3-digit International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes recorded in secondary diagnosis positions on prior medical claims. Two standard measurement tools were used to evaluate patient health status further, the Charlson Comorbidity Index and the Chronic Disease Score. 39,40 In addition, analyses assessed the frequency of coded GI diagnoses during the 12 months before the index prescription. Diagnoses considered for this analysis included peptic ulcer disease (PUD), esophagitis/gastroesophageal reflux disease, ulcer/upper-gi bleed, and gastritis. These diagnoses were identified through medical claims containing the following ICD-9-CM codes: 533.xx (PUD); 530.xx (esophagitis/gastroesophageal reflux disease); 531.xx, 532.xx, 534.xx, and 578.xx (ulcer/upper-gi bleed); and 535.xx (gastritis). Based on the available data, patients were grouped into 4 cohorts based on their use of ns-nsaids or coxibs with or without concomitant use of GPAs. Prescription claims were used to determine concomitant acid-suppressive GPA therapy, defined as initiation of PPI or H 2 RA use up to 14 days after the ns-nsaid/coxib index prescription. In this analysis, H 2 RAs were included in the GPA treatment definition because we assumed that it was a cognitive action taken by prescribers with the presumable intention of preventing subsequent GI events. Analyses also determined the number and percentage of patients with GI diagnoses within 12 months before the index prescription date. Within the ns-nsaid and coxib cohorts, 2 analyses compared the proportion of concomitant and nonconcomitant patients with prior GI diagnoses.

November 2006 ADHERENCE TO GASTROPROTECTIVE THERAPY 1339 Likelihood of Initiating Concomitant Gastroprotective Therapy The 2 analyses first compared the proportion of concomitant PPI/H 2 RA patients between ns-nsaid and coxib cohorts. Logistic regression analyses then modeled the likelihood of initiating concomitant therapy, with the index medication as the primary independent variable of interest and specific risk factors as predictors of secondary interest. 11,31,41 Three risk factors were of particular interest to this study because of their association with increased risk of GI events: patient age older than 65 years, previous ulcer diagnosis, and anticoagulant and/or steroid use. Results were adjusted for patient age and sex. Impact of Adherence on Patient Outcomes The effectiveness of adherence with concomitant GPA therapy on subsequent upper-gi complications was evaluated. These analyses only included PPIs as the appropriate GPA therapy because they are believed to be effective in reducing the incidence of upper-gi ulcers and complications compared with H 2 RAs. 8,12,20,42,43 Adherence to concomitant therapy was determined using a ratio of dispensed days supply of PPI and ns-nsaids or coxibs. The duration of follow-up evaluation could extend for up to 12 months after the index ns-nsaid/coxib prescription date, given that there were no treatment gaps of greater than 120 days. Adherence rates were calculated by normalizing the total days supply of PPI therapy by the total days supply of ns- NSAID/coxib therapy as follows: Adherence % Dispensed PPI days supply 100 Dispensed anti-inflammatory drug days supply Adherence was capped at 100% because the intent was to identify PPI coverage over the course of ns-nsaid/coxib treatment. It was considered a continuous variable ranging from 0% to 100% and also as a categoric variable with 5 levels of adherence: 0% 20% to 80% 100%. A priori, the study hypothesized that the likelihood of adherence to concomitant GPA therapy decreases as the days supply of anti-inflammatory treatment increases. Because adherence might change over the duration of anti-inflammatory treatment with the possibility that patients on therapy for longer durations might have increased rates of nonadherence with time, we evaluated the proportion of patients with PPI adherence of 80% or greater according to the duration of antiinflammatory therapy, measured by the number of index medication refills. The likelihood of adherence was evaluated through multivariate logistic regression models. By using adherence as the dichotomous outcome, models controlled for patient age, hypertension, diabetes mellitus, prior cardiovascular conditions, previous PUD, previous ulcer/upper-gi bleed, number of concomitant medications, and the number of index medication refills. After accounting for prior risk, concomitancy, and adherence, the primary end points of interest examined by the study were PUD (ICD-9-CM code 533.xx), ulcer, and/or upper-gi bleed (ICD-9-CM codes 531.xx, 532.xx, 534.xx, and 578.xx) occurring up to 12 months after the index prescription date. Descriptive analyses determined crude rates of predefined end points based on ICD-9-CM codes; univariate analyses examined the entire sample and 2 analyses compared the rates between ns-nsaid and coxib cohorts. We also examined the predefined GI events for the ns-nsaid and coxib cohorts as a function of adherence. To do so, the number of GI events within each patient cohort was normalized by dividing the sum of events by the cumulative sum of total days supply for the index medication; rates were expressed in patient-years. Rates of GI events per patient-year were plotted against levels of adherence for the ns-nsaid and coxib cohorts. Finally, multivariate analyses modeled the impact of 80% or greater adherence on the likelihood of GI events. The dependent variable was occurrence of PUD, ulcer, and/or upper-gi bleed during the ns-nsaid/coxib treatment period. The ns- NSAID/coxib treatment period was defined as the duration between the initial and final index medication prescription plus days supply for the last prescription or 12 months after the index prescription, whichever occurred first. Adherence was the independent variable of interest; the models also controlled for patient age, sex, and prior GI risk factors (previous PUD, esophagitis/gastroesophageal reflux disease, ulcer/upper-gi bleed, and gastritis) diagnosed within 12 months before the index ns-nsaid and coxib prescription. Logistic models evaluated ns-nsaid and coxib cohorts separately. Based on data from other trials and reports, patients with prior diagnoses of cardiovascular ischemic events are likely to be given aspirin for secondary prophylaxis. 44 47 Because the data could not capture over-the-counter aspirin use reliably, we conducted an exploratory and post hoc analysis using coded cardiovascular diagnoses as a proxy measure for aspirin use to determine its impact on the likelihood of predefined GI outcomes. The analysis compared the rate of GI events among patients with cardiovascular disease diagnosed within 12 months before the index ns-nsaid/coxib date against the rate among patients without diagnosed cardiovascular disease. Cardiovascular conditions included ischemic heart disease (ICD- 9-CM codes 410.xx and 411.xx, excluding 411.1x and 414.xx), angina (ICD-9-CM codes 411.1x and 413.xx), stroke (ICD-9-CM codes 430.xx 438.xx), and peripheral vascular disease (ICD- 9-CM codes 443.8, 443.89, and 443.9). Multivariate logistic analyses modeled the likelihood of GI events in addition to the presence of cardiovascular disease; the model also controlled for patient age, the presence of hypertension and/or diabetes, prior PUD and/or ulcer, the number of concomitant medications during the anti-inflammatory treatment period, and the number of index product refills. Results Patient Sample After all inclusion and exclusion criteria were applied, 144,203 patients were available for analysis (Table 1). Of these, 92,833 (64%) were treated with ns-nsaids and 51,370 (36%) were treated with coxibs. The most common ns-nsaid medications were naproxen and ibuprofen, comprising 37% and 32% of the patient sample, respectively. Other ns-nsaids included nabumetone (8%), diclofenac sodium (6%), etodolac (4%), piroxicam (4%), oxaprozen (3%), and sulindac (2%). All other ns-nsaid products were used by fewer than 2% of patients. Approximately 53% of coxib patients were prescribed rofecoxib and 47% were treated with

1340 GOLDSTEIN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 11 Table 1. Patient Demographics by Index Prescription and Concomitant Therapy ns-nsaids (n 92,833; 64%) Coxibs (n 51,370; 36%) Concomitant Nonconcomitant Concomitant Nonconcomitant Total n 144,203 Patients, n (%) 1312 (1.4 a ) 91,521 (98.6) 1322 (2.6 a ) 50,048 (97.4) 144,203 (1.8) Mean age, y (SD) 48.40 (12.06) 47.04 (11.46) 50.24 (10.86) 50.59 (10.26) 48.32 (9.66) Age, n (%) 19 35 y 176 (13.41) 14,227 (15.55) 120 (9.08) 4080 (8.15) 18,603 (12.90) 36 45 y 332 (25.30) 25,368 (27.72) 278 (21.03) 10,576 (21.13) 36,554 (25.35) 46 55 y 453 (34.53) 31,089 (33.97) 515 (38.96) 19,323 (38.61) 51,380 (35.63) 56 65 y 271 (20.66) 17,184 (18.78) 343 (25.95) 13,449 (26.87) 31,247 (21.67) 65 y 80 (6.10) 3653 (3.99) 66 (4.99) 2620 (5.23) 6419 (4.45) Female, n (%) 815 (62.12) 53,777 (58.76) 833 (63.01) 30,936 (61.81) 86,361 (59.89) Male, n (%) 497 (37.88) 37,737 (41.23) 489 (36.99) 19,110 (38.18) 57,833 (40.11) a The difference in the proportion of concomitant patients between ns-nsaid and coxib cohorts is statistically significant with a P value of.003. celecoxib. Less than 1% of the study population received valdecoxib. For details on health status by index prescription and concomitant therapy, see Supplemental Table 1 (supplementary material online at www.cghjournal.org). Likelihood of Initiating Concomitant Gastroprotective Agent Therapy Only 1.8% (n 2634) of the total sample population initiated concomitant PPI or H 2 RA therapy within 14 days of the index ns-nsaid/coxib prescription (Table 2). Interestingly, coxib patients were more likely to receive GPAs compared with ns-nsaid users. Rates of concomitancy were 2.6% among coxib-treated patients and 1.4% in NSAID-treated patients (odds ratio [OR], 1.82; 95% confidence interval [CI], 1.69 1.96). With respect to GPA therapy, 62% of patients received PPI therapy and 38% were treated with H 2 RAs. Variations were noted based on the index treatment cohort: patients treated with coxibs were more likely to be prescribed PPIs than H 2 RAs (74% vs 26%; P.0001), whereas patients treated with ns- NSAIDs were equally as likely to be prescribed either therapy (50% each). Regression analysis further confirmed that patients treated with coxibs were more likely to initiate concomitant PPI/H 2 RA treatment than patients treated with ns-nsaids (OR, 1.31; 95% CI, 1.26 1.35) (Table 3). Impact of Gastrointestinal Risk Factors on Concomitant Gastroprotective Agent Therapy As shown in Table 2, a significantly higher proportion of concomitant patients within both the ns-nsaid and coxib cohorts had prior GI diagnoses compared with nonconcomitant patients (P.0001). Furthermore, prior GI diagnoses were more common among concomitant coxib users compared with concomitant ns-nsaid users (22.8% vs 12.3%; P.0001). In general and consistent with these results, the multivariate analysis found that patients at increased risk of GI events were more likely to initiate concomitant therapy (Table 3). The probability of concomitancy was 38% higher for patients aged older than 65 years compared with those aged 36 45 years (OR, 1.38; 95% CI, 1.27 1.50), 36% higher for patients with a previous ulcer diagnosis (OR, 1.36; 95% CI, 1.20 1.54), 26% higher for patients with concomitant oral steroid use (OR, 1.26; 95% CI, 1.20 1.33), and 62% higher for patients undergoing concomitant anticoagulant therapy (OR, 1.62; 95% CI, 1.42 1.84). Among ns-nsaid users, concomitancy rates did not vary significantly according to the presence of multiple GI risk factors and ranged from 1.4% among patients with no risk factors to 2.1% among patients with at least 2 risk factors. Similarly, concomitant therapy rates remained consistent across all levels of risk for patients treated with coxibs (no risk factors, 2.6%; 1 risk factor, 2.5%; 2 risk factors or more, 2.4%). Table 2. Prior GI Diagnoses by Index Prescription and Concomitant Therapy ns-nsaids Coxibs Concomitant Nonconcomitant Concomitant Nonconcomitant Patients, n (%) 1312 (1.4) 91,521 (98.6) 1322 (2.6) 50,048 (97.4) PPI prescription, n (%) 656 (50) 978 (74) H 2 RA prescription, n (%) 656 (50) 344 (26) GI events during 12-month preperiod, n (%) PUD 9 (0.7) a 93 (0.1) 17 (1.3) a 96 (0.2) Esophagitis 103 (7.9) a 1147 (1.3) 207 (15.7) a 1136 (2.3) Ulcer/upper-GI bleed 23 (1.8) a 1035 (1.1) 50 (3.8) a 799 (1.6) Gastritis 53 (4.0) a 713 (0.8) 70 (5.3) a 593 (1.2) Any GI events, n (%) 161 (12.3) a 2774 (3.0) 302 (22.8) a 2355 (4.7) a P.0001 vs nonconcomitant therapy.

November 2006 ADHERENCE TO GASTROPROTECTIVE THERAPY 1341 Table 3. Logistic Regression Results: The Likelihood of Initiating Concomitant Therapy Independent variable Reference group OR 95% CI Coxibs ns-nsaids 1.31 1.26 1.35 Age, y 19 35 36 45 y 1.04 0.98 1.10 46 55 36 45 y 1.12 1.07 1.17 56 65 36 45 y 1.18 1.12 1.24 65 36 45 y 1.38 1.27 1.50 Female Male 1.25 1.21 1.30 Previous PUD 2.46 1.81 3.34 Previous esophagitis/gerd 3.78 3.47 4.12 Previous ulcer/upper-gi bleed 1.36 1.20 1.54 Previous gastritis 2.46 2.17 2.78 Previous oral steroid use 1.26 1.20 1.33 Pre-/postanticoagulant use 1.62 1.42 1.84 Table 4. Logistic Regression Results: Predicting Adherence With Concomitant PPI Therapy Independent variable Reference group OR 95% CI Age, y 19 35 36 45 y 1.17 0.78 1.75 46 55 36 45 y 1.43 1.08 1.89 56 65 36 45 y 1.06 0.78 1.45 65 36 45 y 0.81 0.48 1.36 Hypertension 0.87 0.67 1.14 Diabetes mellitus 1.20 0.83 1.75 Cardiovascular condition 1.02 0.67 1.55 Previous PUD 1.06 0.42 2.69 Previous ulcer/upper GI-bleed 1.62 0.84 3.12 Number of concomitant medications 0.90 0.87 0.94 Number of index medication refills 0.97 0.94 0.99 However, regardless of the level of risk, rates of coprescribed GPA therapy remained low. Impact of Adherence on Patient Outcomes For the purposes of evaluating the impact of adherence in reducing the occurrence of clinically significant upper-gi events, we limited our analysis to the concomitant use of PPIs only, resulting in a sample size of 1643 patients: 664 (40%) were treated with ns-nsaids and 979 (60%) were treated with coxibs. As shown in Figure 1, there was a tendency for patients to be less adherent with GPA therapy as the duration of anti-inflammatory treatment increased (as measured by the number of refills of their anti-inflammatory therapies). These results are confirmed in Table 4, which shows that adherence decreases significantly with increasing numbers of index prescription refills (OR, 0.97; 95% CI, 0.94 0.99). The likelihood of adherence also decreases as patients increase the number of any concomitant medications (OR,.90; 95% CI, 0.87 0.94). Recognized risk factors for ulcer complications did not influence the likelihood of adherence. Collectively, 68% of ns-nsaid and coxib patients had adherence of 80% or greater over the entire duration of their days supply of anti-inflammatory drugs. Figures 2 and 3 show the unadjusted rates of GI events per patient-year across increasing levels of adherence. Among ns-nsaid users, the likelihood of GI complications decreases as adherence increases (Figure 2, R 2 0.3088). In comparison, GI event rates remain relatively constant across all adherence levels for coxib patients (Figure 3, R 2 0.0079). Among ns-nsaid users, patients with less than 80% adherence were nearly 2.5-fold more likely to experience upper-gi events during therapy compared with patients with 80% or greater adherence (OR, 2.38; 95% CI, 1.02 5.56) (Table 5). Multivariate analyses confirmed that adherence to PPI therapy did not influence the likelihood of GI injury among the coxib cohort. Other factors found to influence the incidence of GI complications included previous PUD for ns-nsaid patients (OR, 19.62; 95% CI, 3.23 119.37) and previous ulcer/upper-gi bleed for coxib patients (OR, 6.22; 95% CI, 2.75 14.07). The post hoc analysis using cardiovascular diagnoses as a possible proxy for aspirin use found that patients with a previous cardiovascular diagnosis had a significantly higher rate of Figure 1. Percent adherence by number of index medication refills., Adherence 80% or greater;, adherence less than 80%.

1342 GOLDSTEIN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 11 Figure 2. Number of GI events per patient-year by level of adherence in the ns-nsaid cohort. GI complications when compared with patients without any prior cardiovascular risk (10% vs 5%; P.0084). However, the population at risk was small, with only 120 patients having prior cardiovascular disease. Therefore, the inferences that can be made based on these data are limited. Discussion This study was designed to evaluate physician and patient behavior with regard to long-term anti-inflammatory therapy. Specifically, this study evaluated the impact of adherence with GPA therapies on clinically relevant upper-gi outcomes among patients receiving ns-nsaids. Our results show that reductions in adherence with coprescribed GPA therapies in patients taking ns-nsaids are associated with linear increases in the rate of coded end points of upper-gi events, and that lack of adherence increases the rate of upper-gi events 2.5-fold in the population studied. Furthermore, our results indicate that among patients at higher risk for NSAID-associated gastroduodenal injury (age 65 y or older, previous GI complication, or prior oral steroid or anticoagulant use), less than one third are treated with either a coxib or concomitantly with an ns-nsaid plus a PPI or even an H 2 RA. These results are not dissimilar from those reported by Abraham et al. 27 A previous analysis showed that even among subpopulations of patients at higher risk, the level of intervention is inadequate, ranging from 36% of patients with at least 1 risk factor to only 54% with all 3 risk factors. 48 Moreover, even when protective therapy is initiated, it often is suboptimal. Among the patients receiving concomitant therapy, approximately 62% are prescribed the recommended PPIs, leaving nearly 38% to be treated with the less-effective H 2 RAs. Although these data are consistent with findings by other investigators, 27 an additional unexpected observation was seen in our analysis. We noted that more than half of patients receiving a PPI were treated concomitantly with a coxib. Although these results may suggest that when risks are recognized physicians may act more cautiously by prescribing multiple protective therapies, they equally suggest possible overuse of these protective therapies. Although our study shows no advantage of adding a PPI to existing coxib therapy (Figure 3), our findings may not be generalizable to all populations. For example, Rahme et al 49 reported a benefit of adding a PPI to patients receiving coxibs in 2 patient populations, those aged 75 years and older, and those receiving concomitant aspirin. More recently, a preliminary report by Chan et al 50 in patients who had previously sustained an upper-gi ulcer bleeding episode while using NSAIDs, revealed a significant reduction in the rate of recurrent ulcer bleeding in this unique high-risk population receiving esomeprazole 20 mg twice a day plus celecoxib 200 mg twice a day compared with celecoxib alone. Thus, it is possible that concomitant coxib and PPI therapy may offer clinically sig- Figure 3. Number of GI events per patient-year by level of adherence in the coxib cohort.

November 2006 ADHERENCE TO GASTROPROTECTIVE THERAPY 1343 Table 5. Logistic Regression Results: Impact of Adherence on Likelihood of Upper-GI Events ns-nsaids Coxibs Independent variable Reference group OR 95% CI OR 95% CI Adherence 80% Adherence 80% 2.38 1.02 5.56 1.12.61 2.07 Age, y 19 35 36 45 y 2.07.46 9.31 2.27.75 6.83 46 55 36 45 y 1.13.36 3.59 1.92.81 4.55 56 65 36 45 y 1.65.47 5.76.89.32 2.50 65 36 45 y 2.14.36 12.87 1.95.54 7.08 Female Male.71.30 1.66 1.13.62 2.03 Previous PUD N/A 19.62 3.23 119.37 6.04 1.78 20.44 Previous esophagitis/gastroesophageal reflux disease N/A.56.12 2.70 1.54.80 2.97 Previous ulcer/upper-gi bleed N/A 5.75.96 34.44 6.22 2.75 14.07 Previous gastritis N/A.84.14 5.14 2.25.91 5.53 nificant benefits to patients at particularly high risk; however, this was not the population studied in the current analysis. The question then arises whether concomitant GPA therapy is effective in reducing the risk of upper-gi events among patients continuing to use long-term anti-inflammatory treatment. Although endoscopic and outcomes studies have shown that GPA therapy appears to be protective against upper-gi ulcer complications associated with ns-nsaids, 8,11,42 few studies have examined the level of adherence to coprescribed GPA therapy among patients treated with anti-inflammatory products or the relationship between adherence and the likelihood of upper-gi complications. In this study, we were able to examine adherence to concomitant PPI therapy and its impact on patient outcomes. The results revealed that the majority (68%) of patients achieve adherence rates of 80% or greater over the duration of their days supply of ns-nsaids in contrast to the remaining 32% of patients who did not achieve this level of adherence over the duration of the study. It is interesting to note that adherence rates decreased with increasing numbers of index medication (NSAID) refills. Furthermore, the analyses showed that adherence is important when evaluating patient outcomes. Among patients treated with ns-nsaids, adherence of 80% or greater resulted in lower rates of PUD, ulcers, and/or upper-gi bleeds as measured in this database. In addition, among patients treated with coxibs, the likelihood of upper-gi complications remained relatively constant across all levels of adherence at a rate comparable with that observed among ns-nsaid patients with 80% or greater adherence. These data indicate that the protective effect of coxibs is not different from that observed with coprescribed ns-nsaid and PPI therapy as previously described by others. 20 A potential limitation of this study lies within the methodology used to determine adherence. This analysis examined adherence based on the ratio of days supply of PPIs dispensed compared with days supply of anti-inflammatory therapy. Because this was a retrospective study, we did not conduct actual pill counts nor did we monitor drug use directly. As such, our definition is subjective and specific to this study, and results may not be generalizable to other analyses using alternative definitions. Generalizability may be limited further because our study was based on a managed-care database that may not be representative of the overall national patient population using anti-inflammatory treatments. For example, our patient population tended to be younger compared with those in previously reported trials. In addition, the analyses allowed a gap of up to 120 days of anti-inflammatory therapy to account for variability in patient use of their NSAID and/or refill behavior. As in all retrospective database analyses, controlling for patient characteristics that can influence the outcomes of interest are limited to available variables. A specific limitation of our analysis was that coxibs and/or coprescriptions may be given to patients with higher risk factors for GI events known to the physician and/or patient (channeling bias), but unobserved in the data set. One final limitation of our analyses was that we were unable to identify and, as such, control for any potential restrictions (tiered copay, algorithms, prior authorization programs) placed by individual plans and their influence on absolute rates of use. This study provides insight into compliance with anti-inflammatory treatment guidelines and the impact of adherence on patient outcomes in real-world practice settings. Our results show that compliance with treatment guidelines is poor with few high-risk patients receiving GPA treatment in conjunction with long-term anti-inflammatory therapy. Even when risks are recognized, concomitant protective therapy is wholly inadequate. Thus, patients are placed at unnecessarily increased risk of gastroduodenal injury. Increased efforts must be undertaken to educate providers of the risks associated with chronic antiinflammatory therapy and the need to intervene appropriately with coprescription of GPAs and/or to treat with coxibs. Given the decrease in coxib use and the corresponding increase in ns-nsaid use, it is likely that without increased vigilance around the need for GPA use, GI event rates potentially will increase. If coprescribed GPAs are used, such educational activities, combined with further encouragement of patient adherence to concomitant therapy, may result in improved safety and effectiveness of anti-inflammatory treatment. Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of Clinical Gastroenterology and Hepatology at www.cghjournal.org. References 1. American College of Rheumatology Subcommittee on Rheumatoid Arthritis. Guidelines for the management of rheumatoid arthritis: 2002 update. Arthritis Rheum 2002;46:328 346.

1344 GOLDSTEIN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 11 2. American College of Rheumatology Subcommittee on Osteoarthritis Guidelines. Recommendations for the medical management of osteoarthritis of the hip and knee: 2000 update. Arthritis Rheum 2000;43:1905 1915. 3. Wolfe MM, Lichtenstein DR, Singh G. Gastrointestinal toxicity of nonsteroidal antiinflammatory drugs. N Engl J Med 1999;340: 1888 1899. 4. Scheiman JM. Gastrointestinal outcomes: evidence for risk reduction in patients using coxibs. Am J Manag Care 2002;8: S518 S528. 5. Scheiman JM, Fendrick AM. NSAIDs without a prescription: overthe-counter access, under-counted risks. Am J Gastroenterol 2002;97:2159 2161. 6. Blot WJ, Fischer T, Nielsen GL, et al. Outcome of upper gastrointestinal bleeding and use of ibuprofen versus paracetamol. Pharm World Sci 2004;26:319 323. 7. Ekstrom P, Carling L, Wetterhus S, et al. Prevention of peptic ulcer and dyspeptic symptoms with omeprazole in patients receiving continuous non-steroidal anti-inflammatory drug therapy. A Nordic multicentre study. Scand J Gastroenterol 1996;31:753 758. 8. Hawkey CJ, Karrasch JA, Szczepanski L, et al. Omeprazole compared with misoprostol for ulcers associated with nonsteroidal antiinflammatory drugs. Omeprazole versus Misoprostol for NSAID-induced Ulcer Management (OMNIUM) Study Group. N Engl J Med 1998;338:727 734. 9. Yeomans ND, Tulassay Z, Juhasz L, et al. A comparison of omeprazole with ranitidine for ulcers associated with nonsteroidal antiinflammatory drugs. Acid Suppression Trial: Ranitidine versus Omeprazole for NSAID-associated Ulcer Treatment (ASTRONAUT) Study Group. N Engl J Med 1998;338:719 726. 10. Agrawal NM, Caldwell J, Kivitz AJ, et al. Comparison of the upper gastrointestinal safety of Arthrotec 75 and nabumetone in osteoarthritis patients at high risk for developing nonsteroidal antiinflammatory drug-induced gastrointestinal ulcers. Clin Ther 1999;21:659 674. 11. Silverstein FE, Graham DY, Senior JR, et al. Misoprostol reduces serious gastrointestinal complications in patients with rheumatoid arthritis receiving nonsteroidal anti-inflammatory drugs. A randomized, double-blind, placebo-controlled trial. Ann Intern Med 1995;123:241 249. 12. Taha AS, Hudson N, Hawkey CJ, et al. Famotidine for the prevention of gastric and duodenal ulcers caused by nonsteroidal antiinflammatory drugs. N Engl J Med 1996;334:1435 1439. 13. Robinson MG, Griffin JW Jr, Bowers J, et al. Effect of ranitidine on gastroduodenal mucosal damage induced by nonsteroidal antiinflammatory drugs. Dig Dis Sci 1989;34:424 428. 14. Ehsanullah RS, Page MC, Tildesley G, et al. Prevention of gastroduodenal damage induced by non-steroidal anti-inflammatory drugs: controlled trial of ranitidine. BMJ 1988;297:1017 1021. 15. Goldstein JL, Eisen G, Stenson W, et al. Celecoxib significantly reduces gastrointestinal-related healthcare resource utilization compared to NSAIDs: SUCCESS-1 in osteoarthritis (OA) trial (abstr). Arthritis Rheum 2001;44(Suppl):503. 16. Goldstein JL, Silverstein FE, Agrawal NM, et al. Reduced risk of upper gastrointestinal ulcer complications with celecoxib, a novel COX-2 inhibitor. Am J Gastroenterol 2000;95:1681 1690. 17. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA 2000;284:1247 1255. 18. Bombardier C, Laine L, Reicin A, et al. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. N Engl J Med 2000;343:1520 1528. 19. Schnitzer TJ, Burmester GR, Mysler E, et al. Comparison of lumiracoxib with naproxen and ibuprofen in the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), reduction in ulcer complications: randomised controlled trial. Lancet 2004;364:665 674. 20. Chan FK, Hung LC, Suen BY, et al. Celecoxib versus diclofenac and omeprazole in reducing the risk of recurrent ulcer bleeding in patients with arthritis. N Engl J Med 2002;347:2104 2110. 21. Chan FK, Hung LC, Suen BY, et al. Celecoxib versus diclofenac plus omeprazole in high-risk arthritis patients: results of a randomized double-blind trial. Gastroenterology 2004;127:1038 1043. 22. Lai KC, Chu KM, Hui WM, et al. Celecoxib compared with lansoprazole and naproxen to prevent gastrointestinal ulcer complications. Am J Med 2005;118:1271 1278. 23. American Pain Society. Guideline for the management of pain in osteoarthritis, rheumatoid arthritis, and juvenile chronic arthritis. Glenview, IL: American Pain Society, 2002. 24. Lanza FL. A guideline for the treatment and prevention of NSAIDinduced ulcers. Members of the Ad Hoc Committee on Practice Parameters of the American College of Gastroenterology. Am J Gastroenterol 1998;93:2037 2046. 25. Inadomi JM, Fendrick AM. PPI use in the OTC era: who to treat, with what, and for how long? Clin Gastroenterol Hepatol 2005; 3:208 215. 26. Dubois RW, Melmed GY, Henning JM, et al. Guidelines for the appropriate use of non-steroidal anti-inflammatory drugs, cyclooxygenase-2-specific inhibitors and proton pump inhibitors in patients requiring chronic anti-inflammatory therapy. Aliment Pharmacol Ther 2004;19:197 208. 27. Abraham NS, El Serag HB, Johnson ML, et al. National adherence to evidence-based guidelines for the prescription of nonsteroidal anti-inflammatory drugs. Gastroenterology 2005;129:1171 1178. 28. Simon LS. Risk factors and current NSAID use. J Rheumatol 1999;26:1429 1431. 29. Laine L, Bombardier C, Hawkey CJ, et al. Stratifying the risk of NSAID-related upper gastrointestinal clinical events: results of a double-blind outcomes study in patients with rheumatoid arthritis. Gastroenterology 2002;123:1006 1012. 30. Gabriel SE, Jaakkimainen L, Bombardier C. Risk for serious gastrointestinal complications related to use of nonsteroidal antiinflammatory drugs. A meta-analysis. Ann Intern Med 1991;115: 787 796. 31. Shorr RI, Ray WA, Daugherty JR, et al. Concurrent use of nonsteroidal anti-inflammatory drugs and oral anticoagulants places elderly persons at high risk for hemorrhagic peptic ulcer disease. Arch Intern Med 1993;153:1665 1670. 32. Piper JM, Ray WA, Daugherty JR, et al. Corticosteroid use and peptic ulcer disease: role of nonsteroidal anti-inflammatory drugs. Ann Intern Med 1991;114:735 740. 33. Singh G, Rosen Ramey D. NSAID-induced gastrointestinal complications: the ARAMIS perspective 1997. Arthritis, Rheumatism, and Aging Medical Information System. J Rheumatol 1998; 51(Suppl):8 16. 34. Sturkenboom MC, Burke TA, Dieleman JP, et al. Underutilization of preventive strategies in patients receiving NSAIDs. Rheumatology (Oxford) 2003;42(Suppl 3):iii23 iii31. 35. Sturkenboom MC, Burke TA, Tangelder MJ, et al. Adherence to proton pump inhibitors or H2-receptor antagonists during the use of non-steroidal anti-inflammatory drugs. Aliment Pharmacol Ther 2003;18:1137 1147. 36. Rosenow EC III. Patients understanding of and compliance with medications: the sixth vital sign? Mayo Clin Proc 2005;80:983 987. 37. Makaryus AN, Friedman EA. Patients understanding of their treatment plans and diagnosis at discharge. Mayo Clin Proc 2005;80:991 994.

November 2006 ADHERENCE TO GASTROPROTECTIVE THERAPY 1345 38. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med 2005;353:487 497. 39. D Hoore W, Sicotte C, Tilquin C. Risk adjustment in outcome assessment: the Charlson comorbidity index. Methods Inf Med 1993;32:382 387. 40. Von Korff M, Wagner EH, Saunders K. A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992;45:197 203. 41. Singh G. Recent considerations in nonsteroidal anti-inflammatory drug gastropathy. Am J Med 1998;105:31S 38S. 42. Graham DY, Agrawal NM, Campbell DR, et al. Ulcer prevention in long-term users of nonsteroidal anti-inflammatory drugs: results of a double-blind, randomized, multicenter, active- and placebocontrolled study of misoprostol vs lansoprazole. Arch Intern Med 2002;162:169 175. 43. Agrawal NM, Campbell DR, Safdi MA, et al. Superiority of lansoprazole vs ranitidine in healing nonsteroidal anti-inflammatory drug-associated gastric ulcers: results of a double-blind, randomized, multicenter study. NSAID-Associated Gastric Ulcer Study Group. Arch Intern Med 2000;160:1455 1461. 44. Patrono C, Coller B, FitzGerald GA, et al. Platelet-active drugs: the relationships among dose, effectiveness, and side effects: the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy. Chest 2004;126:234S 264S. 45. Patrono C, Bachmann F, Baigent C, et al. Expert consensus document on the use of antiplatelet agents. The task force on the use of antiplatelet agents in patients with atherosclerotic cardiovascular disease of the European society of cardiology. Eur Heart J 2004;25:166 181. 46. Patrono C, Garcia Rodriguez LA, Landolfi R, et al. Low-dose aspirin for the prevention of atherothrombosis. N Engl J Med 2005;353:2373 2383. 47. Antiplatelet Trialists Collaboration. Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ 2002;324:71 86. 48. Kruzikas DT, Walton SM, McLaughlin T, et al. Use of gastroprotective agents (GPAs) among patients receiving nonsteroidal antiinflammatory drugs (NSAIDs) (abstr M986). Gastroenterology 2005;128(Suppl 2):A285. 49. Rahme E, Barkun AN, Toubouti Y, et al. Do proton pump inhibitors (PPIs) infer additional gastrointestinal protection in patients given celecoxib? A retrospective cohort study (abstr). Gastroenterology 2005;128(Suppl 2):A139. 50. Chan FK, Wong VW, Suen BY, et al. Proton pump inhibitor plus a COX-2 inhibitor for the prevention of recurrent ulcer bleeding in patients with arthritis: a double-blind, randomized trial (abstr 732). Gastroenterology 2006;130(Suppl 2):A105. Address requests for reprints to: Jay L. Goldstein, MD, Department of Medicine, University of Illinois at Chicago, 840 South Wood Street (m/c787), Room 1020, Chicago, Illinois 60612. e-mail: jlgoldst@ uic.edu; fax: (312) 413-0342. Supported by Pfizer Inc (NY). J.L.G. is a consultant to and has received travel expenses, educational grants, research grants and contracts, and speaker honoraria from Pfizer Inc, AstraZeneca, TAP, GlaxoSmithKline, Novartis, Takeda, POZEN, and Sucampo; K.B.H. is an employee of Pfizer Inc; S.M.W. is a consultant to and has received travel funding from Pfizer Inc, and has received research support from Takeda; T.P.M. and D.T.K. are employees of NDCHealth, who are paid consultants to Pfizer Inc.

November 2006 ADHERENCE TO GASTROPROTECTIVE THERAPY 1345.e1 Supplemental Table 1. Health Status by Index Prescription and Concomitant Therapy ns-nsaids (n 92,833; 64%) Coxibs (n 51,370; 36%) Health status measures Concomitant Nonconcomitant Concomitant Nonconcomitant Patients, n 1312 91,521 1322 50,048 Payer, n (%) Medicare 95 (7.24) 4698 (5.13) 82 (6.20) 2978 (5.95) Medicaid 38 (2.90) 1461 (1.60) 18 (1.36) 127 (.25) Commercial 930 (70.88) 67,253 (73.48) 802 (60.67) 32,364 (64.67) Self-pay 124 (9.45) 5657 (6.18) 200 (15.13) 5477 (10.94) Other/unknown 125 (9.53) 12,452 (13.61) 220 (16.64) 9102 (18.19) Region, n (%) North 214 (16.31) 10,938 (11.95) 238 (18.00) 8416 (16.82) South 241 (18.37) 13,591 (14.85) 306 (23.15) 10,282 (20.54) Midwest 726 (55.34) 59,505 (65.02) 697 (52.72) 29,109 (58.16) West 131 (9.98) 7487 (8.18) 81 (6.13) 2241 (4.48) Top 10 comorbidities (3-digit ICD-9-CM code), n (%) Hypertension (401) 165 (12.6) 8203 (9.0) 164 (12.4) 6175 (12.3) Metabolic disorders (272) 119 (9.1) 7203 (7.9) 118 (8.9) 5519 (11.0) Joint disorders (719) 78 (6.0) 5620 (6.1) 111 (8.4) 5323 (10.6) Back disorders (724) 72 (5.5) 5686 (6.2) 131 (9.9) 4812 (9.6) Soft-tissue disorders (729) 71 (5.4) 4774 (5.2) 105 (7.9) 4187 (8.4) Osteoarthritis (715) 38 (2.9) 2792 (3.1) 81 (6.1) 3351 (6.7) Diabetes mellitus (250) 66 (5.0) 3513 (3.8) 62 (4.7) 2463 (4.9) Peripheral enthesopathies (726) 30 (2.3) 2719 (3.0) 43 (3.3) 2549 (5.1) Cervical region disorders (723) 34 (2.6) 2452 (2.7) 55 (4.2) 2049 (4.1) Hypothyroidism (244) 49 (3.7) 2265 (2.5) 46 (3.5) 1742 (3.5) Charlson Comorbidity Index, n (%) 0 958 (73.0) 71,768 (78.4) 938 (71.0) 37,216 (74.4) 1 266 (20.3) 16,248 (17.8) 286 (21.6) 10,071 (20.1) 2 63 (4.8) 2773 (3.0) 68 (5.1) 2067 (4.1) 3 13 (1.0) 430 (.5) 17 (1.3) 381 (.8) 4 12 (.9) 302 (.3) 13 (1.0) 313 (.6) Chronic Disease Score, mean (SD) 1.70 (2.5) 1.81 (2.5) 1.80 (2.6) 2.22 (2.8) NOTE. Because of the large sample size, numeric differences achieved statistical significance but did not carry clinical significance.