Impact of Compliance With Proton Pump Inhibitors on NSAID Treatment

, , , , ,
The American Journal of Managed Care, October 2009, Volume 15, Issue 10

Patients with gastroesophageal reflux disease who are compliant with proton pump inhibitor therapy stay on NSAIDs longer than noncompliant patients.


To assess the impact of patient compliance with proton pump inhibitor (PPI) therapy on nonsteroidal anti-inflammatory drug (NSAID) treatment duration and upper-gastrointestinal (GI) complications in patients with gastroesophageal reflux disease (GERD).

Study Design:

Retrospective cohort study.


Study subjects were GERD patients receiving cotherapy with a PPI and a cyclooxygenase-2—selective (COX-2–selective) or nonselective NSAID. Patients compliant and noncompliant with PPI therapy were compared on NSAID treatment duration and incidence of upper-GI events.Kaplan-Meier analysis and a multivariate Cox proportional hazards model were used to compare durations of NSAID treatment, controlling for baseline characteristics. The incidences of GI events were compared using incidence rate and Poisson regression models. The analyses were conducted separately for patients taking COX-2–selective NSAIDs and those taking nonselective NSAIDs.


In both patient groups taking a COX-2—selective agent (n = 12,562; 70.9% compliant) and nonselective NSAID (n = 17,487; 69.9% compliant), mean NSAID treatment duration was significantly longer (84.0 days and 20.8 days longer, respectively) in PPI-compliant patients than in noncompliant patients. Compliance with PPI therapy was associated with a greater reduction in the incidence of GI events than noncompliance in both patients taking a COX-2–selective NSAID (6-fold vs 5-fold; P = .026) and patients taking a nonselective NSAID (8-fold vs 6-fold; P = .002).


In GERD patients receiving NSAIDs, those who were compliant with PPI therapy had a longer NSAID treatment duration and better GI tolerability than those who were noncompliant.

(Am J Manag Care. 2009;15(10):681-688)

This study is the first to examine the effect of compliance with proton pump inhibitor (PPI) therapy on the duration of nonsteroidal anti-inflammatory drug (NSAID) treatment and gastrointestinal (GI) tolerability in patients with gastroesophageal reflux disease.

  • Compliance with PPI therapy while patients are taking cyclooxygenase 2—selective (COX-2–selective) or nonselective NSAIDs is associated with longer NSAID treatment duration than noncompliance with PPI therapy.
  • Compliance with PPI therapy while patients are taking COX-2—selective or nonselective NSAIDs is associated with better GI tolerability than noncompliance with PPI therapy.

Gastrointestinal (GI) adverse events are widely recognized as side effects of nonsteroidal anti-inflammatory drugs (NSAIDs).1,2 When left untreated, NSAID-related upper GI symptoms may lead to early treatment discontinuation,3 decreased quality of life,4,5 and increased healthcare resource use.6-8 In patients with gastroesophageal reflux disease (GERD), NSAIDs can worsen GERD symptoms,9 leading to suboptimal treatment with NSAIDs. Proton pump inhibitors (PPIs) are gastroprotective agents (GPAs) that have been shown to provide effective healing of upper-GI ulcers associated with nonselective NSAIDs.10 Proton pump inhibitors also have been shown to be effective in preventing the recurrence of such ulcers when coprescribed with NSAIDs.11,12

Several studies have reported that the success of an NSAID-GPA coprescription therapy in reducing ulcer complications depends on patients’ treatment adherence. Goldstein et al13 retrospectively evaluated the clinical impact of nonadherence to a GPA among NSAID users and found that nonadherence to the GPA (ie, adherence rates of <80%) was associated with a significantly higher risk of upper-GI ulcers/complications in nonselective NSAID users (odds ratio = 2.4; 95% confidence interval [CI] = 1.0, 5.6). Likewise, van Soest et al14 used data from a primary care electronic medical records database in the Netherlands and reported that the risk of a serious NSAID-related GI complication increased by 16% (95% CI = 2%, 32%) for every 10% decrease in GPA adherence (including histamine-2 receptor antagonist [H2RA] blockers, PPIs, and misoprostol) in high-risk patients who received both nonselective and cyclooxygenase-2—selective (COX-2–selective) NSAIDs. Findings from a retrospective observational study conducted by Sturkenboom et al15 showed that adherence to an NSAID plus a GPA (PPI or H2RA) was low (63%) and decreased over the length of NSAID use.

Although previous studies have investigated the impact of adherence to PPI-NSAID cotherapy on the incidence of GI complications in the general NSAID population, the present study assessed the association among GERD patients, the population susceptible to the GI toxicity of NSAIDs (potentially even COX-2 NSAIDs), and therefore likely to benefit from the gastroprotective effect of PPIs. The objectives of the study were 2-fold. First, we investigated whether compliance with PPI therapy would extend the duration of adherence to NSAIDs. Second, we assessed whether PPI compliance in cotherapy with NSAIDs would reduce the incidence of upper-GI adverse events. For both objectives, COX-2—selective and nonselective NSAID users were analyzed separately.


Data Source

This study uses administrative health insurance claims data from the Ingenix Impact National Managed Care Database, developed by Integrated Health Care Information Solutions (Waltham, MA). This database contains complete medical history data for more than 25 million managed care lives in more than 30 health plans across all US census regions. Data elements include patient demographic characteristics, enrollment records, inpatient and outpatient diagnoses and procedures, and outpatient pharmacy dispensing claims.

Study Design

A retrospective cohort design was used to compare outcomes between PPI-compliant and PPI-noncompliant GERD patients receiving NSAID therapy from January 2000 through February 2005 (Figure). Generic and brand-name prescriptions for esomeprazole, lansoprazole, omeprazole, pantoprazole, and rabeprazole were included as PPI therapies; NSAIDs considered were COX-2—selective (celecoxib, rofecoxib, valdecoxib) or nonselective (eg, diclofenac, diflunisal, etodolac, fenoprofen, flurbiprofen, ibuprofen, indomethacin, ketoprofen, ketorolac tromethamine, meclofenamate, meloxicam, nabumetone, naproxen, oxaprozin, piroxicam, sulindac, tolmetin). Patients were required to meet the following criteria for inclusion: continuous enrollment in a health plan for the time period identified in the study, ≥2 insurance claims with a GERD diagnosis (International Classification of Diseases, Ninth Revision, Clincial Modification [ICD-9-CM] diagnosis codes 530.1x, 530.81, 787.1), ≥90 days of health insurance eligibility before the first GERD claim (GERD index date), age ≥18 years at the first dispensing of a PPI, and ≥1 episode of NSAID therapy with the concomitant use of a PPI after the first GERD claim, where concomitance was defined as an overlap of at least 1 day between NSAID and PPI days of drug supplies. Patients who had used PPIs before their GERD index date were excluded. The unit of observation in the study was the NSAID treatment episode, and only those episodes with concomitant use of a PPI were included in the analysis. For example, if a patient had 2 NSAID treatment episodes, 1 with concomitant PPI use and 1 without PPI use, only the treatment episode with concomitant PPI use was analyzed. In this example, the PPI episode was used to assess PPI compliance status, as described in the section below. The PPI index date was defined as the date of the first PPI dispensing after the GERD index date.

Definition of Compliance With PPI Cotherapy

To measure PPI compliance, patients had to have a continuous period of PPI therapy with at least 2 PPI prescription dispensings. Specifically, a patient had to have (1) at least 1 PPI episode, which is defined as a continuous period of PPI therapy with no interruption of more than 56 days in drug supply, and (2) at least 2 PPI dispensings within an episode, which enabled measurement of patient compliance. Patient-level PPI compliance was calculated using the medication possession ratio (MPR), defined as follows:

MPR = days of PPI supply/days between date of first prescription and drug supply end date of the last refill.

The numerator of this ratio is the total days of PPI prescription drug supplied, including covered days from all refills. The denominator is the intended treatment duration, calculated as the number of days between the first and last refill date plus supply days of the last refill.16 If a patient had more than 1 PPI episode, the MPR was calculated as the sum of the supply days in each episode divided by the sum of the durations in each episode. Using the standard methodology, an MPR >80% was considered to represent PPI compliance, and an MPR ≤80% was considered to represent PPI noncompliance.17,18


Patient demographic information (eg, age, sex) and health plan type (eg, health maintenance organization, point-of-service, preferred provider organization) were assessed on the date of the first PPI claim. ICD-9-CM codes

were used to identify comorbidities associated with GERD,including cough, asthma, sleeplessness, dyspepsia, and dysphagia,19-22 and for history of GI (duodenal, gastric, or peptic) ulcers. The Charlson comorbidity index was calculated as an indicator of general health state.23 Comorbidities related to GERD, history of GI ulcers, and the Charlson comorbidity index were assessed based on claims up to 180 days before the PPI index date. Finally, a compliance behavior indicator, general compliance, was introduced to control for underlying general behavioral differences in medication compliance across the compliant and noncompliant groups to isolate the compliance effect specifically related to PPI therapy. This measure was established based on patients’ compliance with 3 common long-term therapies used to treat chronic diseases (diabetes mellitus, hypertension, and hypercholesterolemia), using the previously described MPR calculation. The general compliance indicator was determined using all data available prior to baseline because it was intended to measure the general behavior of the patient as opposed to the patient’s health status at the beginning of the study. For patients without any of the identified chronic diseases, a separate category of “unknown” was assigned to their general compliance value.

Definition of Outcome Measures

Two outcome measures of interest were compared in the PPI-compliant and PPI-noncompliant groups: NSAID treatment duration and the incidence rate of upper-GI events. Upper-GI events included peptic ulcer disease (ICD-9-CM 533), disease of the esophagus (ICD-9-CM 530) except esophagitis (ICD-9-CM 530.1) and esophageal reflux (ICD-9-CM 530.81), upper-GI bleed (ICD-9-CM 531, 532, 534, 578), and gastritis and duodenitis (ICD-9-CM 535). We specifically excluded esophagitis and esophageal reflux because these were already used to identify the population with GERD.

For each patient, the use of a COX-2—selective or nonselective NSAID was divided into treatment episodes, defined as continuous periods of therapy with no interruption of more than 28 days in drug supplies. Therefore, 1 patient could have more than 1 episode eligible for analysis. For each NSAID treatment episode, the NSAID treatment duration was defined as the number of days from the date of initiation of NSAID treatment to the end of supply days of the last NSAID dispensing.

To measure the impact of PPI compliance on GI tolerability during NSAID use, each treatment episode was divided into 2 periods: the pre-PPI period (the start of the NSAID treatment episode to the first day of PPI dispensing in that episode) and the post-PPI period (the first day of PPI dispensing to the end of the NSAID treatment episode). The incidence rate of GI events before PPI initiation was higher for PPI-compliant patients than for PPI-noncompliant patients. Therefore, the GI incidence rates in the post-PPI period were benchmarked against those in the pre-PPI period by calculating the ratio of the pre-PPI incidence rate over the post-PPI incidence rate (IRR) for both the PPI-compliant and PPI-noncompliant groups. The incidence rate of GI events was defined as the number of GI-related events per person per year, to adjust for varying observation lengths in an observational setting. Multiple GI diagnoses recorded on the same day were considered a single event.

Statistical Analysis

Baseline characteristics of the compliant and noncompliant PPI groups were summarized descriptively. The difference in NSAID treatment duration between PPI-compliant and PPI-noncompliant groups was determined using Kaplan-Meier survival analysis and multivariate Cox proportional hazards models. The multivariate proportional hazards modelincluded the following baseline covariates for adjustment: age, sex, health plan type, GERD-related comorbidities (ie, cough, asthma, sleeplessness, dyspepsia, dysphagia), history of GI ulcer, Charlson comorbidity index, and general compliance. Because general compliance could not be assessed for all patients, we introduced an additional dummy variable for missing data that took a value of 1 when general compliance could be assessed and zero otherwise. The dummy variable permits the inclusion of all patients in the analysis by assigning those with missing values to a separate category instead of deleting them. The expected NSAID treatment duration based on the proportional hazards model was estimated as the area under the resulting survival curve. A 95% CI around this measure was calculated using a bootstrap approach.24

P values for the comparison of IRRs of GI events for PPIcompliant and PPI-noncompliant groups were based on the asymptotic standard deviations of the incidence rates. The IRRs from both PPI-compliant and PPI-noncompliant groups also were compared using a multivariate Poisson regression model, adjusting for the aforementioned baseline characteristics plus indicative variables for PPI compliance (A), pre-PPI versus post-PPI periods (B), and the interaction of A and B. Controls for all baseline characteristics also were included in the model. Because of the logarithm nature of the Poisson model, the respective IRRs for the compliant and noncompliant groups were obtained as exp{A+B} and exp{B}.

Univariate incidence rate analyses were performed using the iri procedure25 of STATA version 9 (StataCorp LP, College Station, TX). All other statistical analyses were performed using SAS software version 9.1 (SAS Institute, Inc, Cary, NC). A 2-sided a level of .05 was used to assess statistical significance.


Patient Characteristics

Table 1A

Table 1B

and present the baseline characteristics of PPI-compliant and PPI-noncompliant groups for the COX-2—selective and nonselective NSAID populations. Of the 12,562 patients taking COX-2–selective NSAIDs who met the selection criteria, 8907 (70.9%) were identified as PPI compliant; of the 17,487 patients taking nonselective NSAIDs who met the selection criteria, 12,092 (69.1%) were identified as PPI compliant. PPI-compliant patients were older than PPI-noncompliant patients in both the COX-2–selective (mean age, 53.8 years vs 50.5 years) and nonselective(mean age, 50.0 years vs 46.4 years) NSAID populations. Patients

were enrolled in a variety of health plan types, though most were in health maintenance organizations and preferred provider organizations. At baseline, the PPI-compliant and PPI-noncompliant groups were comparable in GERD-related comorbidities, except for sleeplessness. In both the COX-2—selective and nonselective NSAID populations, PPI-compliant patients had slightly higher Charlson comorbidity index values than the PPI-noncompliant patients. On average, 102 days and 104 days had elapsed between the GERD index date and the first PPI dispensing in the COX-2-selective and nonselective NSAID populations, respectively.

Impact of PPI Compliance on NSAID Treatment Duration

Table 2

presents the average duration of NSAID treatment in the PPI-compliant and noncompliant groups for both the COX-2—selective and nonselective NSAID cohorts. Overall, the COX-2–selective group had a longer NSAID treatment duration than the nonselective NSAID group, suggesting that patients exposed to cotherapy with a COX-2–selective NSAID and a PPI had better tolerability of NSAID intake than those exposed to cotherapy with a nonselective NSAID and a PPI. Within each NSAID group, the Kaplan-Meier survival analyses yielded a longer NSAID treatment duration in the PPI-compliant group (COX-2 selective, 214.1 days; nonselective, 92.0 days) than in the PPI-noncompliant group (COX-2 selective, 111.9 days; nonselective, 55.0 days). These results were confirmed by multivariate Cox proportional hazards regression analysis, which showed that PPI-compliant patients had significantly longer NSAID treatment duration than PPI-noncompliant patients. PPIcompliant patients took COX-2–selective NSAIDs 84.0 days longer (95% CI = 76.2, 89.1 days); they took nonselective NSAIDs 20.8 days longer (95% CI = 17.6, 23.5 days).

Impact of PPI Compliance on GI Event Incidence Rates

Table 3

shows GI event incidence rates before and after PPI initiation for the PPI-compliant and PPI-noncompliant groups within the COX-2—selective and nonselective NSAID populations. In both populations, the PPI-compliant groups had higher GI event incidence rates before the initiation of PPI therapy compared with their PPI-noncompliant counterparts. Patients in the nonselective NSAID group had a higher baseline GI event incidence rate beforePPI therapy, and thus experienced a steeper decrease in GI complications after the commencement of PPI therapy than the COX-2–selective NSAID population. With both the COX-2–selective and nonselective NSAID populations, post-PPI GI event incidence rates were lower than the pre-PPI rates in both the PPI-compliant and PPI-noncompliant groups, perhaps because of the overall GI-protective effects of PPIs. In the COX-2–selective NSAID population, the PPI-compliant group experienced a 6-fold decrease in GI events versus a 5-fold decrease in the PPI-noncompliant group (compliance/noncompliance ratio = 1.237; P <.001). Similarly, in the nonselective NSAID population, the PPIcompliant group had an 8-fold decrease in GI events versus a 6-fold decrease in the PPI-noncompliant group (compliance/ noncompliance ratio = 1.401; P <.001).

A substantial decrease in GI event incidence rates after PPI initiation among PPI-compliant patients also was observed using multivariate Poisson regression adjustments in both the COX-2—selective (compliance/noncompliance ratio = 1.269; P = .026) and nonselective (compliance/noncompliance ratio = 1.470; P = .002) NSAID populations. Among the control variables, history of dysphagia and ulcer also significantly increased the risk of GI events in both the COX-2 and nonselective NSAID populations.


Noncompliance with PPI therapy among NSAID users with GERD is a significant issue. Approximately 30% of the COX-2—selective and nonselective NSAID users identified in the present study did not comply with their PPI cotherapy, defined as an MPR >80%. This study’s results indicate that compliance with PPI therapy is associated with longer NSAID treatment duration.

To the best of our knowledge, this is the first study that investigated the relationship between PPI compliance and NSAID treatment duration. One may argue that the relationship observed in the present study between PPI compliance and NSAID treatment duration is not causal, but a simultaneous consequence of patients’ underlying behavior. Specifically, patients who are PPI compliant also may be those who are more likely to continue the coprescribed NSAID treatment. We attempted to address this issue by introducing the general compliance indicator in the multivariate models. This variable was positively related to NSAID treatment duration with a significant coefficient at an a level of .05 (data not shown). Nevertheless, even after controlling for users’ general compliance behavior, PPI compliance alone still had an independent, positive, and significant effect on NSAID treatment duration.

We also found that PPI compliance is associated with a lower incidence of upper-GI events, a finding consistent with previous results published by Goldstein et al,13 who found that nonadherence to GPAs was associated with a significantly higher risk of upper-GI ulcers/complications among nonselective NSAID users. However, contrary to this study’s results, Goldstein et al did not observe such a relationship for users of COX-2—selective NSAIDs. The reason for this difference is not clear. It is possible that in the earlier era of the Goldstein et al study, clinicians might have been more inclined to underreport less specific upper-GI complaints in patients using COX-2 inhibitors, possibly because COX-2 inhibitors were promoted during their initial marketing period as having a better GI safety profile than nonspecific NSAIDs. However, it is unlikely that such a reporting bias would extend to more significant upper-GI complications such as GI bleeding.

The finding that compliance with PPI therapy is associated with longer NSAID treatment duration suggests that NSAID treatment may be optimized by the effective alleviation of GI side effects. Moreover, association between PPI compliance and both a reduced incidence of GI events and longer NSAID treatments supports the existing literature indicating that GIevents can lead to discontinuation of NSAID therapy.3 Because PPI compliance is associated with reduced incidence rates of GI events, it is reasonable to infer that PPI compliance also would make NSAID treatment discontinuation less likely.

Because drug claims data do not contain diagnosis information, we attempted to identify reasons for prescriptions of NSAID medications. Examining the 28-day time interval before each NSAID treatment initiation, we found that the top 3 most common diagnoses associated with nonselective and COX-2—selective NSAIDs were chronic pain (14%), osteoarthritis (5%), and rheumatoid arthritis (1%). Results from a subset analysis of patients with chronic pain, osteoarthritis, or rheumatoid arthritis confirmed findings from this study in that PPI-compliant patients used NSAIDs for a longer period of time and experienced larger declines in the incidence of GI events than PPI-noncompliant patients.

In our multivariate Poisson regression analyses for upper-GI incidence rates (Table 3), we found that in addition to PPI compliance, history of GI ulcer and dysphagia were significantly associated with the outcome. It is not clear whether there is a biologic basis for an association between dysphagia and the increased risk of GI events.

This study has several data and methodologic limitations. First, many NSAID episodes were relatively short, thus preventing the evaluation of PPI compliance within these episodes. Patient-level compliance was determined based on all eligible episodes for a given patient. Although this approach retained all NSAID episodes during which PPIs were dispensed, thus avoiding potential selection bias associated with dropping some episodes, it could have introduced some noise in the estimation, making the PPI-compliant and PPInoncompliant groups appear more similar than they actually were. However, given the magnitude of the differences in both NSAID treatment duration and GI event incidence observed between the 2 groups, that should not be a concern.

Second, the general compliance indicator introduced to isolate the respective effects of PPI compliance and patient behavior on NSAID treatment duration could not be assessed for all patients. Thus, it is possible that the 2 effects were not isolated perfectly and that the behavioral effect remained captured in part by the PPI compliance coefficient.

Third, this analysis was subject to the limitations generally associated with administrative claims data such as claims incompleteness and potential inconsistencies in coding. Sources of incompleteness include limited medical history time frame, lack of over-the-counter drug information, lack of information on physicians’ therapeutic intentions, and the coverage overlap between Medicare and private insurance for patients older than 65 years. Related conditions may be miscoded in claims data. For example, miscoding of esophageal disorders such as achalasia, esophageal ulcer, or stricture as GERD could potentially confound our study results. One of the GERD-related comorbidities, sleeplessness, also is susceptible to miscoding. Although sleeplessness due to esophageal reflux is a reasonable marker of GERD activity, sleep disturbances are vague symptoms subject to miscoding, and pain (the underlying indication for the NSAID use itself) could cause sleeplessness as well.

Lastly, the study is limited in that the PPI-compliant and PPI-noncompliant groups were different with respect to age and health conditions. Every effort was made to account for these differences by performing multivariate analyses. Nevertheless, residual confounding still may exist. Despite these limitations, this research has the advantages of relying on real-world data and a large sample size, making the study findings generalizable. Moreover, it uses multivariate adjustment approaches to control for confounding factors such as general compliance of the patients.


Among GERD patients who used COX-2—selective or nonselective NSAID and PPI cotherapy, those who were PPI compliant continued NSAID treatment longer and experienced larger reductions in the incidence of GI events compared with those who were PPI noncompliant.


The authors thank Si-Tien Wang, MSC (Analysis Group, Inc, Boston, MA) for medical writing support.

Author Affiliations: From the Analysis Group, Inc (MSD, AG, HL), Boston, MA; and AstraZeneca (RL, BEL, JAC), Wilmington, DE. Dr Luo is now with Pfizer Inc, Parsippany, NJ. Dr Lewis is now with Esai Inc, Woodland Hills, NJ.

Funding Source: This study was funded by AstraZeneca LP (Wilmington, DE).

Author Disclosures: Dr Duh, Mr Gosselin, and Dr Lohoues are employees of the Analysis Group, Inc, which was paid by AstraZeneca for the development and preparation of this study. Drs Luo and Lewis were employees of AstraZeneca, a manufacturer of proton pump inhibitors, at the time the study was conducted. Dr Crawley is an employee of AstraZeneca and reports owning stock in the company.

Authorship Information: Concept and design (MSD, AG, RL, HL, BEL, JAC); acquisition of data (MSD, AG, HL); analysis and interpretation of data (MSD, AG, RL, HL, BEL, JAC); drafting of the manuscript (MSD, AG, HL, BEL); critical revision of the manuscript for important intellectual content (MSD, AG, RL, HL, BEL, JAC); statistical analysis (MSD, AG, RL, HL); obtaining funding (MSD, RL, JAC); administrative, technical, or logistic support (RL, JAC); and supervision (MSD, AG, RL).

Address correspondence to: Mei Sheng Duh, MPH, ScD, Analysis Group, Inc, 111 Huntington Ave, 10th Fl, Boston, MA 02199. E-mail:

1. Kasper DL, Braunwald E, Fauci AS, et al. Harrison’s Principles of Internal Medicine. 17th ed. Accessed October 6, 2009.

2. Dubois RW, Melmed GY, Henning JM, Laine L. 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(2):197-208.

3. Watson DJ, Harper SE, Zhao PL, Quan H, Bolognese JA, Simon TJ. Gastrointestinal tolerability of the selective cyclooxygenase-2 (COX-2) inhibitor rofecoxib compared with nonselective COX-1 and COX-2 inhibitors in osteoarthritis. Arch Intern Med. 2000;160(19):2998-3003.

4. Wiklund I, Langstrom G, Hawkey C, et al. Upper GI symptoms reduce the quality of life for patients on long-term NSAID therapy. Paper presented at: the European League Against Rheumatism (EULAR); June 18-21, 2003; Lisbon, Portugal.

5. Yacavone RF, Locke GR 3rd, Provenzale DT, Eisen GM. Quality of life measurement in gastroenterology: what is available? Am J Gastroenterol. 2001;96(2):285-297.

6. Chevat C, Peña BM, Al MJ, Rutten FF. Healthcare resource utilisation and costs of treating NSAID-associated gastrointestinal toxicity. A multinational perspective. Pharmacoeconomics. 2001;19(suppl 1):17-32.

7. Moore RA. The hidden costs of arthritis treatment and the cost of new therapy—the burden of non-steroidal anti-inflammatory drug gastropathy. Rheumatology (Oxford). 2002;41(suppl 1):7-15.

8. Herings RM, Klungel OH. An epidemiological approach to assess the economic burden of NSAID-induced gastrointestinal events in The Netherlands. Pharmacoeconomics. 2001;19(6):655-665.

9. Bigard MA, Pelletier AL. Esophageal complications of non steroidal antiinflammatory drugs. Gastroenterol Clin Biol. 2004;28(Spec No 3):C58-61.

10. Hawkey C, Karrasch J, Szczepañski L, et al. Omeprazole compared with misoprostol for ulcers associated with nonsteroidal anti-inflammatory drugs. Omeprazole versus Misoprostol for NSAID-induced Ulcer Management (OMNIUM) Study Group. N Engl J Med. 1998;338(11):727-734.

11. Scheiman JM, Yeomans ND, Talley NJ, et al. Prevention of ulcers by esomeprazole in at-risk patients using non-selective NSAIDs and COX-2 inhibitors. Am J Gastroenterol. 2006;101(4):701-710.

12. Chan FK, Wong VW, Suen BY, et al. Combination of a cyclooxygenase-2 inhibitor and a proton-pump inhibitor for prevention of recurrent ulcer bleeding in patients at very high risk: a double-blind, randomised trial. Lancet. 2007;369(9573):1621-1626.

13. Goldstein JL, Howard KB, Walton SM, McLaughlin TP, Kruzikas DT. Impact of adherence to concomitant gastroprotective therapy on nonsteroidal-related gastroduodenal ulcer complications. Clin Gastroenterol Hepatol. 2006;4(11):1337-1345.

14. van Soest EM, Sturkenboom MC, Dieleman JP, Verhamme KM, Siersema PD, Kuipers EJ. Adherence to gastroprotection and the risk of NSAID-related upper gastrointestinal complications. Aliment Pharmacol Ther. 2007;26(2):265-275.

15. Sturkenboom MC, Burke TA, Tangelder MJ, Dieleman JP, Walton S, Goldstein JL. Adherence to proton pump inhibitors or H2-receptor antagonists during the use of non-steroidal anti-inflammatory drugs. Aliment Pharmacol Ther. 2003;18(11-12):1137-1147.

16. Sikka R, Xia F, Aubert RE. Estimating medication persistency using administrative claims data. Am J Manag Care. 2005;11(7):449-457.

17. Thiebaud P, Patel BV, Nichol MB, Berenbeim DM. The effect of switching on compliance and persistence: the case of statin treatment. Am J Manag Care. 2005;11(11):670-674.

18. Ward A, Ishak K, Proskorovsky I, Caro J. Compliance with refilling prescriptions for atypical antipsychotic agents and its association with the risks for hospitalization, suicide, and death in patients with schizophrenia in Quebec and Saskatchewan: a retrospective database study. Clin Ther. 2006;28(11):1912-1921.

19. Devault KR. Extraesophageal symptoms of GERD. Cleve Clin J Med. 2003;70(suppl 5):S20-S32.

20. Wallander MA, Johansson S, Ruigómez A, García Rodríguez LA, Jones R. Dyspepsia in general practice: incidence, risk factors, comorbidity and mortality. Fam Pract. 2007;24(5):403-411.

21. Dubois RW, Aguilar D, Fass R, et al. Consequences of frequent nocturnal gastro-oesophageal reflux disease among employed adults: symptom severity, quality of life and work productivity. Aliment PharmacolTher. 2007;15;25(4):487-500.

22. Guillemot F, Ducrotté P, Bueno L. Prevalence of functional gastrointestinal disorders in a population of subjects consulting for gastroesophageal reflux disease in general practice. Gastroenterol Clin Biol. 2005;29(3):243-246.

23. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613-619.

24. Efron B, Tibshirani RJ. An Introduction to Bootstrap. London: Chapman and Hall; 1993.

25. StataCorp LP. Survival Analysis and Epidemiological Tables. College Station, TX: Stata Press; 2005:21-77.