Comparative Safety of Nsaids for Gastrointestinal Events in Asia-Pacific Populations: a Multi-Database, International Cohort Study

Home , Loxoprofen

Comparative Safety of NSAIDs for Gastrointestinal Events in Asia-Pacific Populations: A Multi-Database, International Cohort Study

Edward Chia-Cheng Lai,a,b,c Ju-Young Shin,d Kiyoshi Kubota,e Kenneth K. C. Man,f Byung Joo Park,d,g Nicole Pratt,h Elizabeth E. Roughead,h Ian C. K. Wong,i Yea-Huei Kao Yang,a,i Soko Setoguchic,j,k*

Author Affiliations: aSchool of Pharmacy, Institute of Clinical Pharmacy and Pharmaceutical Sciences, National Cheng Kung University, Tainan, Taiwan; bDepartment of Pharmacy, National Cheng Kung University Hospital, Tainan, Taiwan; cDuke Clinical Research Institute, Duke University School of Medicine, Durham, North Carolina, USA; dDepartment of Preventive Medicine, Seoul National University College of Medicine, Seoul, South Korea; eDepartment of Pharmacoepidemiology, University of Tokyo, Tokyo, Japan; fCentre for Safe Medication Practice and Research, Department of Pharmacology and Pharmacy, University of Hong Kong, Hong Kong; gOffice of Drug Utilization Review, Korea Institute of Drug Safety and Risk Management, Seoul, South Korea; hQuality Use of Medicines and Pharmacy Research Centre, Sansom Institute for Health Research, University of South Australia, Adelaide, Australia; iHealth Outcome Research Center, National Cheng-Kung University, Tainan, Taiwan; and jDepartment of Medicine, Rutgers Robert Wood Johnson Medical School, New Brunswick, New Jersey, USA; and kRutgers Center for Pharmacoepidemiology and Treatment Science, Institute for Health, Rutgers University, New Brunswick, New Jersey, USA.

* Authors 3 through 9 are listed in alphabetical order.

Funding/Support: This work was supported by a research agreement between Duke University and Janssen Research & Development, LLC. Dr Kubota was supported by grant 11-2-021 from the Pfizer Health Research Foundation.

Corresponding Author: Soko Setoguchi, MD, DrPH, Institute for Health, Rutgers University, 112 Paterson St, New Brunswick, NJ 08901, USA; telephone: 848-932-6507; email: [email protected].

Key Words: Anti-Inflammatory Agents, Non-Steroidal; Pharmacoepidemiology

Key Points:  Study of nonsteroidal anti-inflammatory drugs (NSAIDs) commonly used in Asia-Pacific countries is limited.  Compared with diclofenac, loxoprofen was associated with a lower risk of gastrointestinal hospitalizations in Korea and mefenamic acid with a lower risk in Taiwan and Korea.  Our study provides lessons for future multinational pharmacoepidemiologic studies in the Asia- Pacific region.

Conflict of Interest Disclosures: Dr. Setoguchi is served as a consultant and receiving a research grant from Pfizer Inc. and Janssen Inc. Dr Kubota was supported by grant 11-2-021 from the Pfizer Health Research Foundation. Prof. Wong reported receiving a research grant from the Pfizer Health Research Foundation to the University of Hong Kong to investigate the use of biologics in Hong Kong. All other authors declared that they have no conflicts of interest.

2847 words Comparative Safety of NSAIDs August 1, 2018 Abstract

Purpose: The safety of nonsteroidal anti-inflammatory drugs (NSAIDs) commonly used in Asia-Pacific countries has had limited study. We assessed the risk of hospitalization for gastrointestinal events with loxoprofen and mefenamic acid compared with other NSAIDs in

Asia-Pacific populations.

Methods: We conducted a cohort study using a distributed network with a common data model in Australia, Hong Kong, Japan, Korea, and Taiwan. We included patients who initiated diclofenac, loxoprofen, mefenamic acid, or celecoxib and followed them until their first gastrointestinal hospitalization, switch or discontinuation of medication, disenrollment, or end of database coverage. We used Cox proportional hazards models to assess hospitalization risk.

Results: We identified 9879 patients in Japan, 70,492 in Taiwan, 263,741 in Korea, and

246 in Hong Kong who initiated an NSAID, and 44,013 patients in Australia, a predominantly

Caucasian population. The incidence of gastrointestinal hospitalization was 25.6 per 1000 person-years in Japan, 32.8 in Taiwan, 11.5 in Korea, 484.5 in Hong Kong, and 35.6 in

Australia. Compared to diclofenac, the risk of gastrointestinal events with loxoprofen was significantly lower in Korea (hazards ratio [HR], 0.37; 95% CI, 0.25-0.54) but not in Japan

(1.65; 95% CI, 0.47-5.78). The risk of gastrointestinal events with mefenamic acid was significantly lower in Taiwan (0.45; 95% CI, 0.26-0.78) and Korea (0.11; 95% CI, 0.05-0.27) but not Hong Kong (2.16; 95% CI, 0.28-16.87), compared with diclofenac.

Conclusions: Compared with diclofenac, loxoprofen was associated with a lower risk of gastrointestinal hospitalizations in Korea and mefenamic acid with a lower risk in Taiwan and

Korea.

2 Introduction

The Asian Pharmacoepidemiology Network (AsPEN) was formed to support the conduct of international pharmacoepidemiologic research and to facilitate prompt identification and validation of emerging safety concerns in Asian countries.1,2 Analyses have been conducted through this research network to build experience in using databases from AsPEN countries using a variety of methodological approaches, including sequence symmetry analysis (SSA).3

AsPEN consists of research groups based in the Asia-Pacific region, including Australia, China,

Hong Kong, Japan, Korea, Singapore, Taiwan, and Thailand.

It is well known that the use of nonselective nonsteroidal anti-inflammatory drugs

(NSAIDs) is associated with higher risks of upper and lower gastrointestinal events.4,5

Cyclooxygenase-2 (COX-2) inhibitors, a selective NSAID, reduce the risk of gastrointestinal events because the inhibition of COX-2 does not affect the synthesis of prostaglandins that protect the gastrointestinal tract.6,7 However, the safety of NSAIDs that are commonly or solely used in the Asia-Pacific region, including loxoprofen and mefenamic acid, has not been well studied. Also, the majority of evidence comes from countries with predominantly Caucasian populations. It is unclear whether ethnic differences between Asian and Caucasian populations play a role in the comparative safety profiles of NSAIDs.7-9

This international, multi-database cohort study aimed to compare the safety of loxoprofen and mefenamic acid with the safety other NSAIDs in terms of the risk of hospital admission for peptic ulcer diseases or upper and lower gastrointestinal bleeding. We compared patients who initiated loxoprofen or mefenamic acid with patients who initiated diclofenac, one of the commonly used and previously studied NSAIDs. We also included celecoxib in the study to benchmark the effect for COX-2 inhibitors with nonselective NSAIDs.

3 Method

Distributed Network Approach With a Common Data Model

We conducted a retrospective, multi-database, international pharmacoepidemiologic study using a distributed network approach10-12 and developed a study-specific common data model consisting of 8 tables, including patient demographic characteristics, eligibility, drug, inpatient encounter, inpatient diagnosis, outpatient diagnosis, inpatient procedures, and outpatient procedures (eMethods). Teams from all participating countries converted their original data structures to the common data model before analyses. The coordinating center generated a systematic SAS program based on the study protocol and distributed it to each site.

The results of the analyses were then returned to the coordinating center for collating.

Participating Databases

Participating countries included Australia, Hong Kong, Japan, Korea, and Taiwan. We included data from the Japan Medical Data Centre insurance claims database (JMDC) for 2005 through 2010; the Korea Health Insurance Review and Assessment Service database (HIRA) for

2006 through 2008; the Taiwan National Health Insurance Research Database (NHIRD) for 2002 through 2008; the Hong Kong Clinical Data Analysis and Reporting System (CDARS) for 2008 through 2012; and the Australian Department of Veterans’ Affairs health care claims database for 2001 through 2012. These databases have been described in detail elsewhere.2

Study Cohort Definitions

We selected patients who initiated NSAIDs (ie, diclofenac, loxoprofen, mefenamic acid and celecoxib). We considered the first dispensed NSAID to be the index medication, and we

4 used the dispensing date as the index date. We included only new users of NSAIDs, which we defined as patients who received no dispensing of an NSAID during the 6 months before the index date. Although use of nonprescription NSAIDs is not common in Asian countries,13 some patients in the study may have used over-the-counter NSAIDs. To minimize exposure misclassification, we further restricted the study population to patients who had certain chronic conditions (ie, diabetes mellitus, hypertension, myocardial infarction or acute coronary syndrome, stroke, and rheumatoid arthritis). These patients are more likely to seek medical care on a regular basis and, therefore, more likely to receive prescriptions for NSAIDs. Also, these patients are at high risk for gastrointestinal complications of NSAID use.14 Furthermore, we excluded patients with a history of cancer (other than non-melanoma skin cancer), human immunodeficiency virus infection, renal failure, liver injury, respiratory failure, or transplant to make the results more generalizable to patients who are chronically ill but not at the end of life.

Finally, we excluded patients with history of liver cirrhosis, esophageal varices, chronic alcoholism, or bariatric or other surgery resulting in gastrojejunal anastomosis, because the presence of these serious upper gastrointestinal diseases may confound events due to NSAID use, and patients with these conditions may be less likely to be prescribed NSAIDs.

Outcome and Follow-up

The outcome of interest was hospital admission with a diagnosis of peptic ulcer disease or gastrointestinal bleeding, which we identified by the presence of International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 531.xx-534.xx, 569.3x,

569.41, 569.8, 578.xx (in Hong Kong and Taiwan) or ICD-10 codes K25x-K28x, K62.5, K62.6,

K63.1-K63.3, K55.20, K55.21, K63.81, K63.89, and K92.0-K92.2 (in Australia, Japan, and

5 Korea) in any position on the inpatient claim. Selection of diagnosis codes was based on previous studies.15,16 The follow-up period for each patient lasted until the patient was hospitalized for a gastrointestinal event, was switched from the index medication, discontinued the medication (ie, had no dispensing for more than 14 days after the end date of the previous dispensing), discontinued enrollment in the databases, reached the end of the study follow-up, or died. We also calculated the number of gastrointestinal adverse event cases by specific gastrointestinal event, drug type, and site.

Covariate and Statistical Analyses

We report means and SDs for continuous variables and frequencies for dichotomous variables. We report the number of events and the incidence per person-year for the outcomes of interest. We estimated hazard ratios (HRs) and 95% CIs using Cox proportional hazards models to assess the risks of hospital admission for gastrointestinal events associated with initiation of loxoprofen, mefenamic acid, or celecoxib compared with diclofenac. We adjusted for demographic characteristics to account for potential confounding. Using inpatient and outpatient prescription data, diagnoses, and procedure data from the 180 days before the index date, we identified up to 500 covariates by high-dimensional propensity score (hdPS)17 and matched on hdPS using 5-to-1 greedy matching to control for confounders and unobserved factors.18-21 We selected hdPS for confounding adjustment for two reasons: First, the hdPS approach is likely to result in less residual confounding than conventional approaches, such as multivariable regression or a propensity score method using prespecified covariates, because hdPS allows data- driven selection of confounders that can be unique to each country. Second, the hdPS approach allows us to maintain data confidentiality when pooling results without sharing identifiable data

6 To evaluate the impact of censoring, we conducted an intention-to-treat analysis in which we considered patients to be exposed to the index medication until the occurrence of events, death, or the end of follow-up, regardless of subsequent changes in exposure. We limited the follow-up period to 1 year for the intention-to-treat analysis. We used SAS version 9.3 (SAS

Institute, Inc) for all analyses. This project was approved by the institutional review board of the

Duke University Health System.

7 Results

We identified 44 013 patients in Australia, 9879 in Japan, 70 492 in Taiwan, 263 741 in

Korea, and 246 in Hong Kong who initiated use of a prescription NSAID. The study population in Australia was predominantly Caucasian; the study populations in the other countries were predominantly Asian. Diclofenac was the most commonly initiated drug in all countries except

Japan, where loxoprofen accounted for 77.% of NSAID initiations. Loxoprofen was only available in Japan and Korea. Mefenamic acid accounted for a small proportion of NSAID initiations in all countries. Table 1 shows the baseline characteristics of the study population.

The mean age was 52.2(±11.4), 62.5 (±13.6) 73.6 (± 6.1), 67.1 (±14.6) and 75.3 (± 10.4) and approximately 56.2%, 48.4%, 36%, 43.9% and 55.1% of patients were male in Japan,

Taiwan, Korea, Hong Kong and Australia, respectively. The highest rate of comorbid condition was hypertension followed by diabetes in all countries; the rate of hypertension in Korea (94.6%) was much higher than that in other countries. The rates of other comorbid conditions and co- medications were generally higher in Taiwan and Korea compared with other countries

(eTable 1). The length of follow up time was longest in Taiwan (205.3 ±347.7 days) followed by

Australia (178.6 ±477.5 days), Hong Kong (98.1 ±208.5 days), and Japan (47.7 ±86.1 days). The mean follow up time was shortest and only 36.4 days (±37.7) in Korea.

The incidence rates of hospital admission for gastrointestinal events per 1000 person- years were 25.6 in Japan, 32.8 in Taiwan, 11.5 in Korea, 484.5 in Hong Kong, and 35.6 in

Australia (Table 2). Compared to diclofenac, the risk of gastrointestinal events with loxoprofen was significantly lower in Korea (HR, 0.37; 95% CI, 0.25-0.54) but not in Japan (HR, 1.65; 95%

CI, 0.47-5.78). The risk of gastrointestinal events with mefenamic acid was significantly lower in

Taiwan (HR, 0.45; 95% CI, 0.26-0.78), and Korea (HR, 0.11; 95% CI, 0.05-0.27) but not in

8 Hong Kong (HR, 2.16; 95% CI, 0.28-16.87), compared with diclofenac initiators. We found that celecoxib initiators had a significantly lower risk of hospital admission for gastrointestinal events compared to diclofenac initiators in Korea (HR, 0.43; 95% CI, 0.24-0.76) and Australia (HR,

0.41; 95% CI, 0.25-0.67) but not in Taiwan (HR, 1.08; 95% CI, 0.96-1.22). The results of unadjusted, demographic characteristics-adjusted hazard ratio and intention-to-treat analyses were largely consistent with primary analyses. The details are presented in the Table 3. We present the number of gastrointestinal adverse event cases by specific end point, drug type, and site in eTable 2. Specifically, we found most of the events were peptic ulcer disease and unspecified site of gastrointestinal bleeding.

9 Discussion

We used five nationwide databases from Australia, Hong Kong, Japan, Korea, and

Taiwan to assess the comparative gastrointestinal safety of NSAIDs that are commonly or solely used in the Asia-Pacific region. We found loxoprofen and mefenamic acid had lower risks of gastrointestinal events requiring hospitalization compared with diclofenac in Korea. To benchmark the effect size the risk of gastrointestinal events, we selected celecoxib as the reference indicator for COX-2 inhibitors known to have a better gastrointestinal safety profile.22

We found that loxoprofen and mefenamic acid may have similar or better gastrointestinal safety profiles than celecoxib.

Loxoprofen is a commonly used NSAID in Japan and Korea but is not marketed in other parts of the world except in Brazil.23,24 Loxoprofen is a pro-drug, which attenuates the inhibition of prostaglandin synthesis in the local gastrointestinal tract and mucosa and avoids topical irritation due to the presence of the free carboxylic group of the NSAIDs and low direct cytotoxicity on mucosal cells.25-27 However, a case-control study in Japan found a similar risk of upper gastrointestinal bleeding with loxoprofen as with aspirin and diclofenac.23 Our study supports this finding, as we observed that loxoprofen appeared to be associated with a higher risk of gastrointestinal events requiring hospitalization than diclofenac in Japan, although the association remained inconclusive because of the small sample size and low statistical power. By comparison, we observed a favorable gastrointestinal risk with loxoprofen in the Korean population. One possible explanation for this difference is that the Korean patients were older, and the benefit of the pro-drug characteristics of loxoprofen were greater in older patients who may be vulnerable to gastrointestinal events. Another possible explanation is that the injection

10 form of diclofenac, which may increase gastrointestinal events and decrease the relative hazard of loxoprofen, has been more frequently prescribed in Korea than in other countries.28

Mefenamic acid is used in all participating countries and most frequently in Taiwan,

Korea, and Hong Kong. Mefenamic acid has been reported to have a lower risk of gastrointestinal bleeding than aspirin in general healthy populations.29 Recently, Chang et al15 conducted a nationwide case-crossover study using a Taiwanese database and found that mefenamic acid had a lower risk of hospitalization for gastrointestinal events than other

NSAIDs, including diclofenac. Our findings support the results of previous studies, though the risk estimates from Hong Kong’s database were unstable because of the small sample size.

However, the potential associations of mefenamic acid with risks of nephrotoxicity and neurotoxicity in elderly patients may also be a concern.30

Celecoxib, a selective COX-2 inhibitor, has been associated with a significantly lower frequency of gastrointestinal adverse events than nonselective NSAIDs.22,31 Our findings in

Korea and Australia are consistent with previous observational studies that celecoxib has lower risk than diclofenac. However, we did not find a significant benefit in terms of gastrointestinal safety in Taiwan. This finding may be a result of uncontrolled selection bias, because use of celecoxib in Taiwan is only covered for patients with certain conditions, including previous history of gastrointestinal ulcer or bleeding, leading to a higher baseline risk of gastrointestinal events in celecoxib initiators.

The distributed network approach using a common data model and executing a systematic program based on a common data structure provided the advantage of controlling the quality of the analyses. This approach has been widely used in other academic networks.1,3,11

However, concerns about data privacy impede the pooling of individual patient data. Our study

11 highlights the importance of evaluating older medicines that are used commonly in Asian countries. The use of more expensive COX-2 inhibitors may increase the economic burden in the health care system, especially for countries with moderate income.32 If less costly loxoprofen and mefenamic acid pose favorable gastrointestinal safety profiles, they could play a role in curbing the high cost of pain control.6,32

This is the first multi-database, international pharmacoepidemiologic study in an Asia-

Pacific population, and implementation of the study holds a number of lessons for future international pharmacoepidemiologic studies in Asia. There were several challenges in the implementation of the study. First, differences among countries in health care systems and medical practice led to potential biases. For example, we observed a wide range of incidence rates for gastrointestinal bleeding events. It has been reported previously that the hospital admission rate for upper gastrointestinal complications in Taiwan may be high due to either the convenience of the health care system or relatively low medical costs.33 This is reflected in our finding that the overall incidence of hospital admission for gastrointestinal events was higher in

Taiwan than in other countries except Hong Kong. The high rate in Hong Kong may be a result of the nature of the hospital-based electronic health records and the insurance system. That is, the selected patients were generally experiencing more severe conditions, giving rise to higher likelihood of adverse events occurrence. In addition, sicker patients may have had more complete information in the hospital-based electronic health records and therefore greater likelihood of being detected.34 However, the number of selected patients in Hong Kong in the study was low, leading to unstable estimates. On the other hand, it has been reported that the persistence of the use of NSAIDs was low, leading to the possibility to underestimation of gastrointestinal events in Korea because of temporal use of NSAIDs and lost follow-up. Second,

12 the validity of the selected ICD-9 and ICD-10 codes in the different databases was uncertain, which might also cause possible misclassification bias of the analyses. Therefore, the study focused on events requiring hospitalization to increase the internal validity of the analyses and to minimize bias. It is noteworthy that the comparisons of characteristics among countries should be careful because the differences could result from underlying causes. Possible reasons, for example, could be include differences in the nature of the databases, medical systems, and the thresholds of diagnosis or use of medications. Despite the variability in characteristics of patients and incidence rates of gastrointestinal events, it should not affect the validity of relative hazards comparing different NSAIDs within countries, because the definitions we used are likely to be highly specific based on existing studies.14,15 Third, as in previous studies in the AsPEN network, there was variability in the products available across countries and the preference and indication of medicines within countries. This is a particular challenge in comparative effective and safety studies across countries with different regulatory conditions. Many other networks such as the Exploring and Understanding Adverse Drug Reactions (EU-ADR) project in Europe and the Mini-Sentinel project in the United States combine databases under similar regulatory frameworks, which mean that in general the indications and the availability of the medicines is more consistent across databases than across Asian countries. We were unable to describe the distribution of covariates. To enhance control for confounding, we allowed data-driven selection of confounders unique to each country using hdPS and did not perform multivariate analyses using predefined covariates. Fourth, we censored data for patients at the time of switching to a different NSAID. However, because the reasons for switching were not available in the databases, it impossible to assess whether the switching is associated with the study end point and the possibility of informative censoring cannot be excluded, which will likely bring the

13 results toward null for NSAIDs that may cause higher rates of gastrointestinal symptoms. From the post hoc analysis, we found a significant number of gastrointestinal bleeding events were coded as unspecified sites representing the clinical nature of bleeding; the site of bleeding may not be identified clinically in many cases. Also, the validity of each gastrointestinal end point has not been examined in Asian databases. Future studies may consider collecting information from hospital medical records to validate these specific gastrointestinal events and confirm the risks on these specific gastrointestinal end point.

In conclusion, this study largely supports previous evidence from countries outside of the

Asia-Pacific region that patients initiating treatment with loxoprofen and mefenamic acid had lower risk of gastrointestinal events compared to patients initiating diclofenac in most but not all countries. In addition, our study provides lessons for future international pharmacoepidemiologic studies in the Asia-Pacific region.

14 Acknowledgments

Funding/Support: This work was supported by a research agreement between Duke

University and Janssen Research & Development, LLC, and a grant from the Ministry of Science and Technology of Taiwan (ID: 106-2320-B-006-025-MY2).

Additional Contributions: The authors thank Dr Chih-Ying Pratt for the preparation works and critical comments to the study.

15 References

1. Asian Pharmacoepidemiology Network (AsPEN). http://aspennet.asia/. Accessed 23

September 2014.

2. Andersen M, Bergman U, Choi NK, et al. The Asian Pharmacoepidemiology Network

(AsPEN): promoting multi-national collaboration for pharmacoepidemiologic research in

Asia. Pharmacoepidemiol Drug Saf 2013;22:700-4.

3. Pratt N, Andersen M, Bergman U, et al. Multi-country rapid adverse drug event

assessment: the Asian Pharmacoepidemiology Network (AsPEN) antipsychotic and acute

hyperglycaemia study. Pharmacoepidemiol Drug Saf 2013;22:915-24.

4. Schaffer D, Florin T, Eagle C, et al. Risk of serious NSAID-related gastrointestinal events

during long-term exposure: a systematic review. Med J Aust 2006;185:501-6.

5. Sostres C, Gargallo CJ, Lanas A. Nonsteroidal anti-inflammatory drugs and upper and

lower gastrointestinal mucosal damage. Arthritis Research & Therapy. 2013;15(Suppl

3):S3.

6. Chaiamnuay S, Allison JJ, Curtis JR. Risks versus benefits of cyclooxygenase-2-selective

nonsteroidal antiinflammatory drugs. Am J Health Syst Pharm 2006;63:1837-51.

7. Chen YF, Jobanputra P, Barton P, et al. Cyclooxygenase-2 selective non-steroidal anti-

inflammatory drugs (etodolac, meloxicam, celecoxib, rofecoxib, etoricoxib, valdecoxib and

lumiracoxib) for osteoarthritis and rheumatoid arthritis: a systematic review and economic

evaluation. Health Technol Assess 2008;12:1-278, iii.

8. Dominick KL, Bosworth HB, Jeffreys AS, Grambow SC, Oddone EZ, Horner RD.

Racial/ethnic variations in non-steroidal anti-inflammatory drug (NSAID) use among

patients with osteoarthritis. Pharmacoepidemiol Drug Saf 2004;13:683-94.

16 9. Fry RB, Ray MN, Cobaugh DJ, et al. Racial/ethnic disparities in patient-reported

nonsteroidal antiinflammatory drug (NSAID) risk awareness, patient-doctor NSAID risk

communication, and NSAID risk behavior. Arthritis Rheum 2007;57:1539-45.

10. Curtis LH, Weiner MG, Boudreau DM, et al. Design considerations, architecture, and use

of the Mini-Sentinel distributed data system. Pharmacoepidemiol Drug Saf 2012;21 Suppl

1:23-31.

11. Carnahan RM, Moores KG. Mini-Sentinel's systematic reviews of validated methods for

identifying health outcomes using administrative and claims data: methods and lessons

learned. Pharmacoepidemiol Drug Saf 2012;21 Suppl 1:82-9.

12. Lai E, Man K, Chaiyakunapruk N, et al. Databases in the Asia-Pacific region: the potential

for a distributed network approach. Epidemiology 2015;26:815-20.

13. Morgan S, Hanley G, Cunningham C, Quan H. Ethnic differences in the use of prescription

drugs: a cross-sectional analysis of linked survey and administrative data. Open Med

2011;5:e87-93.

14. Hippisley-Cox J, Coupland C. Predicting risk of upper gastrointestinal bleed and

intracranial bleed with anticoagulants: cohort study to derive and validate the QBleed

scores. BMJ 2014;349:g4606.

15. Bae S, Kim N, Kang JM, et al. Incidence and 30-day mortality of peptic ulcer bleeding in

Korea. Eur J Gastroenterol Hepatol 2012;24:675-82.

16. Chang CH, Chen HC, Lin JW, Kuo CW, Shau WY, Lai MS. Risk of hospitalization for

upper gastrointestinal adverse events associated with nonsteroidal anti-inflammatory drugs:

a nationwide case-crossover study in Taiwan. Pharmacoepidemiol Drug Saf 2011;20:763-

71.

17 17. Schneeweiss S, Rassen JA, Glynn RJ, Avorn J, Mogun H, Brookhart MA. High-

dimensional propensity score adjustment in studies of treatment effects using health care

claims data. Epidemiology 2009;20:512-22.

18. Rassen JA, Schneeweiss S. Using high-dimensional propensity scores to automate

confounding control in a distributed medical product safety surveillance system.

Pharmacoepidemiol Drug Saf 2012;21 Suppl 1:41-9.

19. Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy

matching techniques. SAS Global Users Group Proceedings 26, Paper 214-26.

http://www2.sas.com/proceedings/sugi26/p214-26.pdf. Accessed 23 September 2014.

20. Curtis LH, Hammill BG, Eisenstein EL, Kramer JM, Anstrom KJ. Using inverse

probability-weighted estimators in comparative effectiveness analyses with observational

databases. Med Care 2007;45:S103-7.

21. Seeger JD, Williams PL, Walker AM. An application of propensity score matching using

claims data. Pharmacoepidemiol Drug Saf 2005;14:465-76.

22. Castellsague J, Riera-Guardia N, Calingaert B, et al. Individual NSAIDs and upper

gastrointestinal complications: a systematic review and meta-analysis of observational

studies (the SOS project). Drug Saf 2012;35:1127-46.

23. Sakamoto C, Sugano K, Ota S, et al. Case-control study on the association of upper

gastrointestinal bleeding and nonsteroidal anti-inflammatory drugs in Japan. Eur J Clin

Pharmacol 2006;62:765-72.

24. Arakawa T, Fujiwara Y, Sollano JD, et al. A questionnaire-based survey on the prescription

of non-steroidal anti-inflammatory drugs by physicians in East Asian countries in 2007.

Digestion 2009;79:177-85.

18 25. Yamakawa N, Suemasu S, Kimoto A, et al. Low direct cytotoxicity of loxoprofen on

gastric mucosal cells. Biol Pharm Bull 2010;33:398-403.

26. Halen PK, Murumkar PR, Giridhar R, Yadav MR. Prodrug designing of NSAIDs. Mini

Med Rev Chem 2009;9:124-39.

27. Yanagawa A, Fukumura T, Matsui H, et al. Possible mechanisms of gastroduodenal

mucosal damage in volunteers treated with nonsteroidal antiinflammatory drugs--the

usefulness of prodrugs. J Rheumatol 1992;19:1075-82.

28. Choi NK, Kim Y, Lee SM, Park BJ. Nonsteroidal anti-inflammatory drugs utilization

patterns among the elderly with osteoarthritis at primary ambulatory care units in Busan

Metropolitan City, Korea. J Prev Med Public Health 2004;37:150-6.

29. Lane AZ, Holmes EL, Moyer CE. Gastrointestinal bleeding in normal subjects: a

comparison of aspirin and mefenamic acid. J New Drugs 1964;4:333-6.

30. Grant DJ, MacConnachie AM. Non-steroidal anti-inflammatory drugs in elderly people.

Mefenamic acid is more dangerous than most. BMJ 1995;311:392.

31. Mallen SR, Essex MN, Zhang R. Gastrointestinal tolerability of NSAIDs in elderly

patients: a pooled analysis of 21 randomized clinical trials with celecoxib and nonselective

NSAIDs. Curr Med Res Opin 2011;27:1359-66.

32. Zeidan AZ, Al Sayed B, Bargaoui N, et al. A review of the efficacy, safety, and cost-

effectiveness of COX-2 inhibitors for Africa and the Middle East region. Pain Pract

2013;13:316-31.

33. Zhao Y, Encinosa W. Hospitalization for gastrointestinal bleeding in 1998 and 2006.

Healthcare Cost and Utilization Project (HCUP) Statistical Brief #65. Agency for

19 Healthcare Research and Quality; Rockville, MD; 2008. http://www.hcup-

us.ahrq.gov/reports/statbriefs/sb65.pdf. Accessed 23 September 2014.

34. Rusanov A, Weiskopf NG, Wang S, Weng C. Hidden in plain sight: bias towards sick

patients when sampling patients with sufficient electronic health record data for research.

BMC Med Inform Decis Mak 2014;14:51.

20 Table 1. Baseline Characteristics of the Study Population Mefenamic Characteristic Celecoxib Diclofenac Loxoprofen Acid Australia No. of patients 14,604 29,409 — — Age, No. (%) 65-84 y 11,635 (79.7) 20,848 (70.9) — — ≥ 85 y 2969 (20.3) 8561 (29.1) — — Men, No. (%) 7792 (53.4) 16,459 (56.0) — — Index year, No. (%) 2001 2661 (18.2) 3159 (10.7) — — 2002 3435 (23.5) 5203 (17.7) — — 2003 2334 (16.0) 4102 (13.9) — — 2004 1709 (11.7) 3639 (12.4) — — 2005 1627 (11.1) 3961 (13.5) — — 2006 805 (5.5) 2279 (7.7) — — 2007 590 (4.0) 1556 (5.3) — — 2008 525 (3.6) 1484 (5.0) — — 2009 372 (2.5) 1214 (4.1) — — 2010 266 (1.8) 1091 (3.7) — — 2011 186 (1.3) 866 (2.9) — — 2012 94 (0.6) 855 (2.9) — —

Hong Kong No. of patients — 231 — 15 Age, No. (%) 18-39 y — 10 (4.3) — 0 60-64 y — 80 (34.6) — 7 (46.7) 65-84 y — 115 (49.8) — 8 (53.3) ≥ 85 y — 24 (10.4) — 0 Men, No. (%) — 104 (45.0) — 4 (26.7) Index year, No. (%) 2008 — 53 (22.9) — 2 (13.3) 2009 — 40 (17.3) — 3 (20.0) 2010 — 53 (22.9) — 5 (33.3) 2011 — 42 (18.2) — 2 (13.3) 2012 — 29 (12.6) — 0

Japan No. of patients 571 1419 7665 224 Age, No. (%) 18-39 y 51 (8.9) 212 (14.9) 1232 (16.1) 36 (16.1) 60-64 y 391 (68.5) 1056 (74.4) 5513 (71.9) 164 (73.2) 65-84 y 111 (19.4) 126 (8.9) 776 (10.1) 21 (9.4) ≥ 85 y 0 4 (0.3) 13 (0.2) 1 (0.4) Men, No. (%) 239 (41.9) 858 (60.5) 4324 (56.4) 130 (58.0) Index year, No. (%) 2005 9 (1.6) 60 (4.2) 193 (2.5) 6 (2.7) 2006 32 (5.6) 151 (10.6) 690 (9.0) 24 (10.7) 2007 22 (3.9) 101 (7.1) 474 (6.2) 12 (5.4) 2008 54 (9.5) 217 (15.3) 999 (13.0) 27 (12.1) 2009 181 (31.7) 389 (27.4) 2044 (26.7) 49 (21.9) 2010 273 (47.8) 501 (35.3) 3265 (42.6) 106 (47.3)

21 Mefenamic Characteristic Celecoxib Diclofenac Loxoprofen Acid

Korea No. of patients 14,414 152,897 68,281 28,149 Age, No. (%) 65-84 y 13,394 (92.9) 143,222 (93.7) 64,771 (94.9) 26,663 (94.7) ≥ 85 y 1020 (7.1) 9675 (6.3) 3510 (5.1) 1486 (5.3) Men, No. (%) 2979 (20.7) 52,619 (34.4) 27,700 (40.6) 11,552 (4.1) Index year, No. (%) 2005 7634 (53.0) 72,879 (47.7) 29,449 (43.1) 12,083 (42.9) 2006 6780 (47.0) 80,018 (52.3) 38,832 (56.9) 16,066 (57.1)

Taiwan No. of patients 16,072 49,987 — 4433 Age, No. (%) 0-18 y 2 (< 0.1) 184 (0.4) — 6 (0.1) 18-39 y 97 (0.6) 2778 (5.6) — 186 (4.2) 40-64 y 3733 (23.2) 26,497 (53.0) — 2060 (46.52) 65-84 y 10,853 (67.5) 17,986 (36.0) — 1903 (42.9) ≥ 85 y 916 (5.7) 1303 (2.6) — 182 (4.1) Men, No. (%) 6056 (37.7) 25,640 (51.3) — 2449 (55.2) Index year, No. (%) 2002 3691 (23.0) 9449 (18.9) — 403 (9.1) 2003 6083 (37.8) 16,203 (32.4) — 811 (18.3) 2004 3603 (22.4) 11,458 (22.9) — 870 (19.6) 2005 1466 (9.1) 6205 (12.4) — 774 (17.5) 2006 691 (4.3) 3295 (6.6) — 569 (12.8) 2007 354 (2.2) 2093 (4.2) — 505 (11.4) 2008 184 (1.1) 1284 (2.6) — 501 (11.3)

22 Table 2. Rates of Gastrointestinal Events Among Patients Initiating a Nonsteroidal Anti- inflammatory Drug Gastrointestinal Events, Country No. (Rate per 1000 Person-Years) Mefenamic Celecoxib Diclofenac Loxoprofen Acid Australia 62 (3.2) 95 (15.1) — — Hong Kong — 31 (479.9) — 1 (663.6) Japan 2 (15.0) 4 (20.2) 27 (28.8) 0 Korea 22 (10.4) 235 (16.8) 40 (5.6) 6 (2.0) Taiwan 536 (38.1) 748 (30.0) — 17 (28.2)

23 Table 3. Cox Proportional Hazards Models of Associations Between Nonsteroidal Anti- inflammatory Use and Gastrointestinal Events Requiring Hospital Admission Compared With Diclofenac Model Celecoxib Loxoprofen Mefenamic Acid Point Estimate Point Estimate Point Estimate (95% CI) (95% CI) (95% CI) Japan Crude hazard ratio 1.12 (0.20-6.17) 1.32 (0.45-3.88) —b Adjusted hazard ratio 2.59 (0.46-14.66) 1.67 (0.56-4.92) —b hdPS-matched hazard ratio —a 1.65 (0.47-5.78) —b Intention-to-treat hazard ratio 1.78 (0.32-10.09) 1.37 (0.47-4.04) —b Taiwan Crude hazard ratio 1.54 (1.38-1.73) —b 0.57 (0.35-0.92) Adjusted hazard ratio 1.08 (0.96-1.22) —b 0.54 (0.33-0.88) hdPS-matched hazard ratio 0.99 (0.86-1.14) —b 0.45 (0.26-0.78) Intention-to-treat hazard ratio 1.13 (1.00-1.27) —b 0.48 (0.29-0.77) Korea Crude hazard ratio 0.58 (0.37-0.90) 0.30 (0.22-0.42) 0.11 (0.05-0.25) Adjusted hazard ratio 0.55 (0.35-0.85) 0.35 (0.25-0.49) 0.13 (0.06-0.29) hdPS-matched hazard ratio 0.43 (0.24-0.76) 0.37 (0.25-0.54) 0.11 (0.05-0.27) Intention-to-treat hazard ratio 0.59 (0.38-0.92) 0.31 (0.22-0.43) 0.11 (0.05-0.25) Hong Kong Crude hazard ratio —b —b 1.26 (0.17-9.54) Adjusted hazard ratio —b —b 2.51 (0.29-21.48) hdPS-matched hazard ratio —b —b 2.16 (0.28-16.87) Intention-to-treat hazard ratio —b —b 1.62 (0.08-31.89) Australia Crude hazard ratio 0.44 (0.31-0.63) —b —b Adjusted hazard ratio 0.44 (0.30-0.62) —b —b hdPS-matched hazard ratio 0.41 (0.21-0.65) —b —b Intention-to-treat hazard ratio 0.99 (0.21-3.74) —b —b Abbreviation: hdPS, high-dimensional propensity score. a Not applicable because of no event. b Not applicable because of no exposure. c The results did not included data from Taiwan and Australia because of considerations of confounding by indication and possible interactions by race, respectively.