Recent Considerations in NonsteroidalAnti-Inflammatory Drug Gastropathy

Gurkirpal Singh, MD, Palo Alto, California

Conservative calculations estimate that approxi-mately 107,000 patients are hospitalized annually fornonsteroidal anti-inflammatory drug (NSAID)-relatedgastrointestinal (GI) complications and at least16,500 NSAID-related deaths occur each year amongarthritis patients alone. The figures for all NSAIDusers would be overwhelming, yet the scope of thisproblem is generally under-appreciated. The Arthri-tis, Rheumatism, and Aging Medical Information Sys-tem (ARAMIS) Post-Marketing Surveillance Program(PMS) has prospectively followed patient status andoutcomes, drug side effects, and the economic im-pact of illness for >11,000 arthritis patients at 8 par-ticipating institutions in the United States and Can-ada. Analysis of these data indicates that: (1) osteo-arthritis (OA) and rheumatoid arthritis (RA) patientsare 2.5–5.5 times more likely than the general popu-lation to be hospitalized for NSAID-related GI events;(2) the absolute risk for serious NSAID-related GItoxicity remains constant and the cumulative riskincreases over time; (3) there are no reliable warningsignals— >80% of patients with serious GI compli-cations had no prior GI symptoms; (4) independentrisk factors for serious GI events were age, pred-nisone use, NSAID dose, disability level, and previousNSAID-induced GI symptoms; and (5) antacids andH2 antagonists do not prevent NSAID-induced gas-tric ulcers, and high-risk NSAID users who takegastro-protective drugs are more likely to have seri-ous GI complications than patients not taking suchmedications. Currently, limiting NSAID use is the onlyway to decrease the risk of NSAID-related GI events.Ongoing ARAMIS research is aimed at developing asimple point-score system for estimating individualrisks of developing serious NSAID-related GIcomplications. Am J Med. 1998;105(1B):31S–38S.© 1998 by Excerpta Medica, Inc.

Gastrointestinal (GI) symptoms are the most com-mon adverse events associated with nonsteroidalanti-inflammatory drug (NSAID) therapy. Be-

cause patients with arthritis are frequent users of NSAIDsand often have other risk factors, this patient populationis particularly at risk for serious GI complications. Tobetter characterize NSAID-related GI complications inpatients with rheumatic disease and to determine meth-ods for reducing their frequency, a series of studies wasbegun several years ago by the Arthritis, Rheumatism,and Aging Medical Information System (ARAMIS).

ARAMIS, which has been funded by the National In-stitutes of Health for the past 25 years, is a prospectiveobservational data bank system that systematically col-lects data on individuals with chronic rheumatic disease.1

Data are reported by both physicians and patients; thedatabase currently contains information on .36,000rheumatic disease patients with .300,000 patient-yearsof follow-up from 17 centers in the United States andCanada.2 Within this larger population are .11,000 ar-thritis patients that comprise the patient population ofthe ARAMIS Post-Marketing Surveillance Program(PMS; Table 1). This program prospectively follows pa-tient status and outcomes, drug side effects, and the eco-nomic impact of illness for consecutively diagnosed pa-tients at each participating institution.

Patient characteristics and study design have been de-scribed in detail elsewhere.3–5 Patients in the ARAMIS-PMS program come from a wide variety of sources, in-cluding academic institutions and urban and communitypopulations, and are generally representative of rheuma-tology patients in North America.1

Diagnosis of rheumatoid arthritis (RA) and osteoar-thritis (OA) is made and documented according toAmerican College of Rheumatology (ACR) criteria. Ad-ditionally, a random 10% subgroup of the Santa ClaraCounty community sample was examined to confirm thediagnosis.

Enrolled patients complete the Stanford Health Assess-ment Questionnaire (HAQ) every 6 months. The Stan-ford HAQ is a validated instrument that evaluates thepatient’s functional ability and quality of life, use ofhealthcare resources and medication, costs of care, andside effects.6,7 The study protocol includes following-upwith nonresponders, contacting patients to fill in missinginformation, and the quality control of questionnaire

From the Department of Medicine, Division of Immunology and Rheu-matology, and the ARAMIS Postmarketing Surveillance Program, Stan-ford University of Medicine, Palo Alto, California.

From a presentation at the National Institute of Diabetes and Diges-tive and Kidney Diseases (NIDDK) conference: Non-Narcotic Analge-sics: Renal and GI Considerations, held in Bethesda, Maryland on April28 –29, 1997.

Requests for reprints should be addressed to Gurkirpal Singh, MD,Stanford University School of Medicine, Department of Medicine, Di-vision of Immunology and Rheumatology, 1000 Welch Road, Suite 203,Palo Alto, California 94303.

© 1998 by Excerpta Medica, Inc. 0002-9343/98/$19.00 31SAll rights reserved. PII S0002-9343(98)00072-2

coding and data entry to ensure accuracy. Follow-up dataare available for 95.4% of patients.

The extensive data available in ARAMIS has providedan unparalleled opportunity to consider some of the out-standing issues relating to NSAID-induced adverseevents: (1) what are the GI side effects of NSAID use; (2)what is the extent of the problem; (3) are there warningsigns for serious GI complications; (4) who is at greatestrisk of serious GI complications; (5) do some NSAIDscause more GI toxicity than others; and (6) do H2 antag-onists and antacids help prevent serious GI complica-tions?

This article will discuss some of those issues using re-sults from the literature and updated findings from theARAMIS studies and will outline some areas of currentand potential future research.

WHAT ARE THE GI SIDE EFFECTS OFNSAID USE?

NSAID-related GI side effects result from inhibition ofprostaglandins, which play a key role in maintaining theintegrity of the gastric mucosa. Although NSAID use isassociated primarily with upper GI events and gastric ul-cers, lower GI problems can also occur, including (1) in-flammation and permeability changes of the intestineand lower bowel; (2) hemorrhage of the ileum, duode-num, and colon; (3) perforation; and (4) stricture forma-tion.8 –11 The adverse GI events associated with NSAIDuse can be broadly categorized into 3 groups: (1) “nui-sance symptoms” (heartburn, nausea, dyspepsia, vomit-ing, abdominal pain); (2) mucosal lesions as seen on en-doscopy or x-rays; and (3) serious GI complications (per-forated ulcers, catastrophic bleeding that requireshospitalization).

In any study of NSAID-induced GI complications, it isimportant to distinguish between the nuisance symp-toms or endoscopic lesions and the serious GI complica-tions. For this reason, endpoints for all ARAMIS studieshave been defined as clinically significant events or le-sions serious enough to result in a hospital stay of at least24 hours. They include GI bleeding, clinically symptom-atic gastritis, ulcers, or gastric outlet obstructions.1 A def-

inite diagnosis was not always possible despite severe GIsymptoms requiring hospitalization. Gastrointestinalevents were determined from the self-administered HAQand confirmed by review of hospital records.

WHAT IS THE MAGNITUDE OF THEPROBLEM?Endoscopic lesions are common and may occur in up to80% of NSAID users.12 Approximately 15% of patientswill have the nuisance symptoms. Using a liberal defini-tion of an ulcer as a break of $3 mm in the gastric mu-cosa, about 25% of patients will have a gastric ulcer, and40 –50% will have either a gastric or duodenal ulcer.However, only 1– 4% of patients will have clinically im-portant GI complications each year.10,13,14 –17 Further-more, reducing the incidence of endoscopic ulcers (e.g.,with misoprostol) does not result in a comparable reduc-tion in clinically significant GI complications.18 Thus,there is little correlation between the occurrence of endo-scopic ulcers and clinically significant events. For this rea-son, the ARAMIS studies reported here have limited thedefinition of GI adverse events to those requiring hospi-talization.

Numerous studies have attempted to correlate NSAIDuse with serious GI complications by determining theratio of NSAID-related events to non–NSAID-relatedevents. Langman et al (1994)19 found an odds ratio of 4.5for NSAID-related peptic ulcer bleeding; Garcia-Rodri-guez and Jick (1994)20 reported a risk ratio of 3.9 forhospitalizations; and a meta-analysis of 40 studies pub-lished over 25 years found an overall odds ratio of 2.74 forserious GI complications and 7.75 for GI surgery.16 Thegoal of current ARAMIS research is to better correlate GIevents with NSAID use by determining adjusted odds ra-tios for specific causes of GI adverse events in patientstaking NSAIDs versus control patients not takingNSAIDs.21

Incidence of GI Complications:ARAMIS Epidemiologic StudyIn this study, GI complications were defined as hospital-izations and deaths due to GI events. Complications wereconsidered to be NSAID-related if the HAQ indicatedNSAID use immediately before the event or if NSAIDswere noted on the hospital discharge summary. The pa-tient population consisted of .4,000 arthritis patientswith about 15,500 patient-years of observation (Table 2).

Hospitalizations. Of the hospitalizations for GI eventsin the RA patient group, 92.5% (124/134) were related toNSAID use, with an annual rate of 1.46% versus 0.27%for patients not taking NSAIDs. The relative risk of hos-pitalization for GI events for NSAID users was 5.49, onlyslightly less than the relative risk of lung cancer for smok-ers (7– 8). The annual hospitalization rate for NSAID-related complications was significantly lower for OA pa-

Table 1. Patients with Rheumatoid Arthritis and Osteoarthri-tis in the Arthritis, Rheumatism, and Aging Medical Informa-tion System (ARAMIS) Post-Marketing Surveillance Program

Rheumatoid arthritis 6,673 patients with 51,241 patient-years of follow-up

Osteoarthritis 4,914 patients with 15,221 patient-years of follow-up

Centers 8*

* Stanford, CA; Santa Clara County, CA; Wichita, KS: Phoenix, AZ;Cincinnati, OH; Baltimore, MD; Saskatoon, Saskatchewan, Canada;Montreal, Quebec, Canada.

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

32S July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B)

tients than for RA patients and the relative risk about halfthat of RA patients. A likely explanation for the lower riskis that OA is a milder disease for which NSAID doses aregenerally lower, OA patients have fewer comorbid condi-tions, and prednisone (known to double the risk forNSAID-related GI events) is rarely used.

The Arthritis Foundation conservatively estimates thatat least 13 million individuals in the United States withOA or RA regularly take NSAIDs (Table 3). Applying theARAMIS data to these figures, the number of potentialhospitalizations for serious GI complications is about107,000 per year. These data correspond well to the fig-ures obtained from the Tennessee Medicaid database.22

At a conservative estimated cost of $10,000 –$15,000 perhospitalization, the total annual cost exceeds $1 bil-lion—a huge drain on national medical resources.

Deaths. Overall, approximately 10% of hospitalizationsfor upper GI bleeding result in death, and 80% of all ul-

cer-related deaths occurred in patients using an anti-in-flammatory agent.23 In the ARAMIS study, 26 deaths (allin RA patients) resulted from GI complications observedin the 12,224 patient-years of exposure to NSAIDs. Ofthese, 19 could be definitely attributed to NSAID use, fora raw annual GI death rate of 0.22% and a relative risk of4.21. Although an incidence of 0.22% per year may soundtrivial, over the 20 –30 years that chronic arthritis patientsmay take NSAIDs, the incremental risk per patient andthe total number of deaths add up very quickly.

Conservative calculations based on ARAMIS data esti-mate that the annual number of NSAID-related deaths(16,500) among patients with definite or probable RA orOA is comparable to the number of deaths from othercommon causes (Table 3; Figure 1).24 NSAIDs causemore than one third as many deaths as human immuno-deficiency virus (HIV) infection and almost as manydeaths as asthma, cervical cancer, and malignant mela-noma combined. If the estimates were to include all pa-tients who take NSAIDs, the numbers would be over-whelming, yet the scope of this problem is generallyunderappreciated.

ARE THERE WARNING SIGNS FORSERIOUS GI COMPLICATIONS?Dyspepsia is a common side effect of NSAID use, but ispoorly correlated with endoscopic lesions or GI bleed-ing.9,10,25 Of patients who develop ulcers or life-threaten-ing GI complications, 50 – 60% will have had no previouswarning signs or symptoms.10,23,26,27

ARAMIS conducted a cohort study1 involving 1,921patients on NSAIDs during their first 2.5 years of obser-vation. Documented events included GI hospitalizations,upper GI endoscopies, NSAID-induced GI side effectsnot serious enough to warrant hospitalization (nausea,vomiting, upper or lower abdominal pain, heartburn, di-arrhea, or blood in stools), and use of GI medications (H2

antagonists, sucralfate, or antacids). Patients who tookmisoprostol were excluded from analysis for 2 reasons:(1) the patient numbers were too small to allow indepen-dent analysis; and (2) the mechanism of action differs

Table 2. Gastrointestinal (GI) Complications in Osteoarthri-tis (OA) and Rheumatoid Arthritis (RA)

ParametersOA

HospitalizationsRA

Hospitalizations

Number of patients 1,283 2,921Person-years of

observation3,234 12,224

Person-years takingNSAIDs

2,199 8,471

Number of GI events 19 134GI events in NSAID

users16 124

Annual incidence onNSAIDs (%)

2.51 5.49

Relative risk on NSAIDs 0.73 (SE 5 0.18)* 1.46 (SE 5 0.13)*Annual incidence of

upper GI events onNSAIDs (%)

0.50 1.27

Annual incidence oflower GI events onNSAIDs (%)

0.23 0.19

* P , 0.001 for RA vs OA.

Table 3. Estimated Annual Nonsteroidal Anti-inflammatory Drug (NSAID)-Associated Gastrointestinal (GI) Hospitalizationsand Deaths in the United States

Diagnosis

NSAIDExposure

(n)

Hospitalizations Deaths

Incidence(%)

Patients(n)

Incidence(%)

Patients(n)

RA 2,000,000 1.5 30,000 0.22 4,400Probable RA* 3,000,000 0.7† 21,000 0.11‡ 3,300OA 8,000,000 0.7 56,000 0.11‡ 8,800Total 13,000,000 107,000 16,500

* Estimated in the community, not under a rheumatologist’s care.† Estimated to be half; presumed milder disease as they are not under a rheumatologist’s care.‡ Estimated from ratio of GI hospitalizations.

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B) 33S

from that of the other GI medications. During the 2.5-year observation period, the overall incidence of NSAID-related GI side effects was 15%, with a 2.2% incidence ofserious GI complications requiring hospitalization. Pa-tients who did have GI symptoms were slightly morelikely to have a serious GI complication than those whohad no warning symptoms (2.7% vs 2.0%), but the con-verse was not true: a staggering 81% (34/42) of patientswho had serious GI complications had no prior GI symp-toms.

WHO IS AT RISK FOR SERIOUS GICOMPLICATIONS?

Since GI side effects are not a reliable warning sign, it isimportant to define risk factors for serious GI complica-tions and to find ways to reduce that risk. A number ofstudies have addressed this issue with fairly consistentresults, identifying as risk factors advanced age,16,22,28 –30

higher NSAID dosages,18,19,22,31 shorter duration of ther-apy,16,18,19 a history of GI problems,18,20,31 and concom-itant corticosteroid20,32,33 or anticoagulant20,34 use. Mostof these studies performed only univariate analyses.

ARAMIS evaluated risk factors for 114 RA patientswho had GI events versus 1,921 RA patients who did nothave GI events. Univariate analysis identified 11 variablesassociated with hospitalization or death at the predictivevisit (before the event): age; disease duration; disabilityindex (0 –3); NSAID dose (% of maximum); number ofcategories of nonrheumatic drugs taken (0 – 8); NSAIDGI side effects (ever); use of prednisone, antacids, or GIprotective drugs; and comorbid conditions (ever).

Of note in this cohort study is the fact that female sexwas not associated with an increased risk, whereas in case-

control studies that is often the case, possibly due to thepredominance of middle-aged women who are takingNSAIDs. Other categorical variables that univariate anal-ysis did not correlate with an increased risk includedwhite race, smoking (ever), and alcohol (ever). Continu-ous variables that were not significantly associated withan increased number of GI events included educationallevel, smoking (packs/day), alcohol (drinks/day), num-ber of concurrent disease-modifying antirheumaticdrugs (DMARDs), and 0 – 6 categories of comorbid con-ditions (ever).

A multivariate model consisting of 5 variables—age,prednisone use, NSAID dose (% of maximum), disabilityindex, and previous NSAID-induced GI side effect—wasidentified by stepwise logistic regression analysis(Table 4). The odds ratios (OR) for these variables can beused to estimate a patient’s risk of hospitalization for thefollowing year.

Although “advanced age” is usually defined as $65years, this is an arbitrary point. It is important to remem-ber that individuals do not suddenly become “at risk” atthe age of 65. Ongoing ARAMIS studies show that the riskincreases steadily by roughly 4% per year increase in age.

The question of whether the risk for NSAID-relatedgastropathy is constant over time has been controversial.One school of thought—the mucosal adaptation theo-ry—is that patients who are going to bleed will do so earlyin therapy and will discontinue taking NSAIDs or die, sothat those who continue are “tolerant,” with a reducedrisk of serious GI complications. The US Food and DrugAdministration (FDA), in the process of deciding on re-vised labeling for NSAIDs, requested an ARAMIS studyto look at this issue. The study followed approximately

Figure 1. Death rate per 100,000 and number of deaths associated with nonsteroidal anti-inflammatory drug (NSAID)-inducedgastrointestinal (GI) damage compared with other causes: United States population, 1994.

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

34S July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B)

1,600 new NSAID users or patients who restarted takingNSAIDs after a washout period of at least 6 months. Theprecise date and time of bleeding was obtained from hos-pital discharge summaries. A Kaplan-Meier survival anal-ysis was performed to determine a continuous approxi-mation of the hazard function, and the results indicatethat throughout most of the 10-year follow-up period,the risk of GI bleed (the hazard function) remained con-stant. Patients on NSAIDs for 5 years have a 5-timesgreater risk of bleeding than do patients on NSAIDs for 1year, whereas the risk after 3 months of NSAID therapy isapproximately one-fourth the risk after 1 year. Thus,from the data, it appears that the body does not adapt tothe insult that occurs with the use of NSAIDs.

The single most important factor that predicts GIbleeding is the duration of NSAID therapy. This factorwas not included in the ARAMIS study because longitu-dinal data (e.g., duration of NSAIDs therapy) is not han-dled well by logistic regression methods. However, ongo-ing ARAMIS research is developing Cox proportionalhazard models based on life-table analysis to incorporatethis risk factor. The ultimate goal of this research is todevelop point-based risk-factor algorithms—a simpleone for patients themselves and a more complex and ac-curate one for physicians—to estimate individual risksfrom NSAID therapy.

ARE SOME NSAIDs MORE TOXICTHAN OTHERS?All NSAIDs cause a full range of GI side effects, althoughthey vary in frequency and severity.10,13,19,20,35– 40 Someof the ARAMIS studies have previously found substantialvariations in the relative overall toxicity of NSAIDs usingthe Toxicity Index (TI).41– 45 A recent study utilized asubset of the TI to evaluate GI toxicity for 6,276 courses of12 NSAIDs during 9,860 patient-years of observation.The GI-TI is a sum of GI symptoms per patient-year ofexposure, weighted by severity and number of hospitaldays and adjusted for risk factors and other key variables(e.g., age, sex, center, drug initiation, disease duration,disability, educational level, duration of NSAID therapy,concurrent medications). Gastrointestinal toxicity, ex-pressed as the GI-TI, varied from a low of ,1 for salsalateto almost 4 for meclofenamate (Table 5). The differencesin GI toxicity only become clinically or statistically signif-icant between the NSAIDs at the upper and lower end ofthe range. The majority of the NSAIDs are in the middlerange from 1.68 –2.0.

As a result of the findings discussed above, the FDA hasrequested that NSAID manufacturers revise their labelingto include the following warnings: (1) serious GI toxicitycan occur at any time, with or without warning symp-toms; (2) the absolute risk remains constant and the cu-mulative probability increases over time, thus increasingthe likelihood of developing a serious GI event with long-

term NSAID use; and (3) the lowest effective dose shouldbe used for the shortest possible duration to minimize thepotential risk, and alternative therapies should be consid-ered for high-risk patients.

DO H2 ANTAGONISTS AND ANTACIDSHELP PREVENT SERIOUS GICOMPLICATIONS?

There are significant differences between the ulcerscaused by Heliobacter pylori and those caused by NSAIDuse. The H. pylori ulcer is a predominantly duodenal ul-cer, associated with low pH, and generally symptomatic.NSAID-induced ulcers, on the other hand, are primarilygastric ulcers, associated with high pH, and usuallyasymptomatic. Acid suppression (e.g., with ranitidine) isvery effective in preventing duodenal ulcers but has littleeffect on gastric ulcers (Figure 2).12,46,47 Similarly, protonpump inhibitors reduce the incidence of duodenal butnot gastric ulcers. Despite this well-established fact, ap-

Table 4. Significant Risk Factors for Hospitalization or Deathin Rheumatoid Arthritis Patients Due to NSAID-Induced Gas-trointestinal Events: Stepwise Logistic Regression

VariableCoefficient

(SE)OddsRatio 95% CI*

Age 0.037 (0.01) 1.04 1.02–1.06Prednisone use 0.587 (0.20) 1.80 1.21–2.66NSAID dose (% of

maximum)0.372 (0.17) 1.45 1.04–2.02

Disability index (0–3) 0.287 (0.18) 1.33 1.03–1.72Previous NSAID GI

side effect0.462 (0.22) 1.59 1.03–2.44

* Confidence interval (CI) for odds ratio (OR); significant when the CIdoes not include 1.00.

Table 5. Relative Gastrointestinal Toxicity of NonsteroidalAnti-inflammatory Drugs (NSAIDs) in Rheumatoid ArthritisPatients: Arthritis, Rheumatism, and Aging Medical Informa-tion System (ARAMIS) Studies

Agent nGI Toxicity Index

(mean 6 SE)

Salsalate 187 0.81 6 0.51Ibuprofen 577 1.13 6 0.29Aspirin 1521 1.18 6 0.18Sulindac 562 1.68 6 0.29Diclofenac 415 1.81 6 0.35Naproxen 1062 1.91 6 0.21Tolmetin 243 2.02 6 0.44Piroxicam 814 2.03 6 0.24Fenoprofen 158 2.35 6 0.55Indomethacin 418 2.39 6 0.34Ketoprofen 259 2.65 6 0.43Meclofenamate 165 3.91 6 0.54

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B) 35S

proximately 34% of the 1,921 individuals in the ARAMISstudy who took NSAIDs were taking concomitant acid-reduction therapy. Of these 657 patients, 58.3% were tak-ing antacids, 23.0% were taking cimetidine, 6.9% weretaking ranitidine, 2.7% were taking sucralfate, and 2.2%were taking famotidine, while 8.5% were taking combi-nations of different GI medications.1 Review of theARAMIS database showed that 75% of GI medicationsco-prescribed with NSAIDS were for prophylaxis, andonly 25% were for treatment. This is a dangerous prac-tice, because these medications may suppress symp-toms—potential warning signs—without reducing therisk of serious GI complications. They may provide a falsesense of security to both physician and patient, encour-aging long-term therapy with higher NSAID doses, whichcould eventually result in serious GI events. Indeed, theARAMIS cohort study1 showed that, of patients who hadnever had any GI side effects, those who were taking pro-phylactic GI medications had about 2.5 times more hos-

pitalizations for NSAID-related GI complications thanpatients not taking GI medications (OR 2.69; 95% CI 51.36 –5.31, P ,0.05; Figure 3). For patients who had pre-vious NSAID-related GI symptoms, the incidence of se-rious complications was similar for patients taking GImedications and patients not taking GI medications (OR0.73; 95% CI 5 0.18 –2.96). The odds ratio was not sig-nificantly different (OR 0.57; 95% CI 5 0.13–2.45) afteradjusting for differences in patient characteristics (i.e.,age, sex, duration of RA, disability index, educationallevel, concurrent use of prednisone).

Using different methods (a case-control study designand a newly designed propensity score to measure thelikelihood of having a gastroprotective drug prescribed),Avorn et al (1996)48 reported similar results in a largegroup of New Jersey Medicaid patients (n 5 3,524 withand 14,096 without hospitalizations). High-risk individ-uals taking histamine antagonist therapy, sucralfate, oromeprazole were about twice as likely (OR 1.7–2.1;

Figure 2. Prevention of nonsteroi-dal anti-inflammatory drug (NSAID)-induced gastrointestinal (GI) ulcerswith ranitidine. Adapted with per-mission from BMJ.12

Figure 3. Gastrointestinal (GI)medication use and subsequent seri-ous nonsteroidal anti-inflammatorydrug (NSAID)-induced GI hospital-izations.

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

36S July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B)

P ,0.05 each) to have serious GI bleeding as those nottaking GI medications. The likelihood of hospitalizationin the different groups compared with low-risk patients isshown in Figure 4 (Jerry Avorn, personal communica-tion).

In clinical trials, the newer GI medications (e.g., miso-prostol, omeprazole) have effectively reduced serious GIcomplications. Future ARAMIS studies will investigatetheir ability to similarly reduce serious GI events requir-ing hospitalization in clinical practice.

Currently, the accepted method for decreasing the riskof serious NSAID-related GI toxicity is to limit NSAIDuse. For OA, acetaminophen is the accepted first-linetherapy when nonpharmacologic methods are insuffi-cient; for RA, disease modifying antirheumatic drug(DMARD)-based strategies should be considered. Neweragents (e.g., tramadol and selective COX-2 inhibitors)may also be useful additions to the physician’s armamen-tarium for the treatment of chronic pain.

As patients in the ARAMIS database continue to beevaluated, research is currently underway to develop asimple point-score system based upon risk factors to es-timate individual risks for developing serious NSAID-re-lated GI complications. Once completed, this systemshould help the clinician quickly determine which pa-tients are at risk for NSAID-associated adverse events andallow them to take the appropriate course of action.

REFERENCES1. Singh G, Ramey DR, Morfeld D, et al. Gastrointestinal tract

complications of non-steroidal anti-inflammatory drugtreatment in rheumatoid arthritis—A prospective observa-tional cohort study. Arch Intern Med. 1996;156:1530–1536.

2. Fries JF, McShane DJ. ARAMIS (The American Rheuma-tism Association Medical Information System): a prototyp-ical national chronic-disease data bank. West J Med. 1986;145:798–804.

3. Fries JF, Ramey DR, Singh G, et al. A reevaluation of aspirin

therapy in rheumatoid arthritis. Arch Intern Med. 1993;153:2465–2471.

4. Singh G, Fries JF, Spitz PW, et al. Toxic effects of azathio-prine in rheumatoid arthritis: a national post-marketing per-spective. (Abstr.) Arthritis Rheum. 1989;32:837–843.

5. Singh G, Fries JF, Williams CA, et al. Toxicity profiles ofdisease modifying antirheumatic drugs in rheumatoid ar-thritis. J Rheumatol. 1991;18:188–194.

6. Fries JF, Spitz PW, Kraines RG, Holman HR. Measurementof patient outcome in arthritis. Arthritis Rheum. 1980;23:137–145.

7. Ramey DR, Singh GS, Fries JF. The Health AssessmentQuestionnaire 1995: status and review. In: Spilker B, ed.Pharmacoeconomics and Quality of Life in Clinical Trials.2nd ed. Philadelphia: Lippincott-Raven; 1996:227–237.

8. Buchanan WW, Brooks PM. Prediction of organ systemtoxicity with anti-rheumatic drug therapy. In: Bellamy N, ed.Prognosis in the Rheumatic Diseases. Boston: Kluwer Ac-ademic Publishers; 1991:403–450.

9. Brooks PM, Day RO. Nonsteroidal antiinflammatorydrugs—differences and similarities. N Engl J Med. 1991;324:1716–1725.

10. Hardin JG, Longenecker GL. Handbook of Drug Therapy inRheumatic Disease. Pharmacology and Clinical Aspects.Boston: Little, Brown and Company. 1992.

11. Paulus HE. Nonsteroidal Anti-inflammatory Drugs. In: KellyWN, Harris ED Jr, Ruddy S, Sledge CB, eds. Textbook ofRheumatology. Philadelphia: WB Saunders Company;1989:765–791.

12. Ehsanullah RS, Page MC, Tildesley G, Wood JR. Preven-tion of gastroduodenal damage induced by non-steroidalanti-inflammatory drugs: controlled trial of ranitidine. BMJ.1988;297:1017–1021.

13. Caruso I, Bianchi Porro G. Gastroscopic evaluation of an-tiinflammatory agents. BMJ. 1980;280:75–78.

14. McCarthy DM. Nonsteroidal antiinflammatory drug inducedulcers: management by traditional therapies. Gastroenter-ology. 1989;96:662–674.

15. Gabriel SE, Jaakkimainen L, Bombardier C. The cost-effec-tiveness of misoprostol for nonsteroidal antiinflammatorydrug-associated adverse gastrointestinal events. ArthritisRheum. 1993;36:447–459.

16. Gabriel SE, Jaakkimainen L, Bombardier C. Risk for seriousgastrointestinal complications related to use of nonsteroi-

Figure 4. Increased risk for hospi-talization with gastrointestinal (GI)medications. (J. Avorn, personalcommunication, 1996.)

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B) 37S

dal anti-inflammatory drugs. A meta-analysis. Ann InternMed. 1991;115:787–796.

17. Paulus HE. Government affairs. FDA Arthritis AdvisoryCommittee meeting: risks of agranulocytosis/aplastic ane-mia, flank pain, and adverse gastrointestinal effects withthe use of nonsteroidal antiinflammatory drugs. ArthritisRheum. 1987;30:593–595.

18. Silverstein FE, Graham DY, Senior JR, et al. Misoprostolreduces serious gastrointestinal complications in patientswith rheumatoid arthritis receiving nonsteroidal anti-inflam-matory drugs. A randomized, double-blind, placebo-con-trolled trial. Ann Intern Med. 1995;123:241–249.

19. Langman MJ, Weil J, Wainwright P, et al. Risks of bleedingpeptic ulcer associated with individual non-steroidal anti-inflammatory drugs. Lancet. 1994;343:1075–1078.

20. Garcia Rodriguez LA, Jick H. Risk of upper gastrointestinalbleeding and perforation associated with individual non-steroidal anti-inflammatory drugs. Lancet. 1994;343:769–772.

21. Singh G, Ramey DR, Terry R, Triadafilopoulus G. NSAID-related effects on the GI tract: an ever widening spectrum.(Abstr.) Gastroenterology. 1997;112:A42.

22. Griffin MR, Piper JM, Daugherty JR, et al. Nonsteroidalanti-inflammatory drug use and increased risk for pepticulcer disease in elderly persons. Ann Intern Med. 1991;114:257–263.

23. Armstrong CP, Blower AL. Non-steroidal anti-inflammatorydrugs and life threatening complications of peptic ulcer-ations. Gut. 1987;28:527–532.

24. National Center for Health Statistics. GMWKI: total deathsfor each cause by 5-year age groups, United States, 1994(unpublished data from NCHS website data warehouse).Hyattsville, MD: Public Health Service, 1997.

25. Larkai EN, Smith JL, Lidsky MD. Gastroduodenal mucosaand dyspeptic symptoms in arthritis patients during chronicnonsteroidal anti-inflammatory drug use. Am J Gastroen-terol. 1987;82:1153–1158.

26. Polisson R. Nonsteroidal anti-inflammatory drugs: practicaland theoretical considerations in their selection. Am J Med.1996;100(suppl 2A):31S–36S.

27. Lichtenstein DR, Syngal S, Wolfe MM. Nonsteroidal antiin-flammatory drugs and the gastrointestinal tract. The dou-ble-edged sword. Arthritis Rheum. 1995;38:5–18.

28. Bjorkman DJ. Nonsteroidal anti-inflammatory drug-inducedgastrointestinal injury. Am J Med. 1996;101(suppl 1A):25S–32S.

29. Longstreth GF. Epidemiology of hospitalization for acuteupper gastrointestinal hemorrhage: a population-basedstudy. Am J Gastroenterol. 1995;90:206–210.

30. Greene JM, Winickoff RN. Cost-conscious prescribing ofnonsteroidal anti-inflammatory drugs for adults with arthri-tis. A review and suggestions. Arch Intern Med. 1992;152:1995–2002.

31. Hallas J, Lauritsen J, Villadsen HD, Gram LF. Nonsteroidalanti-inflammatory drugs and upper gastrointestinal bleed-ing, identifying high-risk groups by excess risk estimates.Scand J Gastroenterol. 1995;30:438–444.

32. Hochain P, Berkelmans I, Czernichow P, et al. Which pa-tients taking non-aspirin nonsteroidal anti-inflammatory

drugs bleed? A case-control study. Eur J GastroenterolHepatol. 1995;7:419–426.

33. Piper JM, Ray WA, Daugherty JR, et al. Corticosteroid useand peptic ulcer disease: role of nonsteroidal anti-inflam-matory drugs. Ann Intern Med. 1991;114:735–740.

34. Shorr RI, Ray WA, Daugherty JR, et al. Concurrent use ofnon-steroidal anti-inflammatory drugs and oral anticoagu-lants places elderly persons at high risk for hemorrhagicpeptic ulcer disease. Arch Intern Med. 1993;153:1665–1670.

35. Lanza FL, Royer GL, Nelson RS, et al. A comparative en-doscopic evaluation of damaging effects of nonsteroidalantiinflammatory agents on the gastric and duodenal mu-cosa. Am J Gastroenterol. 1981;75:17–21.

36. Lanza FL, Nelson RS, Greenberg BP. Effects of fenbufen,indomethacin, naproxen and placebo on gastric mucosa ofnormal volunteers. Am J Med. 1983;75:75–79.

37. Lanza FL. Endoscopic studies of gastric and duodenalinjury after the use of ibuprofen, aspirin, and other nonste-roidal antiinflammatory drugs. Am J Med. 1984;77(suppl1a):19–24.

38. Lanza F, Rack MF, Lyn M, et al. An endoscopic comparisonof the effects of etodolac, indomethacin, ibuprofen,naproxen, and placebo on the gastrointestinal mucosa.J Rheumatol. 1987;14:338–341.

39. Lanza F, Rack MF, Doucette M, et al. An endoscopic com-parison of the gastroduodenal injury seen with salsalateand naproxen. J Rheumatol. 1989;16:1570–1574.

40. Carson JL, Strom BL, Morse ML, et al. The relative gastro-intestinal toxicity of the nonsteroidal anti-inflammatorydrugs. Arch Intern Med. 1987;147:1054–1059.

41. Fries JF, Spitz PW, Williams CA, et al. A toxicity index forcomparison of side effects among different drugs. ArthritisRheum. 1990;33:121–130.

42. Fries JF, Williams CA, Block DA. The relative toxicity ofnon-steroidal anti-inflammatory drugs. Arthritis Rheum.1991;34:1353–1360.

43. Fries JF, Williams CA, Ramey D, Bloch DA. The relativetoxicity of disease modifying antirheumatic drugs(DMARDs). Arthritis Rheum. 1993;36:297–306.

44. Singh G, Williams CA, Ramey Dr, Fries JF. A toxicity indexfor comparison of gastrointestinal toxicity of non-steroidalanti-inflammatory drugs (NSAIDs). (Abstr.) Arthritis Rheum.1993;36(suppl 9):S178.

45. Singh G, Ramey Dr, Morfeld D, Fries JF. Comparative tox-icity of non-steroidal anti-inflammatory agents. PharmacolTher. 1994;62:1750–1791.

46. Robinson MG, Griffin JW Jr, Bowers J, et al. Effect ofranitidine on gastroduodenal mucosal damage induced bynonsteroidal anti-inflammatory drugs. Dig Dis Sci. 1989;34:424–428.

47. Oddsson E, Gudjonsson H, Thjodleifsson B. Comparisonbetween ranitidine and omeprazole for protection againstgastroduodenal damage caused by naproxen. Scand JGastroenterol. 1992;27:1045–1048.

48. Avorn J, Solomon D, Levin R, Lo J. Epidemiologic analysisof prophylactic drug use and NSAID gastropathy. (Abstr.)Arthritis Rheum. 1996;39(suppl):S165.

A Symposium: Non-narcotic Analgesics—Renal and GI Considerations/Singh

38S July 27, 1998 THE AMERICAN JOURNAL OF MEDICINEt Volume 105 (1B)