Exatnining the risks for NSAID-- induced gastropathy

The medical literature describing the gastrointestinal risks associated with chronic nonsteroidal anti-inflammatory drug (NSAlD) therapy is increasing rapidly. In spite of this, clinicians remain uncertain about how to translate this information into clinical practice. C linical decisions regarding the management of adverse reactions to medications have two components. The first involve~ risk to the patient, and the secoml an evaluation of available alternatives for managing that risk. Three research designs have been used to examine this association: ecological studies, case contro l studies and retrospective cohort studies. Although these designs do not provide the strongest evidence for causation, their results point toward the existence of a risk of serious gastrointestinal reactions of between 1.5 and 10 times greater for NSAID users than for nonusers. The wide variation in results is due to multiple factors, including different research designs, study populations and outcome measures. Several subgroups have been suggested to be at particularly high risk. Risk factors include advanced age, female sex, debilitating rheumato id disease, previous gastro intestinal disease, ethanol abuse and smoking. Only in the elderly has there been adequate data to support this associa tion. Although the data regarding females is compelling, this may be confounded by the documented increased NSAlD use among females. Decision analysis is a useful tool for the quantitative examination of the costs and benefits of management alternatives available for dealing with these risks. Research is needed to identify accurately subgroups ofNSAID users at high risk. CanJ Gastroenterol 1990;4(3):108-112

Examen des risques de la gastropathie provoquee par les AINS RESUME: La litterature medicate fait de plus en plus frequemment etat des risques gastro-intestinaux associes aux therapies chroniques (anti-inflammatoires non stero'idiens).Les cliniciens ne savent neanmo ins toujours pas comment traduire ces donnees clans leur pratique.Les decisions cliniques relatives au meJical literature describing gastrointestinal ri sks associated wi t h ch ronic nonsteroidal anti-inflammatory drug (NSA!D) therapy, defin itive recommendations for cl inical practice c1re lc1cking.In order to rnc1 ke sound clinica l decisions regardi ng the management of NSA ID-induced gastrointesti nal effects, the risk to t he patient must fi rst be defi ned; alternatives fo r reducing that risk should then be examined (Figure 1 ).This controversy will be reviewed using t hese two issues as guidelines.disease and its complications.Because both NSAID use anJ the development of ulcers were measured in the population rather than in individuals, it is poss ible that the NSA ID users we re not the ones who developed ulcers or ulcer complications.Ecological studies, therefore, cannot be regarded as strong evidence for cause and effect.Such studies, however, represent an important step in developing the hypothesis of causation.

Rheumauc Disease Unit and Clinical Epidemiology
Two ocher research designs have 110 been useJ to examine the association between NSA!Ds and gastropathy: retrospective cohort and case-control studies ; Figure 2 illustrates the differences between these designs.The Jesign of prospective cohort studies is inc luded for comparison.In a cohort stu<ly, a group of people ( the cohort), al I of whom are ulcer-free, is followed over time to determine which members (NSAID users or nonusers) <levelop ulcers or associated compl icacions.In retrospective cohort studies, NSAID use is identified from pasr records con mined in healLh insurance registries.S uch data arc colleCLed for patient care purpose~ and therefore may 1101 he of sufficient qu::i li1y for rigorous research.Alternatively, in prospective cohort stud ies, the data are collected specifically for the purpose of 1 he srudy and hence many biases are avoide<l, resulting in improved stu<lies examining rhe risk of NSA ID-induced gast rop;ithy.
Case control stu<lies retrospectively compare the frequency of NSAID use in people with and without gascropathy.lf patients with gastropathy were found more likely to be NSAlD users this would constitute some evidence for causation.Case control studies are susceptible ro many more biases than cohort studies.For example, patients receiving NSAIDs are more likely to be investigated fo r the presence of an ulcer than chose not receiving NSAIDs, leading to increased detection of ulcers in chis group.
Tables I and 2 summarize the main study characteristics and relative risks for adverse gastrointestinal events from case control (8-17) and cohort swdies (l 5, [18][19][20][21][22]. Eleven of the l6 studies reviewed show a statist ically significant increased risk of gastric ulcer or its complications for NSAID users compared to nonusers.The considerable variability in the reported relative risks (from 1.05 to greater than LO) is largely due to differences in research design, study population ( eg, pre-paid group practice, elderly, med icaid) and outcome measures (cg, fatal upper gastrointestinal bleed, gastrointestinal hospitalization, chronic gastric ulcer).The majority of studies point to a relative risk for gastrointestinal events two co five times greater for NSA ID users than nonusers.
Overall, the evidence for a causal relationship between NSA lOs and serious gastrointestinal events is strong.
Therefore, the answer to the fi rst question is yes, there is a risk.What is t he magnit ude of risk?The risks of gastrointestinal complications among NSAID users have been estimated from va rious sources.Prospecci vel y collecccc.ldata from a la rge computerized registry indicated risk of gastrointestinal hospitalizminn as l % per year for rheumatoid arthrius patients (23 ).T he fond and Drug Administration has quoted a risk for serious gastrointestin al events of 2 to 4% per year among NSAID users (2 4 ).A recent overview and meta-analysis wh ich combined the rates of gast ro in testi na l complications from over 100 NSA ID cl inical t riab reported an overall complication rate of 2°,{, (25).Estimated ahsolute risb of gastrointestinal complicarium from rhe six coh()rt studies rev iewed va ry from 0.02% to 0. 5<y.L Overall, rhe reported risb vary from two in 1000 LO fou r in JOO.T his variation b due to differences in t he ascertai n ment nf NSA ID exposu re and o utcome nssessment.Wh o is particula rly at ris k ?Studies wh ich examine the types of patienh at risk yielJ somewhat inconsistent fi ndings due to the fact t!"iat most studies cxmnini ng risk facto rs fail to cont ml fur other potenti al confoundcrs.Fnr example, the evidence of an excessive risk in elderly females may he confou nded hy th e documented inc reased NSAID use in th is group ( 6).

EXAMINING ALTERNATIVES FOR REDUCI NG RISKS
Having established the p resence of a cli n ically important risk, the next ster is w examine th e alternatives fo r reducing this risk (Figure l ).Some simple approaches to risk reduction in clude using the minima l dose of NSATD which wi ll control t he symplL)ms, switch ing to non-NSAID analgesia whenever possible, and carefully monitoring elderly patients.The dl'cision to prescribe prophylaxis ts much more difficult becat1se it involves not only potential hencfits hut abo potent ia l riob and additional co~t.Cli nical decisinn nnalysis is useful because it qunntitativcly incorporates such information 111to the clinical decision making proceS5.
Using 'decision tree' methodl)logy, rhe aurhnrs have de,igned a c li nical model representing clinical management dec1s1 on.~for physicians faced with these patien ts.This model can also he adapted for cost-effccnveness an alyses.Preliminary data ind icate that determin ing the cost-effectiveness of proph ylaxis with 7.

SUMMARY
In summary, most of the avai lable li terat ure supports the existence of a causal relationship between ch ronic NSA JD use and t he development of gastric ulcer and its subsequent complications.The magn itude of the risk can he estimated en be bet ween two in 1000 and four in I 00.A lthough several subgroups have been suggested to be at h igh risk, only 111 the elderly and perhaps in females have there been adequate data to support this associiition .Decisi,m analysis is c1 useful tool fo r th e q uantitative examination of th e manageml'nt alternatives available for dea ling with these risks.Ir also represents a model to study I he cost -effecuveness of prophylactic therapy.
Research 1s needed ro defi ne accurately the natural hi,l()ry of NSAID-induced gastmincestmal mucosa!damage and particularly thl' ulcer complication rate, in both the general population anJ various suhgroups.Preventive strategies can then be targeted ro the high risk groups.

-Figure I )
Figure I ) Components of clinical decisions regarding N SA ID-induced adverse grutrnin1estinal events traitement Jes reactions indesirables d 'origine medicamenteuse com portent deux elements.Le premier evalue quels sont les risques pour le patient et le second examine les solutions permertant de composer avec e ux.T rois plans de recherche ont ete utilises pour ecudier cette association: etudes ecologiques, etudes castemoins, etudes re trospectives Jes cohortes.Bienque ces strategies ne constituent pas la fa~on la plus sure d'etablir la relation de cause a effet, les rcsultats indiquent que les usagers d' AINS sont l ,5 a 10 fois plus exposes aux reactions gastro-intest ina les graves que les non-usagers.La la rge variation des resultats est due a des facteu rs multiples.O n suggere que plusieurs sous-groupes etaient formes de suj ets Rate ratio Re/olive risk.users/nonusers: Cl Confidence intervals: GHC Group Health Cooperative, Puget Sound: COMPASS Computer/zed on-line pharmaceutical analysis and survelllonce system: UGI Uppergostrointesttnol PAST PRESENTFigure2) De.1ign of co/ion and case conrro/ qudic1CAN J GN,TROCNTEROL VOL 4 NO 3 MAY 1990

TABLE 2
Rote ratio Odds ratio, users/nonusers: Cl Confidence ,ntervol: GI Gostro,ntestinol Lambert P, Langman MJS.20 years of hospital ad missions for pertic ulcer in England and Wales.case, alcohol use, smoking and severe rhe umatic disease requ ire further study.
. a syntheu c proswglanJin ana logue, hi nges tm a more accurate defin ition of the ulcer complicatinn rate.