The assesslllent of evidence associating nonsteroidal anti ~ inflammatory drugs with complications of peptic ulcerations

A causal relationship is now finnly established between nonsteroidal anti -inflammatory drug (NSAlD) use a nd the occurren ce of peptic ulcer complications. In the United Kingdom, rising NSAID use has been matched by rises in ulcer mortality and perforation rates, particularly in o lder women. It is n ot likely, h owever, that drug use accounts for the entire increase. The reasons why some people deve lop ulcer complications and others do not are poorly understood. It is plausible to propose that oth er factors, such as history of ulcer o r imJigestion, current smo king, and alcoho l consumption, might raise this risk; however, supportive evide nce is lacking. Can} Gastroenterol 1990;4(3 ):91-94

that the use of nonsteroidal antiinflammatory drugs (NSAIDs) will predispose to peptic ulcer complications of bleeding and perforation.The strength of evidence support ing this view va ries, however, as it derives from a variety of sources, including anima l and human experimen tal studies, clinical reports of suspected adverse responses, retrospective studies with o r without controls, and prospective surveillance programs.The value of these findings must therefore be judged by sta nd ard c riteria which can be employed in assessing an y set of evidence bearing upon Jisease causation regardless of the facto rs involved.The basic criteria include: consistency of disease occurrence fo l lowi ng exposu re, strength of association, dose response, t iming, logic and accommodation of con fo unding variables.

DESCRIPTION OF BASIC C RITERIA
Consistency of disease occurrence following exposure: Confirmation of a n associat ion requi res the occurrence of d isease in individuals exposed with at least reasona ble frequency co t he etiological faccor under consideration .In additio n , exposure to this factor should be detected more frequendy in individuals with the d isease under conside rat ion than in contro l individuals without it.The possibility that the etiological association is due co the effect of another associated facto r ( referred to as a 'confounding variable' ) should a lways be examined.For example, in assess ing th e importance of smok ing habits, the confounding influence of concurrent alcoho l con su mption must be considered.Furthermore, it is necessary to conside r whether the control group chosen represents a fai r comparator.Controls should be chosen to be as similar as possible to the disease group except for being free fro m the d isease in questio n.This implies similari ty in age, sex and other factors, such as social class distribution.
Strength of association: The strength of an association is typically expressed as the 'odds ratio' or 're lative risk'.These ratios state how many Limes more likely an individual exposed to a particular factor is to develop a disease than one who is not exposeJ to the sa me factor.Although a raised risk shows association, it does not necessa rily imply causation; raised risks can also be associated with confounding influences.In general, the higher the risk, the greater are the chances thal the relevant factor is ca usa l.
Dose response: In order co measure dose response, the intensity of exposure must be graded.Generally speaking, the greater the intensity of exposure, the greater the risk RELATlNG NSAIDS AND GASTRODUODENAL MUCOSAL DAMAGE Consistency of evidence: A large body of evidence now exisls with regard to NSA!D use and gnstrodmxknal mucosa!damage, deri ving from a variety of sources and with varying stre ngth.Spontaneous adverse reaction reports: NSAID~ have appeared prominently in the spontaneously rcporte<l adverse reaction registers of many counLries and have been responsible for one-quarter to one-fifth of all such reports in Lhe United Kingdom and the United States ( 1-3 ).
These darn ::ire diffi cult Lo assess.Sromaneous reports represent a biased sa mple.These biases lead to reporti ng or, much more frequently, nonreponing, which is poorly underst0od.Reports, however, arc much mo re likely to be generated for new rather than o ld drugs.This of itself in validaLes comparisons between individual agents which are based upon spontaneous reports and, with NSAID-associated gastroinLestinal nd versc events, the occurrences are in commlm for the class.Nevertheless, reports of serio us adverse events of ulcer bleeding, perforation, and death have been so frequen t thal an importa nt clinical problem must be assumed.Complication rates: Little controlled informaLion is available with regard to pcrforntion of ulcers.In the Un ited Kingdom, a higher rnte of prior NSAIO use was found among pariems wiLh perforated ulcers than among inpatient controls in two studies (4,5).O n e of these studies incl uded a suhsrantial admixture nf paLicnts wi1 h ulcer bleeding but focused on thnse whn had heen opera Led upon ( 5 ).Furthermore, deta i Is of Lh e expcri men rn l methods indicaLed Lhat controls were nrn coll ecLed con-temporancously and were not mmched for age.S urve illance studi es in the U niLed Kingdom (6) a nd the Unned St.Hell (7) have suggested tha1 the nsks of ulcer perfora1 ion m,1y have hecn exaggerated.The former sLUd y, however, may nol have included sufficil:nt ly large numbers of elderly people LO detect ulcercomplicalions, whereas in the latter swd y, rhc numhcr of subjeCLli who developed ulcer complicationl, wal, rmher sma ll and Lhc observed dma set included many cases wiLh coding errors.The relative risk, although rnbed at 1.6, had wide confidence intervals, rell ecting the small numhers sLUdied, and the even small er number with NSAID-a~-sociaLcd com pl ,cat ions.Bleeding: Many of Lhe Jara se ts for ulcer bleeding arc uncontrolled.I lowever, data from o ne l,llldy (8), which e mployed matched Cl>mmunity and hospital inpatient controb, suggested that chose with ulcer bleeding, both gaslric and duodenal, were belwccn two and five times as likely to have hecn NSAID users as e ither cont ro l.Risks also differed liLLlc fo r aspirin and nonaspirin NSAIDs.S urveill a n ce o f Medica id script rccipienls in Michigan and Minncsorn revealed a raised relative risk ( l. 5) with a 95% confidence interval of l.l LO 1. 9 for upp er gastrointestinal bleeding from all sources (9).By contrnsr, no mm crial risk for upper gastroi nleslinal hleeding was found at Puget Sound, Washington, USA, in individuals aged 65 or less.Of the 30 1 patients ad miucd to hospital , 14 had filed a prescripl ion for a n NSAID in the previous 90 days; all paLients bleeding from duodenal ulcer in the period under review were excluded (10).Ulcer death: The same raised risk of dcalh from ulcer in NSAID recipient:,, (relaLive rbk 4.7) was found in Lwo North American Mudies; nne from Saskatchewan ( 11 ), and the tither from Tennessee ( 12).This risk, however, was limited to rho:,,c aged 75 and over in the former stud y.Strength of association and dose response: T he strength of an a,sociation b conve ntio na ll y mca,ured hy Lhc relati ve risk or odds raLio, which ,tale~ how muc h more or less likely it is for an CAN J GASTR0ENTEROL VOL 4 Nl) 3 MAY 1990 inJtviJual who possesses a paruculm characterisr rc 11> Jevelnp a specific u1n -dn1on than for ,mL' withou t such a characteristic.W here NSAID use is concemed , this rbk, wh en raised, has been founJ co hi.' of the order of two-tn fou rfo ld.Such a figure, however, docs nm measure the .ihsolute risk.This cannot he obtnined without e ither some accrnnp,mymg measure nf the actual risk in the ordinary popularion or an odds ratio, 1c, a mea~ure of the actual risk m NSA I l) users and the difference compared wn h rhe base population.In Michigan and Minnesota (9), cwocascs of hemorrhage were estim ated en occur for every I 1,000 individual months of tn:aun1::nt, a figure \\'hich corre,ponds with est1m,1te, of n,k in t he U1111ed Kmgdnm (8,13).
Increased risk, c1ccording w rntens11 y or durauon nf exposure w a ,uspccted en ological agent, usually provides confi rmatory ev1dl.'nce of causation.This type of ev 1de11Le, however, b nm avm lab!..:; variaunn 111 the anti -i nflamm atory strengt h and pharmacokinetic pniperues nf individual drugs makes it impossible tu c rente logical compil,1t iom for di rL'Ct comparisom.Timing: A ll available darn sets h:,w been col lectl'd with cnns1deratinn given to drug intake 111 the imml.'diarepreceding time period.It 1s th e refore not poss1hle tu dissect the mformat ion fun her.Logic: NSA !Ds have been comnwn ly associatd with dyspepsia.Furthcrmnre, expernnenrnl otud ics and clinical triab have shown that NSA ID treatment is associated with the appearan ce of upper gastrointcsunnl lesiom, mainly gastric antral in site and varyi ng Ln degree from hemorrhagic spots to 1::rosions or ulcers (14)(15)(16)(17).
What is k ss clear is the basis for the occurre nct: of ulcer comp I icarions, assuming that NSA !l) treatment has a causal role.A th ird or more ofNSA JL) recipients may develop dyspepsia ( 18),   (22,2 3).
Confounding: Curre nt cl inica l evidence suggests that the elderl y are pmt icularly prnne to deve lop NSA ID, assoc iated complicatinm and may a lw he the gn)up most prone to peptic ul-ccrat1on .Most fonna l st ud ies, therefore, account for confounding hy age.Another important fac tor is confounding hy disease indication.There ts ,time reason to bel 1<.'ve that paricnts with rheumawid arth ri tis may he more prone to peptic ulceration th an expected.However, the hulk of NSA ID -assocaated ulcer compli ca tions seem ro occur in individua ls who dn not have rheumatoid arthritis.Disability assoc iated with rheumatoid anhriris has heen associated with liab ility to peptic ulceration, hut multivaria te a na lysis h as not been used to separate disahility from drug treatment (21 ); indeed, they may not he separable by fo rma l st udy.The degree of disability associated with osteoarthritis, the most common indication for NSA I D t reatment, h as nnt been aSsllCiated with susceptibility to ulcer.DISCUSSION There seems good reason to accep t that NSAIL) treatment is frequently assoc iated wi th t h e onset of dyspepsia and tha t peptic ulceration may emue.Further oimplications, however, arc rare.W hat remams unclea r 1s v. h y some 111d1viduab deve lop ulcer compl1canons and others dn not, and wheth er th ere are matenal differences 111 gastric and duodern1I susceptibility to damage.Experi mental stud ies a nd short 1e rm human investigat11ms in ntirmal volu nteers sugge,t th at the ga,tnc ant rum may he the prune , arc 1i damage ( 14-17).Whether mvesugauons Ill healthy volunteer, ( who ,ire usuall y young) c.in be u~ed t1l predict the s1u: and pmtern of damage 111 o !dl.'r people with c h niniL diseases 1s un certa111.In the Unned S tai es t h e hroad assumption tends to he made that danrnge is mainly ga,mc.
Kurata (25) (25,26).Analytical stuJ1es suggest a strong correlation in men and women with :,moking habits (27).Although the differences in findings between diffe rent area~ seem irreconcilable 111 the main, limited agree ment exists in suggesting chat differences in ulcer frequency between men and women may be d isappearing, anJ that NSAI L) use c.m ,lCCl>UnL for a minor progressinn only, perhaps onC•ljWHtcr of the uhserved change.The h,1s1s for 111div1dual susccpuhility i~ unclear.Ulcer complicatiom associated with NSA!D use seem to be particularly prevalent in the elderly.hut NSAlD use increases with age.Armstrong ;iml Blower (5) suggested that NSAl D users might he at particular rbk uf dymg, hut this evidence is difficult to evaluate because cases and contrnb were not age-matched.In another study in Australia, however, there wa~ m, evidence chat NSAID use affected Lhe chances of dying with ulcer disease over and above that nf developing ulcer compl,cat iom (28).
but ulcer complirn tiom seem to occur in less than one in LOOO.NSA ID rreatmcnt inhibits p latel e1 function, which c~iu ld account for th e tendency I ll hemnrrhage.It does not, hl1wever, n.plain ulcer pe1forntion.C linic.1!experience in th e United Kingdom and elsewhere have suggested t hat the occu rrence of dyspepsrn com~lates poorly with l1abil1ty tn ulcer complicatiom ( 19,20 CAN J GASTRllENTEROI VI.Ji 4 Nl1 3 MAY 1990 Peptic ulcer complications Snme clinical studies, hut not al l, have shown thar patients 1n whom compli , cared ulceration develops may have had no prior symptoms.The reasom arc unclea r; thus, we do not know 1f rhe cnmplicaunn nccurs in a newly developed or a n esrnhlbhed ulcer.Neverthe less, th e re seems tll Ix adequate evidence for accepting that NSA ID treatment 1s associated with the development of ulcer crnnplicar1ons a nd tha1 the assm.1.1tion 1s c.iusal.
The importance of the fonner lies in their ability to confuse the nature of the true noxious influence.For example, ma rital srarus can inllue n cc disease occurrence through change in habits.The latter can a lLer risk estimates for the influences under prime consideration.For instance, alcohol and tobacco can independently affecl the occurrence of esop hagea l cancer.
). Ry contrast, tn the (25)es 11llspicalizmion and m,ircalicy rates over the past 30 year,, selfreported ulcer prevalence ha~ ri~en(25).These trend~ may he related to a ~hifL to outpauenc management usmg new drugs.A lternalivcly, United Stales data show a sex ratio corresponding to the rising rate for women found in Lhc Nalional H ealth Imerview Survey have risen in elderly women ( 13).The reas1ms for t he rise were uncle.iram.I could not he ascribed to ei t her smoking habits or changes in NSA l D use, a lthough these could have accoun ted for part of the rise.Despite decr1::as1ng -LANGMAN Uni led