Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Arthritis Rheum. Author manuscript; available in PMC 2012 February 1.
Published in final edited form as:
PMCID: PMC3031757

The Incidence of Gastrointestinal Perforations Among Rheumatoid Arthritis Patients



Gastrointestinal (GI) perforation has emerged as a novel safety concern in relation to medications used for rheumatoid arthritis (RA). However, the incidence and risk factors for GI perforation have not been well characterized in RA patients.


Using administrative databases of a large U.S. health plan, we identified RA patients treated with biologics, methotrexate, oral glucocorticoids, and NSAIDs; additional risk factors were evaluated including diverticulitis. Hospitalized GI perforation was identified using a validated algorithm. Incidence rates and risk factors were evaluated using Cox proportional hazards models.


Among 40,841 RA patients, 37 hospitalizations with GI perforation were identified. The rate of GI perforation among current biologic users concomitantly exposed to oral glucocorticoids was higher (rate=1.12 per 1,000 patient-years [py], 95% CI 0.50, 2.49) than for biologic users who were not glucocorticoid users (rate=0.47 per 1000py, 95% CI 0.22, 0.98) or MTX users using glucocorticoids (rate=0.87 per 1000py, 95% CI 0.36, 2.10). Neither biologics nor MTX were significantly associated with perforation, in contrast to current use of glucocorticoids and NSAIDs together (hazard ratio =4.7, 95% CI 1.9, 12.0) or glucocorticoids alone (hazard ratio=2.8, 95% CI 1.3, 6.1). Diverticulitis also was a strong risk factor (hazard ratio = 9.1, 95% CI 3.1, 26.4). Seventy percent of perforation cases used glucocorticoids, had antecedent diverticulitis, or both.


GI perforationis an uncommon but serious adverse event among RA patients. Because a majority of patients were either glucocorticoid users or had previously-recognized diverticulitis, these individuals should be considered at higher risk.

Keywords: rheumatoid arthritis, tumor necrosis factor antagonist, gastrointestinal perforation, methotrexate, glucocorticoids, NSAIDs


Serious medication-related adverse events are an important issue when selecting appropriate therapies for individual patients who may be at higher risk than others. Gastrointestinal perforation (GIP) is a rare but serious, sometimes fatal, event that has been observed in patients with rheumatoid arthritis (RA). Most literature about GIP is derived from trials of non-steroidal anti-inflammatory drugs (NSAIDs), where perforations of predominantly the upper GI tract have been observed. Less is known about lower GIP, especially in RA patients. Prior literature on GIP events often reported GIP only in aggregate with GI bleeding events, which limits more specific evaluation of the incidence and risk factors for GIP. GIP has become the subject of recent scrutiny in light of a newly approved therapy for RA, tocilizumab (TCZ). During the TCZ development program, 26 cases of GIP were observed in the TCZ-treated patients (at a rate of 2.8 events per 1000 patient-years), and none occurred in the control population (1). A majority of these events occurred in the lower GI tract, in contrast to the NSAID trials where adverse GI events were predominantly upper GI perforations. Some GIPs events occurred in the open-label follow-up phase of the TCZ studies, which inherently lacked a control population, so it was difficult to estimate whether the rate of GIP is similar to or greater than what might be expected in RA patients not treated with TCZ.

In light of a paucity of literature evaluating the incidence and risk factors for GIP in RA patients treated in community settings, we obtained data from a large U.S. health plan to assess the rate of GIP in relation to a variety of medications commonly used for the treatment of RA. We further assessed other risk factors hypothesized to be associated with GIP.


Eligible patient population and observation period

After IRB approval, we used the administrative medical and pharmacy databases of Aetna, one of the largest health insurers in the United States providing medical, pharmacy, dental, life, disability and other employee benefits. Aetna provides medical coverage to more than 18 million individuals and pharmacy coverage to 10 million individuals through numerous employer sponsored and individual plans. We identified RA patients on the basis of International Classification of Disease, 9th edition (ICD-9) codes. Patients were eligible to be under observation after they had one ICD-9 code for RA (ICD9 714.X) from a physician office visit, followed by a prescription or infusion of a tumor necrosis factor-alpha antagonist (anti-TNF)or non anti-TNF biologic or for methotrexate (MTX). Patients were also eligible to be under observation if they had ≥ 2 physician ICD-9 codes for RA separated by >7 days and occurring within six months of one another. Observation time began on the date of the biologic or MTX prescription, or on the date of the 2nd RA diagnosis code. As a comparator cohort, we identified patients with psoriatic arthritis (PsA) with ≥ 2 physician diagnosis codes for PsA separated by >7 days but occurring within six months; these individuals were censored if they subsequently had any physician diagnosis code for RA so as to ensure that the RA and PsA cohorts were distinct.

Patients must have had both medical and pharmacy benefits in the 6 months prior to the index date and throughout follow-up. Individuals age 65 and older were included only if they were enrolled in a Medicare Advantage plan (with a pharmacy benefit) but not if they were commercially insured. Commercially insurance is often a secondary payor to Medicare for adults in this age group, and therefore their claims history may be incomplete. The 6 month ‘baseline’ period was used for covariate assessment for all non time-varying covariates. RA patients with a history of hospitalized GIP or any diagnosis of inflammatory bowel disease (e.g. Crohn’s disease) during the baseline period were excluded. Patients were censored at the time of their first hospitalized GIP, loss of medical or pharmacy benefits, or the end of the study, which began January 1st, 2005 and ended August 31st, 2009.

Drug exposures of interest

Patients were assigned into one of several mutually-exclusive medication categories, which included biologics (i.e. rituximab, abatacept, or anti-TNF agents), MTX, and neither (e.g. other DMARDs including sulfasalazine, hydroxychloroquine, and leflunomide). Non anti-TNF biologics were combined with anti-TNF biologics given sparse data for the non anti-TNF biologic groups. Biologic-treated patients were permitted to use background MTX or other non-biologic DMARDs. Each of the three exposure categories was sub-divided by whether patients were concomitantly exposed to oral glucorticoids or not, yielding a total of 6 exposure categories. Current exposure, defined by the days supply of the medication plus a 90 day ‘extension’ period, was used to characterize exposure for each category. The extension period was added since patients who become ill often stop medications, and using an extension period often captures attributable exposure and has been suggested to reduce bias when assessing acute events (2). For infused drugs, where days supply is not applicable, we assigned current exposure as 30 days to abatacept, 180 days to rituximab, and 56 days to infliximab; the same 90 day extension was added to each. Given low event rates expected, the data did not permit a new user analysis (3).

Outcome of interest

The outcome of interest was hospitalization for GIP. GIP was identified on the basis of inpatient ICD-9 codes from hospital discharge diagnoses (Appendix). The validity of the GIP case algorithm was supported by a validation study which compared the GIP claims-based algorithm against a gold standard of hospital medical record review for 92 RA patients with suspected GIP (4). The positive predictive value of the claims algorithm to identify confirmed GIP events was 94% (95% CI 86%–98%). Because claims data might miss some GIP cases, we conducted a sensitivity analysis that added cases without a discharge diagnosis of GIP but which had two or more inpatient physician diagnoses for GIP on unique days during the hospitalization.

ICD-9 Diagnosis Codes Occurring During Hospitalizations to Identify Gastrointestinal Perforations*

Statistical analysis and covariate selection

After identifying GIP cases, we compared baseline characteristics of cases and non-cases for demographic and risk factors of clinical interest. The incidence of GIP was described for each of the 6 medication exposure groups using a Poisson distribution to estimate 95% confidence intervals. Cox proportional hazards models evaluated the multivariable adjusted rate of GIP in relation to medications and risk factors of interest(shown in Table 1). Age was dichotomized at age 50 given few events in older individuals in this dataset. Diverticulitis was a key risk factors of interest. This condition was assessed during the baseline period, and also as a time-varying covariate during the observation period. Our focus was on evaluating this condition as a risk factor prior to the GIP event, not just at the time of the GIP event, which could result in protopathic bias (5). As a time-varying covariate, we therefore ‘lagged’ the diverticulitis covariate by 90 days in the multivariable models, meaning that if the first occurrence of this condition was identified only within the 90 days prior to the event, it was not considered a previously-recognized risk factor. We create an exposure categories for the 4 mutually-exclusive combinations of glucocorticoids and NSAIDs given the potential for an interaction between them(6, 7). Additionally, we descriptively evaluated use of upper endoscopy and colonoscopy procedures in the 30 days prior to the GIP event among the cases. These procedures were not evaluated as risk factors in the multivariable model because of uncertainty regarding their indication and a lack of information as to whether they were truly antecedent to the perforation event(and thus would be a risk factor), or only part of the diagnostic workup. Analyses were performed using SAS 9.2 (SAS Institute, Cary NC).

Table 1
Baseline* Characteristics of RA Patients with and without Hospitalized Gastrointestinal Perforation


We identified a total of 40,841 RA patients eligible for analysis. During the observation period of 78,384 person-years, we identified a total of 37 hospitalizations for GIP. Of these, 6 (16.2%) occurred in the upper GI tract (duodenum or higher), and the remainder (83.8%) occurred in the lower GI tract. At baseline, patients with a GIP were more likely to be older, have peptic ulcer disease, have diverticulitis or diverticulosis, use oral glucocorticoids, and use proton pump inhibitors (table 1).

Analyses of time dependent exposures indicated that of the 37 GIP cases, 19 (51.4%) were current users of oral glucocorticoids at the time of the event; 4 (10.8%) had diverticulitis and no one had diverticulosis (without diverticulitis) at any time prior to 90 days preceding the GIP event. An additional 8 cases had a diagnosis of diverticulitis first identified within 90 days prior to the GIP event, yielding a total of 12 (32.4%) of the 37 GIP cases that had diverticulitis recognized prior to or proximate to the hospitalization for GIP. Combining glucocorticoid use and diverticulitis as risk factors, 26 cases (70.3%) were current users of glucocorticoids and/or had previously-recognized diverticulitis at the time of the event. A total of 5 (13.5%) cases had an upper endoscopy or colonoscopy procedure in the 30 days prior to the event.

Table 2 shows the crude incidence rates of GIP associated with various RA medications. The two categories with the highest rates of GIP were biologic users concomitantly taking glucocorticoids (1.12 cases per 1,000 person-years) and patients not receiving biologics or MTX (e.g. other DMARD users) who were concomitant steroid users (1.06 cases per 1,000 person-years). An additional 14 cases were identified after adding inpatient physician claims and modestly increased the observed rates. For example, after adding these additional cases, the rate of GIP among current biologic users treated concomitantly with oral glucocorticoids was 1.31 (0.62, 2.74)per 1,000 person-years. The small number of GI perforation events (n=14) in the PsA cohort required that we group the various exposures together. Overall, the rate of GIP in the PsA cohort was 0.22(0.12,0.36) per 1,000 person-years.

Table 2
Incidence of Hospitalized GI perforations among RA Patients Treated with Biologics, Methotrexate, and Oral Glucocorticoids

Multivariable-adjusted associations with GI perforations are shown in Table 3. Neither current use of biologics nor MTX conferred increased risk. In contrast, glucocorticoids, with or without NSAIDs, were a significant risk factor for GIP. Diverticulitis, recognized either during the baseline period or antecedent to the event, was also a strong risk factor.

Table 3
Multivariable-adjusted Risk Factors for Hospitalized GI Perforation


Among commercially-insured RA patients, we found a low rate of hospitalized GIP, which occurred at a rate of 1.12 (95% CI 0.50–2.49) per 1,000 patient-years among individuals who were current users of biologics and oral glucocorticoids. Oral glucocorticoid use was an independently-associated risk factor for GIP, as was diverticulitis (recognized prior to the event). Indeed, 70% of GIP cases had one or both of these two risk factors.

Our results are informative in comparison with the recently reported rates of GIP in the tocilizumab development program, which occurred at a rate of 2.8 (95% CI 1.9–4.1) per 1,000 person-years (1). Similar to our findings, a majority of the tocilizumab-associated GIP cases (24 of 26, or 91%) occurred in the lower GI tract. Little data regarding GIP is available from other clinical trials of other biologics, although this outcome has been observed in patients on anti-TNF therapies(8)and from observational studies of RA patients not on biologics (9). The clinical importance of events that are in the range of 1–3 per 1,000 person-years is debatable; by way of comparison, the rates of hospitalized infections is in the range of 5 per 100 years in most RA cohorts and registries (10). However, prior research suggests that in counseling patients regarding the side effect profile of new medications, fear regarding even very rare events is often judged as important and may impact treatment decisions if the events are serious and unfamiliar (11, 12).

Several observational studies provide additional context for our results. A synthesis of 12 studies conducted both in the U.S. and internationally found that, compared to our rates for RA patients, the incidence of GIP was lower in the general population and occurred at a rate of 0.1 (95% CI 0.04–0.23) per 1,000 person years (13). The rate of diverticular perforations is even lower (14). Another co-occurring event that sometimes accompanies GIP is GI bleeding; studies that have estimated both bleeding and perforation have found one case of GIP for every six to nine cases of GI bleeding (13). In fact, many studies do not report rates of GIP separately and instead group perforation with the more common GI bleeding events (15). NSAIDs appear to be more strongly related to upper GI events compared to lower GI events (16). We found that a majority of the perforation events in our RA population were of the lower GI tract rather than the upper GI tract. A decreasing trend for upper GI bleeding and perforation and an increasing trend for lower GI events has been previously observed and hypothesized to be related to changing patterns of NSAID use, GI prophylaxis with agents such as proton pump inhibitors, eradication of H. pylori infection, and other factors (16). Also concordant with our findings, a prior study that evaluated a combined endpoint of hospitalization for GI bleeding or perforation found that the strongest risk factors for these outcomes were glucocorticoid use (incidence rate ratio [IRR] = 3.1, 95% CI 2.1 – 4.8) and age (IRR = 1.4 per 10 years, 95% CI 1.2–1.8) (17).

The strengths of our study include a large cohort of RA patients treated in real world settings, and the size of the database allowed us to study this relatively uncommon adverse event. Our GIP case definition has been previously-validated against medical record review and has been shown to have very high validity, with a PPV of 94%. Despite these strengths, a few limitations must be recognized. Despite the fact that all outcome events occurred during a hospitalization, commercial administrative data sources typically do not allow for ascertainment of out-of-hospital deaths, which limits understanding of mortality associated with GIP. Previous studies have suggested it is high, up to 24% (14). Because of small numbers of events, we were not able to quantify risk associated with individual biologic agents or non-biologic DMARDs other than MTX. We also did not include in our case definition for GI perforation surgeries for diverticulitis (unless also accompanied by a perforation diagnosis); some of these might have been for acute GI perforation events. Finally, medical records were not available for this analysis. Our claims-based algorithm for GI perforations was supported by a validation study, but we recognize that administrative data provides limited clinical information and thus may misclassify events. Our sensitivity analysis that added inpatient physician diagnoses suggests that the event rates reported in our primary results are conservative, and the actual event rates may be up to 20–30% higher.

In conclusion, our results provide context in evaluating the rates and risk factors for GIP in future safety studies of RA medications. Patients who are not glucocorticoid users nor who have diverticulitis appear to be at very low risk for this event.


Funding: This work was supported by the Doris Duke Charitable Foundation and the Agency for Healthcare Research and Quality (R01HS018517). Some of the investigators receive support from the National Institutes of Health (AR053351: JRC).


Disclosures: JRC: (Research grants & Consulting): UCB, Roche/Genentech, CORRONA; Centocor, Amgen

ED: (Research grants): Amgen

Remainder of coauthors: None


1. Van Vollenhoven RFKE, Furie R, Blesch A, Wang C, Curtis JR. Gastrointestinal Safety in Patients with Rheumatoid Arthritis Treated with Tocilizumab: Data From Roche Clinical Trials. American College of Rheumatology ACR/ARHP Annual Scientific Meeting; Philadelphia, PA. 2009.
2. Dixon WG, Symmons DP, Lunt M, Watson KD, Hyrich KL, Silman AJ. Serious infection following anti-tumor necrosis factor alpha therapy in patients with rheumatoid arthritis: lessons from interpreting data from observational studies. Arthritis Rheum. 2007 Sep;56(9):2896–904. [PMC free article] [PubMed]
3. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003 Nov 1;158(9):915–20. [PubMed]
4. Curtis J, Chen S, Werther W, John A, Johnson D. Validation of ICD-9-CM to Identify GI Perforation in Administrative Claims Data Among Rheumatoid Arthritis Patients. International Society of Pharmacoepidemiology; Brighton, U.K: 2010.
5. Rothman KJ. Epidemiology. New York: Oxford University Press; 2002.
6. Gabriel SE, Jaakkimainen L, Bombardier C. Risk for serious gastrointestinal complications related to use of nonsteroidal anti-inflammatory drugs. A meta-analysis. Ann Intern Med. 1991 Nov 15;115(10):787–96. [PubMed]
7. Fries JF, Williams CA, Bloch DA, Michel BA. Nonsteroidal anti-inflammatory drug-associated gastropathy: incidence and risk factor models. Am J Med. 1991 Sep;91(3):213–22. [PubMed]
9. Sihvonen S, Korpela M, Mustonen J, Huhtala H, Karstila K, Pasternack A. Mortality in patients with rheumatoid arthritis treated with low-dose oral glucocorticoids. A population-based cohort study. J Rheumatol. 2006 Sep;33(9):1740–6. [PubMed]
10. Curtis JR, Jain A, Askling J, Bridges SL, Jr, Carmona L, Dixon W, et al. A comparison of patient characteristics and outcomes in selected European and U.S. rheumatoid arthritis registries. Semin Arthritis Rheum. 2010 Aug;40(1):2–14. e1. [PMC free article] [PubMed]
11. Fraenkel L, Bogardus S, Concato J, Felson D. Risk communication in rheumatoid arthritis. J Rheumatol. 2003 Mar;30(3):443–8. [PubMed]
12. Fraenkel L, Bogardus S, Concato J, Felson D. Unwillingness of rheumatoid arthritis patients to risk adverse effects. Rheumatology (Oxford) 2002 Mar;41(3):253–61. [PubMed]
13. Hernandez-Diaz S, Rodriguez LA. Incidence of serious upper gastrointestinal bleeding/perforation in the general population: review of epidemiologic studies. J Clin Epidemiol. 2002 Feb;55(2):157–63. [PubMed]
14. Morris CR, Harvey IM, Stebbings WS, Hart AR. Incidence of perforated diverticulitis and risk factors for death in a UK population. Br J Surg. 2008 Jul;95(7):876–81. [PubMed]
15. Laine L, Connors LG, Reicin A, Hawkey CJ, Burgos-Vargas R, Schnitzer TJ, et al. Serious lower gastrointestinal clinical events with nonselective NSAID or coxib use. Gastroenterology. 2003 Feb;124(2):288–92. [PubMed]
16. Lanas A, Garcia-Rodriguez LA, Polo-Tomas M, Ponce M, Alonso-Abreu I, Perez-Aisa MA, et al. Time trends and impact of upper and lower gastrointestinal bleeding and perforation in clinical practice. Am J Gastroenterol. 2009 Jul;104(7):1633–41. [PubMed]
17. Wolfe F, Hawley DJ. The comparative risk and predictors of adverse gastrointestinal events in rheumatoid arthritis and osteoarthritis: a prospective 13 year study of 2131 patients. J Rheumatol. 2000 Jul;27(7):1668–73. [PubMed]