General physicians and rheumatologists caring for patients with symptoms of chronic arthritis may be reluctant to abandon using NSAIDs, which help most of these patients, in order to protect the few who may develop serious complications from use of these drugs. On the other hand, gastroenterologists see fewer patients for arthritis but many with serious gastrointestinal complications. Life-threatening events such as perforation or serious hemorrhage from NSAID-induced ulcers, which often develop with little or no warning [6], are a real problem because of the many patients at risk.

Various agents have been used in attempts to reduce the incidence of NSAID-induced gastrointestinal lesions. In one endoscopic study, cimetidine at a dose of 300 mg four times a day showed no benefit in healing NSAID-related lesions compared with placebo, and 400 mg at bedtime provided no benefit in preventing these lesions compared with placebo [7]. Antacids (magnesium-aluminum hydroxide, 10 to 20 mL as needed to a dose as high as 60 mL daily) and sucralfate have recently been reported to reduce dyspeptic symptoms in arthritic patients receiving NSAIDs in whom gastropathic lesions (but not ulcers) were shown endoscopically [8, 9]. The surface-active antiulcer drug sucralfate was ineffective in preventing ulcers in persons receiving NSAIDs [10, 11], and the histamine-2-receptor antagonist ranitidine did not prevent gastric ulcers but did reduce the frequency of duodenal ulcers [12, 13]. Results of preliminary studies indicate that the effects of omeprazole parallel those of ranitidine [14]. A recent study [15] in achlorhydric patients showed that NSAID-induced ulcers develop in the absence of gastric acid. In contrast, several clinical trials have shown that the incidence of endoscopically visible erosions and ulcers associated with NSAID use can be reduced by cotherapy with the synthetic prostaglandin misoprostol [16-19]. However, it was not proved that preventing endoscopic lesions would prevent clinically serious complications of NSAID-induced gastrointestinal ulcers [20].

Our objective was to determine whether concurrent therapy with misoprostol reduces the incidence of serious upper gastrointestinal complications in older patients with chronic rheumatoid arthritis who are taking NSAIDs. Because most patients taking NSAIDs do not routinely have endoscopy, we investigated the incidence of these complications during clinical care as actually practiced. A second goal was to better define which patients were at increased risk for development of serious NSAID-induced upper gastrointestinal complications.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Ambulatory patients at least 52 years of age who had chronic rheumatoid arthritis, defined by American College of Rheumatology criteria [21], and who were expected to be taking 1 of 10 specified NSAIDs at predefined minimum doses for 6 months were sought from practices of family medicine, internal medicine, or rheumatology. Of these practices, 661 in the United States and 3 in Canada enrolled at least one patient between July 1991 and August 1993. For all patients, a medical history was elicited, a physical examination was done, and a modified Health Assessment Questionnaire that included eight items on activities of daily living was administered [22]. Patients were excluded if they had had active peptic ulcer disease within 30 days of study enrollment; were taking or expected to need antiulcer medication (histamine-2 blockers, sucralfate, omeprazole) or any experimental medication during the study; had the Zollinger-Ellison syndrome, pyloric or duodenal obstruction, previous gastric resection or vagotomy, gastroesophageal reflux disease, varices, or cirrhosis; had a history of inflammatory bowel disease, upper gastrointestinal tract malignancies, hepatitis, alcoholism, or bleeding diathesis; were estimated to have a life expectancy of less than 8 months or had do-not-resuscitate status; were women of child-bearing potential; or could not tolerate misoprostol or any prostaglandin. The following were the minimum NSAID doses (mg/d) allowed: aspirin, 2000; diclofenac, 100; flurbiprofen, 200; ibuprofen, 1200; indomethacin, 75; ketoprofen, 150; naproxen, 750; piroxicam, 20; sulindac, 200; or tolmetin, 1200. Patients were allowed to receive more than one NSAID.

Intervention

Eligible patients were randomly assigned to receive either misoprostol or placebo in the form of 200- µgrams tablets from coded bottles supplied by the manufacturer. Patients were randomly assigned in blocks of four, so that in each block, half the patients would receive misoprostol and half would receive placebo. Because blocks were assigned to investigators, patients were randomly assigned within the individual centers. Investigators were not informed about the randomization procedure but were told only that the study was randomized and double-blind. Patients were instructed to begin taking half a tablet with meals and at bedtime each day for 10 days and then, if the drug was tolerated, to increase the dose to a whole tablet four times a day for the rest of the study. If the drug was not tolerated, the patients were instructed to reduce the dose back to half a tablet four times a day. Patients could continue therapy with arthritis disease-modifying agents (such as gold or corticosteroids) and were allowed to take antacids that did not contain magnesium. Patients purchased their own arthritis medications or antacids and reported their consumption when seen monthly for examination and counts of study medication tablets. Physicians were instructed to watch closely for clinical signs of gastrointestinal bleeding or other possible gastrointestinal complications, to inquire about symptoms, and to investigate suspicious episodes by appropriate clinical procedures.

Outcome Measures

All suspicious events, regardless of presumed cause, were to be reported, along with all available patient data, to the study medical officer (HWD) and then to an external review committee. This committee consisted of a gastroenterologist (FS), a rheumatologist (Kenneth Brandt, MD, Indiana University), and an epidemiologist (Marie Griffin, MD, MPH, Vanderbilt University). The committee determined, without unblinding the randomization code, whether the patient had upper gastrointestinal bleeding; other ulcer complications; or a problem such as hemorrhoids, diverticulitis, colon polyps, or cancer. It developed definitions of what were considered to be complications related to NSAID use and categories of such complications. The committee reached consensus and assigned such events to one of the following categories of definite upper gastrointestinal complications:

1. perforated ulcer, proved at surgery;

2. gastric outlet obstruction caused by proven ulceration and stricture, proved by endoscopy;

3. hematemesis, with endoscopically proven gastric or duodenal ulceration or erosion;

4. active or recent visualized bleeding from endoscopically proven ulceration or erosion;

5. melena, with endoscopically proven ulceration or erosion;

6. heme-positive stool, with endoscopically proven ulceration or erosion, plus either a) a decrease in hematocrit of at least 0.05 or b) orthostatic change in the pulse rate (from sitting to standing) of at least 20 beats per minute or decrease in systolic blood pressure of at least 20 mm Hg and a decrease in diastolic blood pressure of 10 mm Hg;

7. hematemesis, without endoscopically proven ulceration or erosion; and

8. melena, with heme-positive stool and without endoscopically proven ulceration or erosion.

The committee also defined categories of events not involving clinically significant bleeding:

9. report of melena with no other data;

10. heme-positive stools, with endoscopically proven ulceration or erosion and without active bleeding; and

11. report of melena and heme-negative stools when the stools were tested.

Sample Size Calculation and Statistical Methods

The placebo group was expected to have an incidence of serious NSAID-induced upper gastrointestinal complications of about 2% per year or about 1% during the 6-month study period. We estimated that the misoprostol group might have 40% to 50% fewer complications and that 60% to 70% of patients would complete the study. Given an less than 0.05 and power greater than 0.80, we planned in the protocol to assess the number of such complications observed, without unblinding the study, after 7500 patients had been followed. We did this to confirm the incidence assumptions and adjust the study size if necessary. We used the Fisher exact test to compare occurrence of events in the treatment groups. To identify risk factors, we evaluated the relation between the occurrence of serious upper gastrointestinal complications and patient characteristics using logistic regression. Supplementary time-to-event analyses were done using the log-rank test and Kaplan-Meier curves.

The study was supported by a grant from G.D. Searle & Co. Data collection and statistical analyses were done by the Quintiles Transnational Corporation.

Results

Top Methods Results Discussion Author & Article Info References

After 2 years of study conduct, the review committee identified 67 "definite" upper gastrointestinal complications, as defined in categories 1 to 8, from 242 suspected events reported among 8843 patients enrolled. The combined event incidence was 0.76% in 6 months, or about 1.5% per year. Because these numbers were deemed sufficient for a statistically significant result if the prestudy assumptions were correct, we discontinued study enrollment. Of the 8843 patients randomly assigned, 4439 had been allocated to placebo and 4404 to misoprostol. Analysis of the prestudy characteristics of the two randomized groups showed that they were similar in all of the variables assessed (Table 1).

Within the first month of treatment, withdrawal because of diarrhea, abdominal pain, or flatulence was greater (P < 0.001) in the misoprostol group (20%) than in the placebo group (15%), but other problems were not. Similar numbers of patients in both groups withdrew from the study after the first month. Overall, 39% of the patients withdrew prematurely; significantly more patients in the misoprostol group (42%) than in the placebo group (36%) withdrew. This difference was attributable to the diarrheal symptoms associated with misoprostol use (Table 2). The increased incidence of these symptoms in some of the patients was consistent with previous experience. Analysis of the prestudy characteristics of patients who withdrew prematurely showed no clear predictive factors, but more patients (19%) in the subset who withdrew because of adverse side effects had a history of peptic ulcer when compared those who completed the trial (13%). The characteristics and risk factors of patients who withdrew from the trial and of patients who completed the study did not otherwise significantly differ between the two study groups or from those of all enrolled patients.

For the 242 suspicious gastrointestinal problems reported by the investigators, the external review committee identified 49 patients with problems in categories 1 to 6 (serious ulcer complications), 18 more with problems in categories 7 and 8 (bleeding, but no ulcer or erosion found or proved), and another 28 with problems in categories 9 to 11 (possible previous but not active bleeding). The committee excluded 147 patients who had alternative causes for the events, primarily various types of lower gastrointestinal bleeding (Table 3).

Patients receiving placebo had more serious ulcer complications confirmed by surgery, endoscopy, or radiographic evidence (events in categories 1 to 6): 16 of 4404 patients receiving misoprostol compared with 33 of 4439 patients receiving placebo (odds ratio, 0.487 [95% CI, 0.268 to 0.886; P = 0.021]), a reduction of 51% (Table 4). The number of perforated ulcers and ulcer-induced gastric outlet obstructions (events in categories 1 and 2) was reduced 10-fold: 1 of 4404 misoprostol recipients compared with 10 of 4439 placebo recipients (odds ratio, 0.101 [CI, 0.013 to 0.787; P = 0.012]). One patient receiving misoprostol had a perforated duodenal ulcer; of the 7 perforations among patients receiving placebo, 4 were prepyloric, 2 were duodenal, and 1 was not specified in the surgical report. Five of the patients with perforations (1 in the misoprostol group and 4 in the placebo group) had no history of ulcers. The time-to-event analysis for events in categories 1 to 6 Figure 1 showed an advantage for misoprostol (P = 0.031).

View larger version (20K): [in this window] [in a new window]   Figure 1. Kaplan-Meier curves for gastrointestinal (GI) events in categories 1 to 6 for the misoprostol and placebo groups. The difference between the curves is significant. The numbers below the horizontal axis are the numbers of patients in the two groups remaining in the study at the times shown. * = the number of patients in each group remaining in the study at 170 days.

When we considered all patients with definite upper gastrointestinal complications (those in categories 1 to 8), 25 of 4404 patients in the misoprostol group had events compared with 42 of 4439 patients in the placebo group. This represented a 40% reduction with misoprostol (odds ratio, 0.598 [CI, 0.364 to 0.982; P = 0.049]); time-to-event analysis showed a supporting trend (P = 0.077). If the less certain bleeding events or probable gastrointestinal complications (those in categories 9 to 11) are included with the complications in categories 1 to 8, 95 events occurred—34 in the 4404 misoprostol recipients and 61 in the 4439 placebo recipients. The number of complications was thus reduced by 44% (odds ratio, 0.558 [CI, 0.366 to 0.851; P =0.007]). Time-to-event analysis showed a difference in favor of misoprostol for complications in categories 1 to 11 (P = 0.016).

We also did analyses to determine the incidences of proven gastroduodenal ulcers. The data were from the 147 patients with gastrointestinal problems but not definite or probable events, those with events in category 10, and those with events in categories 1 to 6. Thirty-seven of these patients had symptomatic but uncomplicated gastroduodenal ulcers Table 3: 11 patients in the misoprostol group and 26 in the placebo group. Among patients with events in category 10, ulcers were found in 6 of 8 patients receiving placebo and in the single patient receiving misoprostol. When this finding is considered together with the finding that 49 patients had serious ulcer complications in categories 1 to 6, 93 patients with ulcers were found: 28 receiving misoprostol and 65 receiving placebo (odds ratio, 0.431 [CI, 0.276 to 0.672; P < 0.001]). Thus, the number of symptomatic ulcers, whether complicated or uncomplicated, was reduced by 57%.

Eighteen patients with bleeding did not have endoscopic or radiographic confirmation of an ulcer or erosion (categories 7 and 8), either because the patient did not consent to the procedure (8 patients) or because no lesion was seen when the procedure was done (10 patients). The incidence of these category 7 and 8 events did not differ between treatment groups. If the minor and less certain bleeding events, those in categories 9 to 11, are considered, there was no statistically significant difference between treatment groups. When all the defined bleeding events are considered, including 38 patients with events in categories 3 to 6 (ulcers), 18 with events in categories 7 and 8 (no ulcer), and 28 with events in categories 9 to 11 (possible previous but not active bleeding), the events occurred in 33 patients receiving misoprostol and in 51 receiving placebo. This 35% reduction in the incidence of bleeding in patients receiving misoprostol was not statistically significant (odds ratio, 0.650 [CI, 0.418 to 1.009; P = 0.062]). Clearly, the reduction in ulcer incidence in this group was more pronounced than the reduction in bleeding events.

Prestudy characteristics of the 67 patients with definite gastrointestinal complications (those in categories 1 to 8) showed that patients with the complications were somewhat older than all patients in the study group (mean age of misoprostol recipients, 73 years; mean age of placebo recipients, 71 years), were more likely to have a history of previous peptic ulcer (24% of misoprostol recipients; 43% of placebo recipients), and were more likely to have a history of gastrointestinal bleeding (16% of misoprostol recipients; 26% of placebo recipients). However, the racial distribution of these 67 patients was similar to that of all study patients entered (Table 1). These findings suggested that age, history of gastrointestinal bleeding, and history of peptic ulcer are possible risk factors.

Analysis of 18 potential risk factors by logistic regression showed that an age of 75 years or older, history of peptic ulcer, history of gastrointestinal bleeding, and history of cardiovascular disease were all significant indicators of increased risk (Table 5). Other possible risk factors, including corticosteroid use, arthritis disability score, and sex, were not significantly correlated with the occurrence of serious upper gastrointestinal complications. Even after adjustment for all four significant risk factors, misoprostol was found to have significantly reduced the incidence of serious upper gastrointestinal complications (P = 0.043). No statistically significant interactions were found between treatment and these four risk factors (P > 0.20). When combinations of these risk factors are considered, the logistic regression model predicts that in 6 months, patients with none of the four factors would have a risk for having one of the defined gastrointestinal complications of only 0.4%; patients with any single factor, a risk of about 1%; and patients with all four factors, a risk of 9%. When age 75 years or older, history of peptic ulcer disease, and history of gastrointestinal bleeding are considered as factors, the risk is 5.3%. When a history of cardiac disease and any two of the other three risk factors are considered, the chance of a serious upper gastrointestinal complication occurring within 6 months while patients were receiving NSAIDs would vary from 3.8% to 4.3%. In addition to age dichotomized at 75 years, age was significant when entered as a continuous variable (P < 0.001) and as a dichotomous variable with a cut point of 65 (P = 0.04). Because 16% of the patients were 75 years of age or older, this cut point was chosen as the primary one.

Among the 67 patients with definite upper gastrointestinal complications, the numbers of patients taking each of the NSAIDs and either misoprostol or placebo were too small to show statistically significant differences. However, they did show similar distributions in the two study groups: 14 complications with naproxen (21%; 6 of 25 patients receiving misoprostol and 8 of 42 receiving placebo), 9 with diclofenac (13%; 3 patients receiving misoprostol and 6 receiving placebo), 6 with piroxicam (9%; 3 patients receiving misoprostol and 3 receiving placebo), 4 with ibuprofen (6%; 1 patient receiving misoprostol and 3 receiving placebo), 3 with flurbiprofen (4.5%; 2 patients receiving misoprostol and 1 receiving placebo), 3 with sulindac (4.5%; 1 patient receiving misoprostol and 2 receiving placebo), 3 with aspirin (4.5%; 3 patients receiving placebo), 1 with indomethacin (1.5%; 1 patient receiving placebo), 2 with tolmetin (3%; 2 patients receiving misoprostol), and none with ketoprofen. Two patients were taking more than one NSAID (3%; 2 patients receiving placebo), and another 2 were taking less than the minimum stated dose (3%; 1 patient receiving misoprostol and 1 receiving placebo). The use of NSAIDs was not reported for 18 patients (27%; 6 patients receiving misoprostol and 12 receiving placebo). The mean doses reportedly taken of each of the 10 NSAIDs were similar in the two study groups and exceeded the minimum doses specified.

Thirty-eight patients died during the study; 8 died within 60 days of their last dose of study medication. The numbers of patients who died in the two groups did not significantly differ (17 of 4404 patients receiving misoprostol and 21 of 4439 patients receiving placebo). Three patients who died (1 receiving misoprostol and 2 receiving placebo) had definite gastrointestinal events; however, only 1 died as a direct result of the event. This patient was an 85-year-old woman in the placebo group who had a bleeding duodenal ulcer and died of massive exsanguination.

Discussion

Top Methods Results Discussion Author & Article Info References

Our study is the first randomized, double-blind, placebo-controlled trial using routine clinical criteria to determine the incidence of serious, NSAID-induced upper gastrointestinal complications in arthritic patients, as detected under ordinary clinical practice conditions without routine endoscopy. The study included mostly elderly patients; more women than men; and a group of chronically ill patients, many of whom were receiving multiple medications and who often had many symptoms and problems. The risk of the enrolled patients was higher than the average risk of the total population of patients with arthritis receiving NSAIDs. This higher risk is attributable to the study patients' greater mean age and to the fact that their underlying disease was more likely to be treated with corticosteroids, gold, or other disease-modifying agents, all of which have their own attendant side effects. However, patients with potentially confounding concurrent diseases were excluded from the study; such patients might have an even higher risk than the selected sample.

The incidence of clinically apparent complications was considerably lower than the incidence of mucosal ulcers or erosions previously reported in serial endoscopic studies of patients taking NSAIDs. This suggests that the practice of waiting for symptoms to appear failed to identify some lesions not producing symptoms or other findings. However, although the relevance of endoscopically proven lesions that do not have clinical manifestations has been questioned [23], serious NSAID-induced ulcer complications may not be preceded by signs or symptoms [2, 6, 24, 25]. It is of interest that the 40% to 50% reductions in the various sets of serious clinical gastrointestinal complications observed in the patients receiving misoprostol in our study are similar to the 47% reduction in endoscopically proven gastric or duodenal ulcers in patients with rheumatoid arthritis or osteoarthritis taking diclofenac who received misoprostol or placebo for 6 months in a similarly designed study [16]. The reduction in complications is also similar to the 57% reduction in ulcers seen in our study. Although the incidence of endoscopically visible lesions in patients taking NSAIDs is much greater than the incidence of clinically apparent gastrointestinal complications seen in our study, the relative reductions of incidence associated with misoprostol treatment were similar.

The finding that a history of cardiovascular disease is a significant risk factor, in addition to increasing age and history of peptic ulcer or history of gastrointestinal bleeding, confirms and extends the results of previous smaller studies [26-28] and analyses [29, 30]. The elderly, especially those with heart disease, do not tolerate the stresses of the gastrointestinal complications of ulcer bleeding, perforation, or obstruction, even if they survive the acute gastrointestinal event. Because there was no statistically significant interaction between treatment and the four risk factors, the 40% reduction in definite gastrointestinal complications caused by misoprostol is maintained in patients with any of the four risk factors or combination of risk factors. Our study did not address the possibly confounding problem of infection with Helicobacter pylori, although it has recently been reported [31] that H. pylori does not increase risk for ulcers in arthritic patients receiving NSAIDs. It has previously been noted that NSAID-induced injury to the gastroduodenal mucosa does not increase susceptibility to H. pylori infection [32], that NSAID-associated gastric ulcers do not require H. pylori infection for their development [33], and that in patients with rheumatoid arthritis but no gastrointestinal symptoms while taking NSAIDs for at least 6 months, mucosal injury is not related to H. pylori infection [34].

Our study involved patients with rheumatoid arthritis, and the results cannot necessarily be extrapolated to the larger population of patients with osteoarthritis who are receiving long-term NSAID treatment. Some investigators have found that risks are similar in both types of arthritis [35, 36]; others have inferred that risks are similar by considering together patients with rheumatoid arthritis and osteoarthritis in studies on NSAID-induced ulcer risks [24, 28]. It has also been stated that patients with rheumatoid arthritis had more ulcers than those with osteoarthritis [37]. Reduction by misoprostol of the incidence of NSAID-induced, endoscopically visualized gastroduodenal ulcers or erosions has been shown in patients with either rheumatoid arthritis or osteoarthritis [11, 17, 18] and in healthy volunteers [19]. There is no clear proof that the type of arthritis leading to NSAID use affects the risk for gastrointestinal complications [38]. In the long-term follow-up of patients with rheumatoid arthritis, however, NSAID-induced mortality plays a relatively minor role in the overall reduction of life expectancy [39].

We observed that misoprostol more clearly reduces the risk for NSAID-induced ulceration than it reduces the risk for gastrointestinal bleeding, especially in patients in whom no ulcer is shown. What caused this difference? It has been shown that aspirin-induced melena or hematemesis of unknown cause was less clearly dose-related than gastroduodenal ulceration [40] and that aspirin can significantly reduce serum thromboxane B₂ levels at much lower doses than that needed for inhibition of gastric prostaglandin production and acute mucosal injury [41]. It is unknown whether these findings may be true for the other NSAIDs.

Our findings add to the previously reported concerns about long-term use of NSAIDs in arthritic patients [23, 28, 30] and raise questions about the cost-effectiveness [42-44] of detecting gastrointestinal complications versus attempting to reduce the number or severity of such problems. Clearly, it is neither cost-effective nor even possible to do prospective periodic endoscopic surveillance in all patients receiving long-term NSAID therapy, even in those at increased risk. Attempts at preventing the problem, or reducing it substantially, by concurrent misoprostol administration also raise questions of cost-effectiveness that deserve further thoughtful consideration. Full discussion of this issue is beyond the scope of our report. The "double-edged sword" aspect of NSAID use in patients with chronic arthritis has recently been emphasized in the rheumatology literature [45].

We conclude that misoprostol administration significantly reduces the incidence of NSAID-induced, serious upper gastrointestinal complications, including perforation, obstruction, and bleeding, in older patients with rheumatoid arthritis.