85
CLINICAL PRACTICE
Upper gastrointestinal bleeding in relation to previous use of analgesics and non-steroidal anti-inflammatory drugs
To assess the risk of upper gastrointestinal bleeding associated with the use of individual non-narcotic analgesics and non-steroidal anti-inflammatory drugs (NSAI Ds), a multicentre study of 875 cases of upper gastrointestinal bleeding and 2682 hospital controls was done. With control for confounding factors, the overall odds ratio estimate for aspirin taken at least once during the week before the first symptom was 7·2 (95% confidence interval 5·4-9·6). Non-aspirin NSAIDs associated with upper gastrointestinal bleeding were diclofenac (7·9 [4·3-14·6]), indomethacin (4·9 [2·0-12·2]), naproxen (6·5 [2·2-19·6]), and piroxicam (19·1 [8·2-44·3]). Paracetamol, propyphenazone, and dipyrone did not increase the risk. A previous history of gastrointestinal bleeding or peptic ulcer did not greatly affect odds ratio estimates, which differed according to sex and were higher for younger than for older patients. However, the incidence of upper gastrointestinal bleeding was higher among the
elderly. Introduction The incidence of upper gastrointestinal bleeding from all is 40-50 per 105 per years Endoscopic studies have shown that non-steroidal anti-inflammatory drugs (NSAIDs) cause gastrointestinal erosive and ulcerative damage. Most of these studies have included aspirin as a reference, and most have concluded that other NSAIDs are safer than aspirin.3 Epidemiological controlled studies have evaluated the risk of upper gastrointestinal bleeding associated with the previous use of aspirin4-14 and paracetamol. 10-14 Clinical and epidemiological studies have also shown an association between the use of non-aspirin NSAIDs and upper gastrointestinal bleeding,14--18 but specific estimates for individual drugs are not available because the lower frequency of use of each individual drug, compared with that of aspirin, limits the statistical power of any study designed to quantify those risks. We have carried out a multicentre case-control study with the aim of causes
quantifying the risk of upper gastrointestinal bleeding associated with the use of the non-narcotic analgesics and NSAIDs in Catalonia and Mallorca. We also measured the incidence of upper gastrointestinal bleeding due to peptic ulcer or acute lesions of the gastric mucosa over 2 months in the population of the island of Mallorca. Patients and methods All patients presenting at the participating hospitals (Hospital Vall d’Hebron, Barcelona; Hospital de Sant Pau, Barcelona; and Hospital Son Dureta, Palma de Mallorca) between Feb 1,1987, and Dec 31, 1988, with haematemesis or melaena and a primary diagnosis of acute upper gastrointestinal bleeding were considered for inclusion. During this period, daily contact with the endoscopists was maintained. Endoscopy was done in all but 3 patients (in whom the diagnosis was made at operation). Only patients with at least one diagnosis of benign gastric ulcer, duodenal ulcer, acute lesions of the gastric mucosa, and erosive duodenitis were selected. We excluded patients with other endoscopic diagnoses (ie, oesophageal varices, anastomotic ulcers, gastric carcinoma, Mallory-Weiss syndrome), those with a history of liver cirrhosis or coagulopathy, people on holiday from other countries, illiterate people, and those who could not be reliably interviewed. Of 938 eligible cases, 38 were not interviewed because they died (1-1%), could not be found within 14 days (1-8%), had psychological difficulties (1-0%), or refused (0-3%). Up to four hospital controls were randomly selected for each case, matched according to centre, time from admission (within 2 months), age (within 5 years), and sex. The controls were selected from patients admitted with acute clinical disorders thought to be unrelated to intake of analgesics or NSAIDs. Their diagnoses on admission were non-alcohol-related trauma (1673), acute appendicitis (382), other abdominal emergencies (115), and elective surgery for non-painful disorders, mostly femoral or inguinal hernias (539). Exclusion criteria were the same as for the cases. Of 2719 eligible controls, 10 (0-4%) refused to be interviewed. A structured questionnaire was administered by specially trained nurses (two) and physicians (two) to the potential cases and controls
ADDRESSES: Unit of Clinical Pharmacology, Department of Pharmacology, Universitat Autònoma de Barcelona, CS Vail d’Hebron, Barcelona, Spain (Prof J-R Laporte, MD, X Carné, MD, X Vidal, MD, V Moreno, MD, J. Juan, RN) Correspondence to Prof J-R. Laporte, Servei de Farmacologia Clinica, CS Vall d’Hebron, 08035
Barcelona, Spain.
86
TABLE I-AGE AND SEX DISTRIBUTIONS OF PATIENTS ACCORDING TO DIAGNOSTIC CATEGORY
ALGM =acute lesions of gastric
mucosa
possible after admission, and always within 14 days. Before the interview, we briefly explained the aims of the study and requested consent. The interview covered general demographic information; clinical course leading to the present hospital admission; previous history, with special emphasis on gastrointestinal (history of ulcer, dyspepsia, or gastrointestinal bleeding), rheumatic, and other painful disorders; alcohol, tobacco, and coffee intake; and drug history, on a daily basis for the 28 days before admission, including information about the dose taken. (Detailed information on the dose-risk relation will be the subject of a separate paper.) To ensure that drug histories were as complete as possible, after an open question about previous use of drugs the patients were questioned about a list of common symptoms often prompting use of non-narcotic analgesics and NSAIDs, then they were asked to recall the trade names of the most popular non-narcotic analgesics, NSAIDs, and drugs used for dyspepsia and ulcer treatment that were read to them. The list was based on marketing data and included names that, in all, accounted for over 90% of sales. In all interviews relatives were allowed to accompany the patient and aid him or her in the recall exercise, but only information confirmed by the patient was collected. From each patient’s clinical history, but with no knowledge of previous drug use, two investigators defmed the index day. For the cases, the index day was the day on which haematemesis, coffee grounds vomitus, melaena, or passage of bloody stools appeared. Among controls, the index day was the day on which the accident occurred, on which symptoms were initially detected or the day of admission for elective surgery. Odds ratio estimates and their 95% confidence intervals were calculated by means of a conditional logistic model. To control simultaneously for potential confounding factors, the following variables were included in the model: number of different drugs as soon as
used by each patient during the 7 days before the index day; history of peptic ulcer expressed in three mutually excluding categories (ie, subjective symptoms, history of confirmed peptic ulcer diagnosis, and history of gastrointestinal bleeding); use of caffeine-containing beverages; smoking habit; alcohol consumption; and intake of
aspirin, paracetamol, propyphenazone, dipyrone, diclofenac, indomethacin, naproxen, piroxicam, other NSAIDs, antacids, or other anti-ulcer drugs. The model also included an interaction term of aspirin use plus paracetamol use, and was based on the investigators’ criteria and statistical considerations rather than on automatic procedures. Age and sex were included in the model only when the analysis was unmatched. Odds ratios were calculated for all patients and for gastric and duodenal patients separately for analgesic use during an exposure timespan window of 7 days before the index day. The categories of gastric ulcer and acute lesions of the gastric mucosa were combined because there were few patients with the latter disorder, and the odds ratio estimates for two disorders did not differ. The reference category was made up of non-exposed cases and controls for this 7 day period. Statistical significance tests for comparisons of multivariate models were based on the difference between log likelihoods. With the aim of ascertaining all cases of upper gastrointestinal bleeding which met the study criteria, all the endoscopy units on the island of Mallorca were contacted daily from Nov 1 to Dec 31, 1988. The population of the island was 543 841.
Results The 900 cases interviewed presented a principal bleeding lesion of which the probable cause was a duodenal ulcer in 465, a gastric ulcer in 313, acute lesions of the gastric mucosa in 69, erosive duodenitis in 28, or could not be determined in 25 patients (more than one lesion on endoscopy; mixed lesions group, table 1). 25 cases and 27 controls were not included in the case-control analysis because they were judged unreliable by the interviewer, leaving 875 cases and 2682 controls for the overall case-control analysis. Patients were grouped in three categories according to their main bleeding site and subgroup case-control analyses were done for duodenal and gastric patients. Patients with gastric bleeding were older than those with duodenal bleeding (mean [SD] 63-1 [15’0] vs 55-9 [174] years; median 64 vs 56 years). The ratio of men to women was fairly constant across the age categories among duodenal patients (4-2); in gastric patients it was 4-8 among those under 60 and 0-9 among those older than 60. Women were generally older than men (66’9[14-1]vs 56-1[16-6] years;
TABLE II-USE OF NON-NARCOTIC ANALGESICS AND NSAIDS IN WEEK BEFORE INDEX DAY
87
TABLE III-USE OF NSAIDS ACCORDING TO AGE AND SEX
median 68 vs 56 years); 69% of women and 38% of men 60 years old (table l). Cases had used more drugs than controls during the week before the index day (2-14 [1’61] vs 1-25 [1.38]; median 2 vs 1); the difference was mainly accounted for by NSAIDs, antacids, and H2blockers (mean 0-98 vs 0-25). Among the controls 15-7% of those with trauma, 18-4% of those with acute disorders, and 13-2% of those undergoing elective surgery had used non-narcotic analgesics and NSAIDs during the week before the index day (distributions adjusted for age and sex). Table n shows the odds ratio estimates for the most commonly used non-narcotic analgesics and NSAIDs of all patients with upper gastrointestinal bleeding. The 95% CIs of the odds ratio estimates for paracetamol, dipyrone, and propyphenazone included 1 -0 in the overall as well as in all subgroup analyses. For aspirin and other NSAIDs, the odds ratio estimates ranged from 4-9 (indomethacin) to 19-1 (piroxicam). Diclofenac and naproxen showed nonsignificant odds ratio estimates for duodenal bleeding, but these were based on small numbers of exposed patients. Of the 875 cases, 480 (55%) had no history of confirmed peptic ulcer or of gastrointestinal bleeding compared with 2334 (87%) of 2682 controls. Since this background might lead to differential use of non-narcotic analgesics and NSAIDs, the odds ratio estimates for each drug were also calculated for this subpopulation. The results were similar were over
(table II). In
our model a previous history of peptic ulcer or gastrointestinal bleeding was strongly associated with the risk of upper gastrointestinal bleeding; the odds ratio was 2-6 (95% CI 1-9-3-5) in patients with a previous history of subjective symptoms, 5-5 (2-8-11-0) in those with a history of confirmed peptic ulcer diagnosis, and 14-5 (7-5-28-2) in those with a history of gastrointestinal bleeding. However, an interaction between the previous history of confirmed peptic ulcer or gastrointestinal bleeding and use of NSAIDs
excluded, and the odds ratio estimate for all NSAIDs differ between patients with and without such a history (7-4 [2’4-23’0] vs 8-0 [3-8-16-6]). The risk of gastrointestinal bleeding from any site associated with the use of aspirin did not vary greatly with age or sex (table III), but for other NSAIDs the risk of gastric bleeding was higher in patients under 60 years than in older patients (78-7 [19-1-325] vs 7-6 [4-0-14-6]; p
CLINICAL PRACTICE
Upper gastrointestinal bleeding in relation to previous use of analgesics and non-steroidal anti-inflammatory drugs
To assess the risk of upper gastrointestinal bleeding associated with the use of individual non-narcotic analgesics and non-steroidal anti-inflammatory drugs (NSAI Ds), a multicentre study of 875 cases of upper gastrointestinal bleeding and 2682 hospital controls was done. With control for confounding factors, the overall odds ratio estimate for aspirin taken at least once during the week before the first symptom was 7·2 (95% confidence interval 5·4-9·6). Non-aspirin NSAIDs associated with upper gastrointestinal bleeding were diclofenac (7·9 [4·3-14·6]), indomethacin (4·9 [2·0-12·2]), naproxen (6·5 [2·2-19·6]), and piroxicam (19·1 [8·2-44·3]). Paracetamol, propyphenazone, and dipyrone did not increase the risk. A previous history of gastrointestinal bleeding or peptic ulcer did not greatly affect odds ratio estimates, which differed according to sex and were higher for younger than for older patients. However, the incidence of upper gastrointestinal bleeding was higher among the
elderly. Introduction The incidence of upper gastrointestinal bleeding from all is 40-50 per 105 per years Endoscopic studies have shown that non-steroidal anti-inflammatory drugs (NSAIDs) cause gastrointestinal erosive and ulcerative damage. Most of these studies have included aspirin as a reference, and most have concluded that other NSAIDs are safer than aspirin.3 Epidemiological controlled studies have evaluated the risk of upper gastrointestinal bleeding associated with the previous use of aspirin4-14 and paracetamol. 10-14 Clinical and epidemiological studies have also shown an association between the use of non-aspirin NSAIDs and upper gastrointestinal bleeding,14--18 but specific estimates for individual drugs are not available because the lower frequency of use of each individual drug, compared with that of aspirin, limits the statistical power of any study designed to quantify those risks. We have carried out a multicentre case-control study with the aim of causes
quantifying the risk of upper gastrointestinal bleeding associated with the use of the non-narcotic analgesics and NSAIDs in Catalonia and Mallorca. We also measured the incidence of upper gastrointestinal bleeding due to peptic ulcer or acute lesions of the gastric mucosa over 2 months in the population of the island of Mallorca. Patients and methods All patients presenting at the participating hospitals (Hospital Vall d’Hebron, Barcelona; Hospital de Sant Pau, Barcelona; and Hospital Son Dureta, Palma de Mallorca) between Feb 1,1987, and Dec 31, 1988, with haematemesis or melaena and a primary diagnosis of acute upper gastrointestinal bleeding were considered for inclusion. During this period, daily contact with the endoscopists was maintained. Endoscopy was done in all but 3 patients (in whom the diagnosis was made at operation). Only patients with at least one diagnosis of benign gastric ulcer, duodenal ulcer, acute lesions of the gastric mucosa, and erosive duodenitis were selected. We excluded patients with other endoscopic diagnoses (ie, oesophageal varices, anastomotic ulcers, gastric carcinoma, Mallory-Weiss syndrome), those with a history of liver cirrhosis or coagulopathy, people on holiday from other countries, illiterate people, and those who could not be reliably interviewed. Of 938 eligible cases, 38 were not interviewed because they died (1-1%), could not be found within 14 days (1-8%), had psychological difficulties (1-0%), or refused (0-3%). Up to four hospital controls were randomly selected for each case, matched according to centre, time from admission (within 2 months), age (within 5 years), and sex. The controls were selected from patients admitted with acute clinical disorders thought to be unrelated to intake of analgesics or NSAIDs. Their diagnoses on admission were non-alcohol-related trauma (1673), acute appendicitis (382), other abdominal emergencies (115), and elective surgery for non-painful disorders, mostly femoral or inguinal hernias (539). Exclusion criteria were the same as for the cases. Of 2719 eligible controls, 10 (0-4%) refused to be interviewed. A structured questionnaire was administered by specially trained nurses (two) and physicians (two) to the potential cases and controls
ADDRESSES: Unit of Clinical Pharmacology, Department of Pharmacology, Universitat Autònoma de Barcelona, CS Vail d’Hebron, Barcelona, Spain (Prof J-R Laporte, MD, X Carné, MD, X Vidal, MD, V Moreno, MD, J. Juan, RN) Correspondence to Prof J-R. Laporte, Servei de Farmacologia Clinica, CS Vall d’Hebron, 08035
Barcelona, Spain.
86
TABLE I-AGE AND SEX DISTRIBUTIONS OF PATIENTS ACCORDING TO DIAGNOSTIC CATEGORY
ALGM =acute lesions of gastric
mucosa
possible after admission, and always within 14 days. Before the interview, we briefly explained the aims of the study and requested consent. The interview covered general demographic information; clinical course leading to the present hospital admission; previous history, with special emphasis on gastrointestinal (history of ulcer, dyspepsia, or gastrointestinal bleeding), rheumatic, and other painful disorders; alcohol, tobacco, and coffee intake; and drug history, on a daily basis for the 28 days before admission, including information about the dose taken. (Detailed information on the dose-risk relation will be the subject of a separate paper.) To ensure that drug histories were as complete as possible, after an open question about previous use of drugs the patients were questioned about a list of common symptoms often prompting use of non-narcotic analgesics and NSAIDs, then they were asked to recall the trade names of the most popular non-narcotic analgesics, NSAIDs, and drugs used for dyspepsia and ulcer treatment that were read to them. The list was based on marketing data and included names that, in all, accounted for over 90% of sales. In all interviews relatives were allowed to accompany the patient and aid him or her in the recall exercise, but only information confirmed by the patient was collected. From each patient’s clinical history, but with no knowledge of previous drug use, two investigators defmed the index day. For the cases, the index day was the day on which haematemesis, coffee grounds vomitus, melaena, or passage of bloody stools appeared. Among controls, the index day was the day on which the accident occurred, on which symptoms were initially detected or the day of admission for elective surgery. Odds ratio estimates and their 95% confidence intervals were calculated by means of a conditional logistic model. To control simultaneously for potential confounding factors, the following variables were included in the model: number of different drugs as soon as
used by each patient during the 7 days before the index day; history of peptic ulcer expressed in three mutually excluding categories (ie, subjective symptoms, history of confirmed peptic ulcer diagnosis, and history of gastrointestinal bleeding); use of caffeine-containing beverages; smoking habit; alcohol consumption; and intake of
aspirin, paracetamol, propyphenazone, dipyrone, diclofenac, indomethacin, naproxen, piroxicam, other NSAIDs, antacids, or other anti-ulcer drugs. The model also included an interaction term of aspirin use plus paracetamol use, and was based on the investigators’ criteria and statistical considerations rather than on automatic procedures. Age and sex were included in the model only when the analysis was unmatched. Odds ratios were calculated for all patients and for gastric and duodenal patients separately for analgesic use during an exposure timespan window of 7 days before the index day. The categories of gastric ulcer and acute lesions of the gastric mucosa were combined because there were few patients with the latter disorder, and the odds ratio estimates for two disorders did not differ. The reference category was made up of non-exposed cases and controls for this 7 day period. Statistical significance tests for comparisons of multivariate models were based on the difference between log likelihoods. With the aim of ascertaining all cases of upper gastrointestinal bleeding which met the study criteria, all the endoscopy units on the island of Mallorca were contacted daily from Nov 1 to Dec 31, 1988. The population of the island was 543 841.
Results The 900 cases interviewed presented a principal bleeding lesion of which the probable cause was a duodenal ulcer in 465, a gastric ulcer in 313, acute lesions of the gastric mucosa in 69, erosive duodenitis in 28, or could not be determined in 25 patients (more than one lesion on endoscopy; mixed lesions group, table 1). 25 cases and 27 controls were not included in the case-control analysis because they were judged unreliable by the interviewer, leaving 875 cases and 2682 controls for the overall case-control analysis. Patients were grouped in three categories according to their main bleeding site and subgroup case-control analyses were done for duodenal and gastric patients. Patients with gastric bleeding were older than those with duodenal bleeding (mean [SD] 63-1 [15’0] vs 55-9 [174] years; median 64 vs 56 years). The ratio of men to women was fairly constant across the age categories among duodenal patients (4-2); in gastric patients it was 4-8 among those under 60 and 0-9 among those older than 60. Women were generally older than men (66’9[14-1]vs 56-1[16-6] years;
TABLE II-USE OF NON-NARCOTIC ANALGESICS AND NSAIDS IN WEEK BEFORE INDEX DAY
87
TABLE III-USE OF NSAIDS ACCORDING TO AGE AND SEX
median 68 vs 56 years); 69% of women and 38% of men 60 years old (table l). Cases had used more drugs than controls during the week before the index day (2-14 [1’61] vs 1-25 [1.38]; median 2 vs 1); the difference was mainly accounted for by NSAIDs, antacids, and H2blockers (mean 0-98 vs 0-25). Among the controls 15-7% of those with trauma, 18-4% of those with acute disorders, and 13-2% of those undergoing elective surgery had used non-narcotic analgesics and NSAIDs during the week before the index day (distributions adjusted for age and sex). Table n shows the odds ratio estimates for the most commonly used non-narcotic analgesics and NSAIDs of all patients with upper gastrointestinal bleeding. The 95% CIs of the odds ratio estimates for paracetamol, dipyrone, and propyphenazone included 1 -0 in the overall as well as in all subgroup analyses. For aspirin and other NSAIDs, the odds ratio estimates ranged from 4-9 (indomethacin) to 19-1 (piroxicam). Diclofenac and naproxen showed nonsignificant odds ratio estimates for duodenal bleeding, but these were based on small numbers of exposed patients. Of the 875 cases, 480 (55%) had no history of confirmed peptic ulcer or of gastrointestinal bleeding compared with 2334 (87%) of 2682 controls. Since this background might lead to differential use of non-narcotic analgesics and NSAIDs, the odds ratio estimates for each drug were also calculated for this subpopulation. The results were similar were over
(table II). In
our model a previous history of peptic ulcer or gastrointestinal bleeding was strongly associated with the risk of upper gastrointestinal bleeding; the odds ratio was 2-6 (95% CI 1-9-3-5) in patients with a previous history of subjective symptoms, 5-5 (2-8-11-0) in those with a history of confirmed peptic ulcer diagnosis, and 14-5 (7-5-28-2) in those with a history of gastrointestinal bleeding. However, an interaction between the previous history of confirmed peptic ulcer or gastrointestinal bleeding and use of NSAIDs
excluded, and the odds ratio estimate for all NSAIDs differ between patients with and without such a history (7-4 [2’4-23’0] vs 8-0 [3-8-16-6]). The risk of gastrointestinal bleeding from any site associated with the use of aspirin did not vary greatly with age or sex (table III), but for other NSAIDs the risk of gastric bleeding was higher in patients under 60 years than in older patients (78-7 [19-1-325] vs 7-6 [4-0-14-6]; p
