Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Current intake of nonsteroidal anti-inflammatory drugs (NSAIDs), studied as a class, is associated with a three- to fivefold increase in upper gastrointestinal complications (UGICs), bleeding or perforation. 1–3 There are extensive epidemiologic data evaluating the risk of UGICs associated with some individual nonaspirin NSAIDs. 4 There is, however, no published epidemiologic study reporting estimates of relative risk (RR) of UGICs for the several NSAIDs that were introduced in the market in the 1990s that claim a better gastrointestinal safety than traditional NSAIDs. 5 Similarly, few studies have evaluated the role of acetaminophen. Among the various preventive strategies of UGICs, use of acid-suppressing drugs in the general population and misoprostol in NSAID users is well accepted. 6–8 A recent study has also found that use of nitrates was associated with a decreased risk of upper gastrointestinal bleeding. 9
We conducted a population-based nested case-control study in the United Kingdom to estimate the RR of UGICs associated with (1) acetaminophen and specific NSAIDs and (2) H2 receptor antagonists, omeprazole, misoprostol, and nitrates.
Subjects and Methods
Patients and Study Design
We conducted a nested case-control study using the United Kingdom General Practice Research Database during the period between April 1993 and October 1998. This study was an extension of a previous one in which the data collection ended in February 1993. 10 The General Practice Research Database is a population-based database in the United Kingdom in which general practitioners store, in office computers, clinical information on their patients including demographics, diagnoses and comments, referral information, and records of all prescriptions issued by them. 11,12 The study population comprised persons 40–79 years of age who had been enrolled at least 2 years with the general practitioner and who were free of cancer, esophageal varices, Mallory-Weiss disease, liver disease, coagulopathies, and alcohol-related disorders at start date. All study members were traced until they met a case definition criterion or one of the above-mentioned exclusion criteria, until they died, or until October 1998, whichever came first.
We identified patients with codes for UGICs and manually reviewed the demographic data and clinical information in their computerized patient profiles. Subjects with any of the exclusion criteria mentioned above in the 2 months after the date of case detection (index date), and subjects with the source of the bleed/perforation in the esophagus or lower gastrointestinal tract were not considered cases. We considered a patient to be a case of UGIC when no exclusion criterion was found, the subject had not been discharged from the hospital in the previous 2 weeks, the specific site of the bleed/perforation was located in the stomach or duodenum or the clinical diagnosis was peptic ulcer, and the patient had been referred to a consultant or admitted to hospital. We classified cases, according to the source of the bleed/perforation, into gastric and duodenal categories. To validate the case status ascertained after review of the computerized information, we sent the general practitioners a questionnaire and a request to send all the information related to that event for a random sample of 100 patients. We received information for 99 patients, with only 1 patient not confirmed as a case of UGIC.
Controls were frequency-matched to cases by age (interval of 1 year) and sex. We randomly selected and an index date from the eligible person-time so that the likelihood of being selected as a control was proportional to the person-time at risk. 13 We applied the same exclusion criteria to the controls as we did for the cases.
We estimated separately the RRs of UGICs associated with acetaminophen, nonaspirin NSAIDs, H2 receptor antagonists (cimetidine, famotidine, nizatidine, and ranitidine), omeprazole, misoprostol, and nitrates (glyceril trinitrate, isosorbide dinitrate, and isosorbide mononitrate). Exposure to these drugs was categorized as “current,” when the supply of the most recent prescription lasted until the index date or ended in the 30 days before the index date; “recent,” when it ended 31–90 days before the index date; “past,” when it ended 91–180 days before the index date; and “non-use,” when there was no recorded use in the 6 months before the index date. We evaluated duration of use adding the periods of “consecutive” prescriptions, defined as an interval of less than 2 months between two prescriptions. Among current acetaminophen users, we calculated the dose-response relation using the following five dose categories, in mg: ≤1,000; 1,001–1,999; 2,000; 2,001–3,999; and ≥4,000. The dose 2 gm stands on its own because it was a commonly used daily dose.
We further divided exposure to nonaspirin NSAID use in the last month into “current,” when the supply of the most recent prescription lasted until index date or ended in the week before, and “intermediate,” when it ended 7–30 days before the index date. Current users were subsequently divided into “current single users” and “current multiple users.” The latter category included patients who received prescriptions for different NSAIDs with their respective supply ending within the month before the index date. Among current single users, we calculated the RR for individual NSAIDs and the dose-response relation using two dose categories: low/medium and high. 10 We classified NSAIDs in two groups according to plasma half-life of the drugs: less than 12 hours and equal or greater than 12 hours. 14 The short-half-life group included aceclofenac, acemetacin, diclofenac, etodolac, fenbufen, fenoprofen, flurbiprofen, ibuprofen, indomethacin, ketoprofen, mefenamic acid, and tiaprofenic acid. The long-half-life group included apazone, meloxicam, nabumetone, naproxen, piroxicam, sulindac, and tenoxicam. Oral presentations of NSAIDs were separated into slow-release form and other forms.
The analysis included 2,105 cases and 11,500 controls. We used unconditional logistic regression to compute estimates of RR and 95% confidence intervals (95% CIs) of UGICs associated with current use of each drug group of interest. We chose non-users of each respective drug group as the reference category. All estimates of RR were adjusted for age; sex; calendar year; smoking; antecedents of upper gastrointestinal disorders; and use of steroids, anticoagulants, nonaspirin NSAIDs, acetaminophen, H2 receptor antagonists, omeprazole, misoprostol, and aspirin.
Overall, use of acetaminophen was associated with a negligible elevated risk of UGICs (RR = 1.3; 95% CI = 1.1–1.5) (Table 1). Pain related to osteoarticular conditions was the indication in 80% of users. We found an increased risk of UGICs among current users of acetaminophen starting at a daily dose of 2 gm and greater, whereas smaller doses did not carry an increased risk. The RR was 3.6 (95% CI = 2.6–5.1) among acetaminophen users of more than 2 gm daily. When we restricted the analysis to individuals who never received a prescription for NSAIDs and who had no recorded antecedents of upper gastrointestinal disorders (including dyspepsia), the corresponding RR for acetaminophen use at doses greater than 2 gm was 5.7 (95% CI = 2.0–16.4). The risk of dose-response increase among acetaminophen users was independent of treatment duration.
Nonsteroidal Anti-Inflammatory Drugs Overall
Current intake of NSAIDs increased the risk of UGICs 4.1 times (95% CI = 3.6–4.8) (Table 2). The risk of UGICs dropped quickly once treatment was stopped. The risk of developing UGICs increased with greater NSAID daily dose. In users of medium daily dose or lower, the RR was 2.4 (95% CI = 1.9–3.1), whereas in users of high daily dose the RR was 4.9 (95% CI = 4.1–5.8). We found a similar RR among new users and among those already on chronic therapy for several months. We compared the risk of UGICs between NSAID users with osteoarthritis, rheumatoid arthritis, or pain-related disorders as indication and found no important difference between indications. The increase in risk was similar for fatal and nonfatal cases of UGICs and did not differ much with the primary site of the lesion. The estimate of RR associated with NSAID use was greater for perforation lesions (RR = 6.6; 95% CI = 4.8–9.0) than for bleeding lesions (RR = 3.7; 95% CI = 3.2–4.3), however.
The risk of UGICs among current NSAID users was multiplied by a factor between 4 and 5 above the baseline rates across all age categories and in both sexes. Figure 1 shows the absolute increased risk associated with NSAID use according to severity of antecedents of upper gastrointestinal disorders. Patients who have had a previous ulcer complication (bleeding/perforation) presented the greatest absolute risk of UGICs when taking NSAIDs, equivalent to an incidence rate between 20 and 30 per 1,000 person-years (assuming a baseline rate of 1 case per 1,000 person-years observed in one of our previous studies). 15
Specific Nonsteroidal Anti-Inflammatory Drugs
The increased risk varied according to the pharmacokinetics (Table 3). NSAIDs with a plasma half-life less than 12 hours (excluding slow-release formulations) had an RR of 3.1 (95% CI = 2.5–3.8), NSAIDs with a half-life greater than 12 hours had an RR of 4.5 (95% CI = 3.5–5.9), and NSAIDs in slow-release formulation presented an RR of 5.4 (95% CI = 4.0–7.1). These differences remained even after stratification of NSAIDs by daily dose.
Table 4 shows estimates of RR for all individual NSAIDs with five or more exposed controls. RRs ranged from 1.1 (95% CI = 0.2–5.1) for fenbufen to 10.3 among users of apazone (95% CI = 3.5–30.0). In the stratified analysis by daily dose, all individual NSAIDs presented an RR less than 4 when administered at low/medium doses. Likewise, all individual NSAIDs presented a greater RR with increasing dose.
Interaction between Acetaminophen and Nonsteroidal Anti-Inflammatory Drugs
Compared with non-users of drugs in either group, the RR for concurrent users of NSAIDs and acetaminophen (2 gm and more) was 13.2 (95% CI = 9.2–18.9). When we restricted the analysis to patients with either osteoarthritis or rheumatoid arthritis, the estimate among concurrent users hardly changed (RR = 14.3; 95% CI = 9.3–22.1). We found essentially no interaction with doses of acetaminophen lower than 2 gm (RR = 4.1; 95% CI = 3.1–5.5).
Table 5 shows the estimates of RR associated with various gastroprotective agents and with nitrates. RRs were 0.6 (95% CI = 0.4–0.9) for omeprazole and 0.6 (95% CI = 0.4–1.0) for misoprostol. Use of H2 receptor antagonists was not associated with a reduced risk of UGICs (RR = 1.4; 95% CI = 1.2–1.8). In the stratified analysis by NSAID use, misoprostol was associated with a reduced risk among NSAID users (there was hardly any use of misoprostol in NSAID non-users), omeprazole showed a similar level of protection among users and non-users of NSAIDs, and H2 receptor antagonists showed no risk reduction in either group. RRs for users of nitrates were 0.9 (95% CI = 0.7–1.1) overall, 1.0 (95% CI = 0.8–1.3) among non-users of NSAIDs, and 0.6 (95% CI = 0.4–1.0) among NSAID users.
In this study, acetaminophen was associated with an increased risk of upper gastrointestinal bleeding/perforation (RR = 3.6) only when taken at daily doses above 2,000 mg. Current intake of NSAIDs increased the risk of having a new episode of UGIC by a factor of 4. The increased risk depended on the doses taken by the patient; all NSAIDs had a greater RR of UGICs at higher doses. After adjusting for daily dose, the estimates of RR for all individual NSAIDs, except those of apazone, tended to converge. Nonetheless, even after allowing for daily dose, NSAIDs with a long half-life or slow-release formulation were associated with a slightly greater RR than NSAIDs with a short half-life.
We found five previously published epidemiologic studies 16–20 that included estimates of RR of UGICs associated with any use of acetaminophen; the estimates ranged from 0.2 through 1.9, with a summary estimate of 1.4 (95% CI = 1.0–2.0). Only one study reported an increased risk of UGICs associated with use of acetaminophen at doses greater than 1,000 mg (RR = 2.6). 19 A recent abstract reported a dose-response relation of acetaminophen with adverse gastrointestinal events. 22In vitro analyses have shown that acetaminophen is a weak nonselective inhibitor of both isoforms of cyclooxygenase in human whole blood assays. 23,24 Our data show a substantial interaction when taking NSAIDs and high-dose acetaminophen (2 gm or more) together. This finding may be attributable in part to an augmentation of cyclooxygenase inhibition, of the same type as found among users of high-dose NSAIDs or multiple NSAIDs, among whom an increased risk is also observed. Acetaminophen (paracetamol) has been traditionally considered a safer alternative than NSAIDs for the pharmacologic treatment of various osteoarticular conditions. 25 The apparent paradox between our findings and those of most of the clinical literature disappears when one takes into account that the majority of these studies included doses lower than 2,000 mg per day in healthy subjects. In addition, endoscopic gastroduodenal injuries among short-term users of NSAIDs are not valid predictors of major clinical gastrointestinal complications inasmuch as the smaller degree of endoscopic injuries observed with enteric-coated aspirin does not translate into a reduced risk of UGICs. 26 An alternative explanation for these findings might be a preferential prescription of acetaminophen to patients at high risk of UGICs. Although we adjusted for antecedents of upper gastrointestinal disorders (including dyspepsia), gastroprotective cotherapy, and several other factors associated with the risk of UGICs, there is potential for incomplete control of confounding. Nevertheless, two lines of evidence argue against this possibility: the clear dose effect found (physicians would be unlikely to prescribe higher doses to patients at greatest risk) and the similar overall results and enhanced dose effect found upon restriction of the analyses to persons without known major risk factors (that is, with no prior upper gastrointestinal disorders and no NSAID use).
In a previous study, Henry et al27 suggested that the pharmacokinetics of NSAIDs contribute in part to the mechanism underlying the risk of serious upper gastrointestinal toxicity. Our data provide some evidence for this hypothesis; NSAIDs with a long half-life or NSAIDs with a short half-life in slow-release formulations were associated with a greater risk of UGICs than were NSAIDs with a short half-life in regular formulations. A common feature of the former type of NSAIDs is the sustained plasma levels they reached either by a slow elimination in the first case or a slow absorption in the second, assuring a longer presence of NSAID concentrations in the target tissue and consequently a more constant cyclooxygenase inhibition.
A recent meta-analysis reported the aggregate estimates for some individual NSAIDs found in the epidemiologic literature. 4 We report here for the first time RRs for NSAIDs such as etodolac, meloxicam, and nabumetone. It has been proposed that these NSAIDs might have a better gastrointestinal safety profile than traditional NSAIDs. 5,28 The alleged mechanisms predicting a lesser risk of UGICs varied according to the specific NSAIDs, among others factors use of a nonacidic prodrug or some degree of selectivity toward cyclooxygenase-2. 29–31 In our study, the estimates for these drugs are compatible with that of the average NSAID effect. Data for these three NSAIDs were relatively scarce, however, resulting in wide confidence intervals, and consequently one cannot exclude with complete certainty that these NSAIDs present an RR substantially lower than 4 (average RR of NSAID class). As in our previous study, with data collected before 1993, 10 apazone was the only NSAID that at the daily doses used (600–1200 mg) was associated with an RR of UGIC distinctively greater than the average NSAID RR.
Confounding by indication is unavoidable in observational research when the studied drug is prescribed for the event of interest, as is the case for gastroprotective agents and UGICs. Users of gastroprotective agents will likely have a higher risk of UGICs, given that gastroprotection was exactly the reason for using the medication. In the current study, control for known risk factors for both use of gastroprotective agents and UGICs (that is, antecedents of gastrointestinal disorders and use of gastrotoxic medications) reduced, but probably did not completely eliminate, this bias. Despite the potential residual confounding, omeprazol and misoprostol were associated with a lower risk of UGICs. We could not find a risk reduction associated with H2 receptor antagonist use. Use of nitrates has recently been associated with a lower risk of upper gastrointestinal bleeding, especially in the subgroup of patients taking NSAIDs. 9 We did not find any reduced risk of UGICs among non-users of NSAIDs. The results among NSAID users, however, are compatible with some protection.
A potential limitation of all studies using computerized prescription data is the underascertainment of over-the-counter drug use. Misclassification of NSAID exposure is expected to be small, as only ibuprofen and, only recently, naproxen could be purchased over the counter. We were able to indirectly assess the magnitude of the misclassification using data from a recent study in the United Kingdom in a similar population, in which the authors interviewed patients to obtain information on all prescribed and self-administered drug intake. 20 They reported a prevalence of 18% in the previous 3 months in the control series, whereas such prevalence was 14% in our study; 25% of their controls were more than 80 years of age, whereas in our study the upper age limit was 79 years. Because the use of NSAIDs is greater among elderly, it is apparent that no major underrecording of NSAID use was present in our data, after allowing for the different age distributions. In the same study, 20 the authors reported a prevalence of 20% for acetaminophen; the corresponding figure in our study was 14%. We did a sensitivity analysis to quantify the impact of nonrecorded drug use. 32 With false negative probabilities beyond 30% (conservative estimate obtained from data presented above), the net impact of nondifferential under-recorded use of NSAIDs or acetaminophen would have been a small underestimation of the excess risk. Moreover, although misclassification of exposures collected prospectively is usually close to nondifferential among cases and controls, we also examined the effects of differential misclassification. Even unrealistically high differential under-recording could not cancel the elevated risks of UGICs found for NSAIDs and high-dose acetaminophen. The negligible impact of missing use of over-the-counter anti-inflammatory drug had been previously reported. 33,34
In conclusion, our findings indicate that acetaminophen used at high dose is associated with an increased risk of UGICs, similar to the one obtained for traditional NSAIDs. We found some evidence that NSAIDs with long plasma half-life and in slow-release formulation carry an additional greater risk over that of NSAIDs with short plasma half-life. Nevertheless, dose is the most important NSAID characteristic in predicting the risk of UGICs. Finally, we confirmed that omeprazole is an effective strategy to reduce the risk of UGICs both in NSAID users and in non-users. Misoprostol was similarly effective among NSAID users, and our data were compatible with a reduced risk of UGICs among users simultaneously taking NSAIDs and nitrates.
We thank the Boston Collaborative Drug Surveillance Program for providing access to the General Practice Research Database, and the general practitioners for their excellent collaboration.
1. Hawkey CJ. Non-steroidal anti-inflammatory drugs and peptic ulcers: facts and figures multiply, but do they add up? BMJ 1990; 300: 278–284.
2. Gabriel SE, Jaakkimaunen L, Bombardier C. Risk for serious gastrointestinal complications related to use of nonsteroidal anti-inflammatory drugs: a meta-analysis. Ann Int Med 1991; 115: 787–796.
3. Bollini P, García Rodríguez LA, Pérez Gutthann S, Walker AM. The impact of research quality and study design on epidemiologic estimates of the effect of NSAIDs on upper gastrointestinal pathology. Arch Int Med 1992; 152: 1289–1295.
4. Henry D, Lim LLY, García Rodríguez LA, Pérez Gutthann S, Carson JL, Griffin M, Savage R, Logan R, Moride Y, Hawkey C, Hill S, Fries JT. Variability in risk of gastrointestinal complications with individual NSAIDs: results of a collaborative meta-analysis. Br Med J 1996; 312: 1563–1566.
5. Wolfe MM, Lichtenstein DR, Singh G. Gastrointestinal toxicity of nonsteroidal antiinflammatory drugs. N Engl J Med 1999; 340: 1888–1899.
6. Hawkey CJ, Karrasch JA, Szczepanski L, Walker DG, Barkun A, Swannell AJ, Yeomans ND. Omeprazole compared with misoprostol for ulcers associated with nonsteroidal antiinflammatory drugs. Omeprazole versus Misoprostol for NSAID-induced Ulcer Management (OMNIUM) Study Group. N Engl J Med 1998; 338: 727–734.
7. Yeomans ND, Tulassay Z, Juhasz L, Racz I, Howard JM, van Rensburg CJ, Swannell AJ, Hawkey CJ. A comparison of omeprazole with ranitidine for ulcers associated with nonsteroidal antiinflammatory drugs. Acid Suppression Trial: Ranitidine versus Omeprazole for NSAID-Associated Ulcer Treatment (ASTRONAUT) Study Group. N Engl J Med 1998; 338: 719–726.
8. Silverstein FE, Graham DY, Senior JR, Davies HW, Struthers BJ, Bittman RM, Geis GS. Misoprostol reduces serious gastrointestinal complication in patients with rheumatoid arthritis receiving non-steroidal anti-inflammatory drugs. Ann Intern Med 1995; 123: 241–249.
9. Lanas A, Bajador E, Serrano P, Fuentes J, Carreño S, Guardia J, Sanz M, Montoro M, Sainz R. Nitrovasodilators, low-dose aspirin, other nonsteroidal antiinflammatory drugs, and the risk of upper gastrointestinal bleeding. N Engl J Med 200; 343: 843–849.
10. García Rodríguez LA, Jick H. Risk of upper gastrointestinal bleeding and perforation associated with individual nonsteroidal anti-inflammatory drugs. Lancet 1994; 343: 769–772.
11. Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. Br Med J 1991; 303: 766–768.
12. García Rodríguez LA, Pérez Gutthann S. Use of the UK General Practice Research Database for pharmacoepidemiology. Br J Clin Pharmacol 1998; 45: 419–425.
13. Walker AM. Case-control studies: sampling the source population. In: Observation and Inference: An Introduction to the Methods of Epidemiology. Chestnut Hill, MA: Epidemiology Resources Inc, 1991; 75–80.
14. Hardman JG, Limbird LE, eds. Goodman and Gilman’s The Pharmacological Basis of Therapeutics. New York: McGraw-Hill, 1996.
15. Pérez Gutthann S, García Rodríguez LA, Raiford DS. Individual non-steroidal anti-inflammatory drugs and hospitalizations for upper gastrointestinal bleeding and perforation. Epidemiology 1997; 8: 18–24.
16. Laporte JR, Carné X, Vidal X, Moreno V, Juan J. Upper gastrointestinal bleeding in relation to previous use of analgesics and non-steroidal anti-inflammatory drugs. Lancet 1991; 337: 85–89.
17. Holvoet J, Terriere L, Van Hee W, Verbist L, Fierens E, Hautekeete ML. Relation of upper gastrointestinal bleeding to non-steroidal anti-inflammatory drugs and aspirin: a case-control study. Gut 1991; 32: 730–734.
18. Nobili A, Mosconi P, Franzosi MG, Tognoni G. Non-steroidal anti-inflammatory drugs and upper gastrointestinal bleeding: a postmarketing surveillance case-control study. Pharmacoepidemiol Drug Saf 1992; 1: 65–72.
19. Savage RL, Moller PW, Ballantyne CL, Wells JE. Variation in the risk of peptic ulcer complications with nonsteroidal antiinflammatory drug therapy. Arthritis Rheum 1993; 36: 84–90.
20. Langman MJS, Weil J, Wainwright P, Lawson DH, Rawlins MD, Logan RFA, Murphy M, Vessey MP, Colin-Jones DG. Risks of bleeding peptic ulcer associated with individual non-steroidal anti-inflammatory drugs. Lancet 1994; 343: 1075–1078.
21. Deleted in proof.
22. Rahme E, Pettitt D, LeLorier J. Dose response curves with time dependent exposure: the case of acetaminophen gastrointestinal side effects. Pharmacoepidemiol Drug Saf 2000; 9: S104.
23. Cryer B, Feldman M. Cyclooxygenase-1 and cyclooxygenase-2 selectivity of widely used nonsteroidal anti-inflammatory drugs. Am J Med 1998; 104: 413–421.
24. Warner TD, Giuliano F, Vojnovic I, Bukasa A, Mitchell JA, Vane JR. Nonsteroid drug selectivities for cyclo-oxygenase-1 rather than cyclo-oxygenase-2 are associated with human gastrointestinal toxicity: a full in vitro
analysis. Proc Natl Acad Sci USA 1999; 96: 7563–7568.
25. Kelly JP, Kaufman DW, Jugelon JM, Sheehan JE, Koff RS, Shapiro S. Risk of aspirin-associated major upper-gastrointestinal bleeding with enteric-coated or buffered product. Lancet 1996; 348: 1414–1416.
26. Dajani EZ. Gastrointestinal toxicity of over-the-counter analgesics. Am J Gastroenterol 1998; 93: 1020–1022.
27. Henry D, Dobson A, Turner C. Variability in the risk of major gastrointestinal complications from nonaspirin nonsteroidal anti-inflammatory drugs. Gastroenterology 1993; 105: 1078–1088.
28. Hawkey CJ. COX-2 inhibitors. Lancet 1999; 353: 307–314.
29. Eversmeyer W, Poland M, DeLapp RE, Jensen CP. Safety experience with nabumetone versus diclofenac, naproxen, ibuprofen, and piroxicam in osteoarthritis and rheumatoid arthritis. Am J Med 1993; 95: 10S–18S.
30. Schattenkirchner M. An updated safety profile of etodolac in several thousand patients. Eur J Rheumatol Inflamm 1990; 10: 56–65.
31. Hawkey C, Kahan A, Steinbruck K, Alegre C, Baumelou E, Begaud B, Dequeker J, Isomaki H, Littlejohn G, Mau J, Papazoglou S. Gastrointestinal tolerability of meloxicam compared to diclofenac in osteoarthritis patients. International MELISSA Study Group. Meloxicam Large-Scale International Study Safety Assessment. Br J Rheumatol 1998; 37: 937–945.
32. Greenland S. Basic methods for sensitivity analysis of biases. Int J Epidemiol 1996; 25: 1107–1116.
33. Drews CD, Greenland S. The impact of differential recall on the results of case-control studies. Int J Epidemiol 1990; 19: 1107–1112.
34. Ulcickas Yood M, Rothman KJ, Johnson CC, Jick SS, Lang J, Wells KE, Jick H. Using prescription claims for drugs available over-the-counter (OTC). Pharmacoepidemiol Drug Saf 2000; 9: S37.