The impact of nonsteroidal antiinflammatory drugs (NSAIDs) on cardiovascular outcomes has received little attention despite the widespread use of this therapeutic class in the general population. The few epidemiologic studies that have explored whether NSAIDs alter the risk of coronary heart disease have found no effect, either beneficial or harmful. 1–3 The pooled relative risk (RR) of coronary heart disease for NSAID current users in these three studies was 1.04 (95% confidence interval [CI] = 0.97–1.12). This lack of effect stands in contrast to the well-established use of aspirin as an effective cardioprotective agent. 4 Similarly, the association between heart failure and NSAIDs has hardly been investigated; two epidemiologic studies have found that users of NSAIDs are at a two-fold increased risk of hospitalization for heart failure. 5,6 Also, NSAIDs have been shown in numerous clinical trials to elevate blood pressure in normotensive individuals and to worsen preexisting hypertension. 7 Underlying hemodynamic compromise and concomitant use of certain antihypertensive medications may enhance the deleterious effect of NSAIDs. There are no clinical trials designed to examine whether NSAIDs have a role in the development of heart failure.
To further explore the cardiovascular effects of NSAIDs we performed a nested case-control study using as cases of heart failure those identified in a recent incidence study of newly diagnosed heart failure in the United Kingdom. 8
Approximately three million residents in the United Kingdom are registered with general practitioners (GPs) who participate in the General Practice Research Database. 9 This scheme consists of almost 2000 GPs who use computers in their offices to record clinical patient information in a standard manner. They have agreed to provide the information anonymously to the Medicine Control Agency and to allow the information to be used for research projects. The recorded information includes demographics, details from general practitioners’ visits, diagnoses from specialists’ referrals and hospital admissions, results of laboratory tests and a free text section. Prescriptions are automatically produced from the computer and recorded on the patient’s computerized file. A previous study using this computerized data source has documented that over 90% of all referrals are entered on the GPs’ computers with a code that reflects the clinical diagnosis. 10,11 A modification of the Oxford Medical Information System classification is used to code specific diagnoses, and a drug dictionary based on the Prescription Pricing Authority drug dictionary is used to record drugs.
Individuals 40–84 years of age at 1 January 1996, with at least 2 years of enrollment with a GP, were identified. We excluded from the source population pregnant women as well as patients with a diagnosis of heart failure or cancer before that time. Patients with cancer were excluded under the assumption that the etiology of heart failure would be different in those patients. The final source population comprised 689,467 individuals. We then followed all source members until the earliest occurrence of one of the following endpoints: a first-time recorded diagnosis of heart failure or cancer or 31 December 1996.
Case and Control Ascertainment
In a previous study we identified a random sample of 1200 potential cases with newly diagnosed heart failure recorded on computer files in 1996 (Figure 1). 8 Briefly, to validate incident cases of heart failure, we sent a questionnaire to the GPs to confirm the diagnosis for all potential cases, define the date of onset, and assess the severity of the initial episode (New York Heart Association [NYHA] criteria), as well as to obtain information on symptoms at presentation and investigations performed. We considered a patient as a case of heart failure when the GP reported dyspnea at presentation, together with at least one of the following criteria: (a) pulmonary edema confirmed clinically or radiographically; (b) peripheral edema and raised jugular venous pressure on clinical examination; or (c) evidence of heart disease (by clinical exam, ECG or echocardiogram). The underlying cause for heart failure was ascertained by manually reviewing patients’ computerized records as well as GPs’ requested information. Patients were then classified into one of the mutually exclusive categories with the following ranking order: coronary heart disease, valvular disease, hypertension, other cardiac disease and other systemic diseases.
We used for the analyses all confirmed cases of heart failure with a known NYHA criteria (N = 857). We then randomly sampled from the source population a group of 5000 controls frequency-matched by age and sex. The date of initial diagnosis of heart failure was used as the index date among the cases. A date during the period of follow-up was randomly generated for each control and used as the index date. Information on the following risk factors was obtained from computerized files for both cases and controls: age, sex, smoking status, body mass index (BMI), units of alcohol consumed per week, comorbidity and drug use.
Drug Exposure Definition
We ascertained the exposure history of drug therapy before the index date for the following drug classes: non-aspirin NSAIDs, acetaminophen, steroids, antihypertensive drugs, nitrates, glycosides, lipid-lowering drugs, warfarin and aspirin. We defined exposure to non-aspirin NSAIDs as “current” when the supply of the most recent prescription lasted until the index date or ended in the previous month, as “intermediate” when the end of the most recent prescription was between 1 and 3 months before the index date, as “recent” when the end of the most recent prescription was between 3 and 12 months before the index date, and as a “past” when the end of the most recent prescription was more than 12 months before the index date. We evaluated duration of use by adding the periods of “consecutive” prescriptions, defined as an interval of less than 2 months between two prescriptions.
Current users were subsequently divided into “current single users” and “current multiple users”. The latter category included patients who received prescriptions for more than one NSAID with the supply of each prescription ending within the month before the index date. Among current single users, we calculated the relative risk for individual NSAIDs and the dose-response relation using two dose categories: low-medium and high. 12
We classified NSAIDs in two groups according to their plasma half-life: less than 12 hours, and equal to or greater than 12 hours. 13 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. We also classified these drugs according to their pharmacologic structures into propionic acid derivatives (fenoprofen, flurbiprofen, ibuprofen, ketoprofen, tiaprofenic, fenbufen, and naproxen), hetoaryl acetic acids (diclofenac), indole and indene acetic acids (etodolac, indomethacin, sulindac, and acemetacin), and others. Information on drug indication was obtained through a blinded review of computerized patient profiles.
Nested Case-Control Analysis
We calculated the odds ratio (assumed to provide a valid estimate of the relative risk) and 95% CIs of heart failure associated with drug use compared with nonuse, using unconditional logistic regression. All estimates were adjusted for age, sex, BMI, smoking, alcohol use, comedication and comorbidity.
Of the 857 patients newly diagnosed with heart failure, 70% were more than 69 years of age and 52% were males. Being overweight, smoking, and having a previous history of heart disease or respiratory disease were associated with heart failure in our population (Table 1).
Fifteen percent of patients with heart failure were current NSAID users compared with 9% of controls, giving a crude estimated RR of 1.8 (CI = 1.5–2.3). The RR was 1.6 (1.2–2.1) after adjusting for known potential confounders. Only adjusted estimates will be presented below unless otherwise specified. The risk was greater during the first month of initiating NSAID therapy (RR = 2.1; 1.4–3.3), decreased with longer duration and dropped quickly once treatment was stopped (Table 2). There was no dose response. However, users of more than one NSAID simultaneously (0.2% of controls) had an RR of 3.6 (1.4–9.5); the RR was 1.6 (1.2–2.1) for users of a single NSAID.
The risk of heart failure was independent from the plasma half-life of the drug. RRs for individual NSAIDs ranged from 1.1 (CI = 0.7–1.8) for diclofenac to 3.4 for indomethacin (1.5–7.9) (Table 2). When we classified these drugs according to the chemical class, the RRs were 1.7 (1.2–2.4) for propionic acid derivatives, 1.1 (0.7–1.8) for hetoaryl acetic acids, 2.4 (1.1–5.3) for indole and indene acetic acids, and 2.6 (1.2–5.4) for other classes.
The most common NSAID indication among controls was osteoarthritis (77%). The estimated RR associated with NSAID use was not notably different among the various indications. However, although based on small numbers, we found an RR of 8.9 (2.6–30.5) associated with use of NSAIDs for pain related to vascular disorders (eg, peripheral vascular disease).
Fifty-two patients with newly diagnosed heart failure (5.5%) died the day of the diagnosis (immediately fatal cases) and 103 (12%) died within the month after the diagnosis (fatal cases). The distribution of cases according to the NYHA functional grades I to IV was 4%, 40%, 34% and 15%, respectively. The risk associated with current NSAID use was elevated for all NYHA grades.
The etiology of the heart failure episode was mainly attributed to coronary heart disease (N = 310), hypertension (N = 190), other cardiovascular diseases (N = 91) and respiratory diseases (N = 121). The proportion of fatal cases was greater for heart failures attributed to coronary heart disease (18%) than for those attributed to hypertension (7%), other cardiac disease (8%) or respiratory diseases (10%). The relative risks associated with NSAID use for heart failure episodes attributed to coronary heart disease, hypertension, other cardiac diseases or respiratory disease were 1.1, 2.3, 2.1 and 1.4, respectively.
The estimated RR associated with current NSAID use was 1.6 (CI = 1.2–2.3) among patients using antihypertensive medications and 1.3 (0.8–2.1) among nonusers of antihypertensive medications. When we stratified the analysis according to specific antihypertensive drugs, the RR associated with NSAID use was 1.4 (0.8–2.3) among users of diuretics, 1.0 (0.5–2.1) among users of beta-blockers, 1.5 (0.8–2.7) among users of calcium channel blockers and 2.9 (1.2–6.9) among users of ACE inhibitors. Among users of ACE inhibitors recently started on NSAID treatment, the RR associated with current NSAID use was 5.0 (1.2–21.2) compared with nonusers of NSAIDs.
The risk of developing heart failure varied depending on whether the patients had underlying vascular medical conditions. Among patients with diabetes, renal failure or hypertension the RR associated with NSAID use was 1.9 (CI = 1.3–2.8); among patients with none of these conditions it was 1.3 (0.9–1.9). As compared with nonusers of NSAIDs with none of the above-mentioned conditions, the RR of heart failure was 1.6 (1.2–2.0) for nonusers of NSAIDs with any of the conditions, 1.3 (0.9–2.0) for users of NSAIDs with none of the conditions, and 2.9 (2.0–4.1) for users of NSAIDs with any of the conditions. Among patients with hypertension, the RR associated with NSAID use was 5.9 (1.8–19.0) for patients with a history of renal failure or diabetes and 1.7 (1.1–2.7) for patients without these conditions.
We found an RR of 1.8 (1.4–2.3) for current users of acetaminophen starting at a daily dose greater than 2 gm, whereas smaller doses did not carry an increased risk (Table 3). The RR was 3.0 (CI = 2.0–4.6) among new users (first month) of doses greater than 2 gm. The risk among new users of high doses was elevated across indications, including osteoarthritis and rheumatoid arthritis. A dose effect was found both for acetaminophen as a single entity and for combination products. The effect observed for acetaminophen at high dose was similar when we excluded NSAID users from the analysis (data not shown). Compared with nonusers of either drug, the RR for concurrent users of NSAIDs and acetaminophen of 2 gm and greater was greater than the RR for each drug individually. However, the effect of concomitant use was no greater than the one expected from adding their independent effects (data not shown).
Current users of NSAIDs had an overall 60% elevated risk of developing heart failure. This risk was two-fold during the first month of starting NSAID therapy and was only slightly elevated among long-term users of NSAIDs. Data are compatible with a minor effect of dose and of plasma half-life on the risk of heart failure among patients recently started on NSAIDs. The risk associated with NSAIDs was higher when the attributed etiology of heart failure was hypertension. Risk was also higher when the patient had a history of hypertension (irrespective of underlying etiology of the heart failure), diabetes or renal failure. These findings are in agreement with two published epidemiologic studies that found a two-fold increased risk of hospitalization for heart failure among NSAID users with an underlying cardiovascular disease. 5,6 Interestingly, although based on relatively small numbers, we found that NSAID use increased by five times the risk of heart failure among patients with hypertension of probable renal origin, whereas NSAID use had a less detrimental effect when the origin of the hypertension was likely extra-renal.
Most of the beneficial and harmful effects of NSAIDs are related to their inhibition of eicosanoids synthesis by inhibition of the isozymes cyclooxygenase (Cox)-1 and -2. Certain prostanoids act as renal arteriole vasodilators preserving renal perfusion, particularly in individuals, such as patients with reduced circulating volume, with high renin/angiotensin states. These patients have an increased biosynthesis of vasodilatory prostaglandins (ie, PGE2 and PGI2) as compensatory mechanism. 14–16 Under normal conditions, prostaglandins play a negligible role in maintaining renal blood flow and preserving sodium/water retention. However, in susceptible individuals, prostaglandin synthesis inhibition by NSAIDs can reduce renal perfusion and increase peripheral systemic vascular resistance. These mechanisms might explain the increased risk of acute renal failure among NSAID users with a previous history of renal disease, 17–23 the elevation of blood pressure in hypertensive patients when taking NSAIDs, shown in clinical trials, 7 and the increased risk of developing heart failure found in NSAID users with an underlying vascular disease. 5,6 Of notice, all these effects shared the same pattern of greater risk during the first days of NSAID intake.
Patients treated with ACE inhibitors appear to present a greater risk of decompensation and consequently of developing heart failure when started on NSAID therapy than patients treated with other antihypertensive drugs. 14,15,24 This observation would be consistent with a pharmacodynamic interaction between NSAIDs and ACE inhibitors involving the prostaglandin-dependent hemodynamic effects of this particular class of antihypertensive drugs. An alternative explanation for our findings would be that ACE inhibitors were prescribed preferentially to patients whose renal function was critically dependent upon intact prostaglandin synthesis. However, in our population, ACE inhibitors were prescribed to elderly patients with renal diseases, diabetes or recent myocardial infarction only slightly more than beta-blockers or calcium channel blockers. Moreover, restriction of the analysis to patients without these conditions gave similar results.
Regarding specific drugs, the RR of heart failure ranged from 1.1 for diclofenac to 3.4 for indomethacin. Interestingly, in two studies evaluating the risk of individual conventional NSAIDs, indomethacin also presented a higher risk of acute renal failure than the other NSAIDs. 17,18 However, the number of current users for many drugs was small in our study, and therefore differences among individual NSAIDs should be treated with caution.
The increased risk observed for users of high-dose acetaminophen might be related to the substantial inhibition of prostaglandin synthesis at doses greater than 2000 mg. In vitro analyses have shown that acetaminophen is a weak nonselective inhibitor of Cox-1 and Cox-2. 25,26 We observed the same pattern of increased risk within the first month of taking high dose acetaminophen as the one among new users of NSAIDs.
Information on drug use was obtained from written prescriptions, reducing the possibility of information bias. The accuracy and completeness of these data have been validated in previous studies. 10 However, a potential limitation of all studies using computerized prescription data is the underestimation of drug use attributable to underascertainment of over-the-counter medications and to noncompliance. Nonetheless, only ibuprofen could be purchased over the counter in the United Kingdom. Moreover, misclassification of exposures collected prospectively is usually close to nondifferential among cases and controls, which would lead to underestimation of the excess risk under most circumstances. We performed a sensitivity analysis to quantify the impact of nonrecorded drug use. With a probability of missing drug use as high as 50%, the net impact of nondifferential underrecorded use of NSAIDs would have been a small underestimation of the RR from 2.3 to the observed crude 1.8. The relatively small impact of missing over-the-counter antiinflammatory drug use had been previously reported. 27
Information on newly diagnosed heart failure was obtained from computer files containing details from GPs’ visits, diagnoses from specialists’ referrals and hospital admissions, results of laboratory tests and a free text section. In addition, as part of the validation process, the GPs confirmed the diagnosis, date of onset, severity, symptoms at presentation and investigations performed of these incident cases, 8 reducing the possibility of false positives. However, because no systematic screening was conducted in this population, there may be an underrepresentation of asymptomatic cases of heart failure. Therefore, results from the current study, which focused on clinically relevant events, might not be generalizable to asymptomatic heart failure. On the other hand, unintended inclusion of patients with undiagnosed heart failure among controls would have yielded a slight underestimation of the RR.
We attempted to control for conditions known to be associated with heart failure. Inclusion of these factors in the model did not substantially alter the results for NSAIDs. Moreover, except in the small subgroup of patients taking NSAIDs for vascular diseases, the risk of heart failure was similar for all NSAID indications, suggesting that the observed association did not result from a specific indication (eg, early symptoms of heart failure). In addition, the lack of association between former NSAID users and heart failure once treatment was stopped suggests that the association was not attributable to a chronic underlying disease. On the other hand, as NSAIDs are contraindicated in patients predisposed to heart failure, our estimate could be underestimating the true association between NSAIDs and heart failure.
The higher risk at the beginning of treatment could be an artifact induced by the use of NSAIDs for the early symptoms of heart failure. However, this possibility can be reasonably ruled out in our study because 93% of NSAID users who developed heart failure within a month after starting treatment were using it for arthritis or pain unrelated to vascular disorders. A more plausible explanation to the duration response might be the existence of a rapid harmful effect of NSAIDs in susceptible patients or an adaptation of the cardiovascular/renal systems to the prostaglandin-dependent vascular effects of NSAIDs in a relatively short timeframe.
The overall incidence of heart failure in our source population (age range, 40 to 85 years) was 4 per 1000 person-years. 8 Given that 9.3% of the control population were current users of NSAIDs and assuming that the estimated relative risk of 1.6 reflects a causal association between NSAIDs and heart failure, the estimated incidence rates among users and nonusers of NSAIDs would be 6.1 and 3.8 per 1000 person-years, respectively. This suggests that between two and three new cases of heart failure per 1000 NSAID users per year might be attributable to NSAID use. The attributable risk would be higher in populations with an elevated risk of heart failure or with special susceptibility to the renal effects of NSAIDs. For example, among individuals 65 years of age and older, with an overall heart failure incidence of 12 per 1000 person-years and over 10% prevalence of NSAID use, between six and seven new cases of heart failure per 1000 users per year could be attributed to NSAID use.
In conclusion, NSAIDs should be used with caution in patients with previous renal failure, diabetes and hypertension, as well as, by extension, any other medical condition involving impaired renal function or vascular hemostasis derangement. In these patients, the risk of developing an episode of heart failure is the greatest in a critical period after the initiation of NSAID therapy.
We are indebted to the Boston Collaborative Drug Surveillance Program for providing access to the General Practice Research Database. We thank Drs. Francisco De Abajo and Carlo Patrono for their thoughtful comments and Marián Castañondo for her inspiration. We thank the general practitioners for their excellent collaboration.
1. García Rodríguez LA, Varas-Lorenzo C, Patrono C. Differential effects of aspirin and non-aspirin nonsteroidal antiinflammatory drugs in the primary prevention of myocardial infarction in postmenopausal women. Epidemiology 2000; 11: 383–387.
2. Ray W, Stein C, Hall K, Daugherty J, Griffin M. Non-steroidal anti-inflammatory drugs and the risk of serious coronary heart disease. Lancet 2002; 359: 118–123.
3. Velentgas P, Cali C, Dreyer N, Verburg KM, Walker A. Risk of myocardial infarction and cardiac death among users of non-steroidal anti-inflammatory medications. Pharmacoepidemiol Drug Saf
4. Antiplatelet Trialists’ Collaboration. Collaborative overview of randomised trials of antiplatelet therapy-I: prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. BMJ 1994; 308: 81–106.
5. Heerdink ER, Leufkens HG, Herings RM, Ottervanger JP, Stricker BH, Bakker A. NSAIDs associated with increased risk of congestive heart failure in elderly patients taking diuretics. Arch Intern Med 1998; 158: 1108–1112.
6. Page J, Henry D. Consumption of NSAIDs and the development of congestive heart failure in elderly patients: an unrecognized public health problem. Arch Intern Med 2000; 160: 777–784.
7. Johnson A. NSAIDs and increased blood pressure. What is the clinical significance? Drug Saf 1997; 17: 277–289.
8. Johansson S, Wallander MA, Ruigómez A, García Rodríguez L. Incidence of newly diagnosed heart failure in general practice. Eur J Heart Fail 2001; 3: 225–231.
9. García Rodríguez LA, Pérez Gutthann S. Use of the UK general practice research database for pharmacoepidemiology. Br J Clin Pharmacol 1998; 45: 419–425.
10. Jick H, Jick S, Dervy L. Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. BMJ 1991; 302: 766–768.
11. Jick H, Terris B, Derby L, Jick S. Further validation of information recorded on a general practitioner based computerised data resource in the United Kingdom. Pharmacoepidemiol Drug Saf 1992; 1: 347–349.
12. García Rodríguez LA, Jick H. Risk of upper gastrointestinal bleeding and perforation associated with individual non-steroidal anti-inflammatory drugs. Lancet 1994; 343: 769–772.
13. Hardman JG, Limbird LE, Molinoff PB, Ruddon RW, Goodman-Gilman A. Goodman & Gilman’s The Pharmacological Basis of Therapeutics. New York: McGraw-Hill, 1996.
14. Mene P, Pugliese F, Patrono C. The effect of nonsteroidal anti-inflammatory drugs on human hypertensive vascular disease. Semin Nephrol 1995; 15: 244–252.
15. Ruoff GE. The impact of nonsteroidal anti-inflammatory drugs on hypertension: alternative analgesics for patients at risk. Clin Ther 1998; 20: 376–387.
16. Whelton A. Nephrotoxicity of nonsteroidal anti-inflammatory drugs: physiologic foundations and clinical implications. Am J Med 1999; 106: 13S–24S.
17. Pérez-Gutthann S, García Rodríguez LA, Raiford DS, Duque Oliart A, Romeu JR. Nonsteroidal anti-inflammatory drugs and the risk of hospitalization for acute renal failure. Arch Intern Med 1996; 156: 2433–2439.
18. Griffin M, Yared A, Ray W. Nonsteroidal antiinflammatory drugs and acute renal failure in elderly persons. Am J Epidemiol 2000; 151: 488–496.
19. Beard K, Lawson D, MacFarlane G. Non-steroidal anti-inflammatory drugs and acute renal disease: a case-control study. Pharmacoepidemiol Drug Saf 1992; 1: 3–9.
20. Perneger TV, Whelton PK, Klag MJ. Risk of kidney failure associated with the use of acetaminophen, aspirin, and nonsteroidal antiinflammatory drugs. N Engl J Med 1994; 331: 1675–1678.
21. Evans J, McGregor E, McMahon A, et al
. Nonsteroidal antiinflammatory drugs and hospitalization for acute renal failure. Q J Med 1995; 88: 551–557.
22. Henry D, Page J, Whyte I, Nanra R, Hall C. Consumption of nonsteroidal antiinflammatory drugs and the development of functional renal impairment in elderly subjects. Results of a case-control study. Br J Clin Pharmacol 1997; 44: 85–90.
23. Whelton A, Stout RL, Spilman PS, Klassen DK. Renal effects of ibuprofen, piroxicam, and sulindac in patients with asymptomatic renal failure. Ann Intern Med 1990; 112: 568–576.
24. Hall D, Holger Zeitler H, Rudolph W. Counteraction of the vasodilator effects of enalapril by aspirin in severe heart failure. J Am Coll Cardiol 1992; 20: 1549–1555.
25. Cryer B, Feldman M. Cyclooxygenase-1 and cyclooxygenase-2 selectivity of widely used nonsteroidal antiinflammatory drugs. Am J Med 1998; 104: 413–421.
26. Warner T, Giuliano F, Vojnovic I, Bukasa A, Mitchell J, Vane J. 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.
27. Ulcickas YM, Rothman K, Johnson C, et al
. Using prescription claims for drugs available over-the-counter (OTC). Pharmacoepidemiol Drug Saf 2000; 9: S37.