Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis

BMJ 2005; 330 doi: http://dx.doi.org/10.1136/bmj.330.7504.1366 (Published 9 June 2005)
Cite this as: BMJ 2005;330:1366

Recent rapid responses

Rapid responses are electronic letters to the editor. They enable our users to debate issues raised in articles published on bmj.com. Although a selection of rapid responses will be included as edited readers' letters in the weekly print issue of the BMJ, their first appearance online means that they are published articles. If you need the url (web address) of an individual response, perhaps for citation purposes, simply click on the response headline and copy the url from the browser window.

Displaying 1-10 out of 29 published

At the BMJ we have two forms of continuing appraisal of articles we have published: Rapid Responses and the annual Post Publication Review Meeting. At the review meeting statistical advisers and BMJ research editors discuss critically how we handled articles which have proved particularly controversial.

Here are the conclusions of our review of our handling of this paper: Julia Hippisley-Cox and Carol Coupland. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis. BMJ 2005;330:1366.

This article gave an accurate and suitably cautious account of this study's findings, strengths, and limitations and warned against inferring cause and effect.

The accompanying BMJ press release was accurate, and also advised caution regarding the possibility of unmeasured confounding.

The criticisms raised in the rapid responses mainly address possible unmeasured confounding, and hence continue the debate but do not negate the findings of the study.

The abridged version of the paper in the BMJ omitted some important details (on absolute risks, Number Needed to Harm or NNH) that would have allowed more accurate and full appraisal of the work, but did also include the same cautious interpreation regarding confounding factors.

Lessons learned and actions we're taking: we are amending the advice to editors who abridge BMJ research papers to ensure that core details (such as absolute event rates, NNH, the main points required by the CONSORT statement) remain in the abridged versions.

Competing interests: I am senior research editor at the BMJ

Competing interests: None declared

Trish Groves, Deputy editor, BMJ

BMJ, BMA House, Tavistock Square, London WC1H 9JR

Click to like:

Hippsley-Cox and Coupland(1)have proposed an interesting association between non-steroidal anti-inflammatory drugs and myocardial infarction. It needs to be remembered that this is a case-control study and the mechanism for the association is not understood, therefore it is difficult to account for all confounding variables. Secondly this is an association and not evidence of causation.

One explanation for the rise in the presciptions for non-steroidal antiinflammatory drugs is that ibuprofen and diclofenac are one of the commonest drugs to prescribe in primary care, and musculoskeltal problems one of the commonest presentations to GPs. Let us summise that a prescription for these drugs is a proxy marker of seeing a GP, in otherwords you are more likely to have a prescription if you have seen a GP for in the last 3 months for what ever reason. It is more likely that patients who have a myocardial infarction have seen their GP in the 3 months prior to the event,compared to the control population, and therefore more likely to have received a prescription for commonly prescribed medications including ibuprofen. An interesting finding but not a cause of myocardial infarction, otherwise we might equally summise that seeing your GP is the casuative problem, and not the prescription.

Yours sincerely,

Dr Martin Wilkinson Associate Dean GP Recruitment and Retention West Midlands Deanery

1. Julia Hippisley-Cox and Carol Coupland Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis BMJ 2005; 330: 1366

Competing interests: None declared

Competing interests: None declared

Martin JB Wilkinson, Associate Dean GP Recruitment and Retention

West Midlands Deanery

Click to like:

I have noticed in the table 1 of the paper of Hippisley-Cox and Coupland that there are more cases using Statin than controls (19.4 % versus 9.4 %). This can introduce a bias in the risk appreciation of myocardial infarction associated with NSAIDS. The consumption of Statin may suggest that a lot of cases had dyslipidemia with an increased risk of myocardial infarction. On the other hand, following the methodology to correlate risks with a drug consumption, it could be taking into account that Statin increases the risk of myocardial infarction. This joke points out the difficulty to ascertain a cause-effect mechanism in retrospective correlation studies.

Competing interests: None declared

Competing interests: None declared

Jacques M. Frey, Retired Professor in Medical Biochemistry

Saint Etienne, 42000, France

Click to like:

Editor,

Even though the methodology used by Hippisley-Cox and Coupland (1) may be called in question, the finding of increased risk of myocardial infarction associated with traditional non-steroidal anti-inflammatory drugs (NSAIDs), especially diclofenac, deserves attention.

The cyclooxygenase (COX) selectivity of diclofenac is roughly equivalent to celecoxib and diclofenac produces minimal antiplatelet effect using typical dosing. (2) Compared with coxibs, traditional NSAIDs achieve particularly high concentration in tissues with low pH, pathologically found in the inflamed or hypoxic tissues. This raises the possibility that prothrombotic potential of diclofenac may be higher than for coxibs. On the other hand, increased thrombotic risk observed with certain preferential COX-1 inhibitors is more difficult to corroborate. For example, antiplatelet efficacy of naproxen is comparable with aspirin and sustained throughout its dosing interval.(2) Likewise high doses of aspirin, the concurrent inhibition of prostacyclin biosynthesis is unlikely to counteract cardioprotective effect of naproxen, demonstrated in several case-control studies. (3,4).

We would like to emphasize that clinical efficacy of naproxen for cardiovascular protection depends substantially on regular daily dosing. Furthermore, several NSAIDs including naproxen may counteract the cardioprotective effects of aspirin either by competitive interaction on platelet COX-1 (5) or, as we have recently suggested, by interference with anti-inflammatory effects of aspirin mediated by COX-2 (6). Hence, marked differences in the compliance , together with other hard-to-control variables such as concomitant over-the-counter aspirin use or interaction with other cardiovascular medicines, could easily explain the apparently conflicting results of various observational studies and randomised controlled trials.

Thus, we conclude that the results of any study involving NSAIDs should be interpreted with caution, taking into account all available evidence in order to avoid so called “tomato effect”, i.e. depriving patients of an effective therapy based on an artificially exaggerated risk of an adverse event.

1. Hippisley-Cox J, Coupland C. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis. BMJ 2005; 330: 1366.

2. Van Hecken A, Schwartz JI, Depre M, De Lepeleire I, Dallob A, Tanaka W, et al. Comparative inhibitory activity of rofecoxib, meloxicam, diclofenac, ibuprofen, and naproxen on COX-2 versus COX-1 in healthy volunteers. J Clin Pharmacol 2000;40:1109-20.

3. Rahme E, Pilote L, LeLorier J. Association between naproxen use and protection against acute myocardial infarction. Arch Intern Med 2002; 162: 1111–15.

4. Watson DJ, Rhodes T, Cai B, Guess HA. Lower risk of thromboembolic cardiovascular events with naproxen among patients with rheumatoid arthritis. Arch Intern Med 2002; 162: 1105–10.

5. Capone ML, Sciulli MG, Tacconelli S, Grana M, Ricciotti E, Renda G, et al. Pharmacodynamic interaction of naproxen with low-dose aspirin in healthy subjects. J Am Coll Cardiol 2005;45:1295-301.

6. Pijak MR, Huzicka I, Gazdik F. Cardiovascular risk of coxibs depends on their COX-2 selectivity and interaction with aspirin [letter]. CMAJ 2005; in press. Epub available at http://www.cmaj.ca/cgi/eletters/172/10/1299.

Competing interests: Dr. Pijak has served as a paid speaker and consultant for the local branches of the following manufacturers of NSAIDs: Pfizer, Merck Sharp & Dohme (MSD) and Fournier Slovakia.

Competing interests: None declared

Michal R. Pijak, Consultant in Internal Medicine, Rheumatology, Allergy and Clinical Immunology

Frantisek Gazdik, Igor Huzicka

1 Department of Internal Medicine, University Hospital, Limbova 5, 83305 Bratislava, Slovakia

Click to like:

Dear Dr Godlee

Here is a copy of letter submitted to Chemist and Druggist in response to the letter from David Pruce (which also appeared on the BMJ site). We happy for you to post this on the BMJ website.

Dear Ms Ellis

Thank you for giving us the opportunity to reply to the letter from David Pruce regarding the study we published in the BMJ on 11th June 2005.

There are several points which need clarification and we hope that we can do this for you.

In his letter, Mr Pruce mentions that “we are concerned that no other possible explanation for the findings was addressed in the paper before concluding that the NSAIDs were a causal factor in myocardial infarction”.

This statement is inaccurate on two accounts. Firstly, in our paper we did not state causality. Rather we stated the following in the abstract

“These results suggest an increased risk of myocardial infarction associated with current use of rofecoxib, diclofenac and ibuprofen despite adjustment for many potential confounders. … This is an observational study and may be subject to residual confounding that cannot be fully corrected for”.

This interpretation was also stated clearly in page 5 and page 6 of the paper. It was also re-iterated this in every single interview with the media. I have attached a copy of the full text article for your information.

The statement that “no other possible explanations for the findings were discussed” was also inaccurate. On page 5 in the ‘Discussion of methods’ section we discussed other possible explanations for our findings as follows:

“This is an observational study and therefore at risk of confounding. For example, some confounding by indication could be present, such as if patients had been prescribed NSAIDs for chest pain that was actually angina…..Similarly, we considered whether channelling may explain our results… Although we have adjusted for a number of confounding variables some residual confounding may result from misclassification of those variables and confounding by unmeasured variables.”.

David Pruce makes a point about “poor matching” of cases and controls. This issue however was not raised in the BMJ Editorial, contrary to what David Pruce states. We matched on age, sex, practice and calendar year. It would be staggering if we had not found differences between cases and controls for the well established risk factors for coronary heart disease (such as diabetes, hypertension, obesity and smoking). These differences point to the internal consistency of the dataset. We adjusted for these additional confounding variables in the analysis which is a well established method for taking account of potential factors and means that factors such as these do not confound the results. In the methods section of the paper we described two additional models which we carried out to further examine the effects of the differences in prior risk factors of any missing data, the results of which are described in the Results section

“We fitted a second model restricted to cases and controls with complete data for smoking and body mass index. We fitted a third model restricted to patients without either diabetes or ischaemic heart disease in order to reduce possible effects of residual confounding”.

We obtained similar results with both of these additional analyses implying that these factors were not the explanation for our results. This is clearly stated in the results and conclusions.

Lastly, the National Electronic Library for Health published an entirely independent critique of our paper who said the following (http://www.nelh.nhs.uk/hth/antiinfdrugs.asp)

“How reliable are the conclusions? This appears to be a well-conducted case-control study. It was based on a large general practice based dataset, with data collected over a four year period. The cases of heart attack were identified without prior knowledge of drug exposure, and were matched with controls on a number of relevant criteria. The statistical analysis was correct, and presents both adjusted and unadjusted results. The adjusted results accounted for a number of potentially important risk factors for myocardial infarction (age, co- morbidity, smoking habits, use antidepressants, statins and aspirin), although this data was not complete for all cases. The authors acknowledge that this study may be at risk of residual confounding. Ibuprofen can be obtained without prescription, and some patients may have been misclassified as not taking ibuprofen; although the authors state that this number is likely to be small. In addition, data on the use of NSAIDs was based on practice prescription records, which do not record whether the patient actually took the drug or for how long. It should also be noted that this study only considered prescribed drugs (people taking prescription painkillers tend to take them at a higher dose and for a longer period than those who buy them over the counter), and that almost 70% of the sample was aged 65 years or over;”

Competing interests: None declared

Competing interests: None declared

Julia Hippisley-Cox, Professor of Clinical Epidemiology & General Practice

Carol Coupland

University of Nottingham

Click to like:

Despite high methodological weaknesses of the study by Hippisley -Cox and Coupland (1), higher risk of myocardial infarction with diclofenac in comparison with other non-steroidal anti-inflammatory drugs (NSAIDs) warrants attention.

Selectivity of diclofenac for cycloxygenase isoforms is roughly equivalent to celecoxib and using typical dosing, diclofenac produces minimal antiplatelet effect. Because of its acidic nature diclofenac preferentially accumulates in hypoxic tissues. In a view of this, the possibility of prothrombotic effect for diclofenac may be higher than for celecoxib. However, similar risk associated with several other NSAIDs, namely naproxen, seems to be unjustified.

Naproxen, unlike diclofenac, was shown as a potent and long-lasting COX-1 inhibitor, which explains its powerful antiplatelet efficacy comparable to that achieved with aspirin and sustained through its dosing interval. Inhibition of PGI2 biosynthesis by naproxen is unlikely to counteract its potential cardioprotective effect, in light of the demonstrated efficacy of aspirin at high doses, having a similar effect on PGI2. This assumption is supported by several case-control studies with naproxen and randomized controlled trials with other traditional NSAIDs, as well. (2,3)

Compliance and regular daily doses are likely to represent the main determinants of the clinical efficacy of naproxen for cardiovascular protection. The apparently conflicting results of randomized clinical trials, like VIGOR, and some observational studies may reflect marked differences in compliance and regular use of a high dose of the drug in the two settings. Moreover, naproxen, like several other NSAIDs may counteract the cardioprotecitve effect of aspirin.(4)

NSAIDs, in addition to pharmacodynamic interaction on platelet COX -1, could also interfere with anti-inflammatory effects of aspirin mediated by COX-2.(5) Absence of significant interaction between aspirin and NSAIDs in studies such as Hippisley-Cox and Coupland could be explained by many uncontrolled factors including over the counter aspirin-containing drug use, interactions with other cardiovascular drugs and compliance with prescribed use.

1. Hippisley-Cox J, Coupland C. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis. BMJ 2005; 330:1366. [Abstract] [Full Text]

2. Rahme E, Pilote L, LeLorier J. Association between naproxen use and protection against acute myocardial infarction. Arch Intern Med 2002;162: 1111–1115.[Abstract/Free Full Text]

3. Brochier ML.. Evaluation of flurbiprofen for prevention of reinfarction and re-occlusion after successful thrombolysis or angioplasty in acute myocardial infarction. Eur Heart J 1993;14: 951-957[Abstract].

4. Capone ML, Sciulli MG, Tacconelli S, Grana M, Ricciotti E, Renda G, et al. Pharmacodynamic interaction of naproxen with low-dose aspirin in healthy subjects. J Am Coll Cardiol 2005;45:1295-301.

5. Pijak MR, Huzicka I, Gazdik F. Cardiovascular risk of coxibs depends on their COX-2 selectivity and interaction with aspirin http://www.cmaj.ca/cgi/eletters/172/10/1299 (accessed 16 May 2005)

Competing interests: Dr. Pijak has served as a paid speaker and consultant for the local branches of the following manufacturers of NSAIDs: Pfizer, Merck Sharp & Dohme (MSD) and Fournier Slovakia.

Competing interests: None declared

Michal R. Pijak, Consultant in Internal Medicine, Rheumatology and and Clinical Immunology

Frantisek Gazdik, Katarina Gazdikova

1 Department of Internal Medicine, University Hospital, Limbova 5, 83305 Bratislava, Slovakia

Click to like:

The most widely used non-steroidal anti-inflammatory drug (NSAID)is aspirin. Administration of low doses of aspirin after a myocardial infarction reduces the risk of re-infarction and death. However, recent studies suggest that long-term aspirin does not reduce (1) and may increase (2,3,4) the rate of cardiovascular events in individuals who have not had a recent cardiovascular event. These studies also suggest that aspirin may increase the proportion of events that occur 'silently' and thus are not admitted to hospital which could increase the risk of sudden death, as observed in several randomised secondary prevention trials (3,4). This combined with the risk of gastrointestinal and intra-cranial haemorrhage (5) and the problems of meta-analysis (3,4), the only basis for long-term aspirin use in patients with known coronary disease, further question whether the medical community has been premature in recommending chronic aspirin therapy so widely. Further data on the effects of NSAIDs, including aspirin, on the risk of cardiovascular death, including sudden death are required.

1. Ridker PM, et al. A Randomized Trial of Low-Dose Aspirin in the Primary Prevention of Cardiovascular Disease in Women. New Engl J Med 2005; 352:1293-1304.

2. Pulmonary Embolism Prevention (PEP) Trial Collaborative Group. Prevention of pulmonary embolism and deep vein thrombosis with low dose aspirin: Pulmonary Embolism Prevention (PEP) trial. Lancet 2000; 355(9212):1295-1302.

3. Cleland JGF. For Debate: Preventing atherosclerotic events with aspirin. BMJ 2002; 324(7329):103-105.

4. Cleland JGF. Is aspirin 'The Weakest Link' in cardiovascular prophylaxis. The surprising lack of evidence supporting the use of aspirin for cardiovascular disease. Prog Cardiovasc Dis 2002; 44:275-292.

5. Nelson MR, Liew DL, Bertram M, Vos T. Epidemiological modelling of routine use of low dose aspirin for the primary prevention of coronary heart disease and stroke in those aged >70. BMJ doi:10.1136/bmj.38456.676806.8F (published 20 May 2005)

Competing interests: None declared

Competing interests: None declared

John GF Cleland, Professor of Cardiology

University of Hull. HU6 7RX

Click to like:

David M Pruce of the Royal Pharmaceutical Society has raised an important issue in noting that the two page press release issued by the BMJ was evidently geared towards getting a headline.

Complaints about sensationalism in the mass media rarely, if ever, acknowledge the role played by medical journals, researchers, funders of research, and the pharmaceutical industry in generating the torrent of press releases which flood onto the journalist’s electronic desk.

The notion of a reporter for the Sun, say, diligently combing through medical journals in order to find some worrying minutiae to distort into banner headlines is complete fantasy. With a few honourable exceptions, reporters simply reproduce what they are told; the fault lies in the people who are doing the telling, not the people who are simply passing it on. The most we can say is that they should have disbelieved what they were being told.

For example, some of the most sensational headlines this week have stemmed from a press release issued by Cancer Research UK which kicks off with: ‘A new international survey reveals that around one in five women would consider having both breasts removed to help reduce their risk of developing breast cancer if told they were at an increased risk of developing the disease.’ 1

Now that statement certainly suggests the need for some serious research; an inquiry into the role played by medical researchers in ratcheting up levels of fear and anxiety in the general public would be a very good place to start. But this has nothing to do with conveying genuine information; it’s a selling tool.

We are all familiar with pseudo surveys; attempts to sell us dodgy investment products, dubious time shares and even more dubious politicians, in the guise of gathering information.

This is a twist on that: the reputable market research company, NOP World, which conducted this survey asserts on its website that

‘NOP World's trusted research team and world-class consultative staff delivers People Knowledge—detailed intelligence about your customers, prospects and employees to help you build winning strategies and market leadership globally.’ 2

and I’m sure it does. It nowhere claims any expertise in scientific research, but then that isn’t needed.

What is being sold in this case is a clinical trial into the use of anastrozole. The pitch is that if people will consider having both their breasts removed then anything by comparison looks good; taking a pill must be so much better than that...

What makes it even more offensive is that nowhere in the press release is there any mention of the potential side-effects of anastrozole. It’s all gain and no pain.

These points should have occurred to the journalists who reproduced chunks of the press release, but the real fault lies with the people who issued it. And neither the general public nor the mass media can do anything about it; the answer lies in the medical profession itself.

Stevie Gamble

1. http://www.cancerresearchuk.org/news/pressreleases/breastremoval_13jun05 2. http://www.nopworld.com/

Competing interests: None declared

Competing interests: None declared

Stevie M Gamble, retired HMIT

EC2Y 8BL

Click to like:

Editor,

The study by Hippisley-Cox and Coupland [1], which reported an increase in the risk of myocardial infarction in patients receiving certain NSAIDS, has been justifiably criticised both in an accompanying editorial in the BMJ [2] and in rapid response letters. Attention has been drawn to the substantial differences between the cases and controls in terms of the recognised risk factors, to the incomplete data relating to known risk factors, and to the possible links between the indications for the prescription of NSAIDS and ischaemic heart disease. Moreover, it seems that all concerned – the authors, the editorial writers and the correspondents – acknowledge that the reliability of the results is uncertain. Why, then, are studies such as these performed? And why are they published?

For those with a ready access to large data-bases, it is a straightforward matter to deliver a nested case-control study – not to mention a relatively painless means of expanding a curriculum vitae. After all, there is no tedious phase of data collection. The study can also be performed rapidly and, hence, in time to participate in the latest medical controversy – a feature which certainly appeals to the journals. Then, of course, there is the added advantage that the study will involve many thousands of patients, thus ensuring the necessary statistical significance – regardless of the clinical relevance – to guarantee publication. It is not difficult, therefore, to understand why these generally worthless studies are both performed and published.

But the disadvantages of this particular type of research are plain to see: the unwarranted distress caused to patients, the confusing messages to clinicians and the bringing of medical research into disrepute, to name but a few. The study by Hippisley-Cox and Coupland [1] is dependent upon the myth that statistical analysis somehow cancels out the errors inherent in drawing causal inferences from a comparison of disparate groups.[3] It is a myth that is now endemic in the medical community and one which should be of much greater concern than the results of this flawed, nested case-control study.

References

[1] Hippisley-Cox J, Coupland C. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control study. BMJ 2005;330;1366-9.

[2] Juni P, Reichenbach S, Egger M. Cox 2 inhibitors, traditional NSAIDS, and the heart. BMJ 2005;330;1342-3.

[3] Penston J. Secondary prevention of heart disease with statins: groups were as different as chalk and cheese. BMJ 2005;330;1208.

Competing interests: None declared

Competing interests: None declared

James Penston, Consultant Physician/Gastroenterologist

Scunthorpe General Hospital, Cliff Gardens, Scunthorpe, North Lincolnshire DN15 7BH

Click to like:

There is nothing that the media like better than to publicise yet another drug or food that increases heart attacks or mortality by 50%. As a result of this, a friend has stopped taking her anti-arthritic drugs although her risk of a heart attack is very small. One would have hoped that reputable medical journals would stop this hysteria by giving an idea of what the absolute increase in risk is for an individual. Unfortunately, the most reputable medical journals fail to do this. The recent article in the BMJ by Hippisley-Cox and Coupland (BMJ 2005; 330: 1366-9) provides a formidable wealth of information about the relative increases in risk of taking anti-arthritic drugs but none about the absolute increase. It is, therefore, of little value in helping us advise patients about the balance of risks and benefits of taking these drugs. Would editors please stop providing the media with the material for scary headlines and insist that absolute risks are always incorporated into papers like this one ?

Competing interests: None declared

Competing interests: None declared

Desmond G Julian, Emeritus Professor

7 Netherhall Gardens, London NW3 5RN

Click to like:

THIS WEEK'S POLL