select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
The research suggested that taking certain painkillers daily over a long period of time may lead to a slightly raised risk of heart attack and stroke.
Prof Sheila Bird, Programme Leader, MRC Biostatistics Unit, said:
“Absolute rates (line 1 on p9) are LOW, as I’ve outlined below – but they were missing from the BMJ Abstract, which is disappointing – and it is a modest increase in these low rates that patients need to offset against the benefit in pain-relief that they get from NSAIDs. Trelle et al. address potential harms but not the corresponding benefits. Doctors and patients need to take both into consideration.
“Network meta-analysis is short-hand for how analysts proceed when faced with a series of RCTs, some of which are head-to-head comparisons, other versus placebo, and inferences are sought on drug D versus drug E (yet D vs E have not been compared directly in RCTs) or on drug I versus placebo when drug I features in head-to-head trials only.
“Immediately, a curious reader will ask: why is no RCT of drug I versus placebo available for analysis? Analysis can be formulated in such a way as to anticipate some bias in the estimation of rate ratios versus placebo if the estimation is indirect only.
“Let’s now consider the rates that Trelle et al. are concerned with. Per 1,000 trial patients, the rates at issue are around 3 (cardiovascular deaths), 4 (stokes), 5 (myocardial infarctions . . . some of which will result in cardiovascular death), 6 (deaths from any cause . . . which include cardiovascular deaths), to 10 (for a composite outcome of non-fatal MI, non-fatal stroke, or cardiovascular death).
“Trelle et al. suggest that a genuine 30% increase in the above rates – such as from 10 placebo to 13 drug X per 1,000 patients experiencing the composite outcome – would be alerting. The 95% credible intervals for all seven non-steroidal anti-inflammatory drugs (NSAID) under review are consistent with 30% but only three of them are consistent with the risk being greater for those on NSAID versus placebo. Moreover, the three include the two NSAIDs on which there is least information, namely:
Ibuprofen with around 5,000 person-years (pys) of observation, nil versus placebo; and Lumiracoxib with around 9,000 pys, but only only 434 of them in two RCTs against placebo (209 pys) and around 3,800 versus Ibuprofen.
“Should this matter? Perhaps: because this pair has little direct evidence vs placebo. However, the BMJ reader is offered many insights to the data, the neat network diagram (Fig 1), estimated 95% credibility intervals for rate ratio (vs placebo, see Fig 2), trial details in Tables 1 & 2 with event totals (from 554 MIs to 1,091 composite events – not all trials reported on each outcome, however) and the revealing drug X versus drug Y comparisons in Figure 3.
“In Figure 3, readers will be struck by the closely similar profiles for Diclofenac versus Etoricoxib and fairly similar for Celecoxib versus Rofecoxib (the first two represented by 25,000 to 30,000 pys, the second pair by half as many pys); by the wide uncertainty which qualifies the comparison of Ibuprofen versus Lumiracoxib; and that 14 out of 21*5 related comparisons of drug X versus drug Y are formally significant.
“Key messages (from Fig 2 & 4) are:
a) that MI rate ratio is significantly raised for Rofecoxib (95% credible interval: 1.3 to 3.6);
b) that stroke rate ratio is fairly precisely estimated and consistent with 1 for Celecoxib (0.6 to 2.1) and Rofecoxib (0.6 to 1.8); and consistent with 1 also for Naproxen (0.9 to 3.3) but Fig 4 suggests that its stroke rate is bordering on being worse than Rofecoxib’s;
c) cardiovascular death rates are significantly raised, but very imprecisely estimated, for Diclofenac (1.5 to 5.0) and Etoricoxib (1.2 to 16); and of concern for Celecoxib (1.0 to 4.6) and Rofecoxib (0.9 to 2.6;
d) when it comes to ‘death from any cause’, the 95% credibility intervals for the four ‘coxibs’ and for diclofenac are not only consistent with 30% increased rate but they clearly exclude that the rate ratio is less than 1/1.3 = 0.77. For Naproxon and Ibuprofen, 95% credible intervals are wide and just fail to exclude 0.77.
“Contrary to Trelle et al., it may be in the interests of doctors and patients, let alone of pharmaceutical companies themselves, to randomize large numbers of patients between NSAIDs that may differ in whether they are associated with increased risk of stroke or of MI. If not, patients and their doctors will have to make their judgements on the indirect and often imprecise evidence that this important network meta-analysis offers.
“Rightly, the authors point out that all seven NSAIDs have centrally-estimated total-mortality rate ratios above one. Could further analyses by NSAID-subclasses add insight?”
Sotiris Antoniou, Consultant Pharmacist in Cardiovascular Medicine at the Royal Pharmaceutical Society, said:
“The studies carried out in this research are based on chronic and maximum doses of ibuprofen, taken for at least a year. Pharmacists have been aware of the small increased risk of cardiovascular events associated with high doses (2400mg daily) of ibuprofen for long term treatment, in line with clinical trial evidence and research to date. As such, it is always recommended to use the lowest effective dose for the shortest duration of treatment.
“The dose used most commonly by people is a low dose of 200mg – 400mg up to three times a day i.e. taken intermittently in line with the individual OTC product licenses. Overall there is no suggestion in these studies that low dose ibuprofen (less than 1200mg daily) is associated with an increased cardiovascular risk.”
Professor Simon Maxwell, Chair of the British Pharmacological Society’s Prescribing Committee and Professor of Clinical Pharmacology at the University of Edinburgh, said:
“This paper adds to the growing body of evidence that regular use of NSAID drugs (e.g. ibuprofen) is associated with a small but measurable excess risk of cardiovascular events.
“While this finding may be a cause of alarm to some it should be seen in context. Most users of these drugs will only take them for a relatively brief duration to treat short lasting episodes of pain and are at minimal risk. Patients with chronic pain may find themselves using such drugs for more prolonged periods. While this study suggests they might be exposed to some excess risk the alternatives may be less acceptable. Living with pain severely impairs quality of life. Alternative painkillers are also associated with very real adverse effects which are arguably less acceptable.
“The advice to the public should continue to be to take this class of drugs only when they are genuinely necessary to control pain and in the lowest effective dose, not only because of any cardiovascular risk but also because of potential effects on the stomach.”
Professor Philip Bath, member of the British Pharmacological Society and Stroke Association, and Professor of Stroke Medicine at the University of Nottingham, said:
“The work by Trelle and colleagues is important because it reminds us that non-steroidal anti-inflammatory drugs (NSAIDs) carry potential disadvantages as well as the benefit of being useful pain killers. Using data from 31 randomised controlled trials, and the statistical technique of network meta-analysis, the paper shows that taking either older NSAIDs, such as ibuprofen and diclofenac, or newer NSAIDs, such as rofecoxib, may increase the risk of stroke, heart attack and vascular death.
“While it is conceivable that the drugs are toxic to blood vessels and their contents, as they are to the kidney, it is also possible that negative effects of stroke and heart attack are because NSAIDs reduce the effect of other beneficial drugs such as aspirin.
“Whatever the cause, physicians and patients need to balance the benefits of using NSAIDs to reduce pain, sometimes excruciating, with the hazards of using these drugs, which can include gastric bleeding and kidney impairment. This balance needs to take account of the availability of other painkillers, including paracetamol and opiates. But if a NSAID is needed, as it often will be, then this new work suggest that naproxen may be the safest drug in people who have previously had a stroke or heart attack.”
Professor Donald Singer, member of the British Pharmacological Society and Professor of Clinical Pharmacology and Therapeutics at the University of Warwick, said:
“This is an interesting but complex study aiming to use new statistical methods to compare cardiovascular risks between non-steroidal anti-inflammatory drugs [NSAIDs] and against placebo across many studies when those contrasts are not available from direct comparative studies.
“There are two main cautions about these findings. First, headline results are given as relative risks. It is very important to be given clear information on absolute rates adjusted for other obvious risks in these older populations studied. Second, the statistical approach used by the authors is at risk of bias: apparent differences in risk between treatments may be due to differences between study groups rather than to differences between drugs.
“For the method in this paper (network meta-analysis) treatment groups should have very similar characteristics. However there are obvious differences among treatment groups. That can be seen from the wide (around 3-fold) difference in stroke rates on placebo treatment in different studies [25 to 82 strokes per 10,000 patient years of treatment]. And, for example, although ibuprofen is flagged as linked to higher stroke risk, the actual risk of stroke in the ibuprofen studies appeared low and within the range for many of the placebo studies [24 and 53 strokes per 10,000 patient years of treatment].
“Furthermore we are not told about adjustments across treatment groups for degree of cardiovascular risk burden as a source of bias and the authors acknowledge that some of their findings are imprecise. Pending further information, it would seem sensible to be cautious and to take into account cardiovascular risk when NSAIDs are prescribed or obtained over the counter.”
Cardiovascular safety of non-steroidal anti-inflammatory drugs: network meta-analysis, Trelle, S. et al, BMJ 2011;342:c7086, published in the British Medical Journal, Tuesday 11 January 2011.