NPC Archive Item: Even more data on cardiovascular risks of NSAIDs: time to review prescribing

10 October 2011

A meta-analysis of observational studies has added to the evidence base confirming an increased risk of cardiovascular (CV) events with coxibs and traditional NSAIDs. The study suggests that diclofenac elevates CV risk at 100 mg/day or less (close to the maximum recommended dose for over-the-counter diclofenac products). It also confirms that naproxen and low-dose ibuprofen are least likely to increase CV risk. Although this study, which has been widely reported in the lay media, adds important details to the picture, the main findings should not be new news to health professionals.

Level of evidence:
Level 2 (limited quality patient-oriented evidence) according to the SORT criteria.

Action
Prescribers should continue to follow our previous advice given in MeReC Extra 30 (November 2007). They should assess the CV and gastrointestinal (GI) risk of each patient individually and carefully consider the balance between benefit and risk before starting treatment with any NSAID. They should continue treatment with NSAIDs for the shortest time and at the lowest dose necessary to control symptoms.

• Low-dose ibuprofen (≤1200 mg/day) is an appropriate first choice NSAID in view of its low risk of GI and CV side effects.
• Low-dose ibuprofen or naproxen 1000 mg/day would appear more appropriate than other NSAIDs for patients in whom CV risk is a significant consideration in decision-making.
• Consider prescribing a proton-pump inhibitor (PPI) with any NSAID to reduce the risk of adverse GI effects, particularly in those who are at high GI risk (this includes anybody aged 65 years or older) and long-term NSAID users. Specific recommendations for co-prescribing of PPIs are provided in NICE clinical guidelines for osteoarthritis, rheumatoid arthritis, and low back pain.
• Although coxibs are associated with a lower risk of GI side effects than traditional NSAIDs, there is no good evidence to support the use of coxibs alone ahead of traditional NSAIDs co-prescribed with a PPI. Coxibs also have a higher CV risk than ibuprofen ≤1200 mg/day or naproxen 1000 mg/day.
• The MHRA advises that patients whose blood pressure is persistently above 140/90 mmHg and inadequately controlled must not receive etoricoxib. Furthermore, high blood pressure should be controlled before starting treatment, and should be monitored for 2 weeks after the start of treatment and regularly thereafter. It also advises that patients receiving etoricoxib are monitored closely for any signs and symptoms of cardiovascular side-effects (e.g. fluid retention, high blood pressure, shortness of breath, or chest pain).

Note also that the MHRA advises that patients at risk of renal impairment or renal failure (particularly elderly people) should avoid NSAIDs if possible. If NSAID treatment is absolutely necessary, then the lowest effective dose for the shortest possible duration should be used to control symptoms. The renal function of such patients should be carefully monitored during NSAID treatment.

Pharmacists and prescribers should consider discussing the possible renal, GI and CV risks and benefits of NSAIDs, including those bought over-the-counter.

Prescribing managers should review local prescribing trends for NSAIDs as suggested in the NPC document Key therapeutic topics – Medicines management options for local implementation

What is the background to this?
Health professionals were first alerted by the MHRA to the CV risks of non-selective NSAIDs such as diclofenac in October 2006, following a meta-analysis (MA) of randomised controlled trials (RCTs) by Kearney et al. The excess CV risk appears to increase with dose and persists throughout treatment. For example, a more recent cohort study found an increased risk within the first weeks of treatment of patients with prior myocardial infarction (MI).

In November 2007, the NPC placed this evidence in the context of the rest of the evidence relating to NSAID safety in MeReC Extra 30. An implementation programme of education events, audits and other local activities followed soon afterwards, instigated and supported by the NPC. This work by the NPC, its therapeutics trainers and by PCT medicines management teams has been associated with a decline in diclofenac use in many parts of the country (see review published in Prescriber). However, there are still wide differences between localities with regard to the total proportion of NSAID prescriptions which were for ibuprofen or naproxen: in the period April to June 2011 this varied from 31% to 72%.

What did this study set out to do?
Previous reviews and MAs have provided limited information on dose effects and relevant patient characteristics. In addition, most of the analyses have involved a few extensively investigated drugs, with little information on some other widely used drugs. The authors of this MA sought to update their previously published review of large observational studies to address these points.

The authors extracted the adjusted risk estimates for individual drugs and for the doses reported from non-randomised controlled observational studies. Thirty case-control studies included 184,946 CV events, and 21 cohort studies described outcomes in >2.7 million exposed individuals. The authors calculated pooled relative risks (RRs) for 11 drugs compared to controls, and were able to stratify the risk by dose for five drugs.

What does this study claim?
The authors state “This review suggests that among widely used NSAIDs, naproxen and low-dose ibuprofen are least likely to increase cardiovascular risk. Diclofenac in doses available without prescription elevates risk.”

Risk estimates were constant with different background risks for CV disease and rose early in the course of treatment. The highest overall risks (for NSAIDs on the market in the UK) were seen with etoricoxib (RR 2.05, 95% confidence interval [CI] 1.45 to 2.88), etodolac (RR 1.55, 95% CI 1.28 to 1.87) and diclofenac (RR 1.40, 95%CI 1.27 to 1.55). The lowest were seen with ibuprofen (RR 1.18, 95%CI 1.11 to 1.25), celecoxib (RR 1.17, 95%CI 1.08 to 1.27) and naproxen (RR 1.09, 95%CI 1.02 to 1.16). CV risks were not statistically significantly increased by piroxicam or valdecoxib, but confidence intervals were wide reflecting the limited data and it should not be concluded that these drugs do not increase the risk of CV events.

When the risks were stratified by ‘high dose’ and ‘low dose’ for five NSAIDs, it was found that the risk associated with rofecoxib, celecoxib and diclofenac was statistically significantly increased at both high and low doses. For diclofenac, this included doses less than 100 mg/day. The risk with ibuprofen was significantly raised only at high doses (in most studies this was more than 1200 mg/day). The risk with naproxen was not significantly increased at either low dose or high dose.

So what?
This study supports what was previously known about the comparative CV risks of NSAIDs but adds important new detail.

It confirms an increased risk of CV events with diclofenac and, importantly, suggests that this risk is likely to be statistically significant and potentially clinically significant (depending on the patient’s baseline risk of CV events), even at comparatively low doses. It also suggests a high relative risk with less-well-studied drugs such as etoricoxib and etodolac. With regard to ibuprofen, it underlines the need to use lower doses (around 1200 mg a day or less) to avoid increasing the risk of CV events. Finally, it confirms that the risk of CV events with naproxen is low or absent at higher as well as lower doses. It has its limitations: the studies were not randomised and included patients from many different populations. Neverthless, the results are in keeping with the accumulating body of evidence.

The study has gained media attention in the lay media (Daily Telegraph, Daily Mail, Daily Express) and at least one newspaper for health professionals (Pulse). Pulse reports BMA Council and GPC member Dr Helena McKeown as saying “Yet again GPs wake up to hear on [BBC Radio 4’s] Today programme that concerned, often stable patients will be requesting face-to-face or telephone consultations about their medication.” She went on to say that she would have wanted practices to have had a chance to be proactive with information.

The issue of whether or not health professionals could or should be warned in advance of any articles published in learned journals that might be picked up by lay media is important. However, it is interesting that this study appears to have been seen, at least in a number of quarters, as new news, when the UK drugs safety regulator first wrote to health professionals about this risk five years ago. Why should this be, and why should there be a more than two-fold variation across England in the proportion of NSAID prescriptions for ibuprofen and naproxen?

All decision-making in healthcare requires the recall, interpretation and application of a large volume of information. The evidence around individual decision-making is covered in more depth in MeReC Bulletin 22:1 and in part 4 of the NPC’s DVD on ‘Making Decisions Better’. The processes humans use to hand large volumes of complex information are the same, whatever its context or type. There is a limit to the amount of information humans are able to use in decision-making and in the face of large volumes of evidence we usually truncate the amount we use, so as to make a ‘good enough’ decision. Furthermore, we tend to favour the intuitive, automatic way of processing this, known as System 1 thinking. This involves the construction and use of mental maps and patterns, shortcuts and rules of thumb (heuristics), and ‘mindlines’ (collectively reinforced, internalised tacit guidelines). These are developed through experience and repetition, usually based on undergraduate teaching, brief written summaries, seeing what other people do and talking to local colleagues, and personal experience. The alternative – System 2 processing – involves a careful, rational analysis and evaluation of the available information. This is requires effort and is time consuming. Both approaches have advantages and disadvantages: System 1 is fast and it would be impossible to practice as a health professional (or indeed as a human being) without it. However, in addition to the potential bias which is introduced because only a subset of the total evidence is known and used, there are many well-described cognitive biases that affect the heuristics and mindlines involved in this rapid decision-making. A major challenge for all practitioners is to pick out the information they need to inform their practice from the daily flood of information they receive. However, even if they succeed in this, the preference all humans have for System 1 thinking makes it hard to modify practice in the light of new information which conflicts with one’s previous (perhaps tacit) assumptions and knowledge. It also makes it difficult to switch into System 2 thinking when this is needed.

Looking at the prescribing data, it would seem that a significant proportion of prescribers have not heard of the increased CV risk associated with commonly used NSAIDs, including diclofenac, or have heard it but either not accepted it or not yet adapted their practice accordingly. By being more aware of the processes by which humans make decisions, practitioners are likely to improve the quality of their decision-making. Practitioners can also develop and improve the way they mange the enormous amount of information with which they are presented, through Information Mastery techniques, as discussed in the NPC’s e-learning resources and a MeReC Briefing. Finally, developing a network of colleagues who adopt a similar approach can be a very effective way to develop one’s practice. However, one must be continuously aware of the potential misleading effect of one’s own and others’ mindlines, and the risk of developing non-evidence-based group views.

Study details
McGettigan P, Henry D. Cardiovascular risk with non-steroidal anti-inflammatory drugs: systematic review of population-based controlled observational studies. PLoS Med 2011;8(9): e1001098. doi:10.1371/journal.pmed.1001098

Design: Systematic review of community-based controlled observational studies. Thirty case-control studies included 184,946 cardiovascular events and 21 cohort studies described outcomes in >2.7 million exposed individuals

Outcomes and results

Summary pooled relative risk (RR)

*NS: not statistically significant

Summary analyses for individual drugs, ranked from highest to lowest pooled RR (y–axis) based on the point estimates. Note: some of the estimates are imprecise, because of sparse data, as reflected in the wide CIs.

Dose response relationship

1. Diclofenac low/high dose: six studies ≤100 mg / >100 mg; two studies <100 mg / ≥100 mg; two studies <150 mg / ≥150 mg
2. Ibuprofen low/high dose: eight studies ≤1200 mg / >1200 mg; one study ≤1600 mg / >1600 mg; two studies <1800 mg / ≥1800 mg
3. Naproxen low/high dose: two studies ≤500 mg / >500 mg; four studies ≤750 mg / >750 mg; four studies ≤1000 mg / >1000 mg

*NS: not statistically significant

Sponsorship: No specific funding was received for this study

Further information can be found on NHS Evidence and in the musculoskeletal pain e-learning section of the NPC website

Please comment on this rapid review using our feedback form.

Make sure you are signed up to NPC Email updates — the free email alerting system that keeps you up to date with the NPC news and outputs relevant to you