NNT: how many is too many?
As GPs face intense pressure to intervene, new figures raise questions over the utility of some common treatments, finds Caroline Price
nnt cover image 3x2
It is one of the thorniest problems that medicine faces.
All the pressure on GPs is to intervene earlier and prescribe more, and within the confines of a 10-minute consultation it can be difficult to evaluate the potential benefits and risks of treatments in a meaningful way.
The subtle pressure of single-disease guidelines and the QOF, and the less-subtle threat of legal action if anything goes wrong, has contributed to a situation where balanced clinical decision-making for GPs is rendered almost impossible.
GPs are often not given the tools to help individual patients make decisions. Medical papers – even in high-impact journals – frequently omit to set out the absolute benefits of treatments, and guidance often fails to include any of the information that GPs need to make a balanced choice.
But recent figures published by NICE may help to shed some light on some common treatments that GPs prescribe. For the first time, the institute has published a list of the absolute benefits of treatments in terms of their ‘number needed to treat’. And in some cases, these numbers are shockingly high.
For example, for every thousand low-risk patients prescribed statins for primary prevention, only a single stroke is prevented per year. Meanwhile, you would need to treat more than a thousand people with antihypertensive therapy per year to prevent one death. This is a far higher NNT than that of many interventions GPs are advised against prescribing due to lack of efficacy.
Of course, there are limitations to how far NNTs can be used to inform care decisions. Studies show patients may need help to understand what they mean1, but it is illuminating see the impact of interventions at a population level.
The chair of the committee that produced the NICE figures, Professor Bruce Guthrie, a GP and professor of primary care medicine at the University of Dundee, says: ‘I don’t think we very often have very detailed discussions about the actual benefit patients might expect when starting a treatment.
‘I think many of us, myself included, just have thresholds above which we treat everybody. If you talk to patients about the actual benefit, some, particularly those at lower risk, really don’t want to take a drug for the rest of their life for that kind of benefit.’
The event that threw this type of treatment decision into sharp focus was NICE’s recommendation in 2014 to offer a statin at a 10% risk threshold (over 10 years) for primary prevention of cardiovascular events.
The move was met by consternation among GPs, many of whom questioned the evidence supporting statin prescribing in otherwise healthy people. A Pulse survey shortly after the NICE decision found 55% of GPs would not personally take a statin, or recommend a family member do the same, at the 10% risk level.
Based on the available trials, NICE’s database shows statins prevent only a single CVD death per 2,000 people treated per year. They prevent a single stroke per 1,000 patients, and a single death per 600 patients. The figure for non-fatal MI is not much better – one for every 500 patients treated.
The accompanying numbers needed to harm do not help the case for treatment. According to the NHS shared decision-making tool3, one in 200 patients per year taking statins suffers from side-effects, including muscle pain and liver problems – although critics claim the actual number may be higher.5
And how the risk is explained can sway patients’ perception of the benefits.
As Dr John Ashcroft, GPSI in cardiovascular medicine in Derbyshire, notes, how GPs present the risk can sway patients considerably. ‘For a person with a one in five chance over the next 10 years of having a heart attack, stroke or sudden death that risk sounds high, and the option to take one statin tablet a day to reduce your risk by more than a third sounds good.
‘But if you’re told that if we treat 40 people with a statin for 10 years to prevent a single death – that doesn’t sound so good. Yet both statistics are true.’As Dr John Ashcroft, GPSI in cardiovascular medicine in Derbyshire, notes, how GPs present the risk can sway patients considerably. ‘For a person with a one in five chance over the next 10 years of having a heart attack, stroke or sudden death that risk sounds high, and the option to take one statin tablet a day to reduce your risk by more than a third sounds good.
And the figures for other treatments are just as high. Professor Guthrie says: ‘Despite all the concern about statins, I think clinicians are often surprised at the relatively small benefit offered by antihypertensives.’
According to NICE data, more than 1,000 people need to be prescribed antihypertensive therapy to prevent one death per year; in excess of 800 need to be treated to prevent one heart attack and nearly 600 to prevent one stroke.
Proponents of such interventions will rightly point out that even with high NNTs, thousands of deaths, strokes and non-fatal MI events will be prevented across the whole at-risk population.
But GPC prescribing lead Dr Andrew Green says: ‘It is always interesting to be reminded that the NNTs of many interventions we are advised to use because of their benefits are higher than those of ones we are advised to avoid because of their ineffectiveness.’
Dr Green cites the example of NICE’s advice to avoid treating sinusitis and otitis media with antibiotics because their benefits do not outweigh the downsides of treatment, despite having more favourable NNTs of around 15-20.
While the reasons for avoiding such treatments are complex – the negligible benefits are balanced by a relatively high risk of side-effects such as diarrhoea and of course the critical problem of rising antibiotic resistance – the comparison is sobering.
Professor Tony Avery, a GP and professor of primary health care at the University of Nottingham, says these NNTs mainly help with stopping treatments: ‘I think the NNT tables can be extremely helpful as a guide to which drugs to consider withdrawing first. For example, antihypertensives for primary prevention might be suitable candidates for withdrawal, whereas ACE inhibitors in advanced heart failure might not.’
And this all comes as the policy environment may be changing, with the medical royal colleges all recently publishing lists of interventions that should be reconsidered in patients, because of limited evidence of benefit, under the Choosing Wisely initiative.6
The recent NICE multimorbidity guideline – from which most of these NNT figures are derived – recommends a reduction in polypharmacy in patients who are on 15 or more medications or who are frail and elderly.
But is there an overall NNT cut-off beyond which GPs should not prescribe? Probably not, says Dr Julian Treadwell, vice-chair of the RCGP standing group on overdiagnosis and overtreatment. He recommends: ‘The thing to do is let the numbers speak for themselves… No one can say what is the “right” NNT – it will vary for the individual patient depending on lots of things.’
And perhaps that is the main message for GPs to take back into discussions with patients. The best kind of care is made possible by having the best evidence and the time to talk it over in consultation.
These NNTs may take us a step closer to better decision-making – but GPs having the time to help patients understand it still feels a long way off.
NNTs – not the complete picture
Relative risks tend to overstate the benefit of interventions, whereas NNTs are a way of GPs understanding absolute risk reductions at a population level.
But this can take some explaining to patients. One recent study underlined that patients dislike NNTs because they can be confusing, pointing out people commonly believed, for example, that an NNT of 17 means that one person is saved but the other 16 have a heart attack.1
NICE’s database also presents drugs’ absolute drug benefits as the number of ‘bad’ events avoided for every 1,000 patients treated. Other studies have favoured the approach used in NICE patient decision aids, which tell patients how many people out of every 100 treated will benefit and how many will not.2
The benefits of preventive treatment will also depend heavily on the patient’s baseline risk. For example, the overall NICE NNTs for antihypertensive therapy apply to populations that were at relatively low risk for cardiovascular disease, with a 10-year risk of around 5%.
The annualised NNT figures published by NICE also fail to address how the small benefits add up over time with longer treatment. So, over 10 years of statin treatment – the median duration of studies from which these data are derived – the NNT to prevent one death is 105.
1 Dickinson R et al. Providing additional information about the benefits of statins in a leaflet for patients with coronary heart disease: a qualitative study of the impact on attitudes and beliefs. BMJ Open 2016;6:e012000. tinyurl.com/info-statins
2 NICE Patient Decision Aid. Atrial fibrillation: medicines to help reduce your risk of a stroke – what are the options? tinyurl.com/NNT-AF
3 NHS. Shared decision making. tinyurl.com/NHS-shared
4 Jefferson T et al. Neuraminidase inhibitors for preventing and treating influenza in adults and children. Cochrane Database Syst Rev 2014;4:CD008965. tinyurl.com/tami-review
5 Abrahamson J et al. Should people at low risk of cardiovascular disease take a statin? Errata. BMJ 2014;348:g3329
6 Academy of Royal Medical Colleges. Choosing Wisely UK. tinyurl.com/aomrc-choosing