Rapid Responses to:

NEWS: Susan Mayor Statins cut cardiovascular events by a quarter in people with diabetes BMJ 2003; 326: 1348-b [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Relative risk and NNT

Nicholas J Devine   (23 June 2003)

Re: NNTs

Matthew L Grove   (24 June 2003)

NNT and costs

douglas n salmon   (25 June 2003)

Seeing what you want to see

Philipp Conradi, David Taylor   (25 June 2003)

Re: NNTs and costs

Matthew L Grove   (25 June 2003)

NNTs, costs, and preventing a quarter of Events

L S Lewis   (26 June 2003)

More on costs

douglas n salmon   (28 June 2003)

Relative risk and NNT 23 June 2003
Nicholas J Devine, GP Stockport SK7 3EP Send response to journal: Re: Relative risk and NNT	When are we going to see a consistant delivery of trial results quoting numbers needed to treat and absolute risk changes reather than the relative risk reductions which hype results and given no useful framework to compare the real benefits of treatments. This article is not a good piece of scientific reporting. Competing interests:   None declared
Re: NNTs 24 June 2003
Matthew L Grove, Consultant Rheumatologist NTGH NE29 8NH Send response to journal: Re: Re: NNTs	I agree with Nicholas Devine - reliance on relative risks for headline rates is exaggerating therapeutic benefit. But Medical journalism had at least improved to the point where Susan Mayor has included the absolute event rates (601 (20.2%) first events in the simvastatin group, 748 (25.1%) in the placebo group) and the total study size (5793). The difference between the % event rates is 25.1-20.2% = 4.9%, which gives a rough NNT of 21. Alternatively, given that the study size was 5793, and assuming a 50:50 split between active and placebo arms (say 2897:2896 patients) the absolute event rates can be used to calculate an NNT of 19.7 with 95% c.i. 13.8-34.6. The first calculation can be done by eye, the second takes 5 minutes with a palmtop NNT calculator. (As an aside, this highlights the statistical "fudging" that regularly goes on in drug trials. If you attempt to deduce the sample sizes of the active and placebo arms from the absolute and % event rates you come up with: 601/.202 = 2975 and 748/0.251 = 2980, for a total study size of 5975. The absolute and/or % event rates have been "adjusted", no doubt to account for confounding variables. Of course, you could just read the original paper in Lancet [no doubt free internet access?] to get the sizes of the study groups - but attempting to reconstruct the statistical method often highlights what has been changed) The final fudge is the attempt to argue that compliance has diluted the effect. This sounds suspiciously like an attempt to avoid an intention -to-treat analysis, and ignores the likelihood that compliance is likely to be even lower in clinical practice (and the benefits even more dilute). Final comment: in self defence, get a palmtop. Competing interests:   NNT buff (I know, it's sad)
NNT and costs 25 June 2003
douglas n salmon, gp partner b20 3he Send response to journal: Re: NNT and costs	In summary, treat 20 diabetics for five years, £40,000 drug costs alone, and prevent 1 event. How does this compare with other interventions ? Competing interests:   None declared
Seeing what you want to see 25 June 2003
Philipp Conradi, GP 10 Sladepool Farm Road Maypole Birmingham B 14 5 DJ, David Taylor Send response to journal: Re: Seeing what you want to see	Dear Sir Re: Mayor S. Statins cut cardiovascular events by a quarter in people with Diabetes. BMJ 2003: 326: 1348 “Seeing what you want to see” McCormack and Greenhalgh once argued that the interpretation and dissemination of results of randomised controlled trials is open to bias (1). This week’s BMJ (2) illustrates this perfectly, featuring an overly favourable comment on the Diabetes sub trial (3) of the Heart Protection Study (4). We would like to question the study’s relevance to General Practice by making the following points. 1) Generalisability - The diabetes sub trial is part of the main Heart Protection Study (4) published in 2002. The researchers in the original article excluded more than two-thirds of 63,603 pre-selected patients. Most of them opted out or were deemed not to be reliably compliant for the trial duration. Patients with a concomitant non-vascular disease were equally not included. 2) Bias - Reading through the study’s design (5) raises the possibility of a non-match between treatment and control group – there is a small but potentially important difference in the drop out rate. We already published our arguments earlier (Doctor Magazine 19.06.2003, page 33-34). 3) Bias - The sub study mentions the use of Metformin in 31% of patients with Type II diabetes but we do not understand if both groups used their medication equally. This is disappointing, as Metformin in UKPDS 34 has been shown to be the only drug to reduce the chance of macrovascular events (1). 4) Merging the boundaries - Both the Heart Protection Study and the diabetes sub trial look at patients with and without pre-existing occlusive disease. This is misleading, as patients with an established cardiovascular complaint have a higher risk and would warrant lipid lowering therapy anyway. We disagree with the impression in the BMJ’s commentary, that the case for using statins in patients without established cardiovascular disease and LDL <3.0 mmol/l is now made; the 3.1% absolute difference between treatment and placebo group is only marginally significant at p=0.05. 5) Why not use NNTs? - Finally, we would like to highlight a general problem with result presentation as found in the HPS, ASCOT-LLA and many others. Relative risk reductions and composite end points are misleading and unsuitable to explain a potential benefit of a proposed treatment to a patient. Both forms of result presentation should not be used in serious medical journals. References: 1) McCormack J, Greenhalgh T. Seeing what you want to see in randomised controlled trials: versions and perversions of UKPDS data. BMJ 200; 320: 1720-1723. 2) Mayor S. Statins cut cardiovascular events by a quarter in people with diabetes. BMJ 2003: 326: 1348. 3) Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of cholesterol lowering with Simvastatin in 5,963 people with diabetes: a randomised placebo controlled trial. Lancet 2003; 361: 2005-16. 4) Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of cholesterol lowering with Simvastatin in 20,536 high risk individuals: a randomised placebo controlled trial. Lancet 2002; 360: 7-22. 5) MRC/BHF Heart Protection Study Collaborative Group. Heart Protection Study of cholesterol lowering therapy and antioxidant vitamin supplementation in a wide range of patients at increased risk of coronary heart disease death: early safety and efficacy experience. European Heart Journal 1999; 20: 7254-741. Competing interests:   interested in real evidence based medicine
Re: NNTs and costs 25 June 2003
Matthew L Grove, Consultant Rheumatologist NTGH, NE29 8NH Send response to journal: Re: Re: NNTs and costs	Well, strictly speaking you can be 95% certain that if you spend £70,000 over five years you should prevent an event. Of course, it might cost as little as £28,000 ... It's actually quite comparable to other trials. For instance, the NNT to prevent a vascular event in the HPS trial was 18 (95% ci 15-23) and to prevent a vascular death was 68 (95% ci 45-138). Secondary prevention is better of course; NNT to prevent vascular death in the 4S trial was 29 (95% ci 18-56). All of these over similar time periods (about 5 years seems traditional for statin studies). Competing interests:   None declared
NNTs, costs, and preventing a quarter of Events 26 June 2003
L S Lewis, GP Newport, Pembrokeshire, SA42 0TJ Send response to journal: Re: NNTs, costs, and preventing a quarter of Events	Actually , all Statin trials (whether 4S, Woscops, HPS or whatever ) all show much the SAME thing, viz: " Statins reduce Coronary events by approx 25% regardless of Cholesterol level.. " But where the trials differ is in the CORONARY RISK of the treatment group, which matters MOST of all to the 'rationing' question ( Highest risk patients should be treated first.. ) In fact, the coronary Risk, the cost of the drug, and the now generalisable [ 25% ] relative risk reduction are the ONLY factors required to calculate the NNT and the Cost-per-event prevented. Thus, assuming Stain cost =£29 per month, then :- Risk 30-50 % ( eg: Known IHD - such as 4S ) : Cost £10,000 - 20,000 Risk 30+ % ( eg Woscops ) cost per event approx £35000 Risk approx 20% ( eg HPS ) cost per event approx £55000 Risk 10-15% ( eg Diabetes HPS subtrial ) Cost per event = £75,000- £100,000 Risk 5% , eg: Women in 4S .. figures too small to reach significance !! Competing interests:   GP Contract quality Targets vs. a 'drug budget'
More on costs 28 June 2003
douglas n salmon, gp b20 3he Send response to journal: Re: More on costs	With between 1.4 and 2.4 million diabetics in the UK ( diabetes uk website ), costs of implementing a 'statin for all diabetics' policy would be of the order of £560 million - £1000 million per year in drug costs alone. Total NHS primary care drug budget is only about £6000 million at present.Is this a serious option ? Competing interests:   None declared