Health Risk: Which Numbers Can You Trust?

by Research Insights0 comments

Share this page:

We often hear about the results of drug trials on the evening news. They’re usually accompanied by risk reduction percentages. These are often misleading, since the percentages quoted are ‘relative’ instead of ‘absolute’.

What difference does it make? A lot. The difference is anything but trivial.

It is an ethical imperative that every doctor and patient understand the difference between absolute and relative risks to protect patients from unnecessary anxiety and manipulation.’

Gerd Gigerenzer, Director, Harding Centre for Risk Literacy

Breast Cancer

When the patent on a drug runs out, drug makers often combine the old drug with another to make a new one.

This extends the profitable life of the older drug, in this case, Herceptin. The results of Roche’s Aphinity trial, published at the 2017 Asco conference in Chicago, showed that combining its breast cancer drugs Herceptin and Perjeta produced a small additional benefit.

After 3 years, 94.1% of the early-stage breast cancer patients in the combo arm of Herceptin and Perjeta had not developed invasive HER2-positive disease, but 93.2% of women taking Herceptin on its own hadn’t either.

The absolute risk reduction was 0.9%. José Baselga, Chief Medical Officer at Memorial Sloan Kettering and consultant to Roche, told the media that the difference translated into a 20% lower chance of developing invasive breast cancer. He said this was ‘very meaningful, on par with many other cancer therapies.’

The 20% was relative risk reduction. Wall St analysts were unmoved, and Roche’s share price fell by 5%. One analyst told the media that the results went over like a ‘lead balloon’, and that investors had expected more than a 1% difference.

Lipitor

health-risk-numbers-trust-lipitor

Statin drug trials have routinely cited numbers like 30% to 50% in relation to risk reduction for heart attacks, strokes and death. But, when you check for absolute risk, these claims melt away.

In the Lipitor example, the drug arm of the clinical trial registered 2 heart attacks per 100 subjects compared to 3 in the placebo group. This resulted in a  relative risk reduction of 33% (the inverse of 2:3). In the Lipitor ad at right, they were able to lift this to 36% risk reduction because the actual figure was 1.9 not 2 (per 100 trialists).

That’s sounds pretty impressive until you consider the absolute risk which was 1%. In other words, 1 heart attack was avoided per 100 people who took the drug for 5 years.

Would you take a drug for 5 years that promises to reduce your risk of a heart attack by 1%?

Probably not. That’s why drug makers quote relative risk numbers. Sadly this crucial difference is often missed in medical reports, media reports and lost on doctors.

Doctors need to understand absolute versus relative risk reduction with statins.

Dr Bernard Freudenthal, British Medical Journal

Stents for Stable Coronary Artery Disease

In Australia alone, some 50,000 operations involving stents are carried out every year, each at a cost in excess of $20,000.

The vast majority of these are performed on people with stable heart disease, despite a number of clinical trials showing that stents offer no advantage over best-practice medical care.

1 in 50 of those undergoing this operation experiences a serious complication such as death, stroke, myocardial infarction, arrhythmia and hemorrhage.

‘Nobody who’s not having a heart attack needs a stent.

Dr David Brown, Right Care Alliance.

NNT – Number Needed to Treat

This is a much better indicator of a treatment’s effects, since it tells us how many people have to be treated to achieve a single positive outcome. A really effective treatment might have an NNT of 1.2 – for every 5 people treated, 4 derive a benefit.

In a meta-analysis of 27 statin trials with 165,149 participants, the published results showed a 21% risk reduction in events such as heart attacks and strokes, or interventions such as heart surgery.

This sounds impressive until we see the NNT of 130 – the number of people who need to take a statin for primary prevention to prevent a single cardiac event. The NNT for preventing a single death is 500.

NNH – Number Needed to Harm

This number tells us how many people have to be treated to cause an adverse event. In one recent trial involving subjects without established heart disease (primary prevention), the NNH for muscle pain (myalgia) was 10, and the NNH for developing diabetes was 50.

With information like that, we can weigh up the pros and cons and make informed decisions. Sadly, NNT and NNH numbers are rarely published or buried in the fine print of clinical trial reports. The reason is pretty obvious: relative risk reduction makes feeble outcomes look impressive, and doctors are much more likely to prescribe drugs or treatments that promise big risk reductions.

Physicians’ views of the effectiveness of lipid lowering drugs and the decision to prescribe such drugs is affected by the predominant use of relative risk reduction percentages in trial reports and advertisements.’

British Medical Journal

Find out more in our blog: The 10 Drugs Boomers Should Not Take.

Get Updates

LIKE TO STAY RIGHT UP TO DATE?

Get free updates to your inbox

References

Asco Fallout: Roche in Damage Control as drug combo disappoints investors, Fierce Pharma
The Stats on Statins: Should Healthy Adults Over 50 Take Them? Scientific American
Statin Drugs Given for 5 Years for Heart Disease Prevention (Without Heart Disease), the NNT
How Statistical Deception Created the Appearance That Statins Are Safe and Effective, Crossfit
Stents for Stable Coronary Artery Disease. The NNT
Heart Stents Are Useless for Most Stable Patients. They’re Still Widely Used.’ New York Times

Kim Brebach

Kim Brebach

Hi, I’m Kim Brebach, boomer, information researcher, technical writer and Joiner of Dots at M&M. In my spare time, I review wines and love to cook.

[]
×
Like an EXTRA 5% OFF?

Claim your discount now

No thanks. I'm happy to pay the offered price. *You can unsubscribe at any time.