Risky Business: The 4 Medical Risk Numbers You Must Know

by Research Insights0 comments

Share this page:

So many medical risk numbers! What do they mean and which ones can you trust? I dig into the 4 most-used and most-misunderstood number when it comes to risk, especially risk reduction.

Main Points

  • There are four main risk numbers you need to know; mostly we only hear about one;
  • If trials quote risk and don’t define which number they use, the difference can be over 10-fold;
  • Number Needed To Treat & Harm are rarely quoted but change the picture entirely;
  • Some long-standing trials have become research bedrock but are anything but robust;
  • It pays to dig into numbers to see which ones they use & what they mean.

Risk Is Not Risk

We use the word ‘risk’ a lot on this website. It’s not a scare tactic. We believe you deserve to know the risks involved with drugs, procedures and even foods which could harm you. That way, you can have informed discussions with your health practitioners and make informed decisions, together.

Yet, with risk, it’s troubling; risk can be defined in so many ways. (Warning: there are maths and formulae coming up but don’t panic; they’re short, explained and vital.)

These are the four medical risk numbers you need to know (see full definitions below).

medical-risk-numbers-the-four

A Hypothetical

Let’s take these main four numbers associated with risk, pull them apart and show you what they mean, using an example.

Let’s create a hypothetical clinical trial. We’ll use a statin drug (for lowering cholesterol) as an example because, later, we have actual clinical trial data for statins. The formulae we use here are from the (US) National Library of Medicine.

Let’s say these were the data cited per 100 triallists for this hypothetical trial. (There would have been many more, probably thousands, but trial data is usually expressed per 100 (percent) for easy comparison, and trials usually run for 5 years.)

medical-risk-numbers-trial-data
Myalgia is severe muscle pain.

Absolute Risk Reduction for Heart Attack (ARR) = 1%

It’s called absolute because it’s the most accurate way to examine the results. In the case above, the formula is:

ARR = (Control – Test) expressed as a percentage

= 3-2 = 1%

In other words, 1 person in 100 benefited from the treatment (avoided a heart attack) compared to the control group, making this an Absolute Risk Reduction of 1.

Relative Risk for Heart Attack(RR) = 0.67

This figure compares just the two figures for Control and Test i.e. risk of treatment compared to no treatment, with no reference to ‘per 100 triallists’. The formula is:

RR = Test divided by Control

= (2/3) = 0.67.

The Relative Risk is 0.67 but this can’t be compared directly with the ARR (which is a percentage). To do this, you need the Relative Risk Reduction (%) which is the number most quoted in clinical trials.

Relative Risk Reduction for Heart Attack = 33%

The formula for this is:

RRR = (1-RR) expressed as a percent

= 1-0.67= 33%

When calculated this way, the Relative Risk Reduction is 33%. This how the sponsor of a drug trial could claim a 33% reduction in risk (of heart attack in this case).

In other words, relative to the control, indeed 33% fewer people had heart attacks (2 compared to 3) but, per 100 trialists, it was only one less person who didn’t have a heart attack. You can see why a drug company would opt to quote the RRR.

Number Needed to Treat (NNT) = 100

These next two numbers you’ll almost never see in clinical trial data, but they’re possibly the most important.

The NNT is defined as ‘how many patients need to be treated in order for one person to benefit.’

The formula for this is:

NNT = 100/ARR

= 100/1 = 100

In other words, in our hypothetical trial, 100 people needed to take the drug before 1 gained a benefit (didn’t have a heart attack).

Number Needed To Harm (NNH) = 100

This is a number you almost never see quoted in clinical trials because it shows results (like myalgia and diabetes) which are not trial targets (heart attacks are) and the NNH is always bad news.

The definition of NNH is: how many people need to be treated in order for one person to have an adverse effect.

NNH    = (100/number harmed)

Using data from our hypothetical trial:

For Myalgia = 100/1 = 100

For Diabetes = 100/1 = 100

So, what does it mean?

So, per 100 people in our hypothetical trial, when NNT and NNH are added, the picture looks quite different:

In other words, 100 people had to take the drug for:

1 to gain a benefit of no heart attack (ARR of 1 and NNT of 100);

1 to be harmed by Myalgia (NNH of 100); and

1 to be harmed by Diabetes (NNH of 100),

yet the drug manufacturer could still claim a 33% reduction of risk of a heart attack.

But you say, this hypothetical data is exaggerated. Real trial data could never be like that. Let’s see.

An independent meta-analysis

Statins have been around since 1987 and, apart from being the highest-selling and most profitable drugs of all time, they’re also the most studied.

One pre-eminent study group, the Cholesterol Treatment Triallists (CTT) Collaboration, hails from Oxford University, England. The CTT was established in 1994 it says: ‘after it was recognised that no single randomised controlled trial of a lipid intervention would have enough participants to enable reliable assessment of mortality outcomes or assess effects in particular types of patient’. In other words, individual trials were too small, so pooling them into meta-analyses (studies of many studies) would give more accurate results.

That makes huge sense, especially if the trial data were made public, including any adverse effects (NNH) but they were not then or now. We’re told that the CTT Collaboration holds the raw data of some 30 statin drug trials, which it can’t share because they ‘remain the property of the trial sponsors‘.

The graph below is what you’ll see if you click ‘findings’ on the CTT Collaboration website. Note that there are no numbers on the Y axis, and the word ‘absolute’ is used, even though the ‘~ 1 in 3 ‘ is the Relative Risk Reduction (33%), not the Absolute Risk Reduction (see analysis later). This is as rigorous as it gets.

Source: CTT Collaboration

By the way, the graph’s title translates to: ‘How taking statins for lowering cholesterol affects heart disease’. As we’ll see below, although the take-home message is ‘no stains mean high risk, some statins mean lower risk, more stains mean even lower risk’, clinical trial data do not support this.

Confused? Maybe That’s The Idea.

The CTT Collaboration findings are independent, assured by its Independent Oversight Panel which met for the first time 23 years after the collaboration was formed. Independence is also assured by its funding, being from groups like the Australian National Heart Foundation, its British equivalent and a host of similar groups. ‘The CTT Collaboration has not received grant funding from industry’, its website says.

Yet, several sources (here, here and here) claim that, according to the CTT’s own admission, the collaboration, its affiliates and funders are up to 80% funded by undisclosed sources, they say, most being statin drug-makers like Merck. Not for me to conclude, but the simplistic, misleading representation of the CTT’s findings (above), the lack of sharing trial data and of collaborating with other researchers, certainly raise some questions.

So, it’s not surprising that we couldn’t get any original trial data with values for the four vital numbers. To get these, we found a huge meta-analysis which combined CTT’s published findings, and those of three other groups which, together, included nearly 300,000 triallists of statin drugs. These data are all from trials of people of low risk; in other words healthy people on whom statins were being tested as a preventative for heart attack, not on populations with previous heart attack histories. These are those all-important four numbers across those four meta-analyses:

medical-risk-numbers-data-explained
Source: Statins in Persons at Low Risk of Cardiovascular Disease

In other words:

4 people in 1,000 didn’t have a heart attack as a result of taking statins (not 1 in 3);

250 had to take statins before 1 benefited; and

20 had to take statins before 1 was harmed by myalgia; 204 had to take them before 1 was harmed by diabetes.

All up, the risk of harm was greater than the risk of benefit, yet a Relative Risk Reduction for statins could still be claimed as 33%.

One question

Would you take a drug for 5 years (or maybe for life) if it only gave you a 0.4% chance of avoiding a heart attack?

Statins are just one drug class with dubious benefits for boomers. Find out more on our blog 10 Drugs that boomers should never take.

Get Updates

LIKE TO STAY RIGHT UP TO DATE?

Get free updates to your inbox

Kim Brebach

Tracey James

Hello, I’m Tracey James, boomer, former scientist, technical writer and Fixer of Things at M&M. In my spare time, I like to walk, swim and garden.   

[]
×
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.