Applying the evidence to individual patients

Trial data shows us what happens over a limited period, sometimes less than 1 year. The absolute risk reduction shown over these short time periods may be very small, perhaps just 1-2%.

But absolute benefits are likely to add up over time. So, for example, a treatment with a trial result showing a 1% absolute risk reduction in a 1 year period might give a 5% absolute risk reduction over 5 years.

• This would also be reflected in NNTs. In this example we would see an NNT of 100 over 1 year, and an NNT of 20 over 5 years.
• So it is important not to “write off” an apparently small absolute risk reduction if it is over a short time period.

However, there is mostly no scientific proof that this increase in benefit will be the case. Some treatments may have their biggest impact over the short term. For example:

• Antiplatelets or beta-blockers after acute MI, where the benefits are to do with stabilising an unstable physiological situation.

It is also important to think about the prognosis of a condition without treatment. For example heart failure with reduced ejection fraction (HF-REF) deteriorates over time if untreated, so the effects of treatment need to be considered against that alternative scenario.

Over the much longer term (say, more than 10 years), things are even more uncertain. For example:

• Does 30 years of anti-hypertensive treatment give someone 10 times the risk reduction as was seen in a 3-year trial?

This seems unlikely. This sort of calculation suggests risk can be reduced to zero with enough time.

There are also what are known as “competing risks”. People develop other illnesses and die from other causes, so they do not benefit from the theoretical risk reduction of the treatment being considered.

If the trial participants were mostly White, North American males aged 60 with a single condition and the research was conducted in the 1980s … can I apply those results to a patient with very different characteristics today?

Again, there is little hard science to guide us (and if there was, it would probably be condition and treatment specific).

As a rule of thumb, it may be best to think more in terms of physiology and baseline risk rather than demographics.

Physiological differences: co-morbidities and metabolic impairment, frailty, polypharmacy will all affect the benefit-risk ratio of a treatment, particularly risk of treatment harms.

• See the co-morbidities and polypharmacy section for more on this.

Baseline risk: results from clinical trials conducted in particular high-risk groups are sometimes extrapolated in clinical guidelines to apply to people at lower baseline risk. This means that a lower-risk person is less likely to benefit from treatment than the trial population. Where this is a particular issue, we have highlighted it on this website.

Very old age and the presence of life-limiting illness: “competing risks” become key. How likely is it that this person will live long enough to benefit from preventive treatment? How likely are they to die from another condition or old age? Will treatment side effects cause suffering in the time they have left?

BUT

Differences in sex, ethnicity, age and geography are probably less important. Unless there is evidence or a good physiological reason to think different demographic groups will respond differently to a treatment, it is probably best to assume that benefits apply to all. The risk of denying effective treatment to someone because their demographics differ from a trial population is probably greater than a theoretical risk from applying evidence from another demographic group to the person in front of you.

Clinical trials tend to recruit healthier people in middle and early old age, exclude people at high risk of treatment harms, and provide high quality follow up to minimise treatment risk.

This may mean your patient is more likely to experience harms than the trial participants were. Unfortunately there is no formula for estimating the risks to an individual based on trial data.

However, looking at the demographics and biological parameters of the study population can give you a guide as to “how different” your patient is from the average in the trials, and how much more at risk they might be from treatment harms.