We must try harder to avoid biological bias in clinical trials | Artificial Intelligence
TO EVALUATE a new drug, you need a clinical trial that’s designed to most clearly reveal its effects. Test it in too variable – or too sick – a group of people, and you are less likely to pick up the powerful effects you hope it is capable of.
That’s just statistics. But such efforts to get the clearest signal have led drug developers to skew clinical trials to one particular group: white people. As many as 86 per cent of participants in drug trials are white, according to one 2014 analysis. This is a problem: a person’s ethnicity can influence how effective or dangerous a drug is, as can their age, gender or weight. Testing a drug on a group that doesn’t represent the wider population means that guidelines on how to use it will then largely apply only to a subset of people.
In the past, many explanations have been put forward for the low numbers of, for example, African Americans in clinical trials. These included lower awareness of trials, low numbers of black biomedical researchers, and a historically justified lack of trust in the US medical establishment.
But it’s time to face the fact that the entry requirements are also against them. Just as women have been excluded from research due to fluctuating hormones, minority volunteers can be rejected from trials due to stats that gauge their health. For example, over-reliance on crude markers for kidney or immune health may explain why so few black men are included in prostate cancer trials, despite the disease being more common in this group (see “Black men are left out of cancer trials because of their biology).
This is of particular concern given that minority groups often experience worse health than the population average. So how can we better represent real populations and still detect benefits of new drugs? Bigger trials or extra trials in subgroups may help, although both will cost more money. Whatever we do, what’s certain is that we must act.
This article appeared in print under the headline “Biological bias”
More on these topics: