Tom Chivers

Tom Chivers is a science writer. His first book, The AI Does Not Hate You is out now.


I had a conversation with someone recently about cancer screening. He had a friend who had been screened for prostate cancer: the friend had the PSA (prostate specific antigen) levels in his blood checked. It was found to be high, and they discovered a tumour. The cancer was removed; a life was, possibly, saved.

Stories like this are compelling. It’s hard to argue, in the face of a person who is alive and who otherwise might not have been, that cancer screening doesn’t work. Plainly it does – for that person, at least. But it is more complex than anecdotes of these kinds suggest.

That’s why I’m only cautiously excited when I hear about a new form of cancer screening that can detect many different kinds, with a blood test, in just 10 minutes. The test – still at an early stage of research, but showing promising results – detects circulating tumour DNA (ctDNA), bits of the cancer that have fallen off the main tumour and are floating around the bloodstream.

It might sound ridiculous to suggest that cancer screening might not save lives, or that it can even kill people. But let me try and explain why that might be the case.

Cancer is a frightening disease. A diagnosis still sounds like a death sentence, and we talk in terms of the need to ‘fight’ it or ‘beat’ it – even though treatment has improved to the point that for many people it is a chronic disease, one that you can live a full life with, for decades, before dying of something else.

But for a long time, treatment was less effective. So there was a great drive to improve early detection, to allow improved treatment. Part of that was the introduction of screening programmes. Screening for breast cancer was introduced in 1987; cervical and bowel screening came later. Prostate cancer screening is available for men over 50 if they ask for it.

It sounds like an unalloyed good: you screen for cancer, and if you have it, you get treated. If you don’t, you don’t. But in fact the evidence that it does good is highly variable. In the case of breast cancer, it’s shaky, and for prostate cancer, it’s essentially non-existent – the NHS says that it has not been shown that “the benefits outweigh the risks” (although the situation for bowel cancer is better). Here’s why.

First, any cancer screening test will have false positives. That’s inevitable. But because cancer is quite rare in the general population, even a low false positive rate will mean that many – possibly most – apparent positive results will be false. If you’ve got a 99% accurate test, you might think that if you get a positive result, there’s a 99% chance you’ve got cancer.

But it’s not as simple as that, because the tests are imperfect, and because the programmes are expensive: spending money on them means not spending it on treating cancers.

First, the imperfect tests. Imagine you have 1,000,000 people you want to test for cancer, and that one in every 1,000 actually has it. You run your 99%-accurate test on all one million of them. Of the 999,000 people without cancer, it will correctly say that 989,010 don’t have cancer; of the 1,000 who actually have cancer, it will correctly say that 990 of them do.

But look! Of the 11,980 people who it said had cancer, 10,990 don’t! You’ve unnecessarily scared 10 people for every one you’ve correctly detected.

These are made-up numbers, but the effect is real. A 2016 meta-analysis by the US government found that 60% of women who have a mammogram every year will have at least one false positive. The PSA test for prostate cancer has a false positive rate of 70%.

Getting a false positive is not harmless – as anyone who knows someone who’s had cancer can tell you, the scans and interventions are exhausting and demoralising. The meta-analysis I just mentioned found higher levels of depression and anxiety among women who’d had false positives than those who hadn’t. Plus the X-rays, operations and biopsies that result are dangerous; they cause a significant number of deaths.

The other thing is that even a true positive is not always a good thing. Prostate cancers, for instance, are often slow-growing, and usually diagnosed late in life. If you are screened, and one is found, it may be that if you hadn’t been, you might have died with the cancer, not of it; you may never have known about it. If, instead, you undergo surgery to remove it, you could die on the operating table. Or you could endure unpleasant radiation or chemotherapy to destroy it, all unnecessary.

Taking these facts into account, the evidence for the benefit of most types of screening is weak. And money you spend on screening is money you can’t spend on treatment. A 2017 study in the Netherlands found that screening programmes have reduced women’s annual risk of death from breast cancer by at most 5%. Improved treatments, on the other hand, have reduced it by 28%. This is a real trade-off that you can’t avoid; a pound spent somewhere in the NHS is a pound not spent somewhere else.

But there is huge public pressure to maintain screenings. “It’s so emotive,” an epidemiologist I know told me. “People think it’s their right, and that it’s safe.”

Politicians love screening programmes because they’re big, eye-catching things they can introduce, and ‘prevention is better than cure’ is the sort of folk wisdom we can all get behind; in October the Department of Health announced plans to catch 75% of cancers at or before stage 2 by 2028. The public hear high-profile examples of cancers detected, such as Stephen Fry’s. But you don’t hear anecdotes about the melanoma treatments not funded because we spend money on screening programmes.

There’s an analogy here, I think, with grammar schools. The evidence for their benefit is weak; although the people who go to them usually do well, they have a negative impact on the schools around them. Taking all the brightest, most motivated pupils away makes the remaining ones harder to teach, and the schools fill up with middle-class children anyway, so they don’t improve social mobility. But the people who go to them are extremely aware of the benefits, and unaware of the costs. So you get a vocal pressure group in their favour.

Something similar is going on here. If a PSA test detected an early-stage cancer in your uncle, you’ll have strong opinions about whether the PSA test is a good thing. Same if your breast cancer was detected in a mammogram. But all the people who die because there wasn’t enough money to fund a new CyberKnife radiotherapy centre, or to make a new immunotherapy drug available, aren’t as obvious to you. And once a screening programme is introduced, it is politically almost impossible to end: the benefits of ending it are widely spread and hard to see, but Aunt Jane dying of a breast cancer that screening might have caught is very obvious.

Some screening definitely has benefits; bowel cancer tests, in particular, are cheap, safe, and can be done at home. And you can reduce your false-positive ratio, and therefore your waste, by targeting screening on at-risk groups – older people, or those with family history or relevant exposure – rather than the population as a whole. For instance, even if you screen young women for cervical cancer every year, it doesn’t protect them as much as screening middle-aged women just once every three years. Clever targeting improves your cost-benefit ratio enormously.

But the introduction of a new way to screen for cancer should not be something we automatically get excited about. There have been extraordinary advances in cancer treatment in recent years; near-miraculous shifts in clinicians’ ability to treat previously deadly diseases such as melanoma. If a new screening programme is introduced – even if it works perfectly – it will mean we have less money to spend on those treatments. That might sometimes be worth it, but the experience so far suggests that we should be wary; and once a screening programme is in place, it will probably never go away.