Measuring success in cancer screening

Breast cancer cells

One of the more polarised debates about healthcare – and one that shows no sign of reaching consensus – is whether screening healthy people for signs of cancer does more good than harm.

This month (breast cancer awareness month, in case you hadn’t noticed) has brought another flurry of accusations that breast cancer screening is either (a) a life-saver that no sane woman should refuse, or (b) a pointless exercise that leads to over-diagnosis, over-treatment and over-anxiety. Also this month, the US Preventative Task Force ruled that prostate cancer screening should not be used, because the benefits are, at best, small, and the test can lead to unnecessary harm. The ruling was met with howls of fury from doctors and patients who claimed the test had saved their lives.

Among the noise, it’s hard to know what and who to believe. For one thing, the estimates of the effects of cancer screening vary widely, depending on how you calculate the figures. I’ve more or less given up trying to follow all the calculations and recalculations of the effect of breast cancer screening. My hunch is that, if there is a protective effect, its a small one.

But an interesting debate in the BMJ (British Medical Journal) this week threw light on some of the controversy. The question sounded innocuous: Should we use ‘all cause’ mortality to judge the success of cancer screening programmes, or ‘disease specific’ mortality? In laymen’s terms, should you look at whether people who take part in cancer screening live longer overall, or should we just look at whether they are less likely to die from the cancer being screened for?

At first glance, that sounds straightforward. If you’re screening for bowel cancer, for example, you expect that to have an effect on how many people die of bowel cancer, not how many die of other things. It would seem unreasonable to expect screening to cut deaths overall.

But wait a bit. Many people who die after being diagnosed with bowel cancer won’t have ‘cancer’ on the death certificate. They might die of pneumonia, while weakened from cancer and chemotherapy. They might die of a deep vein thrombosis, after bowel cancer surgery. Are these deaths caused by bowel cancer? Yes, probably. But they might not be included in your analysis, if you relied on cancer-specific mortality. By not recording deaths like this from people who’d been screened, you’d over-estimate the protective effect of screening.

And what about direct harm from screening? A positive fecal occult blood test often leads to a colonoscopy examination, to find out whether the positive test was caused by cancerous growths. Rarely, a colonoscopy may damage the bowel wall, creating a risk of septicaemia, or if you’re really unlucky, death. It’s not very likely. But if it happened, would it be included as a death from bowel cancer? Probably not, because the person being examined quite likely didn’t have bowel cancer. So this death would not be included in your analysis of how likely bowel cancer screening is to protect you from dying of bowel cancer. That seems wrong to me.

If, by contrast, you look at death from any cause, you have a nice, clean outcome that no-one can be confused by. People are either alive or not. The down-side is that, as deaths from individual cancers only account for a small proportion of overall mortality, you would need big, long-term studies to show any advantage. To date, no cancer screening programme has shown that it can improve all-cause mortality.

It might help to stand back and think about what we’re trying to achieve with cancer screening programmes. Is the idea simply to prevent deaths from one specific cause – in which case we’ll need lots of programmes to test for as many cancers as possible? Or is it to help us to live longer, healthier lives? The first might just be achievable through cancer screening programmes. The second is probably not.

It’s all explained in greater detail in the BMJ articles. NHS consultant physician James Penston says we should measure success by all-cause mortality. Robert Steele and David Brewster, respectively directors of the Scottish Bowel Cancer Screening Programme and the Scottish Cancer Registry, think not. What do you think?

%d bloggers like this: