New Scientist (pg. 19), December 16, 2006


[Rachel's introduction: After 35 years of waging a "war on cancer" maybe it's time to take stock. If progress continues at the present rate, cancer deaths in the U.S. should be entirely eliminated by the year 3508, a little more than 1500 years from now.]

By Ralph Moss

Next week will mark the 35th anniversary of the National Cancer Act, the opening salvo of the US government's battle to eradicate cancer. In those 35 years, the US has spent tens of billions of dollars on cancer research, and we are frequently told that this has won us significant progress. Look at the data, though, and the picture is more confusing. Are we really winning the war on cancer?

In 1971, when the National Cancer Act was signed, we were assured that cancer would be cured by 1976. Since then this and other targets have come and gone, leaving the ultimate goal as distant as ever. Until recently, the National Cancer Institute (NCI) in Bethesda, Maryland, stuck by the astonishing claim that all suffering and death due to cancer would come to an end by 2015, and continues to quote a former director as saying that every benchmark of the 1971 congressional mandate has been achieved.

In the US, the number of people diagnosed each year with cancer stands just shy of 1.4 million, nearly double what it was 35 years ago, and 1 in 2 men and 1 in 3 women are expected to get cancer at some time in their lives. Those who wish to continue on the course we have followed till now naturally present the situation in as favourable a light as possible. We are told that while a cancer diagnosis 35 years ago was inevitably a death sentence, many cancers today are curable. Yet it has been known for 100 years that cancers are generally curable if they can be removed while still in their early stages. When somebody dies of cancer it is usually because it has spread from one site in the body to another, yet over the past 35 years the death rate from most of these metastatic cancers has remained largely unchanged.

Similarly, apologists for current strategies often cite improvements in five-year survival figures as proof of progress. While only half of those diagnosed with cancer in the 1970s survived five years, today two-thirds survive to the five-year mark. On the surface, this sounds like a big step forward, but it ignores a statistical artefact known as "lead-time bias". Thanks to widespread screening, people are often now diagnosed with cancer earlier in the course of their disease than they would have been in the past. However, the natural history of the disease has not changed at all: the time of death is typically the same as it would have been had the disease been diagnosed later.

It is also important to examine what exactly the term "cancer" is being used to describe. The introduction of sensitive screening tests has meant that more patients are being diagnosed with pre-symptomatic forms of the disease. Up to 30 per cent of all breast cancer diagnoses, for example, are now of the pre-cancerous (in situ or non- invasive) type, a large proportion of which -- perhaps the majority, according to some cancer specialists -- might never progress to invasive cancer. Many of these people are being labelled as cancer patients and counted as having been cured of a condition that would never have killed them.

Or take the example of prostate cancer. Over the past two decades, largely as a result of the widespread use of prostate specific antigen (PSA) testing in the US, there has been an enormous upswing in the diagnosis of early-stage prostate cancer. Most of the abnormalities discovered in this way are clinically unimportant, and would rarely progress to become life-threatening malignancies. Obviously, the more people who are labelled as prostate cancer patients, yet who have forms of the disease that are essentially non-invasive and not life threatening, the better survival statistics look.

We also need to examine precisely how cancer statistics are kept. Cancer registries, such as the NCI's SEER database, rely solely on information provided by death certificates. Yet cause of death is often recorded as, for example, pneumonia or liver failure, when in fact the underlying cause is advanced disseminated cancer. Another way in which cancer mortality is underestimated arises from the steady decline in the number of routine autopsies done on patients who die in hospital. In the US, for example, the autopsy rate has plummeted from around 45 per cent several decades ago to approximately 11 per cent today. Some hospitals now perform autopsies on fewer than 5 per cent of patients who die there.

Meanwhile, the unembellished statistics speak for themselves. In 2002, for example, the number of cancer deaths in the US was 557,272. In 2003, it dropped to 556,902, an absolute decline of 370 deaths. Hardly a cause for wild celebration, one would think. Yet this blip on the epidemiological radar screen set off a chorus of self-congratulation in the US media. "It proves that our expectation of continued progress against cancer is well-founded," said Andrew von Eschenbach, then director of the NCI.

Really? If progress continues at this rate, cancer deaths in the US should be entirely eliminated by the year 3508, a little more than 1500 years from now.