Early detection of cancer is broadly seen as key to successful treatment, but many screening programmes are coming under fire for being relatively ineffectual. On 18 March, for example, The New England Journal of Medicine published online two studies of screening programmes for prostate cancer, which look for the biomarker prostate-specific antigen (PSA). One found that the screening programme had no effect on mortality rates for prostate cancer (G. L. Andriole et al. N. Engl. J. Med. 360, 1310–1319; 2009), the other found that screening reduced the rate of death from prostate cancer by 20% but was associated with a high risk of overdiagnosis (F. H. Schröder et al. N. Engl. J. Med. 360, 1320–1328; 2009). Both found that a large number of men received unnecessary medical treatment as a result of the screening programmes.

The early stages of more virulent cancers still elude today's tests.

Hard on the heels of these studies came another that looked at the gene-expression profile of ovarian cancers detected early, and found that they would probably advance more slowly and be less deadly than those detected later. If screening is to succeed, the authors conclude, we need other, better tests that can detect the virulent forms earlier (A. Berchuck et al. Clin. Cancer Res. 15, 2448–2455; 2009). Then earlier this month, researchers reported that screening for ovarian cancer by measuring levels of the protein CA125 and performing transvaginal ultrasound detected mostly advanced cancers, not the early-stage tumours that could presumably be treated more effectively (E. Partridge et al. Obstet. Gynecol. 113, 775–782; 2009).

These studies highlight a problem that has also been seen in other cancers, such as breast and lung: slower-growing cancers are easier to detect, but less likely to be fatal, precisely because they are less aggressive. But the early stages of more virulent cancers still elude today's tests, which themselves can bring risks when false-positive results lead to unnecessary medical interventions.

It is clear that, although some screening programmes are very effective at saving lives — cervical and breast cancer being prime examples — many others are not. This doesn't mean that early detection is worthless, simply that some current methods are not as good as they could be.

Researchers are now trying to improve tests by gaining a better molecular understanding of cancer. They are looking for biomarkers that can distinguish aggressive tumours from slow-growing ones (see http://tinyurl.com/naturecancer). They are studying the cancer genome to devise better screening tests and more targeted therapies (see page 719). And they are learning how to give doctors genetic information to guide the use of drugs in the clinic (see Nature 458, 131–132; 2009).

A better molecular understanding of cancer is unlikely to cure cancer on its own. For example, scientists are still unravelling the complex biology of angiogenesis inhibitors, which once promised to starve tumours to death by choking off their blood supply. Researchers have now found that the drugs can actually increase the spread of cancer in mouse models in certain situations (see page 686). And although the researchers hope that their studies will eventually lead to better angiogenesis inhibitors, the fact remains that every previous cancer 'breakthrough' — be it a targeted therapy or a marker for early detection — has also hit roadblocks.

Given that reality, the best strategy is to continue working on a broad front, improving these methods along with every other intervention in the cancer-fighting tool kit — even those that seem outside a basic biologist's remit. Nobel laureate Roger Tsien at the University of California, San Diego, for example, is trying to devise probes that could light up tumour cells to help surgeons do a more thorough job of removing them. Helping surgeons probably doesn't hold the same appeal for basic biologists as discovering a groundbreaking cancer pathway. But it may be that the most effective treatments for cancer will come, not from some major conceptual breakthrough, but from the steady refinement of the tools we already have.