A simple blood test could tell doctors which women with breast cancer need more aggressive treatment, scientists believe.
Number of cancer cells predicts risk
Women with five or more tumour cells in their blood sample would have aggressive disease needing more treatment.
Scientists says the findings could also lead to more tailored treatments, sparing some women from chemotherapy.
The University of Texas study appears in the New England Journal of Medicine.
Currently it is difficult for a doctor to make an accurate prediction of the likely prognosis for women with advanced breast cancer that has spread from the original tumour.
In the study, the presence of cancer cells in the blood was a better predictor of prognosis than two of the measures already commonly used by doctors - knowing where the cancer had spread to or whether the cancer has oestrogen receptors.
If the tumour has oestrogen receptors, its growth will be fuelled by the female hormone oestrogen.
The authors believe their findings could identify early on those who do need aggressive treatment.
Lead researcher Dr Massimo Cristofanilli said: "The most obvious case is oestrogen-receptor positive disease.
"Some doctors are reluctant to give these women a hormonal treatment at diagnosis, but would rather be safe and give chemotherapy - the most aggressive treatment.
"Utilising this test, we may one day definitely tell oestrogen-receptor positive women if they have a worse prognosis and that chemotherapy is the right approach.
"Or, for those with few or no circulating [cancer] cells, it is safe to go ahead with hormonal treatment alone.
"This is the first time that we can actually stratify metastatic breast cancer patients based on their risk," he said.
They tested the blood of 177 women with advanced disease before and four weeks after starting some form of cancer therapy.
At the first testing, half of the women had five or more circulating tumour cells in a 7.5ml sample of their blood.
These women were more likely than the others to die sooner from their cancer or for their cancer to progress quicker.
At the second testing, after treatment had been started, about a third of the women had five or more circulating tumour cells in their blood.
This suggested some of the women had responded to the treatment.
Again, the women with high levels of cancer cells in their blood had poorer survival times - eight months compared with 18 months on average.
The researchers plan to look at the gene pattern of these circulating tumour cells if it is the same as that of the breast tumour they originate from.
If so, it might be possible to make a decision about treatment in women with advanced disease without having to take a biopsy of the original tumour, they believe.
Dr Michelle Barclay from Breakthrough Breast Cancer said: "This is a small but interesting study and we look forward to seeing the results of further research on a larger scale.
"Predicting a patient's prognosis plays an important role in ensuring they receive the most appropriate treatment.
"Information about the aggressiveness of a woman's breast cancer will help to determine the best treatment options available."
But she cautioned: "This research is at an early stage and determining prognosis at this level will not be generally available for some time."