A blood test could detect the earliest stage of breast cancer, early stage research has suggested.
The test measures several blood markers of cancer
Among 345 women the blood test picked up 95% of cancers, making it much more sensitive than existing tests.
It measures numerous proteins to find a characteristic cancer fingerprint, the Journal of Proteome Research reports.
Detecting cancers earlier is crucial to saving more of the 500,000 people who currently die from breast cancer each year around the world.
In recent years, breast cancer death rates have fallen because of earlier detection and more effective therapy.
Currently, doctors rely on triple testing - breast examination, imaging with mammography and ultrasonography, and biopsy - to pick up these tumours.
However, the study authors, from University College London in the UK and the University of Pennsylvania in the US, point out that there remains a need for more effective screening, especially in younger women where mammography scans are less sensitive.
The blood test they have developed detects very small changes in concentrations of proteins in the blood.
Some of the proteins are specific for breast tissue, while others fluctuate with many types of cancers and are markers of inflammation and new blood vessel growth in response to the tumour.
By changing the combination of proteins the test detects, it could be used to spot other cancers as well as breast cancer, the researchers believe.
They are now exploring how many, and which combinations of protein markers would be needed to screen for several cancers simultaneously in one blood test.
Professor Jasminka Godovac-Zimmermann of UCL said: "Our pilot studies show that using blood samples, breast cancer and several other types of epithelial cancers can be detected with much better sensitivity and specificity.
"This may allow new, less intrusive, safer and much less expensive approaches for the early diagnosis of cancer, for distinguishing malignant and benign cancers, and for monitoring cancer therapy."
In the study, the test correctly picked up 95% of breast cancers and missed 5%.
When cancer was not present, the test again correctly identified this 95% of the time and incorrectly said there was a tumour present 5% of the time.
Stephen Duffy, Cancer Research UK's professor of screening, said: "These are very interesting results which may have far-reaching implications for diagnosing breast cancer in the long term.
"Further research involving an independent group of patients and healthy volunteers is needed to validate these results, and to find out if the test is equally accurate for diagnosing early breast cancer as well as advanced disease."