A newly developed test could dramatically improve doctors鈥� ability to recommend the right type of cancer treatment for their patients.
The test, which involves genetically profiling tumour samples, is 80% accurate in predicting the right chemotherapy to aggressively attack each person鈥檚 cancer, according to researchers.
Anil Potti at the Duke Institute for Genome Sciences and Policy in Durham, North Carolina, US, and colleagues developed the test. They did it by analysing the known levels of gene activation in numerous types of cancer cells commonly used in laboratory tests. These cells have well-recorded responses to various chemotherapy agents, such as Topotecan and Paclitaxel.
Advertisement
A statistical analysis then revealed which patterns of gene activation were associated with the best response to various chemotherapy drugs.
Predictive tests
The team then checked their genetic-profiling predictor using stored tumour samples taken from cancer patients. They measured the levels of gene activation in these samples by looking at the amounts of specific molecules, called messenger RNA, in the cells. The researchers then suggested a chemotherapy drug based on their predictor tool.
Because the stored tumour samples had come from patients who had already undergone various cancer treatments, researchers could easily find out whether their statistical tool would have indicated the correct choice of treatment.
It turned out that the method they developed was 80% accurate in predicting the most appropriate chemotherapy agent.
鈥淚t takes the randomness out of the treatment approach,鈥� says Potti, adding that at present the chances that a patient will respond to a given chemotherapy are generally no better than 鈥渢he flip of a coin鈥�.
Wasting time
A patient who receives the wrong chemotherapy drug 鈥� one that they do not respond to 鈥� can lose valuable time in the fight against their cancer, and suffer unnecessary side effects, such as a weakened immune system.
While Potti鈥檚 team only analysed samples from patients with leukaemia, breast cancer and ovarian tumours, he says that the tool should work to predict the best drugs for other cancers, too. 鈥淚n my mind there鈥檚 no reason it shouldn鈥檛 be applicable beyond these types,鈥� he says.
One of the best aspects of this new statistical approach is that it could be done for relatively little cost, says Potti. 鈥淕ene expression profiling of patients鈥� tumours is already done for prognosis. In theory, this is not adding extra cost because it鈥檚 the same test 鈥� just analysed in a different way.鈥�
Journal reference: Nature Medicine (DOI: 10.1038/nm1491)
