The moves would allow companies to test drugs in people before symptoms appear, offering a better chance of intervening before the disease has destroyed too many brain cells.

Under the proposed research framework, Alzheimer's would be characterised by three factors: evidence of two abnormal proteins associated with the disease - beta amyloid and tau - and evidence of neurodegeneration or nerve cell death, all of which can be seen through brain imaging or tests of cerebral spinal fluid.

It also incorporates measures of severity using biomarkers and a grading system for cognitive impairment.

The hope is that the new definition will help researchers pick better subjects on which to test new Alzheimer's treatments, which may help improve drug companies' search for treatments.

So far, experimental Alzheimer's drugs have had a dismal track record, with more than 100 failures, including most recently a treatment from Merck, while Pfizer said in January it was quitting the field.