Alzheimer's disease probably affects twice as many people as current estimates suggest, experts believe, but sufferers are yet to show symptoms.
Scientists at the Mayo Clinic in Minnesota have been re-evaluating the prevalence of the disease using brain imaging to give a definitive assessment of the number of people affected.
In Britain, it is thought about 850,000 people have Alzheimer's, but most are diagnosed when already showing symptoms.
However, tests on a cohort study of 2,500 people found twice as many as expected had the telltale signs of protein plaques and tangles in the brain that are the marks of the disease, even though they were not yet experiencing dementia.
Dr Cliff Jack, a professor of Alzheimer's disease research at the Mayo Clinic, said the prevalence of the condition was at present "based on clinical assessment... on the question 'do you have dementia?'
"But as a general rule the prevalence of amyloid [protein plaques] and tau [tangles] as denoted by biomarkers is about two times higher," he said.
"Classically defined Alzheimer's undercounts people who have the pathology but do not have symptoms. A lot more people have the disease but do not have symptoms, just like a lot more people have hypertension than have had a stroke."
Failing to diagnose the disease early enough, before symptoms developed, is one of the reasons clinical trials have failed. Several major pharmaceutical companies have pulled out of drugs trials for neurodegeneration following costly failures.
However, Prof Jack believes the drugs do work, but are being given too late to help people already suffering from memory loss because their brain cells have already been destroyed.
"If you have a trial when someone is fully demented it's too late." he said.
"Science has found antibodies which target amyloid, but if a clinical trial which targets amyloid is going to be effective it needs to start before people have symptoms.
"You really have a narrow window to select people in just the right phase."
Trials may also have failed because many people enrolled on them did not actually have Alzheimer's disease, with Prof Jack estimating that around one third of patients were suffering from a different condition.
Last month a study led by Dr Peter Nelson at the University of Kentucky, named a new type of dementia, caused by the build-up a protein called TDP-43. Dubbed "late disease" and with symptoms that mimic Alzheimer's, it was found to be present in the brains of 50 per cent of people over 80 during post-mortem examinations.
Prof Jack said: "The assumption was that if people have a memory problem they had Alzheimer's disease.
"But cracks very quickly appeared in the facade. There are other things that go wrong with the ageing brain.
"You can have a reasonable estimate that if people don't have amyloid but a ton of
neurodegeneration then they have late disease. And the numbers are likely to be high."
He added: "So the notion that every older person with a memory problem has Alzheimer's really falls apart. Probably 20-30 per cent who... are told they have Alzheimer's do not."
Experts hope blood tests will soon be developed to allow screening for Alzheimer's to begin decades before symptoms occur.