The brains of Alzheimer's sufferers contain deposits of a protein called amyloid beta.
These are thought to relate to the brain damage and dementia that are symptomatic of the disease.
During the disease the amyloid beta is increasing and this leads to neurodegeneration.
Three years ago, researchers at Elan proposed that injecting amyloid beta into the blood triggers an immune response that fights the disease in the same way as our system reacts to the foreign antigen (i.e. bacteria) and eliminates it.
No one is yet sure how this works.
Tests were spectacular in mice engineered to develop a form of Alzheimer's.
Amyloid deposits shrank, and the animals' memories improved.
The vaccine was made from a synthetic version of amyloid.
This could have lead to simple problem salvation, but trials in Alzheimer's sufferers were halted in January, when some patients developed symptoms similar to meningitis and encephalitis.
The new finding suggests how this may have come about.
Jucker and his colleagues injected elderly mice with antibodies against amyloid, rather than amyloid itself.
Five months later, the mice had smaller deposits of the protein in their brains.
But they also had many small haemorrhages in cerebral blood vessels.
The mice ‐ and most human Alzheimer's patients ‐ have amyloid deposits in the brain's blood vessels, as well as its tissue.
Clearing out the protein might weaken these vessels.
Vaccinating people before they get Alzheimer's, or in the very early stages of the disease, might help, as amyloid would not have had time to build up, says Richard Harvey, research director of Britain's Alzheimer's Society.
But, if immunized human patients bled, it is surprising that none of them had a stroke, Harvey adds. "I doubt that the bleeding is the whole story".
Bleeding may be a second side effect to put alongside inflammation, agrees neuroscientist Dave Morgan of the University of South Florida.