Ruppin & Reggia (1994) A Neural Model of Memory Impairment in Diffuse Cerebral Atrophy. British Journal of Psychiatry (1995) 166, 19-28.
I've been looking for papers about homeostatic synaptic plasticity in Alzheimer's or other neurodegenerative diseases. Not much at all. Which is odd. Intuitively you would think that homeostatic effects like synaptic scaling must be an important factor in the progression, if not the aetiology, of the disease.
But I did turn up this suggestive neural network paper from 1994.
The authors built an attractor neural network. I'm not 100% sure what that is, but my understanding is that it is a network of interconnected model neurons which tends to move to one of a finite number of stable states. With the appropriate learning rules the network can be trained to "remember" a number of input patterns.
They subjected their attractor network (which had several 100 model neurons) to a gradual loss of neurons or synapses. They found that if they built in a compensatory strengthening of surviving synapses (analogous to synaptic scaling) then memories degraded gracefully rather than suddenly. They also found that remote memories are spared in comparison with recent memories (with or without scaling), and that scaling increases the number of false positives. All these results are reminiscent of the clinical features of Alzheimer's Disease.
Which doesn't prove that synaptic scaling is important in Alzheimer's, of course. But it strongly suggests that it's at least worth taking a look.