Many models of disease progression in Alzheimer's disease (AD) have been proposed to help guide experimental design and aid the interpretation of results. Models focussing on the genetic evidence include the amyloid cascade (ACH) and presenilin (PSH) hypotheses and the amyloid precursor protein (APP) matrix approach (AMA), of which the ACH has held a dominant position for over two decades. However, the ACH has never been fully accepted and has not yet delivered on its therapeutic promise. We review the ACH, PSH and AMA in relation to levels of APP proteolytic fragments reported from AD-associated mutations in APP. Different APP mutations have diverse effects on the levels of APP proteolytic fragments. This evidence is consistent with at least three disease pathways that can differ between familial and sporadic AD and two pathways associated with cerebral amyloid angiopathy. We cannot fully evaluate the ACH, PSH and AMA in relation to the effects of mutations in APP as the APP proteolytic system has not been investigated systematically. The confounding effects of sequence homology, complexity of competing cleavages and antibody cross reactivities all illustrate limitations in our understanding of the roles these fragments and the APP proteolytic system as a whole in normal aging and disease play. Current experimental design should be refined to generate clearer evidence, addressing both aging and complex disorders with standardised reporting formats. A more flexible theoretical framework capable of accommodating the complexity of the APP proteolytic system is required to integrate available evidence.