Start with why seven years is probably an underestimate. The team had roughly a decade of MRI in hand before the scans that settled plaque status [5], and the divergence they report shows up at least seven years back [2]. Subtract, and only about three years of window remain unused [12]. An estimate that runs nearly to the edge of the available data is reporting where the film ran out as much as where the biology began.
The comparison group is built from the instrument under suspicion. "People who did not show plaques" means people in whom amyloid-PET did not show plaques [5], and the study's own premise is that PET may not be sensitive enough to register the earliest amyloid-related processes [3]. Anyone quietly accumulating inside that group makes the two trajectories look more alike, which drags the measured separation toward zero [14]. Bias running that way is the comfortable kind: it makes a positive result harder to obtain.
Now the enrollment consequence. Amyloid-PET is the scan the field currently uses to mark the earliest identifiable signs of the disease [3]. A recruitment rule written on that marker inherits its latency, because you cannot select for a stage your instrument does not report. If structural change is already separating future accumulators from everyone else seven years before the threshold trips [2], and those people are cognitively well-functioning at the time [8], then the population with the longest runway ahead of symptoms is the population a plaque-positive criterion sends home.
A difference in structural change between people who later showed plaques and people who never did [5] is a statement about groups, not a guarantee about any one person's scan. The Oslo release leaves out the participant count, the effect size, and the brain regions involved [13], so the distance between "the curves diverge" and "this scan flags you" is simply not measured in what has been made public.
Anders Fjell's two readings [9] carry different bills. Under the first, the structural change is entangled with plaque accumulation that PET cannot yet see, and amyloid-directed work is arriving late to a target it already had. Under the second, something else moves the brain first, which is why he argues for continuing to develop treatments aimed at other biological processes [10], a point he sharpens by noting how closely the disease is tied to aging and how many biological factors probably feed it [11]. James Michael Roe, who led the work, puts the finding as the earliest signal detected on brain scans to date [6]. My read: the timing claim looks robust and probably conservative, while the causal ordering behind it remains open, and only the second of those is what a drug program can act on.
