The demographic data is shown in . The MTA ≥ 2 group was significantly older (P = 0.001) and had lower MMSE score (P = 0.05) relative to the MTA < 2 group at baseline. Over the 3-year follow-up period, 60% of the subjects with MTA ≥ 2 and 33.8% of the subjects with MTA < 2 converted to possible or probable AD. Within the MTA < 2 group, there were 48 MCIc and 63 MCInc. MCIc with MTA < 2 had larger proportion of ApoE-4 carriers (P < 0.001) and lower MMSE at baseline (P < 0.001) relative to MCInc with MTA < 2. There were no significant education or gender differences in either comparison.
Results from the baseline MTA < 2 versus MTA ≥ 2 comparison are shown in . We hypothesized that relative to MTA < 2 the MTA ≥ 2 subjects will have greater atrophy in CA1 and subiculum and possibly CA2/3. The 3D statistical maps revealed greater left-sided hippocampal atrophy in MTA ≥ 2 versus MTA < 2 (left Pcorrected = 0.08, right Pcorrected = 0.4, corrected for age). The areas where trend-level significant between-group differences were seen showed 10–30% greater atrophy in MTA ≥ 2.
The maps of relative atrophy between MTA = 1, MTA =1, MTA = 2, and MTA > 2 versus MTA 0 are shown in . An animation of the hippocampal morphometric changes with incremental worsening of the MTA score can be viewed as Supporting Information on the journal's website. and the animation show that even when no hippocampal atrophy can be visually appreciated [i.e., MTA = 1, when the only abnormality detectable by the human eye is the mildly dilated choroid fissure, see Korf et al. 
] the hippocampal formation has already undergone structural changes that could be ascertained with an advanced 3D mapping technique such as the radial distance technique. This finding is further supported by supplementary Figure 1
provided as Supporting Information on the journal's website, which documents the presence of significantly greater atrophy in MTA < 1 converters versus MTA < 1 nonconverters. The ratio maps of supplementary Figure 1
document that the majority of these atrophic changes localize to the CA1 and subicular areas.
Figure 2 MTA subgroup comparisons. The top row maps show areas where significant between-group differences were observed in the specific comparisons. The bottom row maps show the quantitative between-group differences (in %). [Color figure can be viewed in the (more ...)
The correlations between MTA rating and hippocampal radial distance can be seen in (bottom row). Significant correlations between radial distance and MTA scores were seen in all hippocampal subfields bilaterally (middle row). After multiple comparison correction the correlations of MTA with left and right hippocampal radial distance remained significant (left pcorrected
= 0.0001, right Pcorrected
= 0.0003), see middle row of supplementary Figure 2
provided as Supporting Information on the journal's website.
In the MTA < 2 baseline versus follow-up comparison, we tested the hypothesis that there will be interim increase in hippocampal atrophy affecting CA1 and subiculum. The 3D statistical maps in supplementary Figure 2
revealed (middle row) more significant left-sided interim hippocampal atrophy (left pcorrected
= 0.008, right pcorrected
= 0.05). The quantitative maps shown in the bottom row of supplementary Figure 2
provided as Supporting Information on the journal's website revealed that areas showing significant change underwent 5–10% atrophy over the course of the study. Changes over time appeared in the subiculum as well as CA1–3.
With the MTA < 2 baseline and follow-up MCIc versus MCInc comparisons, we tested the hypothesis that MCI subjects predestined to progress to AD (MCIc) will have significantly greater atrophy of the subiculum and CA1 at both time points and new CA2–3 atrophy at follow-up relative to MCInc group. The 3D significance maps can be seen in . The hippocampi showed statistically significant between group differences both at baseline (left pcorrected = 0.002; right pcorrected = 0.006, corrected for ApoE4 status) and at follow-up (left=pcorrected = 0.002; right Pcorrected = 0.0003, corrected for ApoE4 status). MCIc showed greater involvement of the subiculum and CA1–3 relative to MCInc at both time points. Statistically significant areas showed 10–20% difference.
Our goal in the MTA < 2 baseline MCInc versus follow-up MCIc comparison was to develop a map sequence that will best show disease progression in MCIc. To accomplish this, we used the MCInc at baseline as a reference group for both the baseline (MCIc at baseline) and the follow-up (MCIc at follow-up) comparison. Thus, the regional increases in hippocampal atrophy from the top row images to the middle row images of represents the absolute disease-associated changes in the MCIc cohort. In addition to progressive involvement of the CA1 and subiculum, the CA2–3 area was progressively involved in the MCIc group. The areas with most pronounced interim atrophy progression showed 3–10% radial distance loss.
Figure 5 These maps illustrate the absolute spread of hippocampal atrophy in MCIc from baseline (top row) to follow-up (second row). The bottom row 3D maps show the interim 3-year hippocampal atrophy in the MCIc group alone (in %). [Color figure can be viewed (more ...)
Cox Proportional Hazards Analysis
Our Cox proportional hazards models showed that atrophy of the right but not left CA1 area was associated with conversion to dementia. In the unadjusted model, a person with 1 mm greater right CA1 radial distance had 50% reduced hazard for future conversion to dementia (hazard ratio HR = 0.5, 95% CI = 0.28–0.91, P = 0.02) relative to those with 1 mm smaller radial distance. After adjusting for age, education and gender the hazard was similar (HR = 0.53, 95% CI = 0.29–1.0, P = 0.048). In models using standardized CA1 and standardized MTA measures (recoded so that lower scores correspond to atrophy) adjusting for age, gender, and education to predict conversion to AD, a one SD decrease in right CA1 was associated with a 34% increased hazard of conversion (HR = 0.74, 95% CI = 0.56–0.998, P = 0.048) and a one SD decrease in MTA is associated with a 68% increased hazard of conversion (HR = 0.59, 95% CI = 0.46–0.78, P < 0.001). However, in our dataset, CA1 radial distance and MTA rating showed different distributions such that one SD decrease in MTA actually corresponds to a 1.44 SD decrease in CA1. A 1.44 SD decrease in CA1 in turn correspond to a 55% increased hazard of conversion.