Evoked potential and reaction time data for the P300 and the TOVA are also displayed in Table 2. The P300 latency was significantly longer (t=4.10, p=.00006) and the amplitude significantly lower (t=3.04, p=.007) in the HM group compared to the NM group. The TOVA RT was significantly longer (t=3.82, p=.00006) in the HM group compared to the NM group. The difference score between the TOVA RT and P300 latency was significantly greater (t=2.74, p<.002) in the HM than the NM group, as was the absolute value of this difference (AbsValDiff, t=2.74, p=.009).
The presence of memory complaints also differentiated one memory variable and several neurophysiologic variables (Table 3). Patients with memory complaints had lower MMSE scores (t=-2.23, p=.027), but did not differ from those without memory complaints on the CNSM, WMS IM, or WMS WM Indexes. Patients with memory complaints were also older (t=2.95, p=.003), showed prolonged latencies (t=2.72, p=.007) and decreased amplitude (t=-2.83, p=.005) on the P300, and showed longer RTs on the TOVA (t=2.72, p=.007). The difference was most marked between the TOVA RT and P300 latency (t=5.41, p=.000001). Comparing the relationship of memory complaints to HM, 40 of the 43 HM patients (93%) complained of memory problems, whereas 102 of the 187 NM patients (54%) presented with memory complaints (χ2=21.9, p=.000001). Significant values were not found for P300AgeDiff, WMS indices, and CNSM indices.
Table 3: Means, SDs, and significance of subjects with and without memory complaints (Cx)
| With Memory Cx | Without Memory Cx | Significance | |
| No. of subjects | 142 | 88 | N/A |
| Age | 57.9 + 13.9 | 52.2 + 14.6 | 0.003 |
| P300 latency | 335.1 + 29.5 | 324.6 + 26.3 | .007 |
| P300 amplitude | 4.1 + 2.4 | 5.0 + 2.3 | .005 |
| TOVA* Reaction Time (RT) | 392.0 + 83.4 | 333.5 + 55.6 | .007 |
| P300–TOVA RT | 58.5 + 78.6 | 9.2 + 56.9 | .000001 |
| Absolute Value of Difference of P300–TOVA RT | 71.0 + 67.5 | 44.4 + 36.5 | .0002 |
| MMSE* | 27.9 + 3.5 | 28.8 + 1.6 | .027 |
*TOVA = Test of Variables of Attention; MMSE = Mini-Mental State Exam.
The presence or absence of memory complaints and HM was also examined by dividing the sample into 4 subgroups and performing one-way analyses of variance (ANOVA). The ANOVA of the P300 amplitude was significant (F=4.46, p=.005), and post hoc analyses showed that the patients with memory complaints and HM showed significantly lower amplitudes than those without memory complaints and NM (p=.002). The analysis of AbsValDiff was significant (F=6.97, p=.00017). Post hoc analyses showed the HM group with memory complaints had a significantly greater absolute value difference between TOVA and P300 (95.41ms) than the NM subgroup with memory complaints (62.8ms, p=.021) and the NM subgroup without memory complaints (43.5ms, p=.00006). ANOVA was significant for the MMSE (F=7.39, p=.0001) with HM patients with memory complaints scoring below NM patients with memory complaints (p=.0002) and NM patients without memory complaints (p=.0001). The ANOVA was also significant for CNSM (F=6.10, p=.001); HM patients with memory complaints scored below NM patients both with (p=.0003) and without (p=.002) memory complaints.
Those patients missing TOVA, CNSM, and MMSE scores (n=57) were removed from the final sample (n=230) and logistic regression modeling was performed on the modified sample (n=173). Logistic regression analysis was used to validate membership in the HM vs. NM groups using P300 latency (e.g., the difference score between the obtained latency and the predicted latency from the age adjustment [obtained latency–(300+age)]) and P300 amplitude, the absolute value of the difference score between the TOVA RT and P300 latency, and scores on the MMSE and CNSM. The best model was significant (F=13.2, p=.000004) and retained the CNSM score (Wald=6.41, p=.011) and the absolute value of the difference score between the TOVA RT and P300 latency (Wald=4.88, p=.025). The MMSE and P300 amplitude were not retained as validators. The classification model correctly validated 13 of 25 HM subjects (52.0%) and 125 of 148 NM subjects (84.5%) for overall classification accuracy of 79.8%. We recognize that our model demonstrates high specificity with lower sensitivity. Because of the clinical interest in determining HM validating MCI or later dementia in patients with memory complaints, the HM group with memory complaints was compared to the NM group without memory complaints with logistic regression entering age, memory scores, and neurophysiologic measures as validators (same as above). This analysis was significant (F=16.4, p=.000001) retaining CNSM (p=.011) and TOVA-P300 difference score (p=.001). The classification model was significant (χ2=28.8, p=.000001) and correctly classified 15 of the 28 HM patients (53.6%) with memory complaints and 72 of the 79 (91.1%) of the NM patients without memory complaints for an overall correct classification of 81.3%.
With regard to the HM subjects, and because of the small sample sizes in Groups 3 (frontal), 4 (temporal), 5 (focal), and 6 (none), only Groups 1 (parietal) and 2 (parietal + temporal/frontal) were compared. The majority of HM patients showed abnormalities in the parietal lobe (Table 4). Group 2 had significantly lower (p=.021) WMS IM scores (96.5, SD=22.5) than Group 1 (112.75, SD=19.6) as well as significantly lower scores (p=.047) on CNSM (74.4, SD=31.5 vs. 101.6, SD=22.3). The 2 groups did not differ on any of the P300 latency or amplitude variables. It is noteworthy that 66 participant subjects from the present cohort were diagnosed with depression. We found that 54 of these patients were classified as NM, while only 12 were classified as HM. We also found that 50 of these patients were classified as having amnestic MCI, while 16 were classified as having non-amnestic MCI. All depressed HM subjects (n=12) have amnestic MCI.
Table 4: Brain regions affected in HM subjects
*Subjects overlapped in categories
| Group Number | Brain Region of Hypometabolism | HM Subjects Affected |
| 1 | parietal | 18 |
| 2 | parietal + temporal/frontal | 16 |
| 3 | frontal | 3 |
| 4 | temporal | 1 |
| 5 | focal | 8 |
| 6 | none | 5 |
As noted earlier, both HM and NM subjects were evaluated for signs of MCI according to 4 domains21; 2 HM subjects showed no signs of MCI. Of the HM subjects, 47% were amnestic, 49% non-amnestic and 4% showed no signs of MCI progression (Table 5). For convenience purposes, subjects were categorized into one table: HM/NM,