Abstracts

RATIONALE: The MMPI-2 is commonly given to patients with temporal lobe epilepsy (TLE), but there is little agreement on the typical profiles or significance of MMPI-2 profiles in TLE. The MMPI-2 profiles of TLE patients are heterogeneous, with no single profile characteristic of TLE. In learning disabilities research, statistical subtyping by cluster analysis has been useful for identifying subtypes of personality and psychopathology, and the relation of these subtypes to cognitive dysfunction and neuropathology. Clustering of MMPI-2 profiles of TLE patients may be similarly informative. Subtypes derived by cluster analysis must demonstrate internal and external validity. Internal validity is demonstrated by replication of subtypes across clustering methods and samples. External validity is demonstrated by meaningful subtype differences on variables not used to develop the typology. The purpose of this study was to determine if internally valid MMPI-2 subtypes could be derived from TLE patients with cluster analysis.
METHODS: Sixty consecutive cases with TLE were reviewed. Forty-three cases had completed neuropsychological testing, including the MMPI-2. A multivariate (density) outlier detection algorithm resulted in deletion of seven patients, leaving 36 for clustering (18 females, 18 males). Clustering was performed with MMPI-2 scales 1, 2, 3, 7, 8, 9, and 0. As clustering based on profile shape was desired, prior to clustering variables were transformed by removing profile elevation and dispersion. An initial cluster solution was developed using Ward[ssquote]s method followed by a k-means relocation pass. Replication was attempted using equal-variance maximum likelihood (EML), unweighted group average linkage (UPGMA), and complete linkage (COM) clustering methods with a k-means relocation pass. Agreement between solutions was evaluated using Rand[ssquote]s statistic.
RESULTS: Internal criteria (pseudo-[italic]F[/italic] and [italic]t²[/italic]) suggested four clusters in the initial Ward[ssquote]s solution. These clusters were replicated perfectly (Rand = 1.0) by EML and UPGMA methods. The COM solution was also very similar, with Rand = .98, and only 1 misclassification. The clusters and their code-types based on mean MMPI-2 profile were: 13 ([italic]n[/italic] = 14); Exaggerated (scale F [gt] 80, [italic]n[/italic] = 8); 278 ([italic]n[/italic] = 7); and 21 ([italic]n[/italic] = 7). The subtypes did not differ on duration of seizure disorder, age at onset of seizures, or WAIS-R IQ measures. The subtypes did differ on age at assessment ([italic]p[/italic] [lt] .05), with the Exaggerated subtype tending to be younger than the 278 subtype (29.4 years vs. 45.9, respectively). The Exaggerated subtype tended to contain more males than females (75% of the cluster); the 21 subtype contained more females than males (71% of the cluster).
CONCLUSIONS: This preliminary study suggests that in patients with TLE reliable MMPI-2 subtypes can be derived using cluster analysis. The 13 code-type was the most frequent (~40%), with 278, 21, and Exaggerated types being about equally frequent (~20%). The age and sex differences between subtypes provide limited support for the external validity of the typology; however, the reliability and external validity of this cluster solution must be assessed in a larger sample of patients.