|
(Continued from page 140)
that a point under consideration would belong to the one multivariate normal distribution or the other. These probabilities are proportional to the densities at the point of interest and they sum to 1. Thus, discrimination analysis can result in an estimate of the strength of classification and not merely a specific classification decision. The data were presented to the statistician in a blinded fashion, with no indication as to the nature of the tests or of the disease. The analyses were done with MINITAB, Release 7 for the PC (MiniTab Inc, State College Pa).
Results
This limited pilot study shows the group results plus standard deviation and probability in the Table for the 14 control subjects and the 12 AD subjects. The experimental group represents AD patients. The results show a significant difference between the two groups for plasma concentrations of aspartate (Asp) (P < .00 1); glutamate (Glu) (P < .003); a-ketoglutarate (a-kg) (P < .001); and glutamate dehydrogenase (GDH) (P < .00 1). However, no significant difference exists between the two groups for plasma concentrations of -y-aminobutyric acid, glutamate decarboxylase, glutaminase, glutamine, and glutamine synthetase. The plasma concentrations of the compounds studied in the control patients are consistent with those in previously published data.6-14 A discriminant analysis number was then obtained. Discriminant function = 66.65 -- 0.41 (Asp) -- 2.39 (GDH) -- 2.59 (a-kg) -- 0.35 (Glu).
This discriminant function may be used by substituting a patient's test results (for each of the four tests needed, the average concentration of three duplicate samples) into the discriminant function. If the result is less than 0, the patient is classified as having AD, otherwise as not having it. If a discriminant function classifies perfectly, it has, of course, the sensitivity and specificity of 100%.
Discussion
The compounds analyzed are normally found in the CNS and throughout the body and exhibit different abilities to cross the blood-brain barrier. The most appropriate analysis would have come from cerebrospinal fluid (CSF) or tissue analysis. However, this type of test is not always practical. The hippocampus, a region concerned with memory and glutamatergic innervation, does not have a well-developed blood-brain barrier; thus, systemic blood concentrations may be similar to tissue concentrations. Glutamate occurs in two metabolic pools, the majority for general cellular metabolism and a smaller, separate pool for the neurotransmitter metabolism. Therefore, these results may not be the exact concentrations in the glutamatergic regions but may only reflect proportional concentrations. Additional information was needed to support the search for either a systemic change15 or a neu (Continued on page 142)
|
|
