David Payne Bennett and Dawson Hedges, Psychology
Introduction
Developing a biomarker for accurate assessment and detection of Alzheimer’s disease was the goal of our research. The P300 component of the event-‐related potential has been indicated to apparently change with the progression of Alzheimer’s disease. As measuring the P300 is non-‐invasive and relatively easy, we sought to investigate abnormalities in the latency of the P300 in probable Alzheimer’s disease compared to healthy controls. Should such abnormalities be present, they would constitute a strong foundation for clinically significant biomarkers.
Methodology
In order to amass a large dataset for analysis, we identified twenty-‐eight peer reviewed source studies that report P300 latency data obtained from either auditory or visual oddball stimulus for a control group and a group diagnosed with probable Alzheimer’s disease (of the non-‐early onset variety). We selected source studies that underwent uniform and similar methodology. The effect sizes of the individual source studies were pooled to yield an overall effect size. Various identifying characteristics, such as age, percent male, disease severity, and years of education, were used in meta-‐regression to attempt to identify the cause of the variation in individual effect sizes.
Results
Of the source studies that met inclusion criteria, the overall effect size for studies using an auditory stimulus was 1.176 (p < 0.001), and 1.155 (p = 0.010) for those that used a visual stimulus. Both are considered large effect sizes. None of the meta-‐regression covariates were able to account for a significant amount of the effect size. In all, these findings indicate that probable Alzheimer’s disease is significantly associated with lengthening of the P300.
Discussion
The twenty-‐eight source studies yielded 965 subjects, a relatively sizable population. The meta-‐analysis definitively indicated that the latency of the P300 component of the event-‐related was significantly longer in probable Alzheimer’s disease than in healthy controls. This extended latency was exhibited at each of the electrode locations (Cz, Fz, and Pz), which suggests that the changes in the brain due to Alzheimer’s disease are not localized, but rather systemic in nature.
Even after adjusting meta-‐analysis results with the Trim and Fill test, which corrects for potential publication bias in meta-‐analyses, the effect sizes were still large. The large overall effect sizes, in comparison with the high inter-‐study variability suggest that increased P300 latency is strongly associated with probable Alzheimer’s disease and is found across different samples of the human population. The high I2 values indicate that much of the difference in effect sizes between studies was due to actual differences, rather than to experimental error. We sought to determine the source of such variation through meta-‐regressions on age, educational attainment, sex, and disease severity. Though not all source studies reported data in all of these categories, none of the four characteristics were found to be associated with P300 latency. The lack of association between the disease severity and lengthened P300 latency was particularly surprising. This could be due, however, to the low number of source studies reporting such data, which could have caused the meta-‐regression to be underpowered. Alternately, the neurophysiological changes that cause the lengthening of the P300 latency could happen early in the disease progression, allowing such latency changes to be present in all subjects with probable Alzheimer’s disease.
P300 latency increases with age, regardless of the presence of neurophysiological diseases, such as Alzheimer’s disease. However, most source studies explicitly stated their use of age-‐matched controls to the Alzheimer’s group. Age-‐matching minimizes neurophysiological differences between the two groups caused by the aging process. Only in one source study (of the twenty-‐eight used) of our meta-‐ analysis was it unclear as to whether the two groups were age-‐matched. This, in light of the large effect sizes we found, indicate that Alzheimer’s disease is associated with the lengthening of the P300 beyond that of the normal aging process.
P300 latency has previously been suggested to reflect the speed of cognitive processing. The lengthening of the P300 latency would then indicate a loss of processing ability. This hypothesis requires further testing. However, the lengthening of the P300 latency is not unique to Alzheimer’s disease, although certain aspects of the lengthening, such as the nature of onset, could be disease-‐ specific. This line of inquiry also requires further investigation.
Conclusion
We found that P300 latency in probable Alzheimer’s disease is significantly longer than that exhibited by healthy age-‐matched controls. The effects size was large across analyses, indicating that actual differences were being discerned, and that such changes are stereotypical for Alzheimer’s disease progression. We suggest that further investigations attempt to ascertain the clinical significance of our findings.
Our findings are currently under peer-‐review for publication in the journal Alzheimer’s Disease & Associated Disorders.