Volumetric assessment included normalization for interpatient variation of head size by dividing hippocampal volume by the total intracranial volume of that particular patient. Percentiles for normalized hippocampal volume had been previously reported. Percentile scores were then converted to a W score, which is a value from a standard normal distribution corresponding to the observed percentile (in a standard normal distribution, the 50th, 5th, and 2.5th percentiles are given by W scores of 0, -1.645, and -1.96, respectively).
A W score of less than or equal to -2.5 would be associated with the 0.6th percentile of hippocampal volume. Sixty-seven of 80 MCI patients in a recent study (Jack et al., 1999) had W scores less than 0 (13 had W scores less than or equal to -2.5, which is 16.25% of the population of MCI patients studied). Assuming outliner variability as previously reported across studies of errors less than 1.9% coefficient of variation, then the number of MCI patients needed to produce sufficient power to find a significant effect would be 30.
This would provide 80% power to detect a difference of 16% as a deviation from .5, with alpha equal to .01. If 10.5% of the 30 MCI patients fall below a W score of -2.5, rather than the expected 16.25%, we would have 80% power to detect an alpha at .01 (h = .45 at 0.6th percentile of the control's standard normal distribution) (Cohen, 1988). Sensitivity (true positives / (true positives + false negatives) and specificity (true negatives/ (false positives + true negatives) calculations were also computed for each method's ability to correctly identify incipient AD, MCI, and controls.
A W score of less than or equal to -2.5 would be associated with the 0.6th percentile of hippocampal volume. Sixty-seven of 80 MCI patients in a recent study (Jack et al., 1999) had W scores less than 0 (13 had W scores less than or equal to -2.5, which is 16.25% of the population of MCI patients studied). Assuming outliner variability as previously reported across studies of errors less than 1.9% coefficient of variation, then the number of MCI patients needed to produce sufficient power to find a significant effect would be 30.
This would provide 80% power to detect a difference of 16% as a deviation from .5, with alpha equal to .01. If 10.5% of the 30 MCI patients fall below a W score of -2.5, rather than the expected 16.25%, we would have 80% power to detect an alpha at .01 (h = .45 at 0.6th percentile of the control's standard normal distribution) (Cohen, 1988). Sensitivity (true positives / (true positives + false negatives) and specificity (true negatives/ (false positives + true negatives) calculations were also computed for each method's ability to correctly identify incipient AD, MCI, and controls.
No comments:
Post a Comment