© 2005 American Thoracic Society
OSA Brain Morphology DifferencesMagnitude of Loss Approximates Age-related EffectsFrom the Authors:We thank Drs. Macey and Harper for their letter in response to our recent report concerning cerebral structure in severe obstructive sleep apnea (OSA) (1). Voxel-based morphometry (VBM) allows automated voxel-by-voxel comparison of brain structure between populations, in this case patients with OSA and healthy control subjects. As discussed in our manuscript, it is desirable to correct statistically for multiple comparisons (2). Importantly, only changes surviving a correction for multiple comparisons are indicative of alterations generally present in subjects affected by the disease of interest. It is therefore not surprising that the changes detected by Macey and colleagues at an uncorrected threshold were not replicated in our sample. We optimized our design to detect differences between patients and control subjects by applying rigorous inclusion and exclusion criteria and the use of a 3-tesla scanner. Nevertheless, after correcting for multiple comparisons, our study did not find evidence of alterations in brain structure in OSA. The letter by Macey and coworkers confirms that this was also the case in their sample. Macey and coworkers argued in their article (3), as in their letter, that since they were able to detect changes in their normal subjects with increasing age that this was a "validation of the methodology." We believe this logic is flawed. Noting that the cited study on the normal effects of aging on brain structure included 465 subjects (4), we would again state that VBM results without correction for multiple comparisons can only validly be interpreted as trends. Such trends may, with greater statistical power, reach statistical significance, or alternatively may not prove to be "real." Macey and Harper further state that we did not include age or handedness as covariates in our analysis. As stated in our article (1), we included age as a covariate (i.e., ANCOVA) to compare subjects with OSA and control subjects. This corresponds exactly to the statistical analysis performed by Macey and coworkers, with the exception that we then corrected for multiple comparisons. For the comparison of initial and follow-up scans in the OSA group, a paired t test was used in which the need for covariates is largely removed by the paired design. The confusion probably arises from the online data supplement of our paper, where a general description of the optimized VBM technique is given. We did not feel that it was justified to include handedness as a covariate in our ANCOVA analysis. Handedness is notoriously unreliable as an indicator of language dominance, with only about a quarter of left-handed subjects showing atypical dominance for language, and atypical dominance occasionally occurring in right-handed individuals (5). It is the language dominance that may affect VBM results, as there may be some asymmetries in the areas subserving these functions (6). There is no reason to expect a difference in the distribution of language dominance between OSA and control subjects. Therefore, not controlling for language dominance could slightly reduce the power of the design, but should not include a bias. Our findings do not negate the evidence for abnormal neural and neuropsychological function in OSA. It may be that the relevant structural alterations in OSA are at a subvoxel level, or that they mainly affect the interplay between neuronal excitation and inhibition, or the functional organization of brain networks. There is certainly a need for further studies to address these interesting questions in detail.
a Institute for Breathing and Sleep, Austin Health, Heidelberg, Victoria, Australia and Brain Research Institute, Heidelberg West, Victoria, Australia FOOTNOTES Conflict of Interest Statement: None of the authors have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. REFERENCES
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
