Literature search and data extraction.
We searched MEDLINE (1946 to February 2012) and EMBASE (1980 to February 2012) databases using terms including “diffusion tensor,” “Parkinson*,” “progressive supranuclear palsy,” “multiple system atrophy,” “corticobasal,” “lewy bod*,” “Richardson*,” “Shy Drager,” “striatonigral degeneration,” “olivopontocerebellar atrophy,” “PD,” “LBD,” “PSP,” “CBD,” and “MSA” combined with Boolean operators as appropriate. There were no language restrictions, and translation was obtained as necessary. All titles and abstracts from the retrieved articles were screened and the full text of those that may be eligible was obtained. Reference lists of identified studies were searched for additional studies. Two independent assessors (C.J.C., K.P.E.) performed the search, reviewing all articles and extracting data. We included studies if they were published as full text articles and used DTI to compare participants with parkinsonian syndromes (Parkinson disease [PD], progressive supranuclear palsy [PSP], multiple system atrophy [MSA], corticobasal syndrome, PD dementia [PDD], or Lewy body dementia [LBD]) with a healthy control group or with a comparison group with a different parkinsonian syndrome. Only studies with more than 5 patients and with a minimum of 6 diffusion-encoding directions were included to be comprehensive while ensuring sufficient reliability. For the parkinsonian syndromes and dementias, a probable diagnosis by standard diagnostic criteria was considered sufficient for inclusion. Patients with MSA with motor features dominated by either parkinsonism (MSA-P) or cerebellar ataxia were included. We excluded studies that did not use formal diagnostic criteria or have participants with established diagnosis, studies including patients undergoing deep brain stimulation, studies analyzing mixed patient groups jointly, for example, including both idiopathic and vascular parkinsonism, as well as duplicate publications. If 2 or more studies contained the same or overlap sample of patients, only the largest relevant study was included.
From each study, we recorded the following data when available: number of patients and controls, mean age, number of males and females in groups, diagnosis, disease duration, and the use of medications. We recorded the main analysis approach used, all brain structures and abnormalities measured, acquisition characteristics, and field strength of the MRI scanner. The method for labeling neuroanatomy differed among studies, thus for the review, we grouped results according to cerebral lobe (frontal, temporal, parietal, and occipital), structure (substantia nigra [SN], putamen, etc.), or tract (corpus callosum [CC], inferior longitudinal fasciculus, etc.). Cerebral lobes, structures, or tracts that contained a minimum of 1 region with a significantly different (p < 0.05) mean anisotropy or MD in group comparison were recorded for each study to provide a neuroanatomical overview of significant differences. TBSS and VBA results reported were taken as significant on corrected p values unless specifically highlighted. The meta-analysis of the SN in PD included studies reporting FA values in or adjacent to the SN as means and SDs and those presenting p values. A conservative estimate of p = 0.049 was assumed if p was reported to be p < 0.05.
Data analysis was performed using Comprehensive Meta-Analysis (version 2.2.048, ©2006; Biostat Inc., Englewood, NJ). Effect size was measured using Hedges' g to correct for bias from small sample size.4
A random-effects model was selected to calculate the pooled mean effect size. Heterogeneity was assessed using Cochran Q
Publication bias was examined by visual inspection of funnel plot asymmetry and applying Egger regression intercept test.