White matter lesions (WML) are commonly due to small vessel disease and are often present in cognitively healthy elderly.¹ Fluid attenuated inversion recovery (FLAIR) imaging as well as diffusion tensor imaging (DTI) are sensitive to these changes.^2,3 Recently, it has been demonstrated that in subjects with WML, mean diffusivity (MD) and fractional anisotropy (FA) are elevated and reduced, respectively, not only in brain regions with WML but also in regions with normal appearing white matter compared to subjects without WML.⁴ However, it has not been fully determined if and to what extent the prevalence of WML in a control group will influence the outcome of statistical analysis of DTI metrics. The purpose of this study was therefore to investigate this presumed methodological issue by comparing MD and FA in the cingulum bundle (CG) and the superior longitudinal fasciculus (SLF) between patients with amnestic mild cognitive impairment (aMCI) and two different control groups of cognitively healthy elderly subjects with different prevalence of WML.Two hundred cognitively healthy elderly subjects (mean age 72.6±4.8, 58% females) and one hundred and thirty-seven subjects with aMCI (mean age 71.4±5.4, 42% females) participated in the study. Inclusion and exclusion criteria have been described elsewhere.^5,6 Briefly, inclusion criteria for the cognitively healthy elderly subjects were: age ≥ 60 years, scoring 28-30 points on MMSE⁷, no subjective cognitive impairment, fluent in Swedish, and no significant neurologic disease or severe psychiatric disease. DTI data were acquired on a Siemens scanner (3 T, standard 12-channel head coil) with full head coverage. For DTI acquisition two b-values (b = 0 and 1000 s/mm²) and 64 encoding directions were used, and the resolution was 2×2×2 mm³. FLAIR imaging comprised 27 slices, resolution = 0.7×0.7×5.2 mm³. MD and FA maps were calculated from the eigenvalues, DTI volumes were registered to MNI152 standard-space using FSL, and whole-brain tractography was generated using a deterministic algorithm (FA threshold = 0.2, angle = 30°) as implemented in TrackVis.^8,9 The CG and the SLF were segmented from the whole-brain tractography using one ‘AND’ gate according to the ICBM-DTI-81 WM labels atlas and one ’NOT’ gate according to JHU WM tractography atlas, both defined in MNI152 standard-space and projected to native-space (Figure 1).^10,11 Using FLAIR images, WML were rated according to the Fazekas scale.² Subjects with Fazekas score of ≥ 2 bilaterally in either the periventricular region or in the deep white matter were said to have WML (Figure 2). This classification was used to subdivide the cognitively healthy elderly subjects into one healthy control group having WML (HC-WML+; n = 83, mean age 73.9±5.1, 65% females) and one not having WML (HC-WML−; n = 117, mean age 71.7±4.3, 53% females). The Student’s t-test was used to compare MD and FA in the CG and SLF between the aMCI group and the HC-WML+ and the HC-WML− groups, respectively.According to the classification used in this study, 42% of the cognitively healthy elderly subjects and 57% of the subjects with aMCI had WML. Table 1 shows the results of the group-wise comparisons. A significant elevation of MD and reduction of FA was found in the right and the left CG and SLF in patients with aMCI compared the cognitively healthy elderly subjects not having WML (HC-WML−). No significant difference in MD or FA was found in neither the right nor the left CG or SLF in patients with aMCI compared to cognitively healthy elderly subjects having WML (HC-WML+).Patients with aMCI have previously been shown to have elevated MD and reduced FA in the CG and SLF compared to controls.¹² In this study, we were only able to reproduce those results when comparing to healthy controls with a low prevalence of WML. The inclusion or exclusion of subjects having WML in the control group thus highly influenced the outcome of the statistical analysis. A limitation of this study could be that, due to the heterogeneous etiology of WML, in some subjects, WML could be an expression of other incipient disease that DTI is sensitive to.¹³ However, we made an effort to minimize this potential bias when designing the inclusion criteria.^5,6When using DTI metrics to quantify differences in the white matter of the brain between a diseased group and controls, the prevalence of WML in the control group will influence the results of the statistical analysis. This should be taken into consideration and be accounted for in the design and interpretation of DTI studies.