White matter lesions (WML) are a common manifestation of cerebral small vessel disease with prevalence in the general population ranging from 11−21% at age 64 to 94% at age 82 [1]. WML are associated with cognitive decline and a larger WML burden is often found in Alzheimer’s disease (AD). Levels of Aβ42, the 42 amino acid isoform of Aβ, in the cerebrospinal fluid (CSF) and amyloid PET have been established as the most specific markers of cerebral amyloid deposition [2] with a strong inverse correlation between CSF levels of Aβ42 and cortical amyloid PET ligand binding [3, 4]. However, a larger volume of WML has previously been associated with lower CSF Aβ42 [5]. Therefore, we investigated if WML affect the relationship between CSF Aβ42 and amyloid deposition measured using [¹⁸F]-flutemetamol.The study comprised 321 individuals from the prospective Swedish BioFINDER study, including 122 cognitively healthy elderly (73.7 ± 4.5 years; 38 % male), 101 participants with subjective cognitive decline (SCD) (70.3 ± 5.5 years; 47 % male) and 98 participants with mild cognitive impairment (MCI) (71.2 ± 5.6 years; 58 % male). The WML volume [mL] was determined with the Lesion Segmentation Toolbox [6] using routine FLAIR and T1-weighted images that were all acquired at the same 3 Tesla Siemens Trio. CSF levels of Aβ42 were determined using an ELISA (INNOTEST). The standardized uptake value ratio (SUVR) of amyloid was measured using [¹⁸F]-flutametamol PET.

Baseline data are given in Table 1.

CSF Aβ42 and a neocortical composite SUVR of [¹⁸F]-flutemetamol were highly correlated (r for = -0.73, p < 0.001). Both were significantly associated with WML volume (r for CSF Aβ42 = -0.164, p = 0.004 and r for SUVR = 0.124, p = 0.028, respectively). Linear regression with the composite SUVR as independent and CSF Aβ42 as dependent variable showed a strong association (standardized beta -0.716, p < 0.001) that was unchanged when the WML volume was added to the model (standardized beta still -0.716, p < 0.001), and the WML volume was not a significant predictor (standardized beta 0.005, p=0.91). Similar results were obtained when CSF Aβ42 was dichotomized as normal or abnormal, cut-off 550 ng/L (standardized beta 0.725 and 0.719, p <0.001 and p <0.001 respectively, without and with WML volume as a predictor); again, WML volume was not a significant predictor (standardized beta 0.06, p=0.13). Also logistic regression with the composite SUVR dichotomized as normal or abnormal, cut-off 1.42 (see below), as dependent and CSF Aβ42 as independent variable showed a strong association (B = -6,996, p < 0.001) that was similar when WML volume was added to the model (B = -7.038, p < 0.001) and again the WML volume was not a significant predictor (B = -0.106, p = 0.42).

In addition, the CSF Aβ42 level did not differ between participants with low and high WML volume when comparing these in the subgroups with normal (n=203) and abnormal (n= 118) composite SUVR (independent samples t-test, p = 0.26 for participants with normal SUVR and p = 0.16 for those with abnormal SUVR) (Figure). The cut-off for abnormal SUVR was established at 1.42 using mixed modeling analysis [3]; the cut-off for abnormal WML volume was determined from the cumulative frequency at 50 % in the total study population.

Results were similar for the total population as well as for each diagnostic group. All analyses were corrected for age and gender.

Here, we show that WML, that are abundant in the elderly population and increased in Alzheimer’s disease, do not influence the relationship between CSF Aβ42 and amyloid deposition determined using [18F]-flutemetamol in spite of a significant correlation between each of these and the WML volume. We conclude that WML do not act as a confound and that CSF Aβ42 and amyloid PET can be used to measure amyloid deposition regardless of individual differences in WML volume.