Risk of Alzheimer’s Disease in Obstructive Sleep Apnea Syndrome: Amyloid-β and Tau Imaging

Participants recruitment

This was a cross-sectional study evaluating AD risk in the community-based Vietnam Veterans. Veterans were recruited via the Older Veterans Psychiatric Program of the Repatriation Hospital, Austin Health and advertisement in magazines and newsletters of Retired Service League and the Vietnam Veterans Association of Australia. The institutional review board of Austin Health, a major metropolitan health service in Melbourne provided ethical approval for the study. Presence of dementia, existing diagnosis of MCI, psychotic and bipolar affective disorder, current substance abuse, and any unstable medical condition that could have impacted cognitive performance or made participation difficult were exclusions. Assessment consisted of medical history, Pittsburgh Sleep Quality Index (PSQI) [22], neuropsychological examination, apolipoprotein E (APOE) ɛ4 status, vascular risk factors, and Aβ and tau imaging. For the analysis of cognitive functions, subjects with traumatic brain injury (TBI) were excluded given its impact on cognitive outcomes. All subjects were screened for a history of polysomnographic diagnosis of OSA and the use of Continuous Positive Airway Pressure (CPAP) mask. This history was corroborated by documentation from the primary care doctor of participants whenever possible. Those with no polysomnographic diagnosis of OSA or symptoms of OSA were the controls. Considering the poor sensitivity of the global score of PSQI in detecting OSA [23], the global score was not used to diagnose OSA but positive responses to the individual items-cessation of breathing during sleep and excess daytime sleepiness-were exclusion criteria for the controls.

Assessment of cognitive functions and vascular risk factors

The neuropsychological examination included Logical Memory subset test 1 and 2 of Wechsler Memory Scale (WMS) – Anna Thompson story only [24], digit span forward and backwards from the Wechsler Adult Intelligence Scale third edition [25], categorical fluency test from the Delis-Kaplan Executive Function System [26], Rey Osterrieth Complex Figure Test (ROCFT) [27] and Trail Making Test parts A and B [28]. Vascular risk factor score was calculated by giving one point to each of the following: Diabetes mellitus, hypertension, coronary artery disease, hypercholesterolemia, body mass index (BMI) above 30, current smoking status, previous history of stroke, and atrial fibrillation. Sum of all points gave cumulative vascular risk.

PET imaging

The participants underwent a 20-min PET scan (4 x 5-min frames of emission data collected) acquired 90 min after a slow IV bolus administration of 250 MBq (±10%) of ¹⁸F-florbetaben and 70 min after the injection of 370 MBq of ¹⁸F-AV1451. Acquisition was performed with a Siemens PET/CT mCT128 and CT attenuation correction was applied. Image reconstruction used the Ordered Subset Expected Maximization (OSEM) algorithm. There was no correction for partial volume effect. ¹⁸F-florbetaben and ¹⁸F-AV1451 PET were both analyzed with Computational Analysis of PET from AIBL (CapAIBL) software developed by the Commonwealth Scientific and Industrial Research Organization (CSIRO) [29]. CapAIBL allows quantitative PET measurements without relying on magnetic resonance imaging [30]. Global Aβ and regional tau burden were calculated by standardized uptake value ratio (SUVR) using cerebellar grey matter uptake as the reference. ¹⁸F-Florbetaben scans were also read visually according to the manufacturer’s instructions by three readers and the classification into negative or positive scan was based on majority results. ¹⁸F-AV1451 regional uptake in three regions was calculated by CapAIBL software: Mesial temporal (amygdala, hippocampus, entorhinal cortex, and parahippocampus); temporoparietal (inferior and middle temporal lobe, fusiform gyrus, posterior cingulate/precuneus, superior and inferior parietal and lateral occipital cortex); and rest of the neocortex. A visual read of ¹⁸F-AV1451 images was not performed as a standard method has yet to be developed.

Statistical analyses

The continuous variables viz., ¹⁸F-florbetaben and ¹⁸F-AV1451 SUVRs and neuropsychological test scores were analyzed using independent t test while the categorical variables, OSA and APOE ɛ4 status, were analyzed by Chi-square test. All tests were two-tailed with 95% confidence interval. Pearson correlation was used to find correlates of both ¹⁸F-florbetaben and ¹⁸F-AV1451 SUVRs and cognitive test scores. Multilinear regression analysis was then performed with explanatory variables found to be correlated and associated with ¹⁸F-florbetaben or ¹⁸F-AV1451 SUVRs and cognitive scores. ¹⁸F-florbetaben and ¹⁸F-AV1451 SUVRs and cognitive scores were treated as the dependent variables. General linear model was used to test the interaction between OSA and amyloid and tau tracer SUVRs and a visually positive amyloid scan in predicting cognitive function.

Cognitive functions

The cognitive functions previously reported to be impaired in OSA include attention, procedural memory and episodic memory [31–33], processing speed [34], spatial memory [33], and executive function [35, 36]. Executive dysfunction has been reported in Vietnam veterans with OSA [35]. Language ability and psychomotor functions remain relatively unaffected. In the present study, subjects with OSA performed poorly on Trail Making Tests A and B. The Trail Making Test A measures visual attention and processing speed whereas Trail Making Test B assesses executive function. Felver-Gant et al. reported impaired performance on Trail Making Test B in OSA [35]. We have replicated this finding in subjects with OSA, but when we controlled for the score of Trail Making Test A, the effect of visual attention and processing speed the significance of Trail Making Test B did not remain indicating that the deficit was actually in the processing speed rather than executive function. The association between OSA and impairment in these cognitive domains was independent of age, APOE ɛ4, and education of the subjects, but not vascular risk burden, which was associated with OSA. The confounding effect of vascular burden was not adequately addressed in previous studies. A review of cognitive deficits in OSA has revealed varying cognitive deficits across studies [37]. The pattern of cognitive deficits varied according to the assessment settings (community cohorts against sleep medicine clinic), treatment with CPAP and age [38, 39]. A study that compared cognitive functions between young and older individuals found more attentional deficits in the older patients, a pattern supported by our study [37].

Sleep apnea and Alzheimer’s disease

Apart from cognitive impairment, existing data suggest an association between OSA and MCI and dementia. A seminal longitudinal study found a two-fold risk for MCI or dementia in patients with OSA over five-year follow-up [20]. Following early observations of association between OSA and dementia further findings have accrued recently supporting the link between sleep disordered breathing and both MCI and dementia [18–20]. A recent study demonstrated annual decline in the level of cerebrospinal fluid (CSF) Aβ₄₂ over a two-year period in cognitively asymptomatic elderly patients with OSA, implying increased risk of AD in this condition [40]. The change in CSF Aβ₄₂ correlated with the severity of OSA independent of APOE. However, there was no association between OSA and increased amyloid uptake on PET scan. The results of the present study show that patients with a polysomnographic diagnosis of OSA have slight but significantly higher global uptake of ¹⁸F-florbetaben compared with subjects without a diagnosis OSA. This association was independent of age and vascular risk factors, but not APOE ɛ4 or BMI. Both BMI and APOE ɛ4 were associated with ¹⁸F-florbetaben SUVR. Previous studies have suggested a negative correlation between BMI and amyloid load [41, 42]. The relationship between dementia and BMI is two-phased, increased risk of dementia was seen with BMI when weight was measured 20 years or more prior to dementia diagnosis and this association was reversed when weight was measured in 10 years or less before dementia diagnosis [43]. The first phase may represent a causal effect of obesity on dementia and the second phase is due to weight loss from metabolic changes arising from damage to medial temporal lobe during long preclinical stage of AD. The mean age of our subjects was 68.36 years suggesting that the participants have not yet reached the stage of declining weight which may commence with increasing amyloid accumulation at any stage later.

There was no significant difference in regional uptakes of ¹⁸F-AV-1451 between the groups. Recently a small study found increased amyloid deposition in elderly patients with MCI and OSA in correlation with oxygen desaturation index [21]. Such a correlation was not found in cognitively normal individuals with OSA [21], but the sample sizes of this study were very small, eight patients with normal cognitive function and five patients with MCI. Our subjects were elderly without an existing diagnosis of MCI.

Interaction between sleep apnea and APOE ɛ4

The mechanism through which OSA is associated with dementia and Aβ deposits is not precisely known. Aβ binding to APOE ɛ4 results in Aβ-APOE ɛ4 complex which is internalized by Very Low-Density Lipoprotein (VLDL) receptor more slowly than Aβ-APOE ɛ2 and Aβ-APOE ɛ3 complexes leading to decreased clearance of Aβ with APOE ɛ4 [44]. In the general population and clinical samples, OSA was found to be associated with the APOE ɛ4 allele [45–47] but other studies and a meta-analysis did not find such an association [48]. We found a significantly increased rate of APOE ɛ4 in OSA in our sample but the significantly increased Aβ tracer retention in OSA was no longer observed when the effect of APOE ɛ4 was controlled. APOE ɛ4 had stronger association with Aβ burden than OSA. The influence of OSA on ¹⁸F-florbetaben SUVR was higher in the presence of APOE ɛ4, than in its absence (Fig. 2). We did not find a significant difference in tau retention between the OSA group and controls but considering that cortical tau deposition follows Aβ accumulation and may be more closely associated with cognitive deficits, or dementia, this is not unexpected as our sample excluded subjects with MCI.

Limitations

While the diagnosis of OSA was made with polysomnographic studies, subjects without a diagnosis of OSA did not have laboratory sleep evaluation. The corollary is that some of the control subjects may have undiagnosed OSA. Therefore, the control group is not really a group without OSA; rather it is a group without symptoms of OSA. In our study we used PSQI, a sleep questionnaire that elicits symptoms of arrested breathing during sleep and daytime sleepiness. One study has reported that PSQI has poor sensitivity (38%) in screening for OSA, but this study used global PSQI score, not the responses to individual questions that are most relevant to OSA [23]. We excluded the participants who reported symptoms of OSA on PSQI. The rate of OSA in our sample (35.2%) is higher than the prevalence of OSA reported in the older general community (17% –24%) suggesting that Australian Vietnam veterans are well monitored for this condition and at low risk for missed diagnosis [49]. Our sample was primarily veterans with military deployment. This may limit generalizability of the above findings. The correlational analysis between nadir oxygen saturation and ¹⁸F-florbetaben SUVR involved a small sample and type II error needs to be considered in this context. It is noteworthy that the subject with highest ¹⁸F-florbetaben SUVR (2.01) had the lowest oxygen saturation (74%). Therefore, an association between oxygen desaturation and Aβ in OSA cannot be ruled out.