Background
Florbetapir positron emission tomography (PET) amyloid positivity (A
+) is a biomarker for fibrillar amyloid associated with a high likelihood of progression to Alzheimer’s disease (AD) dementia [
1]. β-Amyloid (Aβ) accumulation has been linked to brain atrophy [
1‐
3] and cognitive decline in AD [
1‐
7]. However, findings have been mixed regarding whether and how A
+ relates to cognitive dysfunction or hippocampal volume (HV) across the spectrum of normal cognition (NC), early mild cognitive impairment (eMCI), and AD [
4,
8‐
13].
Some inconsistencies in the extant literature relate to disease stage included and concurrent versus longitudinal assessment of outcomes. For example, on one hand, although A
+ may be detectable in NC, it may not be meaningfully related to concurrent cognition until development of eMCI [
2,
14]. On the other hand, A
+ may be strongly linked to retrospective longitudinal decline in NC but less predictive of future decline as the disease progresses to eMCI [
4]. Other potentially explanatory factors include differences in sample size, type of Aβ measurement (e.g., binary positivity versus continuous measures of Aβ load), hippocampal segmentation and correction methodology [
15‐
17], and type of outcome measure employed (e.g., screening measures such as overall scores on the Mini Mental State Examination [MMSE] versus more specific memory measures).
Although traditionally treated as a confound, sex differences in the relationship of A
+ to cognition and HV may also explain conflicting findings. Recent investigations have revealed sex effects on hippocampal atrophy in normal aging, eMCI, and AD, though one study showed this only before controlling for Aβ levels [
18‐
20]. Researchers have also shown that women’s established verbal memory advantage over men [
20‐
23] appears to function as a form of sex-specific cognitive reserve, affording women equal or better cognitive performance compared with men via compensation despite positive biomarkers for AD, including mild to moderate levels of hippocampal atrophy [
24] or fluorodeoxyglucose (
18F-FDG)-PET hypometabolism [
25]. Mechanisms of sex effects on hippocampal atrophy and cognition remain unclear, but as recent reviews and studies suggest, the etiology may include a complex interaction of effects of sex hormones; genetics (e.g., apolipoprotein E ε4 [APOEε4] carrier status); and psychosocial (e.g., differences in stress or coping), demographic (e.g., education), and lifestyle (e.g., exercise, smoking, and alcohol use) factors [
20,
26]. Whether sex-specific reserve in cognition is seen in the face of A
+ remains unclear. It is also unknown whether sex-specific hippocampal reserve exists for women with A
+.
We examined whether sex moderates the effect of A+ on verbal learning and memory and HV in individuals with NC and eMCI. We hypothesized that sex would moderate the relationship of A+ and diagnosis with cognition such that women with NC would show a memory-related advantage over men that persists despite A+ and that women with eMCI would lose that advantage. We further hypothesized that sex would moderate the relationship of A+ and diagnosis with HV such that women with women with A+ and NC would show a neural robustness in hippocampal integrity but that women with eMCI would lose that advantage.
Discussion
In the present study, we examined the moderating effects of sex on the impact of diagnosis and A+ on verbal learning and memory and HV. The main finding was that sex moderated the effects of A+ and diagnosis on verbal learning. In addition, we showed that sex marginally moderated the effects of A+ and diagnosis on verbal delayed recall and that sex marginally moderated the effects of A+ and diagnosis on right HV. In contrast, no sex moderation effects were observed for left HV.
With respect to cognition, our findings specifically suggest that women’s advantage over men in verbal learning—and to a lesser extent delayed recall—was robust to A
+ in NC. Moreover, in eMCI, only women with A
+, and not those with A
−, showed poorer learning—and to a lesser extent poorer delayed recall. These effects were observed after accounting for baseline cognitive status, age, education, and APOE ε4 carrier status. We conceptualize these findings as consistent with A
+ eMCI representing a prodromal AD stage and A
− eMCI as representing suspected non-Alzheimer’s pathophysiology (SNAP). Our findings are consistent with literature showing better verbal memory performance in women [
24] and positing a cognitive or memory reserve advantage for women with fewer prodromal AD traits (i.e., amnestic eMCI but moderate to large HV), but not with more prodromal AD traits (i.e., amnestic eMCI, dementia diagnosis, and small HV) [
25,
26,
46]. Our findings are also partially consistent with a very recent study showing that women with low to moderate Aβ burden (but not high Aβ burden) had better verbal delayed recall than men and that this effect was specific to MCI versus NC or AD. A moderating effect of sex shown in the present study may help to explain some conflicting findings in the extant literature because, depending on sample size and diagnostic stage included, collapsing across sexes may lead to masked or exaggerated findings.
The present result showing that sex moderates the effect of A
+ and diagnosis on learning and memory also has implications for clinical diagnosis of AD in women. Specifically, as has been suggested in the past [
24,
25], memory reserve in women could delay prodromal AD diagnosis even in the face of positive biomarkers such as A
+. However, the present results suggest that longitudinal assessment of the potentially steeper decline in memory for women between NC and prodromal AD, which is absent in SNAP, or combining measures of memory with other biomarkers, possibly with an approach that places heavier weight on biomarkers such as A
+ early on, could increase diagnostic accuracy. This finding could also be relevant for development of therapeutics for AD, both with respect to inclusion criteria for trials (e.g., guidelines including a memory or learning score deficit requirement could exclude women with preclinical AD unintentionally) as well as outcome measures (e.g., the differing trajectories of memory decline in men and women could either exaggerate or mask important findings, depending on group composition, if sex is not considered).
With respect to HV, our present findings suggest moderating effects of sex for right HV. Similar to the pattern of results for the cognitive data, there was no relationship of A+ with right HV in women with NC. For women with eMCI, those with prodromal AD, but not SNAP, showed smaller right HV. The pattern in men was weaker and not significant, but it was similar. No moderating effects of sex were found for left HV.
Taken together, these findings may suggest that women have a neural reserve at the level of the hippocampus, such that hippocampal integrity is robust to effects of A
+ in preclinical stages in women. Importantly, the present results do not imply that women have larger hippocampi and thus more volume to lose. Instead, they would suggest that neural reserve could be defined as a robustness to neurodegeneration, beginning at similar neural volume as men, when adjusting for TIV. Replication in even larger samples, as well as in samples of clinic-typical patients, will be important for understanding whether this is a true example of sex-specific neural reserve or whether findings would be significant in men with larger cohorts. If the latter were true, it might alternatively suggest that A
+ is sensitive to concurrent HV loss in early clinical disease stages but not in NC. Larger cohort replication might also be helpful in determining whether the present lateralized findings might be consistent or whether bilateral effects would emerge. Certainly, left hemisphere effects might be expected, given the literature showing that women with NC and women with eMCI have stronger verbal memory [
20,
24], and our lateralized results deserve further investigation.
Of note, in the present HV analysis, we intentionally employed a residual correction methodology for TIV [
39], based on our specific sample composition as well as on guidelines recently published [
17]. Previous work has suggested that a major source of variability in literature describing assessment of HV sex differences may be lack of [
15], or differing methods for [
16], correcting HV for total brain or intracranial volume. Use of a deliberate statistical approach taking sex into account at all levels may help to reduce or explain contradictory results in the literature relating A
+ and sex to HV across NC and eMCI. Further research is needed to determine what pattern of sex moderation may exist at AD dementia stages at which women have been shown to have more rapid trajectories of decline [
18,
46].
Strengths of the present study include use of a large, well-characterized study sample employing the prodromal AD diagnosis and rigorous control of potential confounding variables. Limitations include lack of longitudinal analysis, which could help to clarify causality, and use of a smaller cohort of individuals with HV data. It was also beyond the scope of the present analysis to explore ways in which HV may itself be a moderator of cognitive decline or to further probe HV at the level of subfields [
47].
Future research is warranted on the longitudinal implications of these findings, as is replication in a larger cohort. In particular, because the present analysis employs clinically defined diagnostic stage groups in which men and women would be expected to express similar clinical symptoms, fine-grained examination of when exactly—or how much—pathology such as amyloid burden leads to cognitive, atrophic, and clinical symptom expression is needed. In addition, further validation and exploration of the currently used modification of the MoCA eliminating the memory component will also be important. Finally, examining the moderating effect of sex on other outcome measures, including hippocampal subfields, nonverbal memory, and resting state functional MRI, may be interesting.
Competing interests
JZKC, JLB, and SJB declare that they have no competing interests. JLC declares having received in kind research support from Avid Radiopharmaceuticals, Teva Pharmaceuticals, and CogState; having done consultation for AbbVie, ACADIA Pharmaceuticals, Accera, Actinogen Medical, Adamas Pharmaceuticals, Alkahest, Alzheon, Anavex Life Sciences, Astellas Pharma, AstraZeneca, Avanir Pharmaceuticals, Axovant Sciences, Biogen Idec, Biotie Therapies, Boehinger Ingelheim, Chase Pharmaceuticals, Eisai, FORUM Pharmaceuticals, Genentech, Grifols, Intra-Cellular Therapies, Iris Pharma, Ionis Pharmaceuticals, Eli Lilly and Company, Lundbeck, Merck, Neurotrope BioScience, Novartis, Nutricia, Otsuka, Pfizer, Probiodrug, QR Pharma, Resverlogix, Roche, Servier, Sunovion Pharmaceuticals, Suven Life Sciences, Takeda, Taisho Toyama Pharmaceutical Co., Transition Therapeutics, United Neuroscience, GE Healthcare, and MedAvante; owning stock in Adamas Pharmaceuticals, Prana Biotechnology, Sonexa Therapeutics, MedAvante, NeuroTrax, and Neurokos; and owning the copyright of the Neuropsychiatric Inventory. In addition, JLC has provided expert witness/legal consultation regarding olanzapine and ropinerole.