Superior Global Cognition in Oldest-Old Is Associated with Resistance to Neurodegenerative Pathologies: Results from The 90+ Study

Participants

The initial participants of The 90+ Study were survivors of the Leisure World Cohort Study (LWCS) [17], an epidemiologic study on the members of a retirement community in Orange County, California. The 90+ Study commenced in 2003 when LWCS participants aged 90 or older on January 1, 2003, were invited to join. A similar invitation was extended on January 1, 2008, and every year thereafter to those turning 90 years old. More recently, volunteers aged 90 years and older who were residents of Orange County, California, and lived within a two hour drive of the study location, were recruited in the study through an open recruitment [18]. While the LWCS participants were recruited regardless of cognitive diagnosis, the volunteers had no or mild dementia. Currently, approximately two thirds of The 90+ Study participants are from the original LWCS and one third are volunteers from open recruitment.

All 1,335 participants (966 from LWCS and 369 volunteers) who agreed to in-person examination were invited to be part of the autopsy program. As of February 2022, 623 participants (308 from LWCS and 315 volunteers) had enrolled in the autopsy program. Of the 623 participants, 481 have died, 452 (94%) came to autopsy, and we have completed the autopsy evaluation and final cognitive diagnosis for 407 participants. Only the 117 participants with normal cognitive diagnosis at the time of death were considered for this analysis (Fig. 1).

OPEN IN VIEWER

Fig. 1

Participant flowchart: The 90+ Study. CIND, Cognitive impairment, no dementia; MMSE, Mini-Mental State Examination; non-SGCP, non-Superior Global Cognitive Performers; SGCP, Superior Global Cognitive Performers.

Standard protocol approvals, registrations, and patient consents

All participants or their designated informants provided consent to participate in the study and the study was approved by the University of California, Irvine (UCI) Institutional Review Board (IRB).

Neuropathological assessment

The UCI pathology team procured the brain specimens, which were fixed in formalin and then sent to the Department of Pathology at Stanford University for inspection, dissection, and (immuno)histopathologic evaluation according to current consensus criteria while blinded to cognitive diagnosis and any other participant information. The neuropathologic features were scored as follows: 1) ADNC (0 = not; 1 = low; 2 = intermediate; 3 = high) was based on the National Institute on Aging-Alzheimer’s Association (NIA-AA) “ABC” score, which incorporates Thal Phase for amyloid-β (Aβ) plaques, Braak staging for neurofibrillary tangles, and Consortium to Establish a Registry for AD (CERAD) staging for neuritic plaques [19, 20]; 2) Cerebral amyloid angiopathy (CAA) (0 = none; 1 = mild; 2 = moderate; 3 = severe) [21]; 3) Microinfarcts (0 = none; 1 = 1; 2 = 2; 3 = 3+) [19, 20]; 4) Atherosclerosis (0 = none; 1 = mild when present at branch points in the circle of Willis; 2 = moderate when present at branch points and elsewhere in the circle of Willis; 3 = severe when present on the cerebral convexity); 5) Arteriolosclerosis (0 = none; 1 = mild for partial replacement of vascular smooth muscle cells; 2 = moderate for full replacement of vascular smooth muscle cells; 3 = severe for concentric thickening of the sclerotic vessel wall); 6) Lewy body disease (LBD) (0 = none/olfactory; 1 = brainstem-predominant; 2 = limbic [transitional]; 3 = neocortical [diffuse]) (2 individuals with LBD only in the amygdala were included in the 0 LBD category) [22]; 7) Hippocampal sclerosis (HS) (0 = absent; 1 = present in either right or left or both) [19, 20]; 8) Limbic-predominant age-related TDP-43 encephalopathy neuropathological change (LATE-NC) (0 = none; 1 = amygdala only; 2 = plus hippocampus; 3 = plus middle frontal gyrus) [23]; 9) Age-related tau astrogliopathy (ARTAG) (0 = none; 1 = occasional; 2 = numerous) [24].

Assessment of comorbid pathology indices

To examine the comorbid neuropatholgic features at different severity levels, we followed the methods from a previous study on the Nun Study and the Honolulu-Asia Aging Study (HAAS) [7] to compute summary comorbid neuropathology indices. Each specific neuropathologic feature was given an index value of 0 for absent/negligible levels, 0.4 for intermediate, or 1 to denote severe levels of pathologic change. The index values for the specific neuropathologic features were then added into an index. As done by the previous study, we chose 0.4 as the intermediate level to distinguish the influence of a single severe pathology (comorbid pathology index 1.0) from 2 or 3 intermediate pathologies (comorbid pathology indices 0.8 and 1.2, respectively). Specific neuropathologic feature index values were assigned as follows: 1) AD index value was 0 for low/not, 0.4 for intermediate, and 1.0 for high ADNC; 2) CAA index value was 0 for was low, 0.4 for mild/moderate, and 1.0 for severe CAA; 3) Microinfarcts index value was 0 for no, 0.4 for 1-2, and 1.0 for 3 or more microinfarcts; 4) Atherosclerosis index value was 0 for no, 0.4 for mild/moderate, and 1.0 for severe atherosclerosis; 5) Arteriolosclerosis index value was 0 for no, 0.4 for mild, and 1.0 for moderate/severe arteriolosclerosis; 6) LBD index value was 0 for none/olfactory LBD, 0.4 for brainstem-predominant/limbic LBD, and 1.0 for neocortical (diffuse) LBD; 7) HS index value was 0 for absent, 0.4 for unilateral, and 1.0 for bilateral HS; 8) LATE-NC index value was 0 for no, 0.4 for amygdala, and 1.0 for hippocampal/cortical LATE-NC; 9) ARTAG index value was 0 for no, 0.4 for occasional, and 1.0 for numerous ARTAG. Additionally, to derive binary severe vascular index values, all vascular neuropathologies, i.e., CAA, microinfarcts, atherosclerosis, and arteriosclerosis, were further recoded as- absent (=0) when the specific vascular pathology index value were 0 or 0.4, and present (=1) when the index value was 1.0.

Four different comorbid neuropathology indices were calculated as follows: 1) Total neuropathology index was the sum of all (i.e., ADNC, CAA, microinfarcts, atherosclerosis, arteriosclerosis, LBD, HS, LATE-NC, and ARTAG) neuropathology index values; 2) Vascular index was the sum of CAA, microinfarcts, atherosclerosis, and arteriosclerosis index values; 3) Severe vascular index was the sum of severe vascular index values for CAA, microinfarcts, atherosclerosis, and arteriosclerosis; 4) Neurodegenerative index was the sum of index values for LBD, HS, LATE-NC, and ARTAG.

Cognitive diagnosis

Full clinical evaluations of participants carried out every six months consisted of a neurologic examination, physical examination, full neuropsychological battery [25], review of medical history, examination of medication containers, and interviews with informants. A final cognitive diagnosis was assigned after death in a multidisciplinary consensus conference using all available information from the longitudinal evaluations, brain imaging when clinically available, and medical records. All cognitive diagnosis assignments were done blinded to pathologic evaluation results. Participants were diagnosed as having normal cognition, cognitive impairment but no dementia (CIND), or dementia. A dementia diagnosis was made according to Diagnostic and Statistical Manual of Mental Disorders 4th Edition diagnostic criteria [26]. CIND was assigned when cognitive or functional impairments were present but did not meet criteria for dementia.

Superior global cognitive performer definition

The neuropsychological battery included the Mini-Mental State Examination (MMSE) to assess global cognition [27]. For this work, we assessed maintenance of superior global cognition until the time of death. Superior global cognitive performers (SGCP) were defined as individuals who had: 1) normal diagnosis in the consensus conference and, 2) an MMSE score of 28 or above in the last visit 12 to 2 months before death. Non- superior global cognitive performers (non-SGCP) were individuals who also had normal diagnosis in the consensus conference but had an MMSE score below 28. This cut-off was based on a previous study on The 90+ Study participants reporting a mean MMSE score of 28 in cognitively normal participants [28]. To limit the possibility of classification error due to poor cognitive performance resulting from terminal illness, we did not consider MMSE scores in the last 2 months of life.

Statistical analysis

We describe participant characteristics and proportion of participants overall and by SGCP status. We calculated the different comorbid indices, examined their distribution by SGCP status, and then categorized the indices into smaller groups according to their distribution, such that each category had sufficient number of SGCP and non-SGCP participants. The total neuropathology index was categorized as <1.8 (reference group), 1.8-<3, 3-<4, and 4+. We did not categorize the vascular index as there was no clear relationship pattern with SGCP status. The severe vascular index was categorized as 0 (reference group), 1, 2, and 3. The neurodegenerative index was categorized as 0 (reference group), 0.4-<1, and 1+. To examine the associations with individual and multiple comorbid neuropathologic features, we calculated odds ratios (OR) and 95% confidence intervals (CI) using multiple logistic regression models adjusting for age at visit (last visit when MMSE was done), sex, and college education (yes/no). For all analyses, our outcome of interest was SGCP status, and we modeled the probability of being a non-SGCP. Therefore, in our results, an OR above 1 would indicate a higher likelihood of being a non-SGCP which also indicates a lower likelihood of being a SGCP. Conversely, an OR below 1 would indicate a lower likelihood of being a non-SGCP which also indicates a higher likelihood of being a SGCP.

Secondary analyses

To assess the robustness of our analysis, we did additional secondary analyses. First, since some of the neuropathologic features had low prevalence in our study population, which could lead to sparse data bias, we conducted exact logistic regressions for all neuropathologic features in relation to superior cognition. Second, we tried two alternative MMSE cut-off points (27 and 29) to define superior cognitive performers. To examine the associations of superior cognition using these alternative cut-offs with individual neuropathologic features, we repeated the multiple logistic regression analyses adjusting for age at visit, sex, and college education to calculate OR and 95% CI. All analyses were performed using SAS 9.4 (SAS Institute; Cary, NC, US).

Distribution of individual and multiple comorbid pathologies by SGCP status

The distribution of neuropathologic features by SGCP status are shown in Table 2. ADNC was not different between SGCP and non-SGCP participants with most participants in both groups having low and intermediate ADNC (66.2% of SGCPs versus 61.3% of non-SGCPs). When separately examining the distribution of the A score, Braak tangle stage, and C scores we found no difference in the SGCP and non-SGCP. Among vascular neuropathologic features, CAA was observed in about half of the participants in both groups. Severe CAA was less common in SGCPs (5.6%) compared to non-SGCPs (16.1%). Microinfarcts were slightly more common in SGCPs (22.5%) compared to non-SGCPs (19.4%). Mild or moderate atherosclerosis was present in more than 70% of participants in both the SGCPs and non-SGCPs while mild or moderate arteriolosclerosis was present in above 85% participants in both groups. The severe vascular pathologic changes were infrequent in both groups and none of the participants in either group had severe arteriolosclerosis. The non-AD neurodegenerative neuropathologic features were less frequent in SGCPs than in non-SGCPs. HS was observed in 4.2% of SGCPs compared to 16.1% of non-SGCPs. LBD was observed in 9.9% of SGCPs compared to 35.5% of non-SGCPs. LATE-NC was observed in 19.3% of SGCPs compared to 32.3% non-SGCPs. HS and LATE-NC co-existed in 4 participants, all of whom were non-SGCPs (not shown in table). Occasional ARTAG was observed in half of the participants in both groups, while numerous ARTAG were observed in 5.6% of SGCPs versus 12.9% non-SGCPs (Table 2).

To examine comorbid pathology indices by SGCP status, we first evaluated their distribution (Fig. 2). The total neuropathology index scores ranged from 0 to 4.8 with SGCPs having lower index scores while non-SGCPs tended to be on the higher index score range, as shown by the red stacked bars in Fig. 2A. Overall, most participants had scores between 1.8 to 2.8 (Fig. 2A). The vascular index scores ranged from 0 to 2.4 but did not show a clear pattern of distribution based on SGCP status (Fig. 2B). The severe vascular index scores ranged from 0 to 3 with SGCPs having lower scores (Fig. 2C). The neurodegenerative index scores ranged from 0 to 2.4 with most SGCPs having lower scores. However, the only participant with a score of 2.4 in the neurodegenerative pathology index was a person categorized as SGCP (Fig. 2D).

OPEN IN VIEWER

Fig. 2

Distribution of the comorbid neuropathology index scores by superior global cognitive performance in the oldest-old. Numbers inside bars denote frequency.

Association of ADNC and total neuropathology index with superior cognitive performance

We used logistic regression models adjusting for age, sex, and education to examine the associations of ADNC, A score, Braak tangle stage, and C score, as well as total pathology index with SGCP status. There was no association between ADNC severity level and SGCP status. The ORs ranged from 0.53 to 1.24, and there was no clear pattern of association with different levels of ADNC severity (Table 2). Similarly, there was no association between the individual ADNC pathologies (A score, Braak stage and C score) with SGCP status (Table 2).

There was a significantly higher likelihood of being a non-SGCP with increasing total neuropathology index score. Compared to the reference category (0-<1.8 score), for the total neuropathology index category 1.8-<3 the OR was 5.87 (95% CI:1.23,28.05), for 3-<4 category the OR was 3.75 (95 CI: 0.62, 22.82), and for 4+ category the OR was 18.31 (95% CI: 2.34,143.19) (Fig. 3A).

OPEN IN VIEWER

Fig. 3

Association between comorbid neuropathology indices and superior global cognitive performance in the oldest-old. A) Association with total neuropathology (ADNC, CAA, microinfarcts, atherosclerosis, arteriosclerosis, LBD, HS, LATE-NC, and ARTAG) index scores. B) Association with severe vascular (CAA, microinfarcts, atherosclerosis, and arteriosclerosis) index scores. C) Association with neurodegenerative (LBD, HS, LATE-NC, and ARTAG) index scores. ADNC, Alzheimer’s disease neuropathological change; ARTAG, Age-related tau astrogliopathy; CAA, cerebral amyloid angiopathy; HS, hippocampal sclerosis; LATE-NC, limbic-predominant age-related TDP-43 encephalopathy neuropathological change; LBD, Lewy body disease; Non-SGCP, Non-Superior Global Cognitive Performers; SGCP, Superior Global Cognitive Performers.

Association between individual and multiple vascular neuropathologic features and superior cognitive performance

Logistic regression models adjusting for age, sex, and education were also used to examine the associations of individual and comorbid vascular neuropathologic features with SGCP status. There were trends that participants with severe vascular pathologic changes were more likely to be non-SGCP (Table 2). For example, the ORs for severe CAA compared to no CAA was 2.77 (95% CI:0.62–12.36), for 2+ microinfarcts compared to no microinfarcts 2.28 (95% CI:0.30–17.52), for severe atherosclerosis compared to no atherosclerosis was 5.72 (95% CI: 0.42–78.78) and no participant had severe arteriolosclerosis. However, none of these ORs reached statistical significance.

Unlike the vascular index, which showed no clear relationship pattern to SGCP status, for the severe vascular index, increasing scores tend to be associated with increased likelihood of being a non-SGCP, although the ORs were not statistically significant (Fig. 3B).

Association between individual and multiple neurodegenerative neuropathologic features and superior cognitive performance

As above, logistic regression models adjusting for age, sex, and education were used to examine the associations of individual and comorbid neurodegenerative neuropathologic features with SGCP status. Compared to participants with no LBD, those having LBD were more likely to be non-SGCP, and the association was significant for limbic LBD (OR = 8.36:95% CI:1.47–47.44) and neocortical LBD (OR = 10.80; 95% CI:1.02–113.82). Participants with HS were more likely to be non-SGCP compared to those without (OR = 5.28; 95% CI:1.10–25.47). Compared to participants with no LATE-NC, for all levels of LATE-NC there was a non-significant but increased likelihood of being a non-SGCP. Compared to participants without ARTAG, those with numerous ARTAG had increased, although non-significant, likelihood of being a non-SGCP (Table 2).

There was a significantly higher likelihood of being a non-SGCP with increasing neurodegenerative index score, with OR = 6.93 (95% CI = 1.66–28.83) in participants with 1+ scores compared to participants with score 0 (Fig. 3C).

Secondary analyses

Given the sparse data for some of the neuropathologic features, we conducted exact logistic regression to confirm the robustness of our findings. For most neuropathologic features, the magnitude of the unadjusted ORs from the exact logistic regression analyses were very similar to the maximum likelihood estimates (MLE) of ORs from unadjusted and adjusted logistic regressions (Supplementary Table 1). The only exceptions were neocortical LBD and HS presence, for which the 95% CI were wider and the ORs slightly attenuated and thus non-significant (Supplementary Table 1).

We also conducted secondary analyses changing the MMSE cut-off for superior performance to 27 and 29 and examined the associations with neuropathologic features. With the lower cut-off of 27, more participants (87 out of 102) were categorized as superior performers and fewer (15 participants) were non-superior performers (Supplementary Table 2). The results for the association of individual neuropathologic features with superior cognition remained mostly unchanged with the following few differences (Supplementary Table 2). For ADNC, the ORs increased in magnitude ranging from 0.9 to 6.6 (versus 0.53 to 1.24 with the original cut-off of 28) but remained non-significant with wider 95% CIs and no clear pattern of association. All levels of CAA were associated with non-significant increased likelihood of being a non-superior performer. For limbic LBD the OR was attenuated (OR = 3.62) and not associated with being non-superior performer. The OR for HS was higher (OR = 7.08) and remained significantly associated with being non-superior performer (Supplementary Table 2).

With the higher MMSE cut-off of 29, fewer individuals (49 out of 102) were superior cognitive performers and more (53 participants) non-superior performers. Even with this cut-off, the results for the associations continued to remain mostly unchanged with the following few differences. Mild (OR = 0.12) and moderate (OR = 0.17) CAA were associated with significantly decreased odds of being a non-superior performer, whereas with the original cut-off of 28 these associations were non-significant. For limbic (OR = 5.25) and neocortical (OR = 3.11) LBD, the ORs were attenuated and not significantly associated with being non-superior performer. For HS, the OR was attenuated (OR = 2.27) and not significantly associated with increased odds of being a non-superior performer (Supplementary Table 3).

Switching the MMSE cut-off values to lower and higher levels resulted in attenuation of some of the ORs, but the direction of association remained consistent. The secondary analyses did not have much of an impact on our overall results and conclusions and suggest the robustness of our findings.

Comparison of autopsied and non-autopsied participants

A substantial proportion of The 90+ Study participants invited to join the autopsy program did not consent to autopsy. Therefore, we compared the autopsied and non-autopsied individuals to explore differences between the two groups. Compared to those who did not consent to autopsy, the autopsy group included fewer women (66.9% versus 74.3%), were slightly younger (93.0 versus 93.8 years), had higher education (51.9% college or higher versus 38.1%), and fewer were LWC participants (49.4% versus 92.4%) (Supplementary Table 4).