Mixed neuropathologies and estimated rates of clinical progression in a large autopsy sample

2.1 Study sample

NACC maintains the Uniform Data Set (UDS) on participants who had been prospectively evaluated and autopsied by an ADC since September 2005. Participants enrolled with any level of cognition and were examined annually in-person using a standard protocol, described in detail elsewhere [20,21]. Neuropathologic data was collected on participants who had died and consented to autopsy. All participants provided written informed consent and institutional review board approval was obtained from all individual ADCs.

This analysis focused on autopsied UDS participants with at least one clinical visit between September 2005 and September 2015. Participants were excluded based on the following criteria: 1.) rare cause of dementia that may conflict with neuropathologic assessment of ADNP or confound clinical conditions, such as Down’s syndrome, autosomal dominant genetic diseases, or frontotemporal lobar degeneration; 2.) missing information on covariates and/or neuropathologic information on ADNP, LBD, or VBI; and 4.) no ADNP, LBD, or VBI but presence of other pathologic burden such as hippocampal sclerosis, Braak stage V–VI with sparse or no neuritic plaques, frequent neuritic plaques but Braak stage 0–II, other major pathologies, or white matter disease. Additionally, in the main analyses we excluded participants without a clinical visit proximal death (e.g. last visit > 2 years prior to death) due to concern that the rate of progression and level of impairment may change closer to death but would have been unobserved. Given these exclusions, 2,046 participants with at least one clinical visit remained for analyses (see Supplementary Figure 1 for sample flow chart). Individuals with only one visit (n=475) were included in analytic models at baseline but did not contribute to longitudinal estimates.

In order to utilize additional information on VBI that was not available for all NACC participants, we conducted a sub-analysis in ADC participants seen at the Oregon Health & Science University (OHSU) (n=211) and University of Washington (UW) (n=82). These two ADCs have a joint agreement as part of the Pacific Northwest Dementia and Aging Neuropathology Group (PANDA) to follow the same neuropathologic assessment protocol, an additional benefit of this sub-analysis. Both ADCs recruit patients seen in clinic for enrollment into the UDS; however, OHSU also recruited participants from a number of cohort studies focusing on healthy aging and described elsewhere [22–25]. Subsequently, we will use the term PANDA ADCs to refer to OHSU and UW ADCs combined.

2.2 Neuropathological features

ADCs follow consensus guidelines but conduct neuropathologic assessments according to their own protocols, which vary between sites. ADNP was defined by Consortium to Establish a Registry for Alzheimer’s Disease scores of neuritic plaque density (none, sparse, moderate, and frequent) [26] and Braak stage for tau neurofibrillary pathology (none, I–II, III–IV, V–VI) [27]. ADNP was defined regardless of a participant’s cognitive status and was categorized semi-quantitatively as: low (no/sparse neuritic plaques & any Braak stage OR any neuritic plaques & Braak stage 0–II), intermediate (moderate/frequent CERAD plaques & Braak stage III–IV), and high (moderate/frequent plaques & Braak stage V–VI). This classification overlaps with the 2012 NIA-Alzheimer’s Association criteria [28]; however, Thal phasing [29] for amyloid plaques was not available for most participants. Assessment for Lewy bodies followed recognized guidelines [30]. LBD was defined as presence of Lewy bodies in any brain region examined. LBD subtype was classified as none, brainstem predominant, limbic (transitional), neocortical (diffuse), or other or unknown region.

Cognitive impairment due to vascular disease is usually considered a result of VBI caused by vessel disorders and other vascular mechanisms [31]. We focused on VBI, as defined by gross and microscopic infarcts, because these are associated with cognitive impairment in other studies [32,33]. VBI was defined as any gross infarcts (small or large artery) or any cortical microinfarcts (infarcts in the cortex only seen microscopically) regardless of age. Some studies suggest multiple VBI, in particular microinfarcts, may be needed to affect cognition [32,34]. Information on number of microinfarcts were not available for most NACC participants. To address this limitation, we abstracted additional data on the number of microinfarcts from neuropathology reports of autopsy PANDA ADC participants. The number (0,1,2,3, or 4 or more) of cortical and subcortical microinfarcts was assessed separately following methods developed in the Honolulu Asia Aging Study [35]. Cerebral amyloid angiopathy, atherosclerosis, and arteriolosclerosis were recorded as none, mild, moderate, or severe. Participants with subcortical leukoencephalopathy or white matter rarefaction were considered to have white matter disease. Hippocampal sclerosis was categorized as present based on neuropathologic diagnosis (forms prior to 2014) or if recorded as unilateral, bilateral, or laterality unknown (forms after 2014). Participants were defined as having a low level of neuropathology (low NP) if they were without ADNP, LBD, VBI, or other major pathologic burden.

2.3 Clinical impairment

Clinical impairment was quantified at each study visit with the Clinical Dementia Rating Scale Sum of Boxes (CDR-SB) [36], a composite measure of the overall level of cognitive impairment and functional disability that is based on clinical judgment and study co-participant report. The CDR-SB ranges from 0–18, with increasing score for increasing impairment. It is a sensitive measure for staging dementia on a continuum from normal to severe dementia in heterogeneous samples [35]. The Mini-Mental State Examination (MMSE) [37], a neuropsychological test of overall cognitive function, was also administered to able participants.

2.4 Other clinical characteristics

Demographic characteristics included in analyses were age, sex, education, race/ethnicity, and ADC. History of comorbidities and clinical judgement of motor and behavioral problems was recorded at each visit. APOE genotyping was performed on consenting participants. APOE ε4 allele status was classified as at least one or none. At all ADCs, either a single clinician or consensus group of clinicians made at each visit a diagnosis of normal cognition, impaired but not mild cognitive impairment (MCI), MCI, or dementia.

2.5 Statistical analyses

3.1 Participant characteristics

Among 2,046 autopsied NACC participants, the mean age at death was 82.1 years (standard deviation [SD]: 10.3), the average interval between last clinical visit and death was 9.1 months (SD: 6.1). There were 1,571 participants with multiple visits (ranging from 2 to 10 visits), including 1,196 participants with 3 or more visits. The median follow-up of those with 2 or more visits was 3.2 years (IQR: 2.0–5.0); 872 participants were followed for 3 or more years. Co-occurrence of ADNP, LBD, and VBI at autopsy was common (Figure 1). Characteristics of participants grouped by ADNP, LBD, and VBI neuropathologies are described in Table 1. There were relatively few participants with LBD+VBI and ADNP+LBD+VBI, so these groupings were not examined separately in analytic models.

3.2 Clinical progression in primary models

On average, participants with ADNP+LBD had the fastest rate of progression followed by ADNP only and then ADNP +VBI (Table 2). However, average trajectories of those with ADNP, regardless of co-occurring pathology were relatively similar upon visualization (Figure 2). Significant negative interactions were present between ADNP and LBD (β: −0.47 points; 95% confidence interval [CI]: −0.76, −0.22; p=0.002) and between ADNP and VBI (β: −0.34; 95% CI: −0.53, −0.10; p=0.003), such that participants with ADNP and co-occurring LBD or VBI had a lower rate of progression than would be expected if each pathology independently (additively) contributed to progression.

In sensitivity analyses, results did not differ substantially with further adjustment for comorbidities and APOE ε4 allele and in unweighted models that did not use IPW to adjust for potential selection bias (Supplementary Table 2). Among those without dementia at baseline, rates of progression were slower overall but the rate of progression diverged further between those with ADNP only and those with ADNP+LBD (Supplementary Table 2), additional adjustment for comorbidities and APOE ε4 allele did not substantially alter results (data not shown). Inclusion of participants without a clinical visit proximal to death did not substantially change rates of progression (Supplementary Table 3). Compared to those with high education; participants with low education had faster progression of low neuropathology, VBI only, and LBD only but slower progression of ADNP with or without VBI or LBD (Supplementary Table 4); trends otherwise remained similar to primary models. Trends were also similar when using the MMSE as the outcome measure (Supplementary Table 5).

3.3 Microinfarcts and clinical progression in PANDA ADCs

Among 293 PANDA ADC participants 10.9% (n=32) had 2 or more cortical microinfarcts and 34.1% (n=100) had 2 or more subcortical microinfarcts. Among participants without ADNP the number of subcortical microinfarcts was associated with a significantly faster rate of progression compared to those with low neuropathology (β: 0.10; 95% CI: 0.02, 0.18; p=0.002) but the number of cortical microinfarcts was not associated with faster progression (β: 0.05; 95%CI: −0.09, 0.26; p=0.6). The rate of progression of those with ADNP and subcortical microinfarcts was lower than would be expected if effects were additive (β for interaction: −0.22; 95% CI: −0.34, −0.08; p=0.003).

3.4 Semi-quantitative measures and clinical progression

Finally, we examined progression by severity of ADNP (none/low, intermediate, and high) and interactions with VBI and cortical LBD (Table 3). In this analysis we focused on cortical LBD, rather than LBD in general since cortical LBD was most strongly associated with progression of all LBD subtypes (Supplementary Table 6). Both Braak Stage and neuritic plaque density were associated with progression (Supplementary Table 7), thus we used a combined measure of ADNP when fitting interactions. Rates of progression were fastest for those with intermediate ADNP+cortical LBD followed by those with high ADNP with or without cortical LBD or VBI. Rates of progression were similar between those with and without VBI when stratified by level of ADNP (Table 3). Trajectories of those with intermediate ADNP+cortical LBD resulted in a similar level of impairment as those with high ADNP at the last visit (Figure 3). We found a suggestive synergistic or positive interaction between intermediate ADNP and cortical LBD, which was not significant (p=0.37; estimates for interactions shown in Supplementary Table 8). The interaction between intermediate ADNP and VBI was not significant (p=0.4). Trends in those with high ADNP were similar to primary models: negative interactions were significant (p=0.003 for high ADNP×cortical LBD and p=0.03 for high ADNP×VBI). In comparison, in a model without interactions between pathologies the difference in progression for high ADNP+cortical LBD compared to high ADNP only was overestimated while difference in rates for intermediate ADNP×cortical LBD compared to intermediate ADNP only were underestimated (Table 3).