Dementia and its associated morbidity and mortality are a growing worldwide health threat, with a prevalence estimated to approximately triple over the next 3 decades as the global population ages. While ADNC is the most common underlying cause of dementia,^2,4 the prevalence of clinical dementia resulting from multiple concurrent pathologies is being increasingly recognized, yet our understanding of the collective impact of mixed pathologies and the specific contributions of each proteinopathy to the overall clinical profile of individual subjects remains relatively limited.^{7,10,13,14,16,17,63} Additionally, identification of subjects who are cognitively “resilient” in the face of documented neurodegenerative pathology further complicates this picture, and a degree of the apparent resilience may be explained by the absence of certain copathologies, and thus may more accurately represent resistance to comorbid pathologies.^64–70 Investigation of less common neurodegenerative pathologies is also hindered by the relative ubiquity of some level of ADNC in the elderly population that may alter the clinical profile of these other pathologies in subtle or overt ways.^9,71 This overlap of neurodegenerative pathologies, at both the individual and population level, has hindered the development of useful clinical biomarkers and treatments based on each proteinopathy, delaying fundamental solutions to this growing social burden.

In this study, we further explored the effects of isolated and mixed pathologies on cognitive function, with a focus on their clinical presentations in a cohort of 1051 cases with select isolated and mixed pathology as well as 4624 cases for multivariate models.²¹ Our analyses of subjects with isolated and mixed pathologies consistently suggested several key findings. Several of the neurodegenerative pathologies investigated have well-defined clinical features (Table 2). As an isolated pathology and in multivariate logistic regression models, ADNC was relatively consistently given a primary clinical diagnosis of AD and had a well-defined amnestic dementia phenotype as well as a number of psychiatric features. Frontotemporal lobar degeneration with TDP-43 and PiD were more defined by disinhibition, bvFTD, and variants of PPA (while lacking an association with amnestic dementia). Lewy body disease was defined by Lewy body syndrome (including a strong association with hallucinations) and Parkinsonism (Figures 2, 5, and 6).^{9,20,58,62,71–75} Subjects with LATE-NC resembled a milder and more restricted version of ADNC, although only ADNC was associated with logopenic PPA and impairment in visuospatial function. In addition, while both ADNC and LATE-NC were independently associated with apathy and disinhibition, ADNC was additionally associated with delusions, hallucinations, anxiety, agitation/aggression, irritability/lability and disturbance in motor function. Previous studies have indicated that while both ADNC and LATE-NC primarily affect cognitive tests associated with logical memory (as well as Boston naming), ADNC affects digit span, trail-making, digit substitution, and animal/vegetable naming tests, while LATE-NC does not, thereby suggesting symptoms and tests that could potentially be used to distinguish these pathologies clinically (Figures 5 and 6; Table 2).^{8,21,28,29,31,61,76,77} Primary age-related tauopathy was not independently associated with significant impairment of cognitive domains or a specific clinical phenotype,^{36,37,39,55,56,78,79} but it is possible that current clinical and neurocognitive testing simply does not adequately assess mild or very specific clinical features that may be present in these disorders that appear to lack distinct clinical features. Emerging machine learning may help to further cluster specific deficits into groups to better define the clinical consequences of these pathologies and aid in future antemortem diagnosis.

In relation to LATE-NC, LBD, and CVD, ADNC appears to represent a more “dominant” pathology with respect to an individual’s clinical phenotype (Figure 3). When various combinations of these pathologies are concurrent with ADNC, the clinical presentation is far more similar to that of ADNC than to the clinical features found in these other isolated disorders. In the cohorts with comorbid ADNC and LATE-NC, subjects were given a primary clinical diagnosis of AD more commonly than subjects with isolated ADNC (Figure 3C), and the clinical features and cognitive testing appeared more exaggerated toward an “ADNC-like” pattern (Figure 3D). Conversely, the more distinctive clinical features of LBD appeared to be somewhat obscured by the presence of ADNC (and less so by LATE-NC and CVD), while the clinical features of ADNC remained intact. These trends became more prominent in subjects with additional comorbid pathologies. Conversely, FTLD-TDP, PiD, PSP, and CBD also appeared to have more “dominant” clinical phenotypes. Unlike subjects with comorbid LATE-NC, LBD, or CVD, subjects with these pathologies and comorbid ADNC have a more mixed clinical phenotype with less association with amnestic dementia, and more correlation with PPA variants, bvFTD, PSP syndrome, and CBD syndrome (Figure 4). It is also possible that clinical and cognitive testing have simply been better tailored to AD (and these frontotemporal dementias), and so are better able to identify features of these pathologies. Clinically, “probable AD” is a relatively nonspecific phenotype characterized by a wide-ranging global cognitive impairment (which includes amnestic dementia as a component), some of which may be independently produced by other pathologies, including LATE-NC (Table 2); a subset of these cases without ADNC at autopsy was given antemortem clinical diagnoses of AD (Figure 2A). Lewy body disease is an historically well-characterized disorder with distinctive clinical features but these distinct features still appeared to be relatively muted in the presence of ADNC (Figure 3D); this finding was previously identified in the interaction of ADNC and LBD.⁹

Given the cognitive effects of these combined pathologies and recent studies that estimate the degree of cognitive impairment that could hypothetically be removed from a population with a “magic bullet” therapy to remove a specific pathologic process,^21,80,81 this becomes an even more important consideration. For example, if a therapeutic could be developed to fully remove ADNC, the portion of the population with concurrent ADNC and LATE-NC may still retain a significant degree of “AD-like” manifestations caused by underlying LATE-NC, including impaired global cognition, impaired memory, orientation, judgment, and visuospatial function, and an amnestic dementia phenotype (Figure 3). Additionally, subjects with isolated LATE-NC have a mean global CDR of 1.6 ± 0.3, statistically equivalent to subjects with concomitant ADNC and LATE-NC (2.0 ± 0.2; P = .2838). This principle will make the development of antemortem biomarkers to identify non-ADNC comorbid conditions (and their extent) increasingly important as personalized medicine begins to develop therapies for these individual disorders.

Previously it was demonstrated that the presence of APOE ε4 alleles was significantly associated with the accumulation of multiple neurodegenerative proteinopathies.^12,15,21 Here, it appears that this effect is driven primarily by the presence of ADNC, that is, there was no significant increase in the frequency of APOE ε4 alleles in any other isolated neuropathology, although the addition of multiple other neurodegenerative pathologies, including LATE-NC and LBD, may be associated with increased frequency of APOE ε4 alleles in combination with ADNC (Table 1).

There are several limitations to this study. While the cohort used was large, most categories of isolated neuropathologies are relatively small (Table 1), and all data were obtained from NACC, which may not be representative of the general population.^23,82 Enrollment of subjects in AD research centers and the NACC cohort generally prioritizes patients with memory and cognitive symptoms, in particular the AD clinical phenotype, with lower rates of inclusion of patients with primary behavioral or motor symptoms. This cohort is also highly enriched for individuals with APOE ε2 alleles (∼44% of the overall cohort).²¹ This cohort also consists of subjects evaluated at multiple different ADRC sites (Table S1) over an extended period of time and may include regional differences in dementia diagnoses.⁸³ To address this limitation, we used multivariate logistic regression analysis to evaluate these clinicopathologic correlations in a larger cohort with various combinations of neuropathological conditions, adjusting for ADRC sites among other variables. While scores of cognitive testing may be relatively objective and reproducible between different settings, the clinical diagnoses and other assessments were performed by many individuals working in different clinical eras (both as individual and consensus assessments), and advancements in imaging and biomarker studies have been used in different institutions at different times. Clinical and neuropathologic diagnostic criteria have also changed over time and new disease entities, such as LATE-NC, which has only been officially defined since 2019, have been recognized; however, this diagnosis is frequently discernable from older recorded TDP-43 data.^21,41,61 The effect of shifting global consensus on some entities on clinical decision-making and data collection is difficult to determine. Additionally, the neuropathologic data available generally assess the unilateral regionality of pathology and not asymmetry or local severity, which may be important in some diseases, particularly in the earlier stages.^{15,56,59,63,78,84–86} Other variables such as FTLD-TDP assessment and TDP-43 distribution do not take into account subtypes, density of the pathology, or detailed location of the pathology beyond a few binary variables. The distinction between LATE-NC and FTLD-TDP remains a controversial issue^61,87–89; FTLD-TDP types A and B appear to overlap with LATE-NC stage 3, and there are a number of more recent recommendations to distinguish between these 2 disease processes.^60,77,90 Given this limitation, subjects included in this study as either LATE-NC or FTLD-TDP from the available NACC data had significantly different frequencies of hippocampal sclerosis, participant ages, primary diagnoses/clinical features, and interactions with ADNC, indicating that these are in fact separate entities, and that our classification was successful in separating most representative cases from each entity.

Despite some limitations, these results corroborate the known cumulative effects of mixed pathologies, further disentangle a variety of interactions between each pathology that contribute to a variety of clinical dementia presentations, and partially support the conceptualization of new dementia nomenclature separating clinical presentation from the underlying pathology.⁹¹ The rational development of disease-modifying therapies and preventative agents requires precise understanding of which pathologies are present in a given individual or population, and developing biomarkers for definitive antemortem diagnoses of individual neuropathological processes requires parallel efforts to understand the interactions between different neuropathologies and the resulting complex cognitive and behavioral phenotypes. The present study represents a significant step forward in understanding the clinical phenotypes associated with specific neurodegenerative pathologic processes and addressing the complex interplay between these disorders.