Mild Behavioral Impairment and Subjective Cognitive Decline Predict Cognitive and Functional Decline

Source population: National Alzheimer’s Coordinating Center (NACC)

Data used in this study were obtained from the NACC database (https://www.alz.washington.edu/). NACC was established by the National Institute on Aging (NIA) and consists of multiple NIA-funded Alzheimer’s Disease Research Centers (ADRCs) recruiting and collecting data on subjects with cognitive function ranging from normal to dementia. The NACC Uniform Data Set (UDS) is a large longitu-dinal dataset that includes demographic and standardized clinical data collected approximately annually. All test centers administered standardized forms and informed consent was collected from all subjects and their informants. Detailed information on the cohort and neuropsychological battery of tests included in UDS is described elsewhere [31–33]. NACC-UDS with a December 2018 data freeze date was used for this study.

Patient groupings

MBI status was derived from UDS using a publi-shed algorithm [34, 35] for transformation of the Neu-ropsychiatric Inventory Questionnaire (NPI-Q) [36] items to MBI domains. Specifically, ten NPS dom-ains from the NPI-Q were used to populate the five MBI domains of decreased motivation (NPI-Q apa-thy/indifference); emotional/ affective dysregulation (NPI-Q depression/dysphoria, anxiety, elation/euph-oria); impulse dyscontrol (NPI-Q agitation/aggres-sion, irritability/lability, aberrant motor behavior); social inappropriateness (NPI-Q disinhibition); and abnormal perception or through content (NPI-Q delusions, hallucinations). To obtain the MBI total score, these five transformed domain scores were added to-gether. As the NPI-Q has a reference range of one month. Thus, to approximate MBI persistence of sy-mptoms criteria, individuals with MBI total score >0 at two consecutive annual visits were classified as MBI positive (MBI+) and their MBI scores were calculated as the average over the interval. Those with no NPS were classified as MBI negative (MBI-) for comparison. To determine subjective cognitive decline, the SCD-Initiative Workgroup criteria [3] were used as a framework and operationalized in NACC as foll-ows: 1) endorsement by participant of a decline in me-mory on the UDS B9 form; and 2) normal cognition.

Figure 2 shows the step-by-step process for participant inclusion/exclusion. All NACC participants from 2005–2018 were initially considered for inclusion. The initial step was to classify based on MBI status (MBI+/-) and those with transient NPS not meeting MBI duration criteria were excluded. Study endpoint was chosen a priori to be 3 years to reflect clinical practice and design of observational cohort studies. This approach provided a concrete time fr-ame to assess change, in order to balance the need to wait long enough to see change, but to also minimize attrition that would accrue due to age-related mortality and other diseases that confound cognitive assessments. We included participants with a follow-up visit ∼3 years after the baseline visit to evaluate the change in Clinical Dementia Rating Scale (CDR® Dementia Staging Instrument) [37] score over time and participants were excluded if they were missing the 3-year study visit. SCD status was then determined and participants with a baseline CDR >0 were excluded. Finally, those with a baseline diagnosis of a psychiatric illness were excluded.

OPEN IN VIEWER

Fig. 2

Flowchart of participants from NACC included for analysis.

Study variables

Baseline variables included age, sex, education, and MBI/SCD category. Our primary outcome measure, the CDR, consists of six domains: memory, ori-entation, judgment and problem solving, community affairs, home and hobbies, and personal care, assessed by objective testing and informant report [37]. In our study, we used the global standard CDR score, which assesses the level of impairment, and ranges in sev-erity from no impairment (CDR 0), questionable im-pairment (CDR 0.5 –corresponding roughly to MCI), mild impairment (CDR 1 –corresponding to mild dementia), to moderate to severe impairment (CDR 2-3). All participants had a baseline CDR score of 0, and we measured a change in cognition and function to a CDR >0 at 3 years.

Standard protocols, registrations, and patient consent

The NACC database itself is exempt from IRB re-view and approval because it does not involve human subjects, as defined by federal and state regulations. However, all contributing ADRCs obtained informed consent from their participants and maintained their own separate IRB review and approval from their institution prior to submitting data to NACC.

Data availability statement

Data is available upon request from the corres-ponding author (ZI).

Statistical analysis

Categorical variables were analyzed with χ² test, and the continuous variables were analyzed using on-e-way ANOVA. We defined patient groups ordinally according to the absence or presence of SCD, MBI, or both at baseline. Cognitive and functional decline was defined as progression to CDR >0 after 3 years. We tested the ordinal association of patient groups and cognitive and functional decline using linear by linear association and Somers’ D [38]. Ordinal approaches permit explicit testing of an ordinal association in the probability distribution for progression to dementia across groups. As our theoretical model posits SCD and MBI as independent axes of dementia risk, the ordinal rank of SCD+MBI- and SCD-MBI+is arbitrary between SCD-MBI- and SCD+MBI+, and therefore both permutations were tested. It is similarly worth noting that the hypotenuse in SCD*MBI space, or dementia risk space, is not expected to be additive according to our theoretical model.

We also computed odds ratios (OR) for cognitive and functional decline using logistic regression with the patient group having neither SCD nor MBI (SCD-MBI-) at baseline serving as the reference group. In this model, we included terms for all variables reaching statistical significance (p<0.05) in the univariate analyses to calculate Adjusted Odds Ratios (AOR). All analyses were conducted in SPSS v24 (IBM Corporation) with α set at 0.05.

Nosology of psychiatric symptoms

Chronic and recurrent psychiatric syndromes are associated with an increased risk of dementia [39, 40]; however, de novo persistent psychiatric symptoms in older age constitute a unique risk marker. In order to restrict our approach to this population, we excluded participants with a psychiatric diagno-sis at baseline as MBI diagnosis is precluded by the presence of a psychiatric disorder. This is important because several large longitudinal cohorts have provided compelling evidence that the age of onset of psychiatric symptoms is a crucial factor in determi-ning the nature of these symptoms. These studies sug-gest that the later in life the onset of psychiatric symptomatology, the more likely these symptoms re-present the early stages of a neurodegenerative process, precede dementia by 5–11 years [41–44], and have a higher progression rate than early onset psy-chiatric syndromes, which are themselves at heightened risk [13]. From a community cohort of 9,931 participants, the emergence of MBI was associated with decline in attention and working memory at 1 year [12]. In psychiatry and neurology specialty clinic samples, incidence of dementia was higher for MBI than other psychiatric disorders [13, 14]. This evidence suggests that chronic and recurrent psychiatric symptoms reflect a psychiatric disorder framed in the context of psychiatric conditions, are sometimes neurodevelopmental in origin, i.e., these conditions are not a consequence of later life neurodegenerative disease, supporting exclusion from our analysis. In contrast to this, late onset psychiatric symptoms may be prodromal or precursor to cognitive decline and dementia in some and are better framed in the context of neurodegeneration. The ISTAART-AA MBI criteria were developed with an appreciation of the difference between later life de novo behavioral changes, and psychiatric disease recurring in later life [9]. Indeed, MBI has been associated with amyloid, tau, and neurodegeneration, and is now harmonized with the biological understanding of AD [30].

SCD

SCD is also represented in Stage 2 AD of the NIA-AA research framework in which there is subjective or objective evidence of subtle decline, not meeting criteria for objective impairment. On the AD continuum, subjective complaints of cognitive impairment, with or without evidence of impairment on cogni-tive testing [45], would be considered evidence of subtle cognitive decline and attributable to the pat-hologic process [30]. Meta-analysis of large longitudinal cohorts has shown that SCD is associated with ORs of 6 of progression to MCI and 2 for progression to dementia over a mean of 4.8 years [6]. In a study of older adults with SCD, ascertained using a composite score of 3 rating scales, 26% were determined to be Aβ+[5]. However, significant inter-site variability in the association between SCD and abnormal cerebrospinal fluid amyloid levels has been attributed to different recruitment approaches and a lack of standardized case definitions and ascertainment [46]. There can be other contributors to subjective complaints of cognitive decline not limited to medical issues, stressors or even medications. Nonetheless, as with MBI, SCD reflects the index clinical manifestation of a neurodegenerative process for some.

Intersection of MBI and SCD

MBI and SCD intersect in some instances. For example, a study of SCD determined that worries about self-perceived functioning were associated with Aβ positivity, rather than subjective cognitive functioning itself [47]. Worries or concerns are in-cluded in the SCD plus criteria, proposed to increase specificity for detecting preclinical AD [3]. Worry is also a component of the MBI affective dysregulation domain, which includes emergent mood and anxiety symptoms. Mood symptoms and SCD have been shown to interact to predict dementia independent of their main effects [48], and similar results have been found examining persistent neuropsychiatric symptoms and MCI [15]. The approach to psychiatric symptomatology in SCD has generally utilized traditional constructs of personality (e.g., neuroticism) and psychiatric conditions [49–51]. However, a change in personality to greater neuroticism (which is a neurodevelopmental construct) can also be framed as the emergence of MBI affective dysregulation, if considered in a neurodegenerative frame of reference [49]. This intersection of MBI and SCD is consistent with both constructs being represented in NIA-AA stage 2 AD.

NPI-Q informant report

The source of information for MBI status in our study was the NPI-Q [36] completed by an informant. The NPI-Q was developed to measure NPS in dementia, and the symptoms as described are relevant to an aging population with neurodegenerative disease. Informant reports have shown to be more reliable assessments of NPS in neurodegenerative disease to minimize the impact of anosognosia [52]. Coincidentally, in another study of SCD, confirmation of decline by an informant was the best predictor of worse cognitive performance and lower gray matter volumes [53]. Anosognosia is also important to consider in the assessment of SCD. The INSIGHT-PreAD study showed that patients with low cognitive awareness (reporting fewer difficulties than their relatives did) showed greater amyloid burden and lower cortical metabolism, compared to the high awareness group [54]. These findings suggest that self-report of symptoms alone, whether cognitive or behavioral, may not be adequate to capture early disease.

Clinical and research implications

Our data indicate that in cognitively normal older adults, the neurobehavioral axis of dementia risk represented by MBI, and neurocognitive axis of demen-tia risk represented by SCD, have complementary associations with the risk of progression to MCI and dementia. As operationalized in our study, MBI appears to be at least as strong a risk factor for progression to MCI or dementia as SCD and the two constructs have overlapping features. The combination of both MBI and SCD was associated with the highest risk (30.9% at 3 years), and this may have clinical utility by identifying a subset of individual at high risk of progression and reducing overmedi-calization of risk markers that in isolation have low specificity.

In addition to screening for subjective and obj-ective cognitive symptoms in older adults, incorporating MBI into clinical assessments may provide complementary information and better risk stratification [55, 56]. Not infrequently, dementia patients are first given a psychiatric diagnosis when presenting with a neuropsychiatric symptom, resulting in delays to treatment [57, 58]. Identifying MBI would prompt clinicians to consider neurocognitive disorders on the differential diagnosis, and flag patients who might benefit from imaging or further workup.

These findings also have clinical trial implications. Research and development costs are higher for AD than other therapy areas due to lower success rates and longer development times [2]. Despite the fact that changes in brain structure and function occur up to 20 years before gross memory impairment [59], screening for preclinical disease is expensive and inefficient. Leveraging the ease of measurement of MBI, in conjunction with SCD, could be an inexpen-sive and scalable method to select patients at highest risk for biomarker positivity and cognitive decline. For dementia prevention trials, combining MBI and SCD could increase yield and improve signal-to-noise ratios for clinical trial screening in order to identify an enriched group for assessment and workup. An associated reduction in screen failures could increase trial efficiency and decrease trial cost.

Limitations

The NACC participants are mostly white, highly educated volunteers seeking care and consultation at urban, university-based centers, and the finding may not generalize to other settings. Our choice of the CDR as the outcome measure was chosen for its clinical relevance, representing cognition-related daily function as a real-world outcome meaningful to patients, family members, and clinicians. However, this outcome does not provide as much detail on cognition as neuropsychological testing which could more accurately describe the performance of the four groups. For MBI case ascertainment, we used the NPI-Q [36] rather than the validated MBI-C [56, 61] which was developed specifically to measure MBI. We used a validated algorithm to convert NPI-Q scores to MBI-C domains and required NPS at 2 consecutive visits to match the MBI criterion of symptom persistence. However, the cutoff of >0 to define MBI+may not provide optimum specificity and risks overestimating the MBI phenomenon. Although 44.5% of the participants had a dementia risk syndrome of either MBI or SCD (or both), 21.8% of these had cognitive and functional decline at 3 years (compared to 5.2% of participants with neither risk syndrome)—a substantial proportion, nonetheless. Additional research is needed to refine the group, balancing the risk of overmedicalization with the need to determine risk with sufficient sensitivity in order to not miss cases of preclinical disease; the MBI-C may help with this need. MBI criteria stipulate that symptoms are not better accounted for by life events. However, life context could not be determined using the data available and there may be a proportion of participants with reactive symptoms, even over the two-visit duration used to determine MBI status. Similarly, MBI criteria require symptoms to have an impact on interpersonal relationships, social functioning and workplace performance. These data were not included in the analysis, and it is possible for some with mild symptoms to have minimal impact in those domains.