Addenbrooke’s Cognitive Examination–Third Edition Predicts Neuropsychological Test Performance

With an increase in the aging population and concomitant rise in the incidence and prevalence of dementia (1), easily implemented screening tools sensitive to cognitive decline are critical in clinical practice. Although many screening tools are available, they vary in sensitivity and specificity. The Addenbrooke’s Cognitive Examination (ACE) test was developed to improve the depth and breadth of examination of specific cognitive domains compared with the Mini–Mental State Examination (MMSE) (2, 3). Now in its third edition (ACE-III) and readily available for clinical and research purposes, the ACE has been found to have high sensitivity and specificity in detecting neurocognitive disorders, including higher sensitivity than the MMSE (4) and perhaps the Montreal Cognitive Assessment (MoCA) (5–11). Administration of the ACE-III is straightforward, relatively brief, and—importantly, given the current concern regarding spread of the COVID-19 pandemic—it can be administered virtually via telehealth with very minimal modifications. Performance is measured by total score and subscale scores for the following five domains: Attention and Orientation, Memory, Fluency, Language, and Visuospatial functions. In addition to severity of neurocognitive disorder (11–13) (i.e., mild vs. major), the ACE-III can help identify cognitive performance patterns characteristic of different types of neurocognitive disorders (4, 14) (e.g., frontotemporal dementia vs. Alzheimer’s dementia).

We investigated the utility of the ACE-III as a brief cognitive screener in a large, U.S.-based clinical sample of older adults referred for neuropsychological evaluation because of concerns about cognitive decline. Specifically, we examined whether ACE-III subscale performances predicted subsequent within-session neuropsychological performance; how ACE-III sensitivity in highly educated individuals (8, 15) compared with the measure’s sensitivity among individuals with lower education levels, because increased cognitive reserve may reduce sensitivity on screening measures (15, 16); and how ACE-III performance compared with standard clinical neuropsychological evaluation as a predictor of neurocognitive disorder. We also examined whether ACE-III scores differed by neurocognitive disorder severity. Overall, this study aimed to expand validation studies from other countries (7, 11, 13), which have found good convergent and construct validity (11) with standardized neuropsychological tests to guide clinicians in effective use of the ACE-III as a screening tool.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

The research was conducted as a retrospective review of medical records. All procedures were reviewed and approved by the Partners Human Research Committee Institutional Review Board. All patients provided written consent.

Patient Population

Patients were referred to the outpatient neuropsychology clinic at Massachusetts General Hospital for a clinical neuropsychological evaluation over a 2-year period. Patients were referred by physicians across several departments, including primary care, neurology, and psychiatry, for an assessment of their neurocognitive status and for diagnostic clarification of a potential progressive neurocognitive disorder. All neuropsychological evaluations were conducted or supervised by licensed clinical neuropsychologists. The evaluations consisted of medical record review, including findings from the neurological examination and imaging (when available), a clinical interview with the patient and in some cases a family member, and administration of a battery of neuropsychological tests. Only patients who completed the ACE-III as part of a full neuropsychological evaluation were included in the study (N=217). All evaluations were conducted in English.

Participating patients were classified as having normal cognition (NC; N=67), mild neurocognitive disorder (mild NCD; N=105), or major NCD (N=45) on the basis of full evaluation data. All patients with mild or major NCD were diagnosed according to criteria in the DSM-5. Additional demographic data are provided in Table 1.

Neuropsychological Test Battery

All patients initially completed the ACE-III and then underwent a comprehensive clinical neuropsychological evaluation, including standardized clinical measures that are commonly used, well validated, and norm referenced. For each test, an individual’s performance (raw score) was converted to a standard score (z score) based on published normative data for peers with the same age and education levels.

ACE-III performance is summarized as a total score and the five subscale scores: Attention and Orientation (e.g., registration of three items, serial 7s; raw scores range from 0 to 18), Memory (e.g., verbal learning, recall, recognition; raw scores range from 0 to 26), Fluency (e.g., letter and animal fluency; raw scores range from 0 to 14), Language (e.g., object naming, comprehension, sentence writing; raw scores range from 0 to 26), and Visuospatial (e.g., cube copy, clock drawing, degraded letter recognition, raw scores range from 0 to 16). Higher ACE-III performance scores indicate better cognition within each domain. Clinical neuropsychological tests that assessed corresponding cognitive domains were included for regression analyses with ACE-III subscale performance. Specifically, scores on the Wechsler Adult Intelligence Scale–Fourth Edition (WAIS-IV) Digit Span Backward (DSB; N=214; raw scores range from 0 to 48, with higher scores indicating better attention and working memory) and Trail Making Test Part B (TMT B; N=172; raw scores are captured in seconds, with faster performance indicating better executive functioning) were compared with the ACE-III Attention and Orientation score. Performance on the Wechsler Memory Scale–Fourth Edition (WMS-IV) Logical Memory (LM) Immediate Recall (N=176; raw scores range from 0 to 50 or 53, with higher scores indicating better learning and recall) and Delayed Recall (N=176; raw scores range from 0 to 39 or 50, with higher scores indicating better learning and recall) was compared with ACE-III Memory performance. Scores on the Controlled Oral Word Association Test (COWAT; N=195; Semantic Fluency [vegetables, N=184]; raw scores are based on number of words produced, with higher scores indicating better verbal fluency) were compared with ACE-III Fluency scores. Boston Naming Test (BNT; N=188; scores range from 0 to 60, with higher scores indicating better confrontation naming) performance was compared with ACE-III Language performance, and scores on the WAIS-IV Block Design (BD; N=178; scores range from 0 to 66, with higher scores indicating better visuospatial functioning) were compared with the ACE-III Visuospatial scores. Although most patients completed all tests, some variability was observed because measures were administered within clinical evaluations.

ACE-III Subscales Predicting Neuropsychological Test Scores

Multivariate multiple regression analyses were used to determine whether ACE-III subscale performance predicted performance on neuropsychological tests measuring similar cognitive constructs. Age was included as a covariate in the models (Table 2). All statistical analyses were performed with IBM SPSS version 24 software.

Exploratory Analysis: ACE-III Performance by Education

Separate multivariate multiple regression analyses were deployed to determine whether level of education also influenced ACE-III subscale scores in predicting performance on neuropsychological tests measuring similar cognitive constructs. Results were split into two groups by years of education (≤14 years of education [N=72], >14 years of education [N=145]) to determine any differences in regression models (Tables 3 and 4). This variable was dichotomized at 14 years according to the education range (7–22 years) of our highly educated sample. This approach was consistent with the educational attainment distribution within the state of Massachusetts: 44%≥16 years and 56%<16 years (17). Age was also included as a covariate in the models.

Predicting Neurocognitive Disorders

Binomial logistic regressions compared ACE-III total score with performance across the neuropsychological measures in predicting neurocognitive diagnosis. Diagnostic group was dichotomized as neurocognitive disorder (mild NCD or major NCD) versus NC. The diagnostic groups differed in age (mild NCD and major NCD>NC); therefore, age was included as a covariate in the model. To summarize overall performance on neuropsychological testing, a composite score was created by averaging z scores across all neuropsychological measures (WAIS-IV DSB, WAIS-IV BD, TMT B, COWAT, Semantic Fluency, BNT, and WMS-IV LM I and LM II). Composite scores were calculated only for individuals who completed all the neuropsychological measures listed above (N=123).

ACE-III Performance by Group

Analyses of variance were used to determine whether diagnostic groups differed in ACE-III subscale and total scores. A Bonferroni correction was used for multiple comparisons (all corrected at p<0.05).

Results

Patient Demographics

Diagnostic groups statistically significantly differed in age (F=9.19, df=2 and 214, p<0.001). A Bonferroni corrected analysis revealed that patients in the NC group (mean±SD age=71.1±7.4 years) were significantly younger than patients in both the mild NCD (75.0±6.1 years) and major NCD (76.0±7.7 years) groups. The mild NCD and major NCD groups did not differ in age. The diagnostic groups did not differ in education (F=0.49, df=2 and 214, p=0.69) or sex (χ²=2.12, df=2, p=0.35). Additional demographic data are provided in Table 1.

ACE-III Subscales Predicting Neuropsychological Test Scores

As shown in Table 2, regression analyses indicated that ACE-III subscale performances significantly predicted performances on within-construct neuropsychological tests. ACE-III Attention and Orientation subscale performance (F=5.89, df=8 and 107, p<0.001) predicted performance on the WAIS-IV DSB (adjusted R²=0.07, p=0.04) and TMT B (adjusted R²=0.43, p<0.001) assessments. ACE-III Fluency subscale (F=4.82, df=8 and 103, p<0.001) predicted performance on COWAT (adjusted R²=0.53, p<0.001) and Semantic Fluency (adjusted R²=0.28, p<0.001) tests. ACE-III Language subscale (F=4.20, df=8 and 69, p<0.001) predicted BNT performance (adjusted R²=0.67, p<0.001). ACE-III Memory subscale scores (F=2.91, df=8 and 99, p=0.006) predicted performance on both WMS-IV LM I (adjusted R²=0.21, p<0.001) and LM II (adjusted R²=0.38, p<0.001). The ACE-III Visuospatial subscale (F=3.74, df=8 and 70, p=0.001) predicted WAIS-IV BD performance (adjusted R²=0.29, p<0.001). ACE-III subscales also predicted neuropsychological test performance in different construct domains, such as the Fluency subscale being predictive of performance on the WAIS-IV DSB, TMT B, BNT, and WMS-IV LM I tests. Similar relationships were observed with the Language, Memory, and Visuospatial subscales predicting performance within different constructs. This relationship was not observed with the Attention and Orientation subscale.

Exploratory Analysis: ACE-III Performance by Education

Separate regression analyses indicated that ACE-III subscale performances significantly predicted neuropsychological test performance within the same construct for those with >14 years of education (Table 3) but were less predictive for individuals with ≤14 years of education (Table 4). For individuals with >14 years of education, the ACE-III Attention and Orientation subscale score (F=5.79, df=8 and 70, p<0.001) predicted performance on WAIS-IV DSB (adjusted R²=0.13, p=0.013) and TMT B (adjusted R²=0.55, p<0.001) assessments. ACE-III Fluency subscale scores (F=3.81, df=8 and 66, p=0.001) predicted performance on COWAT (adjusted R²=0.51, p<0.001) and Semantic Fluency (adjusted R²=0.28, p<0.001) tests. ACE-III Language subscale scores (F=4.20, df=8 and 69, p<0.001) predicted BNT performance (adjusted R²=0.68, p<0.001). ACE-III Memory subscale scores (F=3.43, df=8 and 63, p=0.002) predicted performance on both WMS-IV LM I (adjusted R²=0.30, p<0.001) and LM II (adjusted R²=0.46, p<0.001). ACE-III Visuospatial subscale scores (F=3.74, df=8 and 70, p=0.001) predicted WAIS-IV BD (adjusted R²=0.35, p<0.001).

For individuals with ≤14 years of education, ACE-III Fluency subscale scores (F=1.40, df=8 and 20, p=0.268) significantly predicted performance on COWAT (Letter Fluency; adjusted R²=0.66, p<0.001) but not Semantic Fluency (adjusted R²=0.09, p=0.249) tests. ACE-III Language subscale performance (F=1.808, df=8 and 23, p=0.127) predicted BNT performance (adjusted R²=0.61, p<0.001).

Predicting Neurocognitive Disorders

A logistic regression was performed to determine the effects of age, ACE-III total score, and the neuropsychological composite score on neurocognitive diagnosis (neurocognitive disorder was defined as 0 and NC as 1). The logistic regression model was statistically significant (χ²=65.93, df=3 and 123, p<0.001). The model explained 57% (Nagelkerke R²) of variance in neurocognitive diagnosis and accurately classified 78% of cases. Higher ACE-III total score (β=0.13, p=0.01) and neuropsychological composite score (β=2.2, p<0.001) were associated with an increased likelihood of being classified as having NC.

Overall ACE-III Performance by Group

ACE-III total and subscale scores significantly differed among the three diagnostic groups: ACE-III total (F=122.20, df=2 and 214, η_p²=0.53), ACE-III Attention and Orientation (F=68.86, df=2 and 214, η_p²=0.39), ACE-III Fluency (F=54.31, df=2 and 214, η_p²=0.34), ACE-III Language (F=24.35, df=2 and 214, η_p²=0.19), ACE-III Memory (F=74.89, df=2 and 214, η_p²=0.41), and ACE-III Visuospatial (F=53.03, df=2 and 214, η_p²=0.33) (all omnibus p<0.001). Post hoc analyses revealed that the NC group had higher ACE-III total scores (90.9±5.7) than the mild NCD (82.4±8.5) and major NCD (65.4±11.5) groups, with the mild NCD group having significantly higher ACE-III total scores than the major NCD group (all p<0.05). The NC group also had higher ACE-III subscale scores (Figure 1) than the mild NCD and major NCD groups, with the mild NCD group having significantly higher ACE-III subscale scores than the major NCD group (all p<0.05).

Discussion

This is the first study in a U.S.-based population to examine the utility of the readily available ACE-III cognitive screening tool for predicting subsequent neuropsychological test performance. The results indicate that ACE-III subscale performance was highly predictive of neuropsychological performance on clinical evaluation. Consistent with previous research, ACE-III subscale scores most strongly predicted performance on neuropsychological tests that measured the same cognitive functions (11). Nevertheless, ACE-III subtests also significantly predicted performance on neuropsychological tests measuring different constructs, a finding that is consistent with existing literature. The measures used to assess different cognitive domains often draw upon multiple basic or subordinate processes. For example, previous research shows that scores on the ACE-III Memory subscale are highly correlated with performance on language measures as well as memory measures (4). One reason for this observation may be that the memory measures on the ACE-III are verbal and thus may affect more than one domain. Finally, some shared variance is expected, given that each patient completed all measures.

The relationship between ACE-III performance and neuropsychological test performance showed some education-based differences. The ACE-III predicted performance on more within-construct neuropsychological tests for individuals with higher (>14 years) versus lower (≤14 years) educational attainment. Specifically, for individuals with lower educational attainment, ACE-III subscale scores predicted performance only on measures of phonemic fluency and confrontation naming. This finding may reflect greater variability in educational experiences among individuals with lower versus higher education due to psychosocial factors (e.g., limited academic opportunities). Further, cognitive reserve may be more heavily recruited on neurocognitive testing by those with higher compared with lower educational attainment (18). In clinical practice, these results suggest that extended neuropsychological testing is important for individuals with lower levels of education, because the ACE-III alone may not fully capture the nature or extent of cognitive difficulty. The results are somewhat unexpected, because those with higher education or higher cognitive reserve are thought to require more in-depth or more challenging testing to detect deficits. Further research may help to elucidate this surprising result.

Indeed, for all patients, a comprehensive neuropsycho-logical evaluation was a markedly better predictor of neurocognitive disorder than overall ACE-III performance (i.e., total score). Specifically, age and ACE-III total score had comparable predictive power (i.e., older patients and those with lower ACE-III total scores were more likely to have a neurocognitive disorder), a result that is consistent with previous research (6, 10, 19). However, overall neuropsychological performance was more accurate in predicting mild NCD or major NCD than either age or ACE-III total score. Thus, although the ACE-III is a sensitive cognitive screening measure, a comprehensive neuropsychological evaluation provides the best predictive utility in identifying neurocognitive disorders. In addition to those with lower educational attainment, patients who are younger, have milder symptoms, or present with complex differential diagnostic questions (e.g., possible mixed neurodegenerative pathology, comorbid psychiatric concerns) may particularly benefit from neuropsychological evaluation, which should be a focus of future study. Nonetheless, our results suggest that ACE-III subscales predict neuropsychological test performance and can serve as an important screening tool in clinical practice.

As expected, ACE-III total score discriminated patients with NC from those with mild NCD and major NCD, and patients with mild NCD performed better than those with major NCD, confirming other reports of this screening tool’s sensitivity to the presence and level of neurocognitive disorders (4, 8, 14). This pattern was also consistent for all ACE-III subscales. Groups differed most on the Memory subscale score. Although this finding may reflect that most patients were referred for concerns regarding memory problems, it is consistent with previous research suggesting that the ACE-III Memory subscale has the greatest sensitivity for detecting neurocognitive disorders (11).

Our study was limited in that it was a retrospective analysis of a heterogeneous clinical sample referred for neuropsychological evaluation because of concerns about cognitive function. Additionally, diagnoses of NC, mild NCD, or major NCD were made before and independently of the current study but were considered in clinical history, medical records, previous neurological workup, and the entire neuropsychological evaluation, including ACE-III performance. It will be important to replicate these findings in a clinical sample in which a diagnosis of neurocognitive disorder is made independently of ACE-III performance. Future research should also determine whether ACE-III scores correspond to measures of functional independence (i.e., activities of daily living questionnaires) and biomarkers for neurodegenerative disease. Direct comparison with other established screening measures (e.g., MoCA and Mini-Cog) (20) is also needed to guide decision making about the most optimal screening tool for use in clinical settings. Our analyses examining the role of education influencing predictive utility of the ACE-III were also limited. The exploratory analyses addressed only educational influence on subscale scores and not total score. Future research in a U.S.-based population should examine whether ACE-III total score remains predictive of cognitive performance after accounting for education, as well as sensitivity and specificity analyses of total score.

Conclusions

Altogether, the ACE-III is an easily administered and relatively brief screening measure that enables detection of neurocognitive disorders and also differentiates patients on the basis of disorder severity. In practice, implementation of the ACE-III when cognitive concerns are first noticed in the primary care or neurology setting may lead to earlier identification and intervention. However, our results also suggest that screening tools such as the ACE-III cannot always be substituted for a comprehensive neuropsychological evaluation, but it may offer a wider breadth of cognitive functioning than the MMSE or MoCA before a referral for neuropsychological evaluation. These referrals for further workup and treatment may be most appropriate for patients who have lower educational status, are younger, or have complex symptom presentations.