ACE
ACE-R
ACE-III
M-ACE
MMSE
Orientation: time
5
5
5
4
5
Orientation: place
5
5
5
5
Registration
3
3
3
3
Attention/concentration (serial 7 s, DLROW)
5 (best performed task)
5 (best performed task)
5 (serial 7 s only)
5
Memory: recall
3
3
3
3
Memory: anterograde memory (name and address)
28
19
19
14
Memory: retrograde memory
4
4
4
Verbal fluency: letters and Animals in 1 min
14
14
14
7 (letters or animals in different versions)
Language: naming
12
12
12
2
Language: comprehension
8
8
7
4
Language: repetition
5
4
4
1
Language: reading
2
1
1
Language: writing
1
1
2
1
Visuospatial abilities: intersecting pentagons
1
1
1 (intersecting lemnisci)
1
Visuospatial abilities: wire (Necker) cube
1
2
2
Visuospatial abilities: clock drawing
3
5
5
5
Perceptual abilities: dot counting
–
4
4
Perceptual abilities: fragmented letters
–
4
4
Total score
100
100
100
30
30
In the index study [4], ACE proved acceptable to patients and relatively quick to administer (ca. 15 min). A patient group (n = 139, of 210 screened, excluding patients with dual pathology, depression, and non-degenerative, non-vascular pathology) was examined, of whom most had dementia (115; non-dementia = 24), along with a control group (n = 127; education-matched individuals attending orthopedic or gynecology clinics and their spouses, and members of the Medical Research Council subject panel).
At cut-off scores of 88/100 and 83/100, ACE was reported to have good sensitivity and specificity for identifying dementia (0.93 and 0.71; 0.82 and 0.96, respectively), figures which compared favorably to the MMSE at a cut-off of 24/30 (0.52 and 0.96, respectively).
Mathuranath et al. [4] observed that patients with AD and FTD showed significant differences on performance of different components of the ACE: orientation, attention and memory were worse in AD patients, whilst letter fluency, language and naming were worse in FTD patients. This scoring pattern was translated into an index reported to be useful for the differentiation of AD and FTD, the (V + L)/(O + M) or the VLOM ratio, given by the formula:
For the ACE, the maximum scores for each of these components gave a ratio of 42/17. A VLOM ratio >3.2 showed sensitivity of 0.75 and specificity of 0.84 for the diagnosis of AD compared to non-AD. A VLOM ratio <2.2 showed sensitivity of 0.58 and specificity of 0.97 for the diagnosis of FTD versus non-FTD [4].

6.2.2 Addenbrooke’s Cognitive Examination-Revised (ACE-R)
The Addenbrooke’s Cognitive Examination-Revised (ACE-R) [5] was a development of the earlier ACE which also incorporated the MMSE, but had clearly defined subdomain scores. Like the ACE, the overall ACE-R score was 100 (Box 6.1), from which domain scores for attention and orientation, memory, fluency, language and visuospatial abilities could be generated (Box 6.2). Test reliability was very good as judged by its internal consistency (Cronbach alpha coefficient = 0.8).
Box 6.2 Domain Scores of ACE-R and ACE-III
Attention and orientation | 18 |
Memory | 26 |
Fluency | 14 |
Language | 26 |
Visuospatial | 16 |
Total score | 100 |
In the index study [5], ACE-R proved acceptable to patients and relatively quick to administer (ca. 15 min). The cohort examined (n = 241; dementia 142, mild cognitive impairment [MCI] 36, controls 63) was selected using exclusion criteria as for the ACE study (psychiatric disorder, mixed pathology, non-neurodegenerative disease process). At cut-off scores of 88/100 and 82/100, ACE-R was reported to have good sensitivity and specificity for identifying dementia (0.94 and 0.79; 0.84 and 1.00, respectively). MCI group performance fell between that of controls and AD patients.
As with the ACE, a subscore was derived from the ACE-R, the VLOM ratio, which was reported to be helpful in differentiating AD from FTD. The same criteria were applied for calculating the VLOM ratio (although not explicitly stated, the maximum score for each of these components in the ACE-R gives a ratio of 40/17). ACE-R VLOM ratio >3.2 showed sensitivity of 0.74 and specificity of 0.85 for the diagnosis of AD compared to non-AD; whilst VLOM ratio <2.2 showed sensitivity of 0.58 and specificity of 0.95 for the diagnosis of FTD versus non-FTD [5]. The findings were therefore similar to those with the VLOM ratio derived from the ACE.
6.2.3 ACE-III, ACEapp
ACE-III [6] was developed to expunge the MMSE items in ACE and ACE-R (Box 6.1). Up until 2001, MMSE was freely available, but in that year the copyright was acquired by Psychological Assessment Resources which terminated the free availability of MMSE [9, 10], hence the necessity to remove MMSE items. In ACE-III, these MMSE items were substituted like for like as far as possible, for example the intersecting pentagons were replaced with intersecting lemnisci, resulting in the same domain scores as for ACE-R (Box 6.2). Internal reliability was high (Cronbach alpha coefficient = 0.88).
In the index study [6], the cohort examined (n = 86; AD 28, FTD 33, controls 25) found ACE-III to be acceptable and it was relatively quick to administer (ca. 15 min). ACE-III and ACE-R were highly correlated (r = 0.99), and at the previously recommended cut-off scores ACE-III was both highly sensitive and specific (at 88/100: 1.00 and 0.96 respectively; at 82/100: 0.93 and 1.00 respectively). ACE-III cognitive domains correlated significantly with standard neuropsychological tests.
ACE-III has also been made available as an i-pad based app, which is available cost-free via iTunes and at acemobileorg@gmail.com. The automated scoring and the clear instructions are designed to reduce errors in administration and scoring.
6.2.4 Mini-Addenbrooke’s Cognitive Examination (M-ACE)
The Mini-Addenbrooke’s Cognitive Examination (M-ACE) [7] was developed from the longer ACE-R and ACE-III instruments by using Mokken scaling analysis in 117 dementia patients. The resultant M-ACE comprises tests of attention, memory (7-item name and address), letter fluency, clock drawing, and memory recall, scored out of 30 (Box 6.1). Internal reliability was high (Cronbach alpha coefficient = 0.83).
In the index study [7], the cohort examined (n = 242) was heterogeneous with respect to diagnosis (AD 28, behavioral variant FTD 23, primary progressive aphasia 82, corticobasal syndrome 21, controls 78). Two cut-offs were identified: ≤25/30 had high sensitivity (0.85) and high specificity (0.87); and ≤21/30 had high specificity (1.00) and hence a score almost certain to have come from a dementia patient. M-ACE was more sensitive than the MMSE, and less likely to have ceiling effects [7].
6.3 ACE Translations
The excellent performance of the various iterations of the ACE has prompted translation into a number of languages [11–55] (Table 6.1). These translations have facilitated the examination of ACE performance in a large number of independent patient cohorts.
Table 6.1
Reported translations of the various Addenbrooke’s Cognitive Examinations (ACEs)
Language | ACE | ACE-R | ACE-III | M-ACE |
---|---|---|---|---|
Arabic | Al Salman et al. [11] | |||
Cantonese | Wong et al. [12] | |||
Chinese | Fang et al. [13] | Wang et al. [submitted] | ||
Czech | ||||
Danish | Stokholm et al. [17] | |||
Dutch | Robben et al. [18] | |||
French | Bastide et al. [21] | |||
German | Alexopoulos et al. [22] | Alexopoulos et al. [23] | ||
Greek | Konstantinopoulou et al. [24] | |||
Hebrew | Newman [25] | |||
Hungarian | Kaszas et al. [26] | |||
Italian | ||||
Japanese | Yoshida et al. [29] | |||
Korean | Heo et al. [32] | Kwak et al. [33] | ||
Lithuanian | ||||
Malayalam | ||||
Persian | Pouretemad et al. [39] | |||
Portuguese | ||||
Spanish | Matias-Guiu et al. [54] | Matias-Guiu and Fernandez-Bobadilla [55] |
6.4 Systematic Reviews, Meta-analysis, and Independent Cohort Studies
A systematic review of studies of both ACE and ACE-R published up to April 2010 [56] identified 45 studies in all, of which 9 [4, 5, 57–63] were deemed suitable for review following the authors inclusion/exclusion criteria (translated versions were excluded). It was concluded that both ACE and ACE-R were capable of differentiating between patients with and without cognitive impairment, but that the evidence base on distinguishing dementia subtypes and MCI was lacking [56].
6.4.1 ACE
A meta-analysis of the accuracy of ACE in the detection of dementia and mild cognitive impairment [64] identified 29 studies published up to May 2013, 13 using the English version [4, 57–62, 65–70] and 16 using translated versions [8, 17, 19, 20, 22, 25, 26, 29, 32, 36, 37, 39, 45–48], of which 5 studies met the authors’ specified inclusion/exclusion criteria [4, 17, 29, 47, 60] for meta-analysis.
The sensitivity and specificity of the ACE to identify dementia compared with mixed subjects without dementia were 0.969 (95 % CI = 0.927–0.994) and 0.774 (95 % CI = 0.583–0.918) respectively. In a setting where the prevalence of dementia may be approximately 25 %, such as primary care or general hospital settings, the overall accuracy of the ACE would be 0.823, with a positive predictive value of 0.589. Thus ACE was not recommended for use in low prevalence settings. In the setting of a dedicated memory clinic where the prevalence of dementia may be approximately 50 %, the overall accuracy of the ACE would be 0.872, with a positive predictive value of 0.811. Thus ACE was recommended for use in high prevalence settings [64].
6.4.2 ACE-R
A meta-analysis of the accuracy of ACE-R in the detection of dementia and mild cognitive impairment [64] identified 31 studies published up to May 2013, 16 using the English version [5, 63, 71–84] and 15 using translated versions [11, 18, 21, 23, 24, 30, 31, 33, 40–42, 49 (included in error), 51–53], of which 5 studies met the authors’ specified inclusion/exclusion criteria [5, 23, 30, 31, 72, 73] for meta-analysis.
The sensitivity and specificity of the ACE-R to identify dementia compared with mixed subjects without dementia were 0.957 (95 % CI = 0.922–0.982) and 0.875 (95 % CI = 0.638–0.994) respectively. In low dementia prevalence settings (25 %), the overall accuracy of the ACE-R would be 0.895, with a positive predictive value of 0.719. In high dementia prevalence settings (50 %), the figures for ACE-R accuracy and positive predictive value would be 0.916 and 0.885 respectively. Thus the ACE-R would have good utility at 25 % prevalence and excellent properties at 50 % prevalence [64].
A systematic review and meta-analysis of cognitive tests to detect dementia [85] included 12 studies of ACE-R [5, 12, 13, 21, 23, 24, 27, 31, 33, 40, 51, 75] and found a pooled sensitivity of 0.92 (95 % CI = 0.90–0.94) and pooled specificity of 0.89 (95 % CI = 0.84–0.93). Of the 11 screening tests reviewed in this meta-analysis, ACE-R was the best alternative to MMSE, along with the Mini-Cog [85].
6.4.3 ACE-III
Aside from the index study [6], few studies of ACE-III have been published at time of writing [54, 86, 87], but all confirm its utility for the identification of dementia.
In a cohort (n = 59) of elderly patients (age 75–85 years) attending a memory clinic, Jubb and Evans found excellent accuracy for the detection of dementia, but suggested a lower cut-off (<81/100) was preferable to the published cut-offs at medium and low prevalence rates, with sensitivity of 0.79 and specificity of 0.96 in their patient group [86].
In a study of ACE-III for the diagnosis of early-onset dementia (<65 years), a patient group (n = 71: AD 31, primary progressive aphasia 11, behavioral variant FTD 18, posterior cortical atrophy 11) was compared with healthy controls (28) and subjective memory impairment (15). At the specified ACE-III cut-off of 88/100 ACE-III distinguished early-onset dementia from healthy controls with high sensitivity (0.915) and specificity (0.964), and also from subjective memory impairment with high sensitivity (0.915) and specificity (0.867) [87].
In patients assessed in an in-patient stroke rehabilitation setting, median time to complete ACE-III was found to be 18 min (range 10–35 min) [88].
6.4.4 M-ACE
Aside from the index study [7], few other studies of M-ACE have been published to date.
Using a Spanish translation in a cohort of mixed dementia patients and controls (n = 175) with relatively low educational experience, Matias-Guiu and Fernandez-Bobadilla [55] found that a cut-off of 16/17 had optimal sensitivity (0.867) and specificity (0.870) for the diagnosis of dementia.
In pragmatic studies in a dedicated secondary care cognitive disorders clinic, M-ACE cut-off of ≤25/30 had excellent sensitivity for diagnosis of dementia (1.00) and MCI (1.00) but with limited specificity (0.28, 0.43 respectively), whereas at the lower cut-off of ≤21/30 sensitivity was reduced (0.92, 0.77) but with improved specificity (0.61, 0.82 respectively) [89]. These findings were reproducible in an independent cohort [90].
6.5 Diagnostic Utility
6.5.1 Normative Studies
6.5.2 Dementia and Cognitive Impairment
Perhaps the first objective in any clinical assessment of patients with cognitive complaints is to determine whether they suffer from dementia or from lesser degrees of cognitive impairment which may be variously denoted as mild cognitive disorder, cognitive impairment no dementia, or mild cognitive impairment. Although the latter term may be used broadly for any etiology of cognitive impairment not meeting criteria for dementia, some authorities reserve it for a more restrictive sense, specifically a precursor state for AD (henceforward designated MCI; see below, at Sect. 6.5.4). Hence, the performance of ACEs on this diagnostic question is examined first, prior to differential diagnosis from depression (below, at Sect. 6.5.3) and diagnostic utility for various dementia subtypes. Generally, the idiom of clinical practice revolves around the assessment of patients with cognitive complaints of unknown etiology, rather than groups preselected by diagnosis with or without a control group, as occurs in initial “proof-of-concept” diagnostic test accuracy studies [92]. Hence, pragmatic studies of the ACEs are considered first.
A pragmatic prospective study of the ACE conducted in consecutive new patient referrals to a cognitive function clinic (n = 285; dementia prevalence = 0.49) over a period of 42 months found ACE to be easy to use with very few patients failing to complete the test [60, 65]. ACE scores and MMSE scores were highly correlated (r = 0.92) [65]. Using the ACE cut-offs specified in the index paper (88/100 and 83/100) [4], test sensitivity for the diagnosis of dementia was high (1.00 and 0.96 at 88/100 and 83/100 respectively) but specificity was less good (0.43 and 0.63 respectively), considerably less impressive than those documented in the index study (see above, at Sect. 6.2.1). Using an arbitrarily chosen lower ACE cut-off of 75/100 [66], justified on the basis that, unlike the index study, this pragmatic study did not include a normal control group and hence was more representative of day-to-day clinical practice, ACE sensitivity and specificity were both greater than 0.8, as was positive predictive value (PPV; Table 6.2). Area under the receiver operating characteristic curve (AUC ROC), a measure of diagnostic accuracy (see Chap. 2, at Sect. 2.4.3), was 0.93 (95 % confidence intervals 0.90–0.96) [60].
Table 6.2
Diagnostic accuracy of ACE for diagnosis of dementia: summary of results (with 95 % confidence intervals) at various ACE cut-off scores
ACE cut-off | <88/100 | <83/100 | <75/100 |
---|---|---|---|
Test accuracy | 0.71 (0.66–0.76) | 0.79 (0.75–0.84) | 0.84 (0.80–0.88) |
Sensitivity | 1.00 | 0.96 (0.93–0.99) | 0.85 (0.79–0.91) |
False positive rate | 0.57 (0.48–0.65) | 0.37 (0.29–0.45) | 0.17 (0.11–0.23) |
Specificity | 0.43 (0.35–0.42) | 0.63 (0.55–0.71) | 0.83 (0.77–0.89) |
Youden index (Y) | 0.43 | 0.59 | 0.68 |
False negative rate | 0 | 0.04 (0.01–0.07) | 0.15 (0.09–0.21) |
Positive predictive value (PPV) | 0.63 (0.57–0.69) | 0.71 (0.65–0.78) | 0.83 (0.77–0.89) |
False alarm rate | 0.37 (0.31–0.43) | 0.29 (0.22–0.35) | 0.17 (0.11–0.23) |
Negative predictive value (NPV) | 1 | 0.95 (0.90–0.99) | 0.85 (0.79–0.91) |
Predictive summary index (PSI) | 0.63 | 0.66 | 0.68 |
False reassurance rate | 0 | 0.05 (0.01–0.09) | 0.15 (0.09–0.21) |
Positive likelihood ratio (LR+) | 1.77 (1.53–2.04) | 2.59 (2.10–3.21) | 5.14 (3.54–7.45) |
Negative likelihood ratio (LR−) | 0 | 0.06 (0.05–0.07) | 0.18 (0.12–0.26) |
Diagnostic odds ratio (DOR) | ∞ | 45.5 | 28.6 |
Positive utility index (UI+) | 0.63 adequate | 0.68 good | 0.71 good |
Negative utility index (UI−) | 0.43 poor | 0.60 adequate | 0.71 good |
Area under the receiver operating characteristic curve (AUC ROC) | 0.93 (0.90–0.96) |
Although changing test cut-offs from those defined in index studies is frowned upon as a potential source of bias [93], nevertheless other studies have also found lower ACE cut-offs to be necessary to maximize diagnostic utility, for example in a rural Spanish patient cohort with low educational level [47].
A pragmatic prospective study of the ACE-R conducted over 36 months (n = 243; dementia prevalence = 0.35) found ACE-R easy to administer, with very few patients failing to complete the test [63, 72]. ACE-R scores and MMSE scores were highly correlated (r = 0.90). Initial results using the ACE-R cut-offs specified in the index paper (88/100 and 82/100) [5] showed excellent sensitivity for dementia (1.00 and 0.96 at 88/100 and 82/100 respectively) but poor specificity (0.48 and 0.72 respectively), much poorer values than those documented in the index study (see above, at Sect. 6.2.2). Using a lower ACE-R cut-off of 75/100, as previously used with ACE [60], sensitivity and specificity were both greater than 0.9 and PPV approached this value (Table 6.3) [63].
Table 6.3
Diagnostic accuracy of ACE-R for diagnosis of dementia: summary of results (with 95 % confidence intervals) at various ACE-R cut-off scores
ACE-R cut-off | <88/100 | <82/100 | <75/100 |
---|---|---|---|
Test accuracy | 0.72 (0.63–0.81) | 0.83 (0.76–0.90) | 0.91 (0.85–0.97) |
Sensitivity | 1 | 0.96 (0.90–1.0) | 0.91 (0.83–0.99) |
False positive rate | 0.52 (0.39–0.65) | 0.28 (0.16–0.40) | 0.09 (0.02–0.17) |
Specificity | 0.48 (0.35–0.61) | 0.72 (0.60–0.84) | 0.91 (0.83–0.98) |
Youden index (Y) | 0.48 | 0.68 | 0.82 |
False negative rate | 0 | 0.04 (−0.02–0.1) | 0.09 (0.01–0.17) |
Positive predictive value (PPV) | 0.62 (0.51–0.73) | 0.75 (0.63–0.86) | 0.89 (0.81–0.98) |
False alarm rate | 0.38 (0.27–0.48) | 0.25 (0.14–0.37) | 0.11 (0.02–0.19) |
Negative predictive value (NPV) | 1 | 0.95 (0.89–1.02) | 0.92 (0.85–0.99) |
Predictive summary index (PSI) | 0.62 | 0.70 | 0.81 |
False reassurance rate | 0 | 0.05 (−0.02–0.1) | 0.08 (0.01–0.15) |
Positive likelihood ratio (LR+) | 1.93 (1.49–2.49) | 3.44 (2.23–5.32) | 9.86 (4.26–22.8) |
Negative likelihood ratio (LR−) | 0 | 0.06 (0.04–0.09) | 0.09 (0.04–0.22) |
Diagnostic odds ratio (DOR) | ∞ | 57.2 | 102.9 |
Positive utility index (UI+) | 0.62 adequate | 0.72 good | 0.81 excellent |
Negative utility index (UI−) | 0.48 poor | 0.68 good | 0.84 excellent |
Subsequently sensitivity and specificity of ACE-R was examined at all cut-off values and an optimal cut-off defined by maximal correct classification accuracy for the differential diagnosis of dementia/not dementia (=73/100). At this cut-off, results were similar to those in the initial analysis with cut-off 75/100 (Table 6.4 , left hand column). Area under the ACE-R ROC curve was 0.94 (95 % confidence intervals 0.91–0.97) [72, 73].
Table 6.4
Diagnostic accuracy of ACE-R and MMSE for dementia: summary of results (with 95 % confidence intervals) of ACE-R and MMSE assessments
Cut-off | ACE-R ≥73/100 | MMSE ≥24/30 |
---|---|---|
Test accuracy | 0.89 (0.85–0.93) | 0.82 (0.77–0.87) |
Sensitivity | 0.87 (0.80–0.94) | 0.70 (0.60–0.80) |
Specificity | 0.91 (0.86–0.95) | 0.89 (0.84–0.94) |
Youden index (Y) | 0.78 | 0.69 |
Positive predictive value (PPV) | 0.83 (0.75–0.91) | 0.77 (0.67–0.86) |
Negative predictive value (NPV) | 0.93 (0.89–0.97) | 0.85 (0.79–0.90) |
Predictive summary index (PSI) | 0.76 | 0.62 |
Positive likelihood ratio (LR+) | 9.21 (5.65–15.0) = moderate | 6.17 (3.91–9.73) = moderate |
Negative likelihood ratio (LR−) | 0.14 (0.09–0.24) = moderate | 0.34 (0.21–0.53) = small |
Diagnostic odds ratio (DOR) | 63.7 (39.1–103.9) | 18.4 (11.6–29.0) |
Positive utility index (UI+) | 0.72 good | 0.54 adequate |
Negative utility index (UI−) | 0.85 excellent | 0.76 good |
Area under the receiver operating characteristic curve (AUC ROC) | 0.94 (0.91–0.97) | 0.91 (0.88–0.95) |
A prospective study of 122 patients referred to a cognitive clinic (dementia prevalence = 0.67) found sensitivity and specificity for dementia diagnosis of 0.85 and 0.80 at ACE-R cut-off of 84/100. Misclassification was noted in individuals with high levels of education, focal executive dysfunction, significant vascular disease, medical comorbidities and polypharmacy [75].

Stay updated, free articles. Join our Telegram channel

Full access? Get Clinical Tree

