Abstract
INTRODUCTION
Limb apraxia (LA) is a syndrome of impaired motor cognition that has been defined as the inability to carry out skilled motor acts despite preserved elementary sensorimotor abilities [1]. Current models of apraxia distinguish between two functionally distinct components: a conceptual and a production praxis system [2, 3]. The conceptual system involves meaningful gestures, object use, and associative knowledge [3] and is tested by asking patients to use or match tools, knowledge about body functions, or to produce or understand emblems (symbolic gestures) [4–8]. Conversely, the praxis production system is mainly concerned with the perception, programming, and execution of spatiomotor information of body parts and their topographical relationship; it can be assessed by imitation of nonsymbolic hand positions or hand movements [9]. Although often studied in stroke patients, LA is also found in neurodegenerative disorders such as Alzheimer’s disease (AD) where it constitutes an important non-memory cognitive impairment. Previous studies have found various degrees of LA in AD, dependent on the setting, cohort characteristics, and assessment procedures [5, 10–14]. Most prior studies were concerned with the characteristics of apraxia subtypes, cognitive correlates, or the presence of apraxia among different dementias [4, 14–19]. Also, many studies used extensive or combined test batteries [10, 20]. In contrast, only few studies used bimanual motor tasks [21–23], and the role of bimanual gesture imitation in dementia is only inadequately characterized. The main goal of the study was to determine the diagnostic performance of a new assessment procedure requiring bimanual gestures of variable complexity. Bimanual was preferred to unimanual praxis because its complexity requires higher amounts of spatial processing and executive control [21, 24]. For convenience, an impairment to produce bimanual nonsymbolic gestures and movements was termed bimanual apraxia (BA). We wanted to learn how often BA can be found in first-time visitors of a memory clinic, and what its predictor variables are. Since mild cognitive impairment (MCI) [25] denotes a large cohort in outpatient clinics and is often considered as a preclinical state of AD, we were also interested to see how patients with MCI performed on this task.
MATERIALS AND METHODS
Participants
The study was approved by Ethics Committee of the Medical University of Innsbruck; all patients gave their informed written consent to participate. We included 45 non-institutionalized patients with AD, 38 participants with MCI, and 50 sex-, age-, and education-matched healthy normal controls (NC). Study participants were recruited consecutively from the outpatient memory clinic and were categorized clinically. Patients with confounding neurological (e.g., stroke, trauma) and psychiatric disorders were excluded from the study. AD patients were included if (1) they met the core clinical NINCDS-ADRDA criteria for probable AD with amnestic presentation [26]; (2) had a knowledgeable caregiver to obtain information about the patient’s disease and activities of daily living; (3) had normal motor functions as documented by clinical investigation; and (4) had no visual impairment or verbal comprehension deficits that could affect cognitive testing. Clinical Dementia Rating (CDR) [25] was used to provide a global evaluation of dementia severity. The Disability Assessment for Dementia scale (DAD) [27] measured functional abilities in activities of daily living, including basic, instrumental, and leisure activities. Depression was assessed with the 30-item version of the Geriatric Depression Scale (GDS) [28]. MCI patients were recruited according to NIA-AA criteria [29]. MCI participants were nondemented persons with reported cognitive decline (patient or proxy), neuropsychological test performance ranging <1.5 SD below age and education based norms and normal basic activities of daily living as reported by their caregivers. Based on their cognitive profile [30], MCI subjects were further classified as amnestic single (n = 15), amnestic multiple domain (n = 12), or non-amnestic single MCI (n = 11). Standard neuroimaging (MRI or CCT) was used to exclude subjects with major concomitant cerebrovascular disease or lobar focal atrophy. Longitudinal evaluation including clinical assessment, informant questioning, and neuropsychological testing was used to document progressive cognitive decline.
Neuropsychological background testing
Patients and NC completed the standardized German version of the CERAD-Plus test battery [31] including subtests tapping global cognition (MMSE), semantic and phonemic verbal fluency, verbal episodic memory (learning, retrieval and recognition), constructional praxis, figural memory, confrontation naming (Boston Naming Test, BNT), psychomotor speed and cognitive flexibility (Trail making test A and B). A CERAD total score was also calculated [32].
Bimanual gesture imitation
Assessment included bimanual imitation of hand positions and finger configurations (Interlocking Finger Test, ILF) and imitation of hand movements (Alternating Hand Movements, AHM, and Bimanual Rhythm Tapping, BRT). To reduce perceptual complexity and avoid ‘mirror errors’, patients sat next to the investigator during the assessment. Participants were asked to start their imitation after an ample demonstration of each task. Sufficient time, but no feedback or help was given for their response. Self-corrections were allowed and the best patient performance was scored. Each subtest was scored separately and a composite sum score was also calculated. Patient responses were videotaped for subsequent evaluation. A detailed description of gesture imitation is provided in the Supplementary Material.
Statistical analysis
SPSS 22 was used for statistical analysis. Video recordings were scored independently by the raters and an inter-rater reliability for apraxia scoring was calculated for a subset of the cohort (45 AD patients, 21 MCI subjects, 20 normal controls). Scorings were computed using intraclass correlation coefficients with a two-way mixed model with average measures of absolute agreement [33]. Values between 0.75 and 1 were judged as good or excellent. Groups were compared by either one-way ANOVA for normally distributed continuous variables or the Kruskal-Wallis test. Normal distribution of each variable was determined with the Kolmogorov-Smirnov test. Receiver operating characteristic (ROC) curves were calculated to obtain optimal cut-off values and diagnostic accuracy of the individual tests. A Spearman correlational analysis with Bonferroni correction was performed to screen for important demographical, clinical, and neuropsychological variables representing possible predictors of apraxia. Only variables which were correlated with the apraxia subtests at or below p = 0.003 were considered for further analysis. This resulted in 3 clinical and 9 neuropsychological predictor candidates which were entered in a stepwise multiple linear regression analysis. Two models were employed. Model 1 evaluated disease variables (diagnosis, disease duration, and CDR). Model 2 included all Model 1 variables plus the selected scores of the CERAD plus battery (MMSE, semantic and word fluency, BNT, constructional praxis, word list recall and constructional praxis recall, Trail making testA and B).
RESULTS
Group characteristics
Apart from education (AD patients had fewer school years), the groups were well matched (Table 1). The AD group was widely distributed as to its education (range 8 to 17 years), disease severity (CDR range 0.5 to 3), disease duration (range 12 to 96 months), functional abilities (DAD range 4 to 40), and global cognition (MMSE range 14 to 30). Significant group differences emerged on CDR, DAD, the MMSE, the CERAD-Plus subtests, and global score (NC >MCI >AD, respectively).
Bimanual gesture imitation
AD and NC differed significantly on all imitation subtest, as well as AD and MCI, whereas NC and MCI differed only on the BRT (Table 2). Three errors classes were observed: spatial errors (wrong hand orientation, finger configuration, closing-in), executive errors (perseverations, intrusions), and movement errors (incorrect starting position or temporal sequence; problems with simultaneous, alternating or bilateral movement).
The scores of the three subtests were significantly and positively intercorrelated (ILF with AHM: rho = 0.5, ILF with BRT: rho = 0.41, BRT with AHM: rho = 0.41, all at p < 0.001).
Inter-rater agreement
Inter-rater agreement was high with intraclass correlation coefficients of 0.89 (95% CI, 0.76–0.94), 0.89 (95% CI, 0.83–0.93), 0.87 (95% CI, 0.79–0.92), and 0.92 (95% CI 0.77–0.96) for the ILF, AHM, BRT, and sum score, respectively.
Diagnostic test accuracy and prevalence of BA
The diagnostic performance of all apraxia subtests and the total score is listed in Table 3. Imitation was classified as impaired if the score was 2 SD below the mean of the NC group. Sensitivity to detect AD ranged between 0.5 and 0.7, whereas specificity values were higher (0.9 to 0.98). A less rigorous criterion using cutoffs of 1.5 SD below control mean increased sensitivity modestly, but produced many false positive classifications in the NC and MCI cohort. A group comparison revealed that significantly more AD patients than control subjects and MCI participants were impaired (ILF: χ2 = 39.3; AHM: χ2 = 28.8; BRT: χ2 = 51; Sum Score: χ2 = 61.3, all at p < 0.01, respectively). Conversely, only the BRT revealed a significant difference between MCI and NC subjects (p < 0.05). The ROC analysis indicated good diagnostic accuracy of the apraxia battery to detect BA in AD. Areas under the curve were particularly high for the BRT, the ILF and sum score when comparing AD and NC participants but decreased for the NC versus MCI and AD versus MCI comparisons (Table 3, Fig. 1).
Predictor variables of BA
Model 1 (clinical variables) of the regression analysis explained 29% to 42% of variance (Table 4). Diseases severity (CDR) was a significant predictor of the ILF and BRT, whereas diagnosis significantly predicted all subtypes of BA. In Model 2 (combined clinical and cognitive variables), two neuropsychological variables (Trail making test B, constructional praxis recall) also emerged as predictors. However, Model 2 did not expand the variance explained by Model 1. Age, gender, education, disease duration, and other cognitive variables were no predictorsof BA.
DISCUSSION
Apraxia is a common finding in patients with neurodegenerative disorders; its appearance can be used to suspect or corroborate a dementia diagnosis. A new clinical assessment procedure was introduced to study imitation of bimanual motor functions. The term bimanual apraxia (BA) was chosen to describe the combined motor, spatial, and executive impairment of bimanual gesture imitation. To study its diagnostic usefulness, the new test was administered to subjects attending a memory clinic. The test appeared suitable for a clinical environment: it is short, easy to administer, well accepted by patients, does not require extensive rater training, and is reasonably sensitive and specific. Inter-rater reliability was good to excellent. Furthermore, it requires little language comprehension and is not based on premorbid knowledge regarding object use or emblems, abilities which are culture-dependent and person-specific. All three apraxia subtests and the sum score are highly specific in detecting BA in AD; AD was best discriminated from NC by rhythm tapping and the interlocking finger task.
A group comparison showed that BA was frequent in the AD group, whereas it was found in a smaller proportion of MCI patients [34, 35], and rarely in the NC cohort. Thus, BA appears to be an important clinical disease marker of AD [5, 34], whereas it is less appropriate to discriminate MCI from AD, or MCI from normal aging. Patients had probable AD with amnestic presentation [26] and were mildly to moderately impaired, as often found in an outpatient clinic. Approximately two-thirds of the AD cohort presented with BA; this percentage outnumbered previous studies testing for nonsymbolic movements [5, 37]. This discrepancy probably results from the application of complex, bimanual instead of unimanual gestures. Bimanual gestures require advanced visuospatial analysis and gestural construction (orientation and spatial relations of hands, fingers and body in the ILF); a second component (most relevant for AHM and BRT) comprises intermanual coordination of alternating, synchronous, temporally coupled, sequential, or dual (moving while counting) movement features. These features are largely dependent on executive control functions such as planning, monitoring, inhibitory control, and working memory. Furthermore, bimanual motor functions require interhemispheric coordination. The lack of support functions is also reflected by the large number of spatial and executive errors which appeared during movement imitation, in addition to movement errors. Due to task complexity, increased amounts of control and neural connectivity are mandatory [13, 38–40], which makes bimanual imitation more challenging and a useful diagnostic tool [21].
BA and its correlations deserves further notice. In our sample, BA had no age, gender, or education effects, but was mainly predicted by diagnosis and disease severity; also, cognitive variables were only weak predictors. Thus, BA is mainly predicted by the presence of AD pathology, and only marginally by certain patterns of cognitive impairment [22]. Lack of significant correlation with other cognitive variables and the fact that the three apraxia scores were significantly intercorrelated suggests that BA is a separate impairment which can be distinguished from other cognitive deficits and needs specific assessment procedures.
Although our results suggest that the impairment to produce bimanual nonsymbolic gestures and movements is often associated with AD and is a useful assessment procedure in patients with suspected dementia, the study also has weaknesses. Since cognitive background testing was limited, other more specific cognitive predictors of BA in the area of spatial, executive, or attentional processing remain to be elucidated. Furthermore, the study cannot make clear whether BA is associated with specific structural abnormalities, with a particular clinicopathological subtype of AD [41, 42], or with distinct performance profiles that cannot be accounted for by disease severity [43]. Since this is a pilot study, the role of BA in AD and its impact on daily functioning is still unclear. Finally, though BA is highly prevalent in AD, its disease specificity remains to be shown in future comparisons with other dementias.
DISCLOSURE STATEMENT
Authors’ disclosures available online (http://j-alz.com/manuscript-disclosures/16-0680r2).