Abstract
Background:
Agraphia is a typical feature in the clinical course of Alzheimer’s disease (AD).
Objective:
Assess the differences between AD and normal aging as regards kinematographic features of handwriting and elucidate writing deficits in AD.
Methods:
The study included 23 patients with AD (78.09 years/SD = 7.12; MMSE 21.39/SD = 3.61) and 34 healthy controls (75.56 years/SD = 5.85; MMSE 29.06/SD = 0.78). Both groups performed alphabetical and non-alphabetical writing tasks. The kinematographic assessment included the average number of inversions per stroke (NIV; number of peaks in the velocity profile in a single up or down stroke), percentage of automated segments, frequency (average number of strokes per second), writing pressure, and writing velocity on paper.
Results:
A total of 14 patients showed overt writing difficulties reflected by omissions or substitutions of letters. AD patients showed less automated movements (as measured by NIV), lower writing velocity, and lower frequency of up-and-down strokes in non-alphabetical as well as in alphabetical writing. In the patient group, Spearman correlation analysis between overt writing performance and NIV was significant. That means patients who had less errors in writing a sentence showed a higher automaticity in handwriting. The correctness of alphabetical writing and some kinematographic measures in writing non-alphabetical material reached excellent diagnostic values in ROC analyses. There was no difference in the application of pressure on the pen between patients and controls.
Conclusion:
Writing disorders are multi-componential in AD and not strictly limited to one processing level. The slow and poorly automated execution of motor programs is not bound to alphabetical material.
INTRODUCTION
Writing is an important means of communication, which deserves closer assessment in case of neurodegeneration. Possibly, the assessment of writing disorders, of fine motor movements, and of typical patterns of breakdown may contribute to the early diagnosis of Alzheimer’s disease (AD) [1–7]. The exact measurement of symptoms may reveal subtle and subclinical motor abnormalities that are below the threshold of clinical detection [8].
At a very general level, impairments of writing may be classified as “central agraphias”, affecting both oral and written spelling, and “peripheral agraphias”, affecting only the written output. Central agraphias arise from damage to one or more components in generating the spellings of familiar or unfamiliar words [9–11]. The correct case, font, and shape are assigned to the letters and finally appropriate motor programs are activated [12, 13]. Two effector-independent levels of representation have been proposed [9]: the first one consisting in a long-term store of allographs from where the respective allographic forms are retrieved, the second one being a graphomotor pattern which specifies direction, relative size, position, and order of strokes, but not their absolute size and duration, or how they will be effected. The graphic code would provide the specific neuromuscular instructions to execute writing movements [12]. Specific models of handwriting have been proposed [14].
Several studies converge on the view that agraphia is a typical feature in the clinical course of dementia. In AD, an early lexical-semantic memory deficit would lead to deficits in spelling irregular words (resulting in surface dysgraphia [1]); in later stages, the application of specific sublexical phoneme-to-grapheme conversion rules should also deteriorate affecting all stimuli [5]. As Forbes et al. [3] have pointed out, the writing impairment in AD would be multi-componential and reflect the cortical deterioration experienced in the course of the disease. In the milder stages, patients would show a semantic impairment reflected by few information conveyed in written output. This semantic impairment would be associated with a secondary milder impairment in phonological processing. In the course of the disease, central deficits would be accompanied by difficulties at the peripheral level, as problems with letter formation and stroke placement [3]. Other studies have reported mixed findings on agraphia in AD [15]. In a group of AD patients, some patients showed severe agraphia, while others had minimal or no writing deficits. Lambert et al. [7] found various agraphia syndromes in a group of AD patients, including the selective damage to one of the central or peripheral components, as well as difficulties at multiple writing levels. Their findings do not support the assumption of a typical progression of writing disorder or a typical breakdown of writing routines [7]. It has also been proposed that AD patients’ writing deficits result predominantly from a reduction of general cognitive resources and only marginally from deficits in specific spelling sub-components [2].
In AD, peripheral components of the writing process have been less studied than central aspects [16]. In cognitively demanding tasks, such as copying the details of a check into the appropriate places, in-air-time was longer and pressure was lower in the two patient groups (AD, mild cognitive impairment (MCI)) than in healthy controls [16]. In another study using complex tasks, the assessment of kinematographic features of writing and drawing has been found to be a useful and objective addition to the clinical assessment of patients with cognitive impairment [17]. The complexity of tasks and their demand on cognitive functions may be critical. Complex handwriting tasks rely not only on kinesthetic and perceptual-motor components but also require different cognitive components such as visuospatial abilities, monitoring, planning, and working memory, which are known to be affected also in the earlier stages of dementia. Thus, cognitive difficulties may have compromised the overall writing process and may have altered the kinematographic measures. However, alterations of motor writing performance have also been described in simple tasks (writing a series of the cursive letter “l”; [18]; drawing concentric superimposed circles [8]), showing a loss of fine motor performance and less regular movements in patients with MCI or mild AD.
In this study, we aim at investigating the performance of AD patients with slight cognitive deficits and the performance of healthy older people in a series of handwriting tasks (alphabetical and non-alphabetical). This will help us gaining insight into the possible differences between normal and pathological aging.
We hypothesized that patients in mild stages of AD show difficulties in handwriting in terms of less automation and slower writing. In particular, we were interested to assess whether the difficulties would be limited to alphabetical material or whether the patients would also have deficits in simple non-alphabetical writing movements [8]. The latter case would point to a deficit at a peripheral level, such as activating neuromuscular instructions or executing writing movements, which are not tied to alphabetical writing. Furthermore, we speculated that writing disorders in dementia are multi-componential and less modular than after focal lesions and expected lower automation in the patients with overt writing deficits. We assessed kinematographic parameters of AD patients’ writing separately for those patients who have higher-order writing deficits (omission or substitutions of letters, ill-formed graphemes) and for those who have no overt handwriting deficit. We assumed that writing competence and kinematographic measures relate to the global cognitive status of the person [19]. As education is a proxy of cognitive reserve [20], we hypothesized that highly educated patients show less writing difficulties. Finally, we were also interested as to whether writing deficits and kinematographic measures have a diagnostic value and help identifying patients affected by AD. We assessed the diagnostic value of central writing parameters and of kinematographic measures.
MATERIAL AND METHODS
Participants
Participants were outpatients consecutively recruited from our Memory Clinic where they were referred to because of memory problems. Patients were assessed using standard radiological, neurological, and neuropsychological test procedures. The diagnosis of AD was based on history taking, psychometric testing, neuroimaging (cerebral MRI and FDG-PET), and follow-up investigations. All fulfilled the criteria specified for probable AD [20]. The following exclusion criteria were applied: An additional diagnosis which could possibly influence handwriting (e.g., essential tremor), a history of neurological illness (stroke, head trauma, tumor, etc., other than probable AD), a history of psychiatric illness (e.g., major depression), a history of drug abuse (e.g., alcohol), severe other diseases which may compromise cognition or handwriting, developmental learning disorders, and formal education less than 8 years.
Twenty-three right-handed patients were included in the study. All patients (13 women) performed the Mini-Mental State Examination (MMSE) at the same day as the assessment of handwriting. All patients and all controls performed at least once a standard neuropsychological assessment (CERAD plus battery [21]) as part of the clinical diagnostic procedure. The patients’ mean age was 78.09 years (SD = 7.12; range 67–89), mean educational level 11.17 years (SD = 2.53; range 7–17), and mean MMSE [22] score at the day of writing assessment 21.39 (SD = 3.61; range 14–27).
A group of healthy controls (n = 34; 18 women; 33 right-hander, 1 ambidexter) matched in age and education also performed the writing assessment and the neuropsychological tasks. Controls were recruited among acquaintances, neighbors, and relatives of collaborators. Controls performed an extensive neuropsychological assessment (including the CERAD plus battery) in order to exclude cognitive decline and incipient dementia. Controls had an interview regarding their medical history. People with neurological diseases, psychiatric diseases (present or past), or diseases which may compromise cognition and handwriting were excluded from the study. The controls’ mean age was 75.56 years (SD = 5.85; range 67 to 90), mean educational level was 11.68 (SD = 2.64; range 8 –18), and mean MMSE score 29.06 (SD = 0.78; range 27–30). While age (p = 0.16; Mann-Whitney) and education (p = 0.53) did not differ between patients and controls, the MMSE score was significantly lower in the patient group (p < 0.0001).
Ethic approval
This study was approved by the local ethic committee (approval number: 1262/2019).
Assessment of handwriting
The basic elements of the handwriting trajectory are up and down strokes, which show smooth and single peaked velocity profiles approaching a bell-shape [23]. The analysis of the up and down strokes allows an estimation of the degree of movement automation [24] by assessing velocity and acceleration. Handwriting movements were recorded with a commercially available digitizing tablet (Wacom Intuos Pro; Wacom Europe GmbH, Krefeld, Germany) with a cable-free writing stylus. The tablet records the position of the tip of the writing stylus. The positional data and the force exerted on the tablet (indicating the writing pressure) were transmitted to a personal computer and were subsequently analyzed with the CSWin program (CSWin 2007; MedCom, Munich, Germany). This program has been specifically designed to analyze handwriting and has been applied in numerous studies on handwriting [24]. Sampling frequency of CSWin is 200 Hz, accuracy is 0.05 mm in both x and y directions. The program registers time and spatial positions and calculates velocity and acceleration curves using nonparametric kernel estimation (CSWin manual; 2016). In the present investigation, we used the standard filter as implemented in CSWin (see page 28; CSWin manual, 2016; [24]). CSWin segments the written trace in subsequent up and down strokes. The analysis of handwriting is based on the vertical axis of the writing movement. In the present investigation, we analyzed the following kinematic handwriting parameters as specified in CSWin output: Average number of inversions per stroke (NIV; the average number of peaks in the velocity profile in a single up or down stroke), percentage of automated segments (percentage of segments showing NIV = 1); frequency (average number of strokes per second), pressure (N; force that was exerted onto the tablet by the tip of the writing stylus), and writing velocity (mm/s) on paper. Velocity, frequency, and pressure are relatively stable parameters across different cognitively demanding tasks and NIV is considered a hallmark of automation. Note that the program calculates more parameters which are not analyzed here.
Procedure and tasks
Participants were asked to write on the digitizing tablet in an area (size A4) marked with red dots. They could not see their own handwriting and thus had no visual feedback. Writing without visual cues proved to be particularly difficult for AD patients [18]. They were instructed to use their normal handwriting and were asked to complete 11 tasks of handwriting as implemented in the standard assessment of CSWin. In order to avoid misunderstandings, we used both verbal and written instructions and showed examples in the handwriting of the examiner before each task. After the instruction of the respective task, the example was removed. The time of recording was limited and varied between 3 s and 20 s in the different tasks (as specified in the CSWin program). The standard assessment includes: 1) Writing a sentence to dictation (‘Die Wellen schlagen hoch’ (The waves are surging high); 20 s); 2) Writing cursive “l” (repeating two connected “l”; 10 s); 3) Repeating up-and-down strokes with hand movements (3 s); 4) Repeating faster up-and-down strokes with hand movements (3 s); 5) Repeating up-and-down strokes with finger movements and stable hand position (3 s); 6) Repeating faster up-and-down strokes with finger movements and stable hand position(3 s); 7) Drawing loops, diameter approximately 5 cm (3 s); 8) Faster drawing of loops (3 s); 9) Drawing loops with less pressure (3 s); 10) Drawing loops with eyes closed (3 s); 11) Scribbling (3 s). For the present investigation, we analyzed tasks 1, 2, 4, 6, and 8. We chose these tasks, as we considered their results as more reliable than others (task 4, 6, 8, are repetitions of tasks 3, 5, and 7, respectively). We did not analyze tasks 9 to 11, as some participants had difficulties to follow the instructions.
Analysis
In a first step, we analyzed the correctness of alphabetical writing. We scored one point each when the following conditions were fulfilled: 1) The sentence was complete; 2) No errors in capitalization; 3) No omissions of letters within a word; 4) No substitutions of letters; 5) No errors in the form of the graphemes; 6) 16 or more correct letters in the given time interval of 20 s. The maximum writing score was 6. Participants who reached a writing score of 4 or lower were classified as agraphic. Since we did not classify different forms of central agraphias, this is only a raw classification. Individually used forms of graphemes in the participant’s fluent handwriting were not counted as errors.
In order to simplify the analysis and test our hypotheses, we calculated average scores for alphabetical and non-alphabetical writing tasks. For the alphabetical writing scores, we calculated the average between parameters (NIV, frequency, pressure, percentage automated, velocity) of task 1 and task 2. For the non-alphabetical writing scores, we calculated the average between parameters of task 4, 6, and 8. Two patients had no data in task 8 (failure in recording); for these patients we computed the average of task 4 and 6.
Since part of the data in the patient group was not normally distributed (tested with Shapiro-Wilks-tests), we used nonparametric tests. We compared the performance in alphabetical and non-alphabetical tasks between AD patients and the control group by Mann-Whitney tests. Subsequently, we split the AD group into two subgroups (agraphic and non-agraphic patients, see above). We then compared the three groups (controls and two AD groups) by Kruskal-Wallis tests, followed by Mann-Whitney tests when the results of Kruskal-Wallis tests were significant.
A correlation analysis with Spearman’s rho was performed between writing score and measures of automation in the AD patient group. Note that we did not include measures of velocity here. The writing score could be influenced by writing velocity (e.g., the number of omissions could be higher with slow writing velocity). In order to avoid confounds, we limited the analysis to the grade of automation. A further correlation analysis was performed between the writing assessment and MMSE score, age, and years of education score in the patient group. We also performed a ROC analysis to assess whether writing assessment and kinematic parameters have a diagnostic value.
RESULTS
Errors in writing a sentence
Out of 23 patients with AD, 12 did not complete the sentence, 5 had errors in capitalization, 6 had omissions of graphemes within a word, 3 selected wrong graphemes, 13 showed errors in the form of the graphemes, and 12 did not produce 16 or more correct letters in the given time interval (maximum possible 21 letters). When one point was attributed for the absence of each of the six above mentioned criteria, patients reached a median writing score of 4 (interquartile range 2–5), controls reached a median writing score of 6 (interquartile range 6-6; Mann Whitney test p < 0.0001). Fourteen patients (61%) were classified as agraphic (score 4 or lower; MMSE median 20, interquartile range 19–23), 9 patients (39%) as non-agraphic (score 5 or 6; MMSE median 23, interquartile range 19–25). Note that neither age (p = 0.36), education (p = 0.82, Kruskal-Wallis tests), nor gender (p = 0.96; Chi square test) differed between agraphic AD patients, non-agraphic AD patients, and controls. Kruskal-Wallis tests showed significant differences between the three groups in the MMSE (p < 0.0001). Mann-Whitney tests yielded significant differences between both patient groups and controls (p < 0.0001, respectively), but not between the two patient groups (p = 0.439).
Kinematographic analysis
The kinematographic analysis yielded significant differences between AD patients and controls in most parameters (see Table 1). Patients showed less automated movements (as measured by NIV) in non-alphabetical as well as in alphabetical writing. The writing velocity was lower, as was the frequency of up-and-down strokes in alphabetical and non-alphabetical writing. The writing pressure did not differ between AD group and controls. The comparison between agraphic patients, non-agraphic patients and controls (Table 2) evidenced significant differences between groups (Kruskal –Wallis tests). Subsequent Mann-Whitney tests showed differences between agraphic patients and controls in measures indicating automation (NIV; percentage of segments with NIV = 1), and measures indicating velocity (writing velocity on paper; frequency of up-and-down strokes), but not in writing pressure. The differences were found in alphabetical and non-alphabetical tasks. The comparison between the two AD groups (agraphic and non-agraphic) and between non-agraphic patients and controls yielded no significant differences.
Kinematographic analysis: AD patients and controls
Median scores and quartile ranges for the kinematographic analysis. Mann-Whitney tests showed significant differences between AD patients and controls. Significant p values (< 0.05) are marked with*.
Kinematographic analysis –AD patients (with/without agraphia) and controls
Median scores and quartile range for the kinematographic analysis. P values for Kruskal Wallis tests. Subsequent Mann-Whitney tests showed differences between agraphic patients and controls. * indicates significant group differences between AD patients with agraphia and controls. For exact values, see Supplementary Table 1. Other comparisons yielded no significant results. For the Mann-Whitney tests, we consider p values lower than 0.0167 as significant (correction due to multiple comparisons).
Spearman correlation analysis between writing score and NIV in the patient group revealed a significant negative correlation between writing score and NIV in alphabetical writing (p = 0.044). That means, patients who had less errors in writing a sentence showed a higher automaticity in handwriting alphabetical material. The correlation between writing score and NIV in non-alphabetical material did not reach significance (p = 0.07).
Spearman correlation analysis between MMSE score and kinematographic measures in the patient group yielded no significant correlations. Neither years of education nor age did correlate with any writing measure in the patient group.
A ROC analysis was performed to assess the diagnostic value of writing scores and of kinematographic parameters (distinguishing between AD patients and controls). The overall writing score (see above for scoring) reached the highest value and had an AUC of 0.845. Other scores related to specific writing errors did not pass the 0.80 threshold which indicates excellent discrimination [27]. In non-alphabetical writing, velocity of writing on paper had an AUC of 0.811, the average NIV had an AUC of 0.804 and the percentage of automated segments had an AUC of 0.808. Other parameters reached lower values.
DISCUSSION
The present study assessed handwriting in a group of patients with mild AD. A total of 61%of the AD patients showed writing difficulties, such as omissions, substitutions, or ill-formed graphemes. The kinematographic analysis yielded significant differences between the AD group and controls in most parameters. In non-alphabetical as well as in alphabetical writing, patients showed less automated movements, a lower writing velocity on paper and a lower frequency of up-and-down strokes. A subgroup analysis showed significant differences between agraphic patients and controls in all measures except writing pressure.
Our results are in line with a previous study where AD patients and controls had to draw superimposed circles [8]. Movements of AD patients were significantly less automated, accurate, and regular. Our results differ from those reported previously in a study [16] using rather complex copying and writing tasks. Diverging from our results, writing velocity did not differ in this study between AD patients and controls [16]. Some studies reported that patients applied significantly lower pressure while writing than did healthy participants [16], while others reported significantly higher writing pressure in AD patients. Here, pressure did not differ between groups. Furthermore, in-air time consistently differentiated among controls, AD, and MCI in a previous study [16]. This measure was not analyzed here as we used simple non-demanding tasks. Tasks in the different investigations put differential demands on patients’ cognition and in-air preparation time.
One may question whether difficulties in finding the correct form of the graphemes (selecting graphemes, activating grapheme patterns) cause the slower and less coordinated execution of writing movements. We think that this hypothesis is highly implausible. A general slowing and loss of automation was observed not only for alphabetical material, but also non-alphabetical material. The slow and poorly automated execution of motor programs was not bound to alphabetical material.
A significant correlation was found between writing score and NIV in writing a sentence in the patient group. This points to the fact that writing disorders are multi-componential and not strictly limited to one processing level in AD. Note that we did not include measures of velocity in this analysis. Since omissions of letters (included in the writing score) could be related to reduced velocity, we limited our analysis to NIV as an important measure indicating automation. Our results indicate that writing disorders in AD are not modular and do not concern single processing steps—instead we assume that agraphia concerns not only central levels of processing but also the motor execution.
Spearman analysis between MMSE score and kinematographic measures in the patient group yielded no significant correlations. This result differs from previous studies [8] where significant correlations were found between cognitive level and kinematic handwriting parameters, indicating poorer motor coordination in cognitively impaired patients. The fact that we found no significant correlation between MMSE and kinematographic measures may be due to the specific sample assessed or possibly to the small sample size. Interestingly, years of education did not correlate with any writing measure in the patient group. Thus, cognitive reserve [25] as measured by the extent of formal education [26] does not seem to support better and more fluent writing performance in AD.
Different patterns of brain atrophy have been described in AD, suggesting that different anatomical patterns of degeneration might exist within the same disorder [27, 28]. Poulakis et al. [28] suggest there is substantial heterogeneity in AD that has an impact on how patients function and progress over time. We hypothesize that different brain atrophy patterns within our AD group may be responsible for impaired writing in some but not all cases. Overall, the analysis of anatomical lesion sites in patients with peripheral agraphia discloses a widespread pattern. Lesions due to stroke, tumors, or degenerative syndromes often concern the left posterior region, in particular the left parietal lobe [13, 30] and the left frontal lobe [31, 32] including the left premotor cortex [33]. We speculate that patients with peripheral agraphia showed more pronounced left parietal or left frontal atrophy. However, we cannot test this hypothesis as patients in the present investigation had their MR scans on different scanners thus preventing a statistical analysis of brain atrophy patterns. It has also been shown that a subgroup of pure AD patients develop parkinsonian symptoms as a result of neuronal loss in the basal ganglia, indicating a prominent subcortical involvement [34]. These symptoms include rigidity and subclinical tremor which may remain unnoticed in the clinical examination. These symptoms may well influence kinematographoic measures in alphabetical and non-alphabetical writing.
The lack of anatomical data is a limitation of the present investigation, which may be overcome in a prospective future study. A further limitation is the relatively small number of patients in the present study.
It has been hypothesized [8, 17] that the handwriting and drawing performance may be early indicators of brain dysfunction. Thus, assessment of fine motor function during handwriting and drawing using a digitizing tablet and a pressure-sensitive pen could be a useful resource in the clinical setting [17]. In our investigation, different writing scores reached an excellent area under curve (AUC) in the ROC analysis. Both, the correctness of writing (no omissions, substitutions, or grapheme errors) and the motor performance proved to have high sensitivity and specificity. The best values were found for the overall writing score and for the automation and regularity in the motor performance in non-alphabetical writing (velocity and average NIV). As our results suggest, detailed analysis of very simple handwriting tasks (such as drawing loops and performing up- and down strokes) may offer insights into processes of motor regulation. Using such simple tasks allows the assessment of motor performance without the effects of cognitive impairment in more demanding tasks. Possibly, fine motor alterations are detectable earlier in the course of the disease when other disorders are not yet present. However, this hypothesis remains to be tested. Finally, it should be noted that a kinematographic analysis is easily applicable in clinical practice and that the assessment of parameters related to motor performance may be a useful addition to neuropsychological assessment and neurological diagnostic procedures.
DISCLOSURE STATEMENT
Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/21-0279r1).
