Internally vs. externally triggered movements in patients with major depression

Abstract

Background:

Psychomotor retardation is a prominent clinical feature of major depression. While several studies investigated these deficits, differences between internally and externally triggered response selection and initiation are less well understood. In the current study, we delineate internally vs. externally driven response selection and initiation in depression and their relation to basic psychomotor functioning.

Methods:

20 inpatients diagnosed with a (unipolar) major depression and 20 closely matched healthy controls performed a computerized motor paradigm assessing differences between internally and externally cued movements. Psychomotor performance and basic memory functions were assessed using a neuropsychological test-battery. To examine within group homogeneity a multivariate clustering approach was applied.

Results:

Patients featured a global slowing of internally and externally cued response selection compared to controls, as well as impairments in basic psychomotor functioning. Yet, basic motor speed was preserved. Furthermore, patients were more severely impaired when movements involved internal response selection. The data-driven clustering revealed two patient subgroups, which both showed psychomotor disturbances, while only one featured slowing of response selection.

Interpretation:

The results suggest a differential rather than a global psychomotor slowing in major depression with specific impairments of visuospatial and attentional processing as cognitive aspects of psychomotor functioning. As found for depression, in Parkinson’s disease internally cued movements are more severely affected than externally cued reactions. Both may therefore be caused by dopaminergic deregulation due to frontostriatal deficits. Finally, multivariate clustering of behavioral data may be a promising future approach to identify subtypes of psychomotor or cognitive disturbances in different patient populations.

1. Introduction

Psychomotor disturbances (PMD) are considered a common symptom of major depression that is often already evident in the first clinical contact with a patient. Together with frequently observed cognitive deficits, psychomotor retardation form the core of non-affective symptoms of depression [1,2]. These non-affective symptoms may severely impact on patients’ psychosocial functioning [3–5] but also pose challenges to treatment regimens that requite active participation of the patient, in particular psychotherapeutic intervention. Hammar and Ardal [4] reviewed the literature on cognitive functioning in major depressive disorder over the past decade and concluded that depression is not only associated with cognitive impairments in the acute phase of illness. Rather these cognitive, in particular attentional and mnemonic deficits [6–8] may persist despite recovery from affective symptoms [9,10]. In case of psychomotor symptoms, however, the relation to depression severity remains in question [11]. In a recent meta-analysis McDermott and Ebmeier [12] reported significant correlations of depression severity with disturbances not only of episodic memory and executive functions but also processing speed. They furthermore examined the influence of psychomotor retardation on cognitive deficiencies by comparing psychomotor speed related (timed) and unrelated (untimed) subtests of cognitive functions. From this comparison, they concluded that psychomotor retardation is not responsible for depression severity related cognitive deficits.

In a pioneering review on psychomotor symptoms in depression, Sobin and Sackeim [1] emphasized the significance of motor disturbances as a core symptom of this disorder and concluded that these are a strong marker for the melancholic subtype. Based on this work, Schrijvers et al. [13] reviewed subsequent studies of PMD and distinguished three subdomains of psychomotor functions: speech, gross and fine motor activity.

The construct ‘psychomotor’ necessarily encompasses cognitive and motor aspects of motor control [i.e., planning, programming and execution (cf. [13])]. To investigate which aspects are specifically affected in major depression several studies employed computerized drawing tasks manipulating the cognitive effort necessary for task completion. The cognitive load is rather small in simple line drawing [14–17] and increases with complexity of the template in figure copying [15,16,18–20]. In this approach, movement duration represented the motor aspect while the time to initiate drawing was considered the more cognitive aspect of fine motor activity. Overall, these studies demonstrated that both the motor and the cognitive aspect of psychomotor functioning are slowed in medicated as well as non-medicated depressed patients, regardless of age. There is also good evidence for a depression related slowing in simple choice reaction tasks, which was repeatedly demonstrated using visual cued response selection [21–23].

Some of these features of psychomotor slowing in depression resemble bradyphrenia in Parkinson’s disease [24–26]. A main characteristic of parkinsonian bradykinesia is the difficulty to initiate movements in absence of an external cue rather than a general motor slowing [26,27]. This prompted Rogers [28] to propose the hypothesis of dopaminergic deregulation due to frontostriatal deficits as a general mechanism for psychomotor changes in psychiatric and neurologic diseases such as (melancholic) depression, schizophrenia, dementia or Parkinson’s disease [29]. Major depression was repeatedly associated with disturbances of the dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex (ACC) and the basal ganglia (BG) [28,30,31]. Supporting Roger’s hypothesis, the investigation of internally specified and externally cued movements again resemble prefrontal hypoactivation in Parkinson’s disease [32] and catatonic schizophrenia [33]. Based on these considerations, the aim of the current study was now to delineate PMD for internally vs. externally driven response selection and initiation as aspect of (fine) motor control in major depression. Furthermore, cognitive and motor aspects of PMD should be largely independent of current disease severity [12]. To test these hypotheses, subjects completed a computerized response selection paradigm and further assessment of basic psychomotor functioning in the fine motor domain.

2. Materials and methods

Twenty patients with unipolar depression [diagnosis according to the ICD-10: F32.X, F33.X [34]] as well as twenty age-, sex-, handedness and (own and parental) education matched healthy controls were enrolled in our study. Depressive symptom severity was diagnosed with the ICD-10 and the German version of the revised Beck Depression Inventory-2 [BDI-2 [35]]. All examined patients were inpatients in the RWTH Aachen University Hospital at the time of testing and diagnosed by the attending doctors with major depression (F32) or an acute episode of a recurrent depressive disorder (F33) according to ICD-10 [34]. Any comorbid psychiatric or neurological diseases including alcohol or drug abuse was an exclusion criterion. It was assessed by the attending psychiatrist (in patients) and additionally by a structured clinical interview using the German version of the SCID inventory [36] in patients and also in controls to exclude a history of psychiatric and neurologic diseases. The minimum age for inclusion was 18 years. As we did not introduce an age limit in order to also include older patients, we screened for dementia using the Mini Mental State Examination [MMSE (Folstein [37]] with a cut-off at 23. Benzodiazepine medication was an exclusion criteria, because it is well known to induce cognitive and psychomotor impairments [38]. Socio-economic status was assessed using a structured interview. A comprehensive description of the patient group including antidepressant medication and dose is given in Table 1. All but one patient were right-handed according to the Edinburgh handedness inventory [39]. All subjects including healthy controls gave informed written consent to the study protocol, which had been approved by the local ethic committee of the RWTH Aachen University Hospital.

Table 1.

For all 20 patients age, gender, education (in years at school and university), parental education, diagnose according to ICD-10, self-reported symptom severity according to the revised Beck Depression Inventory (BDI-2), type and dosage of medication at the time of examination and subgroup cluster are listed (SNRI: serotonin-norepinephrine reuptake inhibitor; SSRI: selective serotonin reuptake inhibitor; NDRI: norepinephrine–dopamine reuptake inhibitor; MAOI: monoamine oxidase inhibitor; TeCA: Tetracyclic antidepressant; TCA: Tricyclic antidepressant). Diagnoses of F32.x indicate a first major depressive episode, those of F33.x a major depressive episode of a recurrent depressive illness. The last digit codes the (clinical) severity with F3x.0 indicating a mild, F3x.1 a moderate, F3x.2 a severe episode without and F3x.3 a severe episode with psychotic symptoms. Cluster refers to the groups the patient was assigned to in the subsequent cluster analysis (cf. Fig. 1).

Age	Gender	Education (a)	Parental education	Diagnose (ICD-10)	BDI-2	Medication (mg/d)				Patient cluster
						SNRI/*SSRI	TeCA/*TCA	Atypical-/*antipsychotics	Anticonvulsants/*Hypnotics
18	Male	11	16	F32.0	25	Nil				2
19	Female	12	9	F33.2	23	[NDRI: Bupropion 300]				1
26	Female	10	8	F32.1	48	Duloxetine 60	Mirtazapine 7.5			1
32	Female	12	18	F33.1	37	Venlafaxine 300	*Trimipramine 50	Quetiapine 800, Risperidone 4*Pipamperone 40		2
33	Female	10	9	F32.1	23	Venlafaxine 225		Quetiapine 200		1
36	Female	12	8,5	F32.1	43	*Citalopram 20	Mirtazapine 15			1
40	Female	9	9	F32.0	16	Nil				1
42	Female	17	13	F32.2	2	*Citalopram 40				1
46	Female	10	–	F33.2	37	Duloxetine 120			Pregabalin 300, *Zopiclone 7,5	1
47	Male	10	7.5	F33.1	29	Duloxetine 60	Mirtazapine 45		Pregabalin 225	2
48	Female	18	9	F33.1	31	Duloxetine 120	Mirtazapine 15			2
50	Male	9	8	F32.1	31	Duloxetine 60			Pregabalin 900	1
52	Male	9	8	F32.3	34	Venlafaxine 150		Risperidone 1		1
52	Female	10	10	F33.1	25	Duloxetine 120			Pregabalin 50	2
53	Male	11	-	F33.2	18	Duloxetine 60			Pregabalin 150, *Zolpidem 5	1
55	Male	12	9	F32.1	6	Venlafaxine 150				1
56	Female	11	11	F32.1	12	Venlafaxine 225				2
67	Male	8	8	F33.1	27	[MAOI: Tranylcypromine 70]	Quetiapine 300	Lamotrigine 150, *Zopiclone 7,5	1
69	Female	8	8	F33.1	12	Duloxetine 60		*Pipamperone 40		1
71	Female	8	8	F32.3	17	Venlafaxine 225	*Trimipramine 50	Risperidone 3		2

First, all participants were screened with a structured interview assessing socio-economic status and excluding comorbidity in patients as well as history of psychiatric and neurologic diseases in controls. Then, qualified subjects completed our computerized motor paradigm assessing differences between internally and externally specified reactions. A brief training ensured adequate task comprehension. The paradigm consisted of two parts lasting 16min each. Subsequently, a neuropsychological test battery was conducted, followed by the self-report questionnaire of depressive symptoms (BDI-2) and a short debriefing. During examination participants and patients in particular were allowed to pause in between single tests taking as much time as needed. Altogether data acquisition took between 1.5 and 2 h per subject.

2.1. Motor paradigm

The experimental motor task consisted of unilateral button presses performed with the right or left index finger assessing differences between internally and externally specified reactions in three different conditions: (1) a Free choice of button presses with the left or right hand at a self-chosen point in time, (2) a Timed choice task, when the time of movement was cued by a visual stimulus but the hand to be moved was chosen by the subject, or (3) a Reactive task when laterality and time of movement were both cued by a visual stimulus.

2.1.1. Free choice: self-timed movement selection (free timing/choice of hand)

In the ‘Free’-condition the movements were entirely self-initiated. The subjects were instructed to press one of the two buttons at any self-chosen time. Every response was immediately followed by a 3.5 s visual feedback consisting of an arrow pointing to the side of the button-press. During the feedback no further responses were allowed to prevent sequential finger tapping. When training the subjects, they were explicitly instructed to vary the inter-stimulus intervals as well as the hand used in order to prevent rhythmic responses or any kind of movement routine. The time intervals between single responses were recorded on-line and subsequently used as inter-stimulus intervals (ISI) for the visual cued responses in the other two conditions. Likewise, the frequency of right and left button-presses was fed back as visual cues triggering a lateralized response in the ‘Reactive’-condition.

2.1.2. Timed choice: spatial choice at a cued time-point (external timing/choice of hand)

In the ‘Choice’-condition, stimuli consisted of arrows pointing to both sides presented for 3.5 s. The task was to respond as fast as possible by pressing either the left or right button. Subjects were free in choosing the side of response, but should vary between left and right sided responses. The ISIs and thus the number of button presses from the preceding ‘Free’-condition were presented in a random sequence to assure comparability of motor responses timing between conditions.

2.1.3. Reactive: reaction to a lateralized stimulus (external timing/cued hand)

In contrast to the ‘Choice’-condition, responses in the ‘Reactive’-condition were fully predetermined by the visual cue. Subjects had to react as fast as possible to a single-headed arrow pointing to the left or right by pressing the corresponding button. Like in the ‘Choice’-condition, ISIs and lateralization of responses were matched to the preceding ‘Free’-condition.

In summary, each ISI generated by a subject in the ‘Free’-condition was subsequently used to trigger one response both in the subsequent ‘Choice’- and ‘Reactive’-condition. The experiment consisted of 2 × 12 blocks of conditions, and lasted approximately 33 min. By randomizing ISIs in the ‘Choice’-condition and ISI and number of left and right responses (independently) in the ‘Reactive’-condition, anticipation confounds with respect to cue sequences were avoided, while comparability across conditions was preserved. For each condition 8 blocks of 60 s duration were presented in sequences of either 1 (‘Free’)–2 (‘Choice’)–3 (‘Reactive’) or 1–3–2 in the same pseudo-randomized order. The entire experiment lasted approximately 30 min including 5 s breaks between the blocks.

2.2. Psychomotor and cognitive assessment

All subjects completed a test battery assessing basic psychomotor and cognitive functioning.

2.2.1. Finger tapping

Subjects were asked to perform tapping movements as rapidly as possible for 10 s using the left or right index finger. Median number of taps from 3 trails per hand (separated by short breaks to prevent muscular fatigue) was used as the test score assessing basic motor speed.

2.2.2. Pointing movements

Subjects performed rapid horizontal pointing movements between two spots 30 cm apart using the right or left index finger [cf. CAPSIT Parkinson’s disease test battery [40]]. Subjects were instructed to perform the movements as quickly and accurately as possible in three trials of 10 back-and-forth pointing movements. Median number of 3 trails per hand represented basic motor coordination.

2.2.3. Trail making test

The two classical versions of the Trail Making Test [41] were used to assess attention as well as visuomotor speed (TMT-A) and cognitive flexibility (task-switching in TMT-B). The task consisted in consecutively connecting numbered circles (TMT-A) or switching between numbers and letters (TMT-B). Longer times indicate poorer performance.

2.2.4. Digit span subtest of the Wechsler Adult Intelligence Scale [42]

The verbal reproduction of an auditory presented digit span forwards and backwards were measured as markers for immediate memory (digit span forward, DS-F) and working memory performance (digit span backward, DS-B).

2.2.5. Multiple-choice vocabulary intelligence test

For an estimate of crystalline intelligence, a multiple-choice vocabulary intelligence test [MWT-B [43]] was used. The task consisted in making one actual word among four pseudo-words with increasing difficulty without time limitation.

2.3. Data analysis

All measurements were analyzed for group differences by means of a Wilcoxon–Mann–Whitney two-sample rank-sum test using MATLAB (Mathworks, Natick, MA). This non-parametric test was chosen as data (in particular the raw test-scores) did not fulfill the criteria for parametric testing, i.e., normal distribution. Due to their greater robustness against outliers, medians rather than means of test scores and measured values are reported. To test for relationships between psychomotor and cognitive performance Spearman’s rank correlation coefficients were computed over all obtained measures. The test scores were also correlated with depressive state (symptom severity) as reflected by the ICD-10 diagnosis [34] and the BDI-2 [35].

Possible influences of antidepressant medication on test results were estimated by computing correlations between test scores and on-off state of each drug. To examine homogeneity within the patient group a spectral reordering approach (Johansen-Berg et al. [44]) was applied to assess whether patients may be clustered into subgroups with distinct cognitive-motor performance. This approach involves first to compute a cross-correlation matrix of the measures of cognitive and motor performance obtained for the individual patients. The matrix is then reordered to minimize the weight of cross-correlation values off the diagonal, hereby forcing highly correlated patients close towards each other. Clusters may then be identified in the reordered matrix as sets of patients whose cognitive-motor profiles were strongly correlated with each other and weakly with the rest of the matrix. The association between patient subclusters with diagnosis (first episode, recurrent episode) or with the different types of antidepressant treatment was assessed with Fischer’s exact test.

3. Results

3.1. Clinical characteristics

Patients and healthy controls were not different in terms of age (median: patients 47.5 years vs. controls 49 years, rank-sum test: p = 0.787), handedness (lateralization quotient: 95.7 vs. 100, p = 0.733), education (10.0 years vs. 10.5 years, p = 0.466) and parental education (9.0 years vs. 9.0 years, p = 0.829). Among the 20 patients, half were diagnosed with recurrent depression, the other half with an isolated (or first) depressive episode. The majority (12) were treated for moderate depression, followed by severe (6) and mild (2) depression. Predictably, median scores in the standardized self-report scale of depressive symptoms BDI-2 were evidently higher in patients [25.0 (inter-quartile range: 16.0)] than in controls [0.0 (IQR: 1.5); p < 0.0001].

3.2. Motor paradigm

In both groups, the proportion of right and left button presses was approximately balanced in the ‘Free’- (median proportion of right choices: patients 53.3% vs. controls 54.5%) and ‘Choice’-condition (55.2% vs. 55.1%). Importantly, the proportion of left/right responses did not differ between groups in either condition or between conditions in either group (p > 0.483 for all comparisons). Error rates could only be obtained in the ‘Reactive’-condition. There was no significant difference in the proportion of incorrect or missed responses between groups indicating equally good task performance [median error rates for patients 2.0% (IQR 6.6%) and for controls 1.2% (IQR 1.9%); p = 0.407].

Our experimental paradigm demonstrated a marked psychomotor slowing across all task conditions in the patient group (Table 1). In the ‘Free’-condition, patients initiated less responses by themselves than controls (medians: patients 83.5, controls 105.5; p = 0.006) and hence were slower to initiate new button presses (p = 0.016). In the ‘Choice’-condition, the slowing in the patient group was highly significant (medians: patients 588 ms, controls 407.5 ms; p = 0.0003). Patients also showed slower response times in the ‘Reactive’-condition (medians: patients 500.5 ms, controls 450.5 ms; p = 0.048). The fact that task performance, as indicated by error rates, was clearly not different between groups indicated that this slowing did not reflect a speed-accuracy-tradeoff. A group-by-condition interaction (p = 0.001) was found as patients with depression were significantly slower to respond in internally cued than in externally cued reactive trials (median difference: −42.0 ms), whereas control subjects were faster in the same comparison (median difference: +46.5 ms).

Finally, investigating the trial-by-trial variation in reaction times in the ‘Reactive’-condition revealed that there was a higher within-subject variability in the patient group (standard deviation [SD]: 104 ms) than in healthy controls (SD: 71 ms; p = 0.006). The same was true for the ‘Choice’- (SD: patients 156.5 ms, controls 126.5 ms; p = 0.023) and the ‘Free’-condition (SD: 1765.5 ms, 909.5 ms; p = 0.008). That is, depressive patients were less stable in their performance over the course of the 30 min experiment than controls even though no indication of gradual decline in performance was found suggestive of accelerated fatigue by inspection of the individual reaction time developments.

3.3. Psychomotor and cognitive assessment

Patients showed lower scores in all tests than healthy controls (p < 0.008 for all comparisons, see Table 2) except for finger tapping, hence, basic motor speed was the only variable not reduced in major depression (p = 0.203). With increasing cognitive involvement in the psychomotor tests, depressive patients took longer to perform pointing movements (p = 0.002) and were significantly slower to complete either version of the trail making test. They also showed a significantly lower digit span for both forward and backward reproduction, indicating short-term memory deficits. In spite of the fact that patients and controls were well matched (pair-wise matching of age and education ± 2 years) and did hence not differ with respect to age, own and parental education, the MWT-B as an estimate of crystalline intelligence was lower in the patient group (25.0 vs. 31.0 correct answers out of 37 items; p < 0.001). This apparent difference in intelligence can partially be explained by attention and especially memory deficiencies in major depression, in particular since only in patients we found a significant correlation of the MWT-B with immediate memory performance (r = 0.56; p = 0.016) assessed with DS-F.

Table 2.

Results obtained from the test of psychomotor and cognitive functions (RT: response times; SD: standard deviation; ms: milliseconds). For each cell, median (across the diagnostic group) and interquartil-range (IQR) are provided.

Psychomotor functions	Depressed patients	Healthy controls	p-Value
‘Reactive’-condition
Median Response Times (ms)	500.5 (IQR: 183.5)	450.5 (IQR: 82.0)	0.048*
SD across trials (ms)	104.0 (IQR: 44.0)	71.0 (IQR: 36.5)	0.006*
Errors (%)	2.0 (IQR: 6.6)	1.2 (IQR: 1.9)	0.407
‘Choice’-condition
Median Response Times (ms)	588.0 (IQR: 176.5)	407.5 (IQR: 92.5)	0.000**
SD across trials (ms)	156.5 (IQR: 45.0)	126.5 (IQR: 37.0)	0.023*
‘Free’-condition
Median Number of Responses	83.5 (IQR: 27.3)	105.5 (IQR: 18.0)	0.006*
Median Response Times (ms)	2053.5 (IQR: 1439.0)	1282.0 (IQR: 907.0)	0.016*
SD across trials (ms)	1765.5 (IQR: 1287.5)	909.5 (IQR: 916.0)	0.008*
‘Reactive’ – ‘Choice’ in RT (ms)	−42.0 (IQR: 110.5)	46.5 (IQR: 54.0)	0.001**
Tapping (Basic motor speed)	46.5 (IQR: 7.5)	47.7 (IQR: 11.0)	0.203
Pointing (Motor speed/Coordination)	9.0 (IQR: 3.8)	6.0 (IQR: 2.3)	0.002**
TMT-A (Visuomotor speed, Attention)	27.2 (IQR: 16.8)	17.5 (IQR: 6.3)	0.000**
TMT-B (Cognitive flexibility, Attention)	60.0 (IQR: 50.3)	38.0 (IQR: 16.0)	0.001**
Cognitive functions
DS-F (Immediate memory)	6.0 (IQR: 1.8)	7.0 (IQR: 2.5)	0.006*
DS-B (Working memory)	5.0 (IQR: 2.0)	7.0 (IQR: 3.0)	0.007*
MWT-B (Crystalline intelligence)	25.0 (IQR: 5.8)	31.0 (IQR: 6.0)	0.000**

TMT-A/B: trail making test version A/B; MWT-B: multiple-choice vocabulary test; DS-F: digit span forwards; DS-B: digit span backwards.

^*Significance was assessed using a Wilcoxon–Mann–Whitney two-sample rank-sum test p < 0.05.

^**Significance was assessed using a Wilcoxon–Mann–Whitney two-sample rank-sum test p < 0.005.

3.4. Correlation with symptom severity and diagnosis

Among performance in the motor paradigm, only variation in reaction times in the ‘Choice’-condition showed a significant correlation with disease severity according to ICD-10 (r = 0.46; p = 0.039). That is, patients with worse clinical state performed less stable in selecting a reaction. Furthermore, reduced performance in the DS-B (testing working memory) was correlated with the severity of self-reported depressive symptoms as measured by the BDI-2 (r = −0.53; p = 0.019). The only significant correlation with the two diagnoses (first episode, recurrent episode) in our sample was found also for DS-B (r = −0.60; p = 0.007). This indicates that patient who suffered from a recurrent depressive episode and those who reported increased symptom severity had more severe working memory deficits. In that context, it is important to note, that there was no significant difference in self-report of symptom severity (BDI-2) between patients with an acute depressive episode and with a recurrent depressive disorder (rank-sum test; p = 0.493).

3.5. Correlation between tests

When correcting for multiple comparisons across the number of performed analyses, only the positive correlation between the time taken to complete the two versions of the trail making test (TMT-A, TMT-B; r = 0.72; p < 0.001) remained significant.

At an uncorrected level of p < 0.05, positive correlations between response times in the ‘Reactive’-condition and those in the ‘Choice’- (r = 0.64; p = 0.003) and ‘Free’-condition (r = +49; p = 0.031) was observed along with a trend towards a correlation between the latter variables (r = +42; p = 0.067). A negative correlation between finger tapping performance and response times in the ‘Reactive’-condition (r = −.49; p = 0.027), indicated that patients with faster basic motor speed featured lower reaction times and were hence faster to respond. To follow up this potential confound, a post hoc test was performed to control for the influence of basic motor speed on group differences in the experimental task. More specifically, finger tapping performance was introduced as a covariate in the analysis of group differences with respect to errors and response times in all three conditions. These subsidiary analyses replicated all results obtained above, suggesting that inter-individual differences in basic motor speed did not influence the group comparison.

3.6. Identifying clusters of patients

While the results above were based on correlating psychomotor and cognitive measurements over patients, the final exploratory analysis was based on the correlation between patients over all obtained scores. Spectral reordering of the computed cross-correlation matrix revealed a subdivision of the patient group into 2 distinct clusters, i.e., subgroups, consisting of 13 and 7 patients, respectively (Fig. 1).

Fig. 1.

(Original) Cross-correlation matrix obtained by the hierarchical cluster analysis (color coding of correlation between patients) linking patients on base of their overall similarity in psychomotor and neuropsychology measurements. (Recordered) Spectral reordering of the cross-correlation matrix [57] yields two clusters of higher inter-correlation between the first 13 patients in Cluster 1 and the next 7 in Cluster 2.

While performance of patients within each subgroup was closely correlated, correlation with the respective other group was low. Notably, these two subgroups did neither differ with respect to age (p = 0.832), own or parental education (p = 0.801; p = 0.774) nor depression severity (ICD-10 mean: cluster 1.38 = 1, group 2 = 1.14; p = 0.393; BDI-2: 23.0, 25.0; p = 0.893) or number of antidepressant agents (p = 0.620). The clustering seemed thus not a mere reflection of age, education, clinical status or quantity of medication. There also was no association between subgroup and diagnosis (first episode, recurrent episode) according to Fischer’s exact test (p = 0.642). When the deficits of each group were delineated by comparison to the healthy controls, both groups featured significantly lower scores in almost all neuropsychological tests (p < 0.039 for all comparisons but for DS-F p = 0.06 in the smaller group, see Table 2), while error rate and basic motor speed were not different from the controls in either group. However, only the larger cluster showed pronounced PMD evident by slower response times in the ‘Choice’-condition (p < 0.001) as well as higher inter-subject variability in the ‘Choice’-(p = 0.004) and the ‘Reactive’-condition of the motor paradigm (p = 0.001). Therefore, patient of this subcluster were significantly slowed in the internal selection of a reaction relative to purely reactive trials (median difference: −64.0 ms) (Table 3).

Table 3.

Results obtained from the comparison of psychomotor and cognitive performance of healthy controls with each of the two patient subgroups derived from the clustering approach (cf. Fig. 1; RT: response times; SD: standard deviation; ms: milliseconds). For each group, median (across the diagnostic group) and interquartil-range (IQR) are provided.

Psychomotor	Healthy controls	Patient cluster 1	p-Value	Patient cluster 2	p-Value
‘Reactive’
Median RT (ms)	450.5 (IQR: 82.0)	487.0 (IQR: 197.8)	0.135	514.0 (IQR: 168.0)	0.072
SD (ms)	71.0 (IQR: 36.5)	105.0 (IQR: 30.0)	0.004**	85.0 (IQR: 49.3)	0.159
Errors (%)	1.2 (IQR: 1.9)	2.8 (IQR: 9.8)	0.283	1.4 (IQR: 3.7)	0.945
‘Choice’
Median RT (ms)	407.5 (IQR: 92.5)	624.0 (IQR: 132.8)	0.000**	507.0 (IQR: 178.8)	0.114
SD (ms)	126.5 (IQR: 37.0)	169.0 (IQR: 24.8)	0.001**	123.0 (IQR: 39.3)	1.000
‘Free’
Median Responses	105.5 (IQR: 18.0)	82.0 (IQR: 18.0)	0.008*	92.0 (IQR: 44.5)	0.346
Median RT (ms)	1282.0 (IQR: 907.0)	1809.0 (IQR: 1234.8)	0.022*	2298.0 (IQR: 2073.8)	0.134
SD (ms)	909.5 (IQR: 916.0)	1627.0 (IQR: 1212.5)	0.016*	1905.0 (IQR: 1760.0)	0.063
‘Reactive’ – ‘Choice’	46.5 (IQR: 54.0)	−64.0 (IQR: 115.3)	0.000**	38.0 (IQR: 41.8)	0.695
Tapping	47.7 (IQR: 11.0)	46.5 (IQR: 6.9)	0.294	46.5 (IQR: 8.1)	0.321
Pointing	6.0 (IQR: 2.3)	7.3 (IQR: 3.8)	0.039*	10.1 (IQR: 2.1)	0.000**
TMT-A	17.5 (IQR: 6.3)	27.3 (IQR: 12.8)	0.000**	25.8 (IQR: 20.6)	0.002**
TMT-B	38.0 (IQR: 16.0)	59.0 (IQR: 51.8)	0.002**	76.0 (IQR: 54.9)	0.011*
Cognitive
DS-F	7.0 (IQR: 2.5)	6.0 (IQR: 2.0)	0.009*	6.0 (IQR: 2.5)	0.061
DS-B	7.0 (IQR: 3.0)	5.0 (IQR: 2.5)	0.033*	5.0 (IQR: 1.5)	0.022*
MWT-B	31.0 (IQR: 6.0)	24.0 (IQR: 7.3)	0.001**	25.5 (IQR: 7.0)	0.023*

TMT-A/B: trail making test version A/B; MWT-B: multiple-choice vocabulary test; DS-F: digit span forwards; DS-B: digit span backwards.

^*Significance between groups was assessed using a Wilcoxon–Mann–Whitney two-sample rank-sum test p < 0.05.

^**Significance between groups was assessed using a Wilcoxon–Mann–Whitney two-sample rank-sum test p < 0.005.

3.7. Medication and correlation with test performance

The respective medication of every patient is documented in Table 1. Only 2 patients did not receive any medication while the majority was treated with a combination of antidepressants. The most common activating drug were serotonin–norepinephrine reuptake inhibitors (SNRI) in 14 of 20 patients, that is 8 received Duloxetin [average daily dose (ADD): 83 mg], 6 Venlafaxine (ADD: 213 mg) and additional 2 patients received the selective serotonin reuptake inhibitor (SSRI) Citalopram (ADD: 30 mg). 6 patients were treated with rather sedating Tetracyclic (TeCA) or Tricyclic antidepressants (TCA), more precisely 4 patients received Mirtazapine (ADD: 21 mg) and 2 patients Trimipramine (ADD: 50 mg). Moreover, 6 patients were treated with diverse antipsychotics and 6 with anticonvulsants of which 5 patients received Pregabalin (ADD: 325 mg). Importantly, there was no statistical association between subgroups and any type of medication (Fischer’s exact test, SNRI/SSRI: p = 1; TeCA/TCA: p = 0.122; Antipsychotics: p = 1; Anticonvulsants: p = 1; Hypnotics: p = 0.521).

When analyzing the statistical relationship between medication and test results, no association was found between medication and performance in the motor paradigm. In contrast, the treatment with SNRIs/SSRIs featured a negative correlation with time for pointing movements (r = −0.46; p = 0.041) indicating faster basic motor coordination with SNRI medication. TeCAs/TCAs prescription correlated negatively with finger tapping (r = −0.46; p = 0.044) and positively with pointing performance (r = 0.52; p = 0.019) suggesting slower basic motor speed and coordination with TeCAs/TCAs treatment in 6 patients. Finally, administration of typical or atypical antipsychotics correlated positively with both trail making test scores (A: r = 0.66; p = 0.001; B: r = 0.57; p = 0.009) and negatively with DS-B (r = −0.45; p = 0.044). This suggests that antipsychotic treatment affected cognitive functioning, i.e., attention and working memory in 6 patients.

4. Discussion

The present study investigated psychomotor functioning in the fine motor domain in patients with unipolar major depression. Comparing internally and externally cued response selection and initiation, patients were particularly slowed when movements involved internal movement selection. A data-driven clustering approach revealed a tendency towards two distinct subgroups. While both featured consistent impairments in fine motor functioning, only one subgroup showed significant disturbances in response selection and initiation.

4.1. General and specific psychomotor disturbances

In our experimental motor paradigm we replicated the finding that depressed patients are significantly impaired in reaction speed and response selection based on external cues [21–23]. Moreover, the internal selection of responses (‘Choice’-condition) was markedly slowed in relation to matched controls. Finally, depressed patients also showed significantly fewer self-initiated movements (‘Free’-condition), suggesting that not only the choice of a movement but also the drive to initiate movements is reduced. This congruency in all three aspects, which is echoed by the correlation of the respective response times, thus conforms to the known general impairment of psychomotor functions in major depression [2,13].

In contrast, finger tapping performance was not significantly impaired in our patient sample, which indicates that there was no deficit in basic motor speed. Yet, slowed finger tapping performance was found previously in depressed patients [45]. This divergence to our results may be attributable to the administration of SNRIs or SSRIs in 16 out of 20 patients in our sample. It has been shown that with this medication finger tapping performance does not differ from healthy controls [46] and even that psychomotor functioning generally ameliorated significantly with SSRI treatment [47,48]. Taken together with slowed reaction times and the increased time taken to complete the pointing movements and the trail making tests, this points to specific impairments of cognitive rather than motor aspects of psychomotor functioning. This corresponds well to known impairments of visuospatial and attentional processing in major depression [6,49–52]. Even though antipsychotic medication affected cognitive functioning and in particular attention, excluding the respective 6 patients form the comparison still revealed inferior performance in the trail making tests.

4.2. Deficient response selection and initiation

The most prominent finding of our study was that patients were markedly more slowed in internally triggered (non-cued) than in externally triggered (visually cued) reactions. The same symptoms are found in Parkinson’s disease [27,53]. This specific deficit of (fine) motor control may be attributed to known disturbances in cognitive control centers, like the DLPFC, the rostral cingulate zone (RCZ) of the ACC and the BG in major depression [28,30,31]. In line with this interpretation are several studies associating the very same areas with internal selection and initiation of movements in the healthy brain [54–57]). Consequently, our results give some support to the general symptom hypothesis [28,29] presuming that psychomotor changes in Parkinson’s and (melancholic) depression are caused by dopaminergic deregulation due to frontostriatal deficits. Analyzing event-related potentials in severely depressed patients, Schrijvers [58] recently demonstrated a close relationship between psychomotor retardation and impeded action monitoring. Based on this finding, it was suggested that the same frontostriatal areas form a part of a network of higher-order executive systems involved in response selection, motor control and action monitoring. In this framework, the RCZ is proposed to account for selection of actions [59], which is supported by very recent imaging data provided by using the same motor paradigm as in the present study [56]. Hence, RCZ function should be specifically affected in depressed patients with deficient response selection and initiation.

4.3. Patient subcluster with differential psychomotor disturbances

When patients were clustered by their test performance, a tendency towards two subgroups of patients may be identified. Intriguingly, this subdivision was primarily driven by differences in the motor paradigm. Whereas both subgroups showed impairments in fine psychomotor functioning and short-term memory, only one patient cluster featured significant disturbances of internal selection and initiation of movements. Likewise, Pier [15] demonstrated PMD in patients with major depression using computerized drawing tasks. When differentiating between patients with and without melancholic features in the same study, both groups showed impairments of cognitive aspects of psychomotor activity, whereas only melancholic patients also featured slowing of motor aspects. In this context, it may be argued that internal cueing of responses represents a motor aspect of psychomotor functioning [24,28]. Accordingly, Rogers [22] reported marked deficits specifically in response selection in depressed patients with melancholic features. As melancholia was not assessed in our patient sample, it may only be speculated that the clustering result is a reflection of the presence of melancholic features in the subgroup with more pronounced PMD. In favor of this assumption is that psychomotor retardation is common in depression with melancholic features [11] and even is considered a diagnostic marker for melancholic depression [2,13].

4.4. Psychomotor disturbance, cognitive impairment and clinical state

In the literature, memory deficits are very consistently linked to depression severity [12,50]. Correspondingly, we found increased working memory deficits in patients reporting increased symptom severity (BDI-2) and also with a recurrent depressive episode. Furthermore, severely depressed patients (1CD-10) were less stable in performing internal response selection (‘Choice’-condition). According to Schrijvers [13], this association between clinical state and PMD is primarily observed in melancholic depression. Given the rather coarse classification of objective clinical state by 1CD-10 in this study, the relationship between deficient response selection and disease severity needs further investigation.

In summary, the current results replicate known (fine) PMD in patients with major depression and moreover indicate specific impairments of visuospatial and attentional processing as cognitive aspects of psychomotor functioning. In the present study we demonstrated for the first time more severely affected internally triggered than externally triggered response selection with both being slowed in major depression. Finally, a cluster analysis revealed two subclusters of PMD in our patient sample. Therefore, multivariate clustering of behavioral data may be a promising future approach to identify subtypes of cognitive or psychomotor impairments in patient populations.