Introduction

Subthalamic (STN) deep brain stimulation (DBS) is an effective treatment option in Parkinson disease (PD). Conflicting reports point to a potential detrimental effect of STN-DBS on cognition (Daniels et al. 2010; Smeding et al. 2006; Volkmann et al. 2010; Weaver et al. 2009; Williams et al. 2010; Witt et al. 2008; York et al. 2008; Zangaglia et al. 2009, 2012; Odekerken et al. 2016; Combs et al. 2015). Concomitantly, the natural course of PD has been shown to be associated with the development of dementia of different degrees in 17.4% after 5 years following diagnosis (Pedersen et al. 2017) and 83% following 20 years after disease onset (Hely et al. 2008). Furthermore, an incidence rate of 10% for dementia in PD patients (Aarsland et al. 2001; Marder et al. 1995) and a prevalence of 30% following 8 years of disease onset was reported in a systematic review (Aarsland et al. 2005). In contrast, a longitudinal study showed a minor degree of dementia in PD of 13% following 7 years (Balzer-Geldsetzer et al. 2011).

Numerous predictors at baseline for dementia in PD have been identified, including age, disease duration, mild cognitive impairment (MCI), orthostatic hypotension, REM sleep behaviour disorder (RBD), motor impairment, male sex, hallucinations, and axial symptoms (Aarsland et al. 2001, 2003; Marder et al. 1995; Anang et al. 2014, 2017; Hely et al. 2008; Domellof et al. 2015; Cereda et al. 2016).

Motor and nonmotor fluctuations appear after 5–10 years following onset of PD. STN-DBS has been shown to improve quality of life in PD patients, ameliorating motor fluctuations or therapy-refractory tremor (Deuschl et al. 2006; Schuepbach et al. 2013). Growing knowledge about cognitive and psychiatric effects of STN-DBS is essential to adequately advise patients suffering from disabling PD with and without cognitive dysfunction, potentially eligible for surgery.

Methods

Between 1997 and 2006 all patients (n = 104) suffering from PD at University Hospital Berlin Charité, who were treated via continuous bilateral STN-DBS were included after 6.3 ± 2.2 years (range 3.6–10.5 years) surgery. Inclusion criteria were the diagnosis of PD following the United Kingdom Parkinson’s Disease Society Brain Bank Criteria, bilateral implantation of STN-electrodes, written informed consent for neuropsychological investigation according to the declaration of Helsinki and permission of local ethics commission. Exclusion criteria comprised atypical PD, implantation of unilateral electrode, pallidal or thalamic target localisation, explantation of neurostimulation system and missing presurgical neuropsychological data.

All postsurgical available patients were reassessed by neuropsychological and motor test batteries with identical presurgical tests.

Investigation of motor function was performed by Unified Parkinson Disease Rating Scale, motor subscore (UPDRSm) and Hoehn and Yahr Stage (Fahn et al. 1987; Hoehn and Yahr 1967) in the best on-condition of stimulation and medication.

The assessment of global cognition using modified Mattis Dementia Rating Scale (mMDRS) (Mattis 1988) and the five cognitive domains (i.e., memory, executive function, language, attention, working memory) via Rey Auditory Verbal Learning Test (RAVLT), semantic and phonemic fluency, Stroop-Test, Trail Making Test (TMT), Digit span [for detailed description of neuropsychological tests see (Lezak 1995)] was performed postsurgically in all available patients. Due to missing presurgical data of the visual memory part (4 points), the maximal MDRS score comprised 140 points in all patients (= mMDRS).

Cognitive conditions of each patient were classified as no or mild cognitive impairment or dementia according to level-1 criteria of movement disorders society (Emre et al. 2007; Litvan et al. 2012) in pre- and postsurgical states.

Impairment of each cognitive tests was defined as performance below threshold 16th percentile or more than 1 standard deviation compared to normative age-matched data presurgery or intraindividual data pre- versus postsurgery. Individual deterioration was analysed as change/DBS year for each cognitive domain and for global cognition score and was correlated with age at surgery, disease duration, DBS-years, and preoperative baseline levels of each cognitive domain.

Affective state was assessed using multidimensional mood state questionnaire (MDMQ) (Steyer et al. 1997), Beck Depression and Anxiety Inventory (BDI, BAI) (Beck 1987; Beck et al. 1988), Snaith–Hamilton-Pleasure-Scale (SHAPS-D) (Snaith et al. 1995), Bech-Rafaelsen-Mania Scale (BRMS) (Bech et al. 1979), Montgomery and Asberg Depression Rating Scale (MADRS) (Montgomery and Asberg 1979), Brief Psychiatric Rating Scale (BPRS) (Flemenbaum and Zimmermann 1973) and quality of life (QoL) by Parkinson’s Disease Questionnaire (PDQ 39) (Peto et al. 1995) and the 36-item Short-Form Health Survey (SF 36) (Bullinger 1996).

Statistical analysis

Longterm outcomes were compared to baseline using the Wilcoxon signed-rank test for matched pairs. Statistical tests were two-tailed and were adjusted for multiple testing according to Bonferroni, and the level of significance was calculated as p ≤ 0.001(Kim 2014). Spearman’s ρ coefficient for ordinal and metric data was analysed, for nominal data exact Fishers test was used. Next, variables that showed a potential association in the correlation analysis (p < 0.2) were used in the logistic regression analysis, with conversion to dementia as the dependent variable.

Logistic regression was used to calculate odds ratio (OR) with 95% confidence interval (CI) for PD-D with and without adjustment for age at surgery, sex, and education. Statistical analyses were performed with the SPSS® Statistics, version 22.

Results

Preoperative neuropsychological results were available in 79/104 of the patients. Thirty-seven of these patients at age of 67.6 ± 6.9 years (range 53–81 years) with a disease duration of 17.1 ± 5.1 years (range 6–28 years) were additionally assessed after 6.3 ± 2.2 years (range 3.6–10.5 years) postsurgery (Table 1) via neuropsychological and motor test batteries. In the other 42 patients follow-up assessments were not available due to death (n = 21), loss of patient contact (n = 9; mean age 74.2 ± 7.9 years, mean disease duration 21.6 ± 5.2 years) or patients’ refusal of retesting (n = 12; mean age 76.1 ± 14.2 years, mean disease duration 17.9 ± 6.1 years) (Fig. 1).

Table 1 Patients characteristics pre- and postsurgery
Full size table
Fig. 1
figure 1

Study design

Full size image

At a follow-up of 6.3 ± 2.2 years in 37/79 patients motor function measured by UPDRSm worsened by trend post- versus presurgery (p < 0.01) by a difference of 7.1 ± 14.1 points (i.e., 36.2%) and Levodopa equivalent dose (LED) showed a trend of sustained reduction of 39.1% compared to presurgical levels (p < 0.01). Importantly, all patients had at least one replacement of their implantable pulse generator during an observation period (October 2018), suggesting continued benefit of motor symptoms via STN-DBS.

Presurgical state

Presurgically, 24.4% of patients (n = 9/37) showed normal cognitive functioning (PD-NC), while 75.7% showed mild cognitive impairment (PD-MCI; n = 28/37) according to level 1 criteria (Litvan et al. 2012). None of the patients displayed dementia (PD-D) presurgically (Fig. 2a).

Fig. 2
figure 2

a Cognitive classification of all patients as absent (PD-NC) or mild cognitive impairment (PD-MCI) or dementia (PD-D) according to level-1 criteria of movement disorders society (Emre et al. 2007; Litvan et al. 2012) in pre- and postsurgical states. b Significant correlation of Mattis Dementia-Ratings Scale-Change/DBS-year and disease duration (r = 0.42; p = 0.017, Spearman’s-ρ correlation coefficient), rhombs represent each individual patient score. Note that the rapid decline in one patients most likely reflects long disease duration (19 years) since no clinical signs for atypical PD were present in this patient. c Non-significant correlation of Mattis Dementia-Ratings Scale-Change/DBS-year and age (r = − 0.09; p = 0.62, Spearman’s-ρ correlation coefficient), points represent each individual patient score

Full size image

Age, disease duration, LED, motor function (med-off, med-on), levodopa-response, educational level did not differ in PD-MCI versus PD-NC patients, respectively. PD-MCI and PD-NC also did not differ with respect to affective states and QoL.

Presurgery, PD-MCI versus PD-NC patients differed in premorbide verbal intelligence measured by MWT-B (p = 0.008) by trend, a German equivalent of National Adult Reading Test. Preoperative digit span forward (p = 0.008) and backward (p = 0.008), and memory subscore of mMDRS (p = 0.04) were lower by trend in PD-MCI patients compared to PD-NC patients.

Postsurgical state

Postsurgically, 18.9% (n = 7/37) showed no cognitive impairment, 40.5% (n = 15/37) displayed MCI and 40.5% (n = 15/37) developed dementia (Fig. 2a) according to level 1-criteria (Emre et al. 2007). Thirteen of 28 (46.4%) presurgically identified PD-MCI patients converted to PD-D following 6.3 DBS years. In contrast, the conversion rate to dementia of PD-NC patients was 2/9 (22.2%).

Notably, a trend to association between preoperative mild cognitive impairment and conversion to dementia during the postoperative observation period was observed with respect to age at baseline, sex and education (adjusted OR, 10.8, 95% CI 1.0–119.0, p = 0.052).

In comparison to the non-demented PD patients (PD-NC/MCI) PD-D patients showed longer disease durations (20.2 ± 4.8 years vs. 15.0 ± 4.1 years, PD-NC/MCI versus PD-D; p < 0.01) and more severe motor impairment (UPDRSm 36.1 ± 12.9 versus 23.7 ± 7.8, PD-NC/MCI versus PD-D; p < 0.01) by trend. Additionally, PD-D patients showed more pronounced deterioration of UPDRS motor subscores (bradykinesia, p = 0.012; rigidity, p = 0.018; axial symptoms—summary score of speech, posture, gait and postural stability, p = 0.032) by trend. LED, presurgical levodopa-response as well as age, age of disease onset and DBS-years did not significantly differ between non-demented and demented patients.

In general, there was a deterioration of various cognitive domains (global cognition, memory, language) in all STN-DBS treated patients 6.3 years after surgery (cf. Table 2), while partial working memory was preserved and slightly improved by trend (digit span forward, p < 0.01).

Table 2 Cognition, affective state and quality of life pre- and postsurgery
Full size table

Specifically, PD-D patients worsened significantly compared to PD-MC/NC patients with respect to global cognition (mMDRS, p < 0.001), comprising the domains (1) memory (RAVLT, immediate and delayed recall), (2) executive function (TMT B), (3) language (semantic and phonematic fluency), (4) attention (TMT A), while also working memory did not differ between the two groups (digit span, data not shown).

Overall, patients showed a global cognitive change/DBS years measured by mMDRS of − 1.6 ± 3.8/140 points/DBS year (range − 2.2 to 19.8 points/DBS year), which inversely correlated with disease duration (r = − 0.4, p = 0.02), but not with age at surgery (Fig. 2b/c).

Various cognitive changes/DBS years of single testing domains correlated with disease duration: (1) executive function: TMT B (r = 0.6, p = 0.03), TMT B–A (r = 0.5, p = 0.05); (2) language: semantic fluency “supermarket” (r = − 0.5, p < 0.01), phonemic fluency “F” (r = − 0.5, p = 0.03); (3) longterm memory: verbal memory, RAVLT total score (r = − 0.4, p = 0.03), RAVLT free recall (r = − 0.6, p < 0.01), RAVLT delayed recall (r = − 0.6, p = 0.01); (4) attention TMT A (r = 0.5, p = 0.02); (5) for working memory digit span backward (r = − 0.4, p = 0.05).

Finally, the conversion to dementia (PD-D) according to level-1 criteria of movement disorders society is also positively correlated (Spearman’s ρ r = 0.4, p = 0.01) with disease duration. Furthermore, logistic regression analysis showed significant relation beetween development of dementia and disease duration with respect to age at postsurgical investigation (adjusted odds ratio 1.2, 95% CI 1.0–1.4, p = 0.02). Development of dementia did not correlate with hallucinosis and levodopa-response.

Investigation of affective states in all PD patients showed a trend of postsurgical deterioration of mood measured by patient-related score BDI (p = 0.02), in contrast to the patient-related score MDMQ (p = 0.2) and physician-related score MADRS (p = 0.7). Anxiety (BAI), hedonia (SHAPS-D) and positive psychiatric symptoms (BRMAS, BPRS) did not change postsurgery compared to presurgery (Table 2).

Mood evaluated via BDI positively correlated with MADRS (r = 0.78, p < 0.01) and PDQ 39 (r = 0.69, p < 0.01). Particularly, there was no association between mood and disease duration, LED, motor function and global cognition (data not shown).

Postsurgery, QoL worsened by trend assessed via PDQ 39 (p < 0.01), although not reflected by SF 36 (SF 36 physical summery subscore, p = 0.1, SF 36 mental summery subscore, p = 0.7, Table 2).

QoL investigated by PDQ 39 negatively correlated with SF 36 mental summary subscore (r = − 0.8, p < 0.01), the global cognition score mMDRS (r = − 0.5, p = 0.02), and positively associated with the mood score MADRS (r = 0.7, p < 0.01), motor function scales UPDRSm (r = 0.5, p = 0.03), including the UPDRS motor subscores bradykinesia (r = 0.5, p = 0.03), the summary subscore for axial symptoms (r = 0.7, p < 0.01) and isolated gait (r = 0.67, p < 0.01) as well as Hoehn and Yahr Stage (r = 0.6, p < 0.01).

PD-D patients did not differ from PD-MCI/NC-patients with respect to affective state and QoL (data not shown).

Mortality rate

Approximately 27% (21/79) of all patients died at mean age of 71.5 ± 5.7 years following a mean disease duration of 16.3 ± 5.3 years and mean chronic STN-DBS of 5.4 ± 2.4 years. Causes of death were not available. 9/79 (11.3%) of patients were not available.

Discussion

This observational “real life” study of 37 initially non-demented Parkinson patients provides longterm data on cognition, affective state and QoL of Parkinson’s patients treated with STN-DBS up to 10.5 years.

Study limitations of the present study comprise a missing control group (i.e., a comparable drug-treated group), precluding differential analyses of effects of STN-DBS and natural disease course. Furthermore the high drop out rate of 53% might have biased the present results since PD dementia may be associated with reduced study participation (n = 12/79) and death (n = 21/79). However, this bias is difficult to avoid in real-life studies despite home visits as performed in the present study, but may certainly alter the real dementia prevalence and the correlation between dementia and age. Thus, in 2 other studies on STN-DBS and development of PD dementia dropout distinctly higher rates of 76% or even 87% have been reported (Kim et al. 2014); Fig. 1, approximately 25/103 patients, i.e., 76% drop out rate at 5–7 years post-DBS (Merola et al. 2014) (Fig. 1; 25/189 patients with completed evaluation, i.e., 87% drop out rate 5-year post-DBS).

Following approximately 6 years of STN-DBS and a disease duration of 17 years, 41% of all patients developed either dementia or MCI, whereas 18% continued to not be affected by cognitive deficits. These data comply with the natural disease course of PD in medically treated patients with a dementia rate of 48% following 15 years, which rises to 83% after 20 years of PD duration (Aarsland et al. 2001; Hely et al. 2008).

The presurgical prevalence of MCI of 75.7% after 11.3 years disease course seems to be relatively high in our study population contrasting for instance 25–30% after 6.8 years PD duration and 43% after 6 years in non-surgically treated PD patients, respectively (Aarsland et al. 2010; Pigott et al. 2015).

The presurgical prevalence in other DBS studies is heterogeneous and varies between 23% and 14.2 years (Merola et al. 2014) and 63.1% and 10.6 years (Kim et al. 2014) PD duration, which likely reflects differential selection criteria of the different DBS studies.

In the present study population presurgically identified PD-MCI patients converted to dementia in 46.6% of the cases after an observation period of 6 years. Presence of MCI non-significantly correlated (trend; p = 0.052) with development of dementia. In contrast, only 22.2% of PD-NC patients developed dementia in the investigated study population following a disease duration of 17.1 years. Thus, presurgically existing MCI may predict a conversion to dementia of PD patients, complying with other studies assessing PD patients with (Kim et al. 2014; Merola et al. 2014) and without neurostimulation (Anang et al. 2014, 2017).

Interestingly, the annual deterioration of global cognition measured via mMDRS of 1.6/140 points/DBS-years and subsequently the conversion to dementia correlated with disease duration, but not with age at surgery.

Natural history studies of PD patients identified several predictors of conversion to dementia such as age, disease duration, male sex, baseline RBD, hallucinations orthostatic hypotension, axial symptoms like particular postural and gait disturbances, color discrimination ability, and MCI (Anang et al. 2014, 2017; Aarsland and Kurz 2010; Marder et al. 1995; Hely et al. 2008; Domellof et al. 2015; Cereda et al. 2016; Aarsland et al. 2001, 2003). Additionally, recent studies of PD patients with STN-DBS (Aybek et al. 2007; Smeding et al. 2011; Kim et al. 2014) partially provided similar predictors for global cognitive decline such as age, hallucinosis, impaired executive function, or lower levodopa-response after 1- or 3-year post-STN-DBS. Except for poor presurgical cognitive performance, these factors including age were not identified in the present analyses. However, we can not exclude a bias for instance with respect to age due to differential selection criteria for STN-DBS and/or older deceased patients, who could not be reassessed.

Overall, a deterioration of various cognitive domains was observed in all patients (PD-NC/MCI) postsurgery. Demented versus nondemented patients performed worse in multiple cognitive domains: executive function, language, memory and attention, less pronounced working memory. The progression of each of the multiple domain deficits correlated with disease duration and not with age of PD patients.

Mood and QoL slightly declined by trend postsurgery, whereas declined QoL positively correlated with diminished global cognitive function, mood and motor function (especially axial symptoms), that underlies the impact of nonmotor and motor symptoms on QoL (Martinez-Martin et al. 2011; Perez-Lloret et al. 2017; Schrag et al. 2000; Rahman et al. 2008).

The mortality of 27% (maximal 38% including lots of follow-up) in our study population (n = 79) is lower compared with previously reported epidemiological studies, amounting to 38%and 66% following 10 and 15 years PD disease duration (Hely et al. 2008), possibly reflecting selection criteria of patients evaluated for surgery.

Motor function, represented by UPDRSm showed by trend a progressive worsening over the observational period of 3.6–10.5 years, and complying with previous reports (Merola et al. 2011, 2014; Rizzone et al. 2014; Castrioto et al. 2011). LED was persistently diminished by trend postsurgery, likely due to beneficial DBS-effects, levodopa-resistent symptoms and/or adverse effects of increased levodopa dosage, that often induces hallucinosis or orthostatic dysregulation in severely affected PD patients. Thus, despite the slightly pronounced motor impairment of PD-D-patients versus PD-NC/MCI patients, the LED did not differ in the subgroups, possibly due to potential side effects of levodopa or levodopa-resistent symptoms.

Demented patients suffered by trend from a longer disease duration and were more affected by trend by axial symptoms, rigidity and bradykinesia compared to non-demented patients, which likely reflects the natural history of PD (Hely et al. 2008; Coelho and Ferreira 2012; Anang et al. 2014).

In summary, lack of a matched control group precludes final conclusions regarding possible influence of STN-DBS on disease course with respect to conversion to dementia. However, the present data complies with previous studies on cognitive decline in medically treated PD patients. The gradual deterioration of global cognitive impairment measured by MDRS and the conversion to dementia was associated with disease duration, but not with age of patients.