Abstract
Background:
Subthalamic nucleus deep brain stimulation (STN-DBS) successfully controls the motor symptoms of Parkinson’s disease (PD) but has associated cognitive side-effects.
Objective:
Establish the short- and long-term cognitive effects of STN-DBS in PD.
Methods:
Both the short-term and long-term effects of STN-DBS on cognition were examined through evaluation of the controlled studies that compared patients with STN-DBS to unoperated PD patients, thus controlling for illness progression. We also reviewed the literature to identify the factors that influence cognitive outcome of STN-DBS in PD.
Results:
The meta-analysis of the short-term cognitive effects of STN-DBS revealed moderate effect sizes for semantic and phonemic verbal fluency and small effect sizes for psychomotor speed and language, indicating greater decline in the STN-DBS operated than the unoperated patients in these cognitive domains. The longer-term STN-DBS results from controlled studies indicated rates of cognitive decline/dementia up to 32%; which are no different from the rates from the natural progression of PD. Greater executive dysfunction and poorer memory pre-operatively, older age, higher pre-operative doses of levodopa, and greater axial involvement are some of the factors associated with worse cognition after STN-DBS in PD.
Conclusion:
This evidence can be used to inform patients and their families about the short-term and long-term risks of cognitive decline following STN-DBS surgery and aid the team in selection of suitable candidates for surgery.
Keywords
INTRODUCTION
Cognitive deficits, such as executive dysfunction, commonly occur early in the course of Parkinson’s disease (PD) [1, 2]. Approximately 27% of patients develop mild cognitive impairment (MCI) [3], which is a risk factor for PD dementia (PD-D) [4, 5] with a long-term prevalence of 80% [6–8].
Several randomized controlled trials have established deep brain stimulation of the subthalamic nucleus (STN-DBS) as an effective treatment of the motor symptoms of PD, with concomitant and sustained improvement of quality of life (QoL) [9, 10]. Cognitive impairment and depression are among the non-motor symptoms that have a negative impact on the QoL of PD patients [11]. Therefore, as the main aim of surgical treatments is to improve the QoL of PD patients, it is important that these do not produce adverse cognitive effects. Pre-operative cognitive deficits may persist or even be exacerbated after surgery and hence undermine the positive impact of improvement of motor symptoms with surgery on QoL [12–14].
In this review, we have three main aims. The first is to undertake a meta-analysis of the short-term results of the controlled studies which compared changes in cognitive function before and after STN-DBS surgery with an unoperated control group of PD patients. The results of the controlled studies provide a clearer picture of the effects on cognition of STN-DBS surgery distinct from the progression of PD. Second, we evaluate the long-term effects of STN-DBS on cognitive function, by also examining the results of the controlled studies which have followed up patients from 2 to 9 years after STN-DBS surgery. Finally, to aid clinicians in their selection of suitable candidates for surgery, we reviewed the factors that have been shown to influence the cognitive outcome of STN-DBS in PD.
METHODS
Study selection
For the meta-analysis, studies investigating cognition in PD patients with STN-DBS, assessed before surgery and up to 12 months after surgery compared to matched unoperated PD patients assessed at baseline and 12 months later were included. For the second aim, we reviewed controlled studies investigating the long-term effects of STN-DBS on cognition prior to surgery and two or more years after surgery. The literature searches were conducted through PubMed and ISI Web of Science electronic databases for the years 2000 to 2022. Keywords including “Parkinson’s disease”, “subthalamic nucleus”, “deep brain stimulation”, “neuropsychology”, “cognition”, “verbal fluency”, “memory”, “executive function”, “attention” “processing speed”, “visuo-spatial”, and “language” were used. For the second aim of the review, the additional keyword “long-term effects” was used. The articles identified by the search were obtained, screened and if they met eligibility criteria (see below) were included to address the aims of this review. Each of the eligible controlled studies were independently reviewed by two of the investigators (MF, FL). The abstract and the full paper were read and discussed.
Study eligibility criteria
The study eligibility criteria for the meta-analysis were: 1) patients with idiopathic PD with bilateral STN-DBS; 2) PD control group; 3) interval or ratio data; 4) for both groups reporting neuropsychological data before surgery/baseline and up to 12 months after surgery follow-up; 5) use of at least one standardized neuropsychological test; 6) sufficient report of study results to allow for an effect size to be calculated; 7) missing data which could be calculated with previously identified methods in the Cochrane handbook.
The eligibility criteria for the long-term follow-up studies were: 1) patients with idiopathic PD with STN-DBS; 2) PD control group; 3) reporting neuropsychological data at least on one test before surgery and two years or more after surgery.
Data coding
The authors independently extracted the following information from the articles that were included in the meta-analysis: 1) number of patients in both study groups (STN-DBS and PD control); 2) exclusion criteria; 3) confirmation of target by microelectrode recording or MRI in the STN-DBS group; 4) verification of electrode placement via radiographic means in the STN-DBS group; 5) stimulation parameters in the STN-DBS group; 6) assessment time points; 7) assessment measures; 8) medication type; 9) summary statistics needed for computation of effect sizes. Neuropsychological tests were categorised into the following nine domains: cognitive screening; attention and concentration; executive function; psychomotor speed; learning and memory; visuospatial skills; language; phonemic verbal fluency; and semantic verbal fluency. Although most neuropsychological tests assess multiple cognitive domains, each test was assigned to the domain it reflects most according to clinical practice. When individual coding was achieved both authors met to resolve any disagreements. Phonemic and semantic verbal fluency were regarded as separate domains, because these were most consistently found to decline after surgery and may also be affected differentially.
Data analysis
The present analysis used random effects meta-analytic models. To compare change from baseline between the two patient groups (STN-DBS vs. PD control) the effect size, that is, Cohen’s d was calculated for each study. A negative effect size indicates that patients in the STN group had a larger decline on that cognitive domain at follow-up compared to the PD control group. An effect size of 0.20 reflects a small effect, 0.50 reflects a moderate effect, and 0.80 reflects a large effect [15]. The average weighted effect size, corresponding 95% confidence interval and variance were then calculated. A 95% confidence interval not including zero indicated a significant effect. To investigate whether the study samples had a common underlying effect size, the homogeneity of the effect size was computed for each cognitive domain using the Cochran’s Q and the I2. This was done to get information about the cohesiveness of each of the cognitive domains. All analyses were performed using Review Manager 5.2 (Nordic Cochrane Centre, Copenhagen, Denmark).
RESULTS
Short-term effects of STN-DBS on cognition: Meta-analysis of controlled studies
After initial screening the database search yielded a total of 25 journal articles describing controlled studies, comparing cognition of PD patients with STN-DBS with a matched unoperated PD group. Bilateral, concurrent STN-DBS was performed on all patients in all studies except one, by Morrison et al. [16], in which 2/17 patients underwent a staged surgery. Studies were excluded from analysis for the following reasons: significant sample overlap with a more extensive study that is included in the meta-analysis (n = 3; [17–19]); looked at unilateral instead of bilateral STN-DBS (n = 2; [20, 21]); follow-up assessment at 24 months (n = 2; [22, 23]); inclusion of patients who had unilateral pallidotomy prior to STN-DBS surgery (n = 1; [16]); different neuropsychological tests were administered in the DBS and control groups (n = 1; [24]), insufficient quantitative report of data (n = 3; [25–27]); results of patients with STN-DBS and GPi-DBS were not analysed separately (n = 1; [28]).
Table 1 summarizes and provides details of the 28 studies that were identified from the literature search. Of these 13 studies were considered suitable and were included in the meta-analysis (PD with STN-DBS N = 410; PD unoperated control N = 440), which are starred in Table 1. Supplementary Table 1 presents the particular cognitive domains considered and the specific neuropsychological tests that were used across the studies to tap each of the cognitive domains covered by the meta-analysis.
Short-term effects of STN-DBS on cognition based on controlled studies arranged by year of publication (earliest first)
PD, Parkinson’s disease; STN, subthalamic nucleus; GPi, internal segment of the globus pallidus; B, bilateral; U, unilateral; RAVLT, Rey Auditory Verbal Learning Test; ROCF, Rey-Osterrieth complex figure; TMT, Trail making test; WAIS, Wechsler adult intelligence scale; TOH, Tower of Hanoi; WCST, Wisconsin card sorting test; VF, verbal fluency; DRS, Dementia Rating Scale; NART, National adult reading test; RMT, Recognition memory test; BTA, Brief test of attention; HVLT-R, Hopkins verbal learning test revised; BNT, Boston naming test; VFDT, Visual form discrimination test; JLOT, Judgement of line orientation test; STDS, Standardized test of direction sense; OMOT, Odd man out test; MMSE, Mini-Mental Status Examination; PASAT, paced auditory serial addition task; AVLT, Auditory verbal learning test; PWL, Paired word learning; BVMT, Brief visual memory test; LMT, Logical memory task; CBTT, Corsi’s block tapping test; RM47, Raven’s matrix 47; PAL, Paired associate learning; NMCS, Nelson modified card sorting test; RBANS, Repeatable battery of neuropsychological status; WMS-III-R, Wechsler memory scale; RBMT, Rivermead behavioural memory test, RNGT, Random Number Generation Task, SCOPA-cog, Scales for Outcome in Parkinson’s disease Cognition; LCIG, levodopa/carbidopa intestinal gel treatment; NR, not reported. *Articles that met criteria to be included in the meta-analysis.
The results of the Cochrane’s Q and I2 tests and the average random effect sizes, their variances and 95% confidence intervals are shown in Supplementary Table 1. Tests of heterogeneity were significant for the domains of executive function (I2 = 58%), psychomotor speed (I2 = 73%) and language (I2 = 71%). The remaining tests of heterogeneity were not significant.
Random effects analysis yielded statistically significant differences in change from baseline to follow-up between STN-DBS and control patients on several cognitive domains (see Fig. 1). With the exception of semantic verbal fluency (d = –0.64; 95% CI = [–0.88, –0.4]) and phonemic verbal fluency (d = –0.46; 95% CI = [–0.61, –0.31]) which showed moderate effect sizes, the effect sizes in the other domains ranged from –0.31 to –0.14 and were small by Cohen’s criteria [15] (psychomotor speed (d = –0.31; 95% CI = [–0.51, –0.12]), language (d = –0.31, 95% CI = [–0.58, –0.04], cognitive screening (d = –0.22; 95% CI = [–0.43, –0.01]), attention and concentration (d = –0.2; 95% CI = [–0.43, –0.01]). Therefore, while in all these domains the STN-DBS group had greater decline at follow-up compared to the PD control group, only the effect sizes for the semantic and phonemic verbal fluency are notable (see Fig. 1).
Forrest plot showing the effect sizes for the different cognitive domains.
To date, several meta-analytic reviews investigating the effects of STN-DBS on cognition in PD have been published [43–47]. These differ widely in their focus, comprehensiveness, methodology and the quality of the studies included in the meta-analysis. For example, in their meta-analysis, Elgebaly et al. [47] only included four randomized controlled trials of STN-DBS vs. GPi-DBS, while Wang et al. [45] only focused on seven articles related to four randomized controlled trials comparing STN-DBS and GPi-DBS. Martinez-Martinez et al. [48] solely focused on the effects of DBS on a single cognitive domain: executive function and included some studies with GPi-DBS as well as STN-DBS in their meta-analysis. The early meta-analysis of Parsons et al. [44] included patient case studies (often lacking controls), looking at cognitive measures before and after surgery, as well as the acute effects of stimulation by comparing cognitive performance with STN-DBS on vs. off. Thus, the results of the meta-analysis do not distinguish the effects of surgery and the acute effects of stimulation. Furthermore, in the intervening years, an additional number of better controlled studies have been published (see Table 2). In a more recent meta-analysis, Combs and colleagues [43] compared the cognitive impact of STN-DBS and GPi-DBS. However, most studies that were included in this review investigated either patients with STN-DBS or GPi-DBS, and only three studies provided a direct comparison of the effects of these two surgical targets on cognition. Moreover, with the exception of Xie et al. [46] and the present meta-analysis, other meta-analyses did not control for the effect of natural disease progression on cognition by including studies with unoperated PD control groups that were matched for relevant demographic variables. Therefore, it is unclear how much of the cognitive impairment is driven by STN-DBS vs. disease progression. Furthermore, both patients with unilateral and bilateral STN-DBS were included in some meta-analyses [43, 47]. Despite variations in the scope and methodology of these meta-analyses, they, including the current meta-analysis, allow several common conclusions. First, there is consensus across the meta-analyses that STN-DBS is not associated with major cognitive deficits. Second, there is agreement that decline in verbal fluency is the most consistent deficit following STN-DBS in PD [15–18, present meta-analysis]. Third, there is some evidence that the effect sizes for the decline in semantic fluency are somewhat larger than for phonemic verbal fluency [15, present meta-analysis]. Finally, there are smaller effect sizes for other cognitive domains such as psychomotor speed [16, present meta-analysis], language/verbal functions [16, present meta-analysis], and executive function [46, 48].
Long-term effects of STN-DBS on cognition: Effects from controlled studies
Table 2 lists the controlled studies which have reported the long-term cognitive effects of STN-DBS in PD two to nine years after surgery. The control groups were either PD patients who had had DBS surgery of the GPi or unoperated PD patients. However, a couple of these controlled long-term studies had methodological limitations [23, 25]. Zaganglia and colleagues [25] identified no difference between the STN-DBS and PD control groups in terms of cognitive decline three years after surgery. However, they tested patients who had STN-DBS 1, 6, 12, and 36 months after surgery, whereas the PD control group was only followed up 36 months after baseline. Possible deficits associated with STN-DBS could have been weakened due to practice effects as a result of repeated assessments in the first year after surgery. At two years follow-up, Williams and colleagues [23] reported that 32% of the STN-DBS group had dementia compared to 16% in the control group. However, in the Williams et al. study [23] patients in the STN-DBS group were significantly older, had longer disease duration and higher levodopa dosage at baseline compared to the PD control group, which may have confounded the outcome, since it has been shown that baseline levodopa dosage and age correlate with post-surgical cognitive decline [19, 49–51]. Also, pre-operative cognitive status influences post-operative cognitive status. For example, in one of the studies with the largest samples and long-term follow-up, the sample included cognitively intact PD patients and a subgroup of patients who had MCI prior to surgery [52]. Analysis showed that after five years, more than 40 per cent of the patients developed MCI and 32 per cent developed dementia. Two of the long-term studies [53, 54], which reported cognitive follow-ups of patients with STN-DBS respectively 8 or 6–9 years after surgery included unoperated PD control groups.
Across these studies listed in Table 2, the percentage of PD patients developing cognitive decline or dementia following STN-DBS surgery was highly variable, ranging from 0% [41, 55–57] to 32% [23, 54]. An even higher figure (84% of 64 patients with PD) showed decline on a composite cognitive score three years after STN-DBS has also been documented [58]. However, this inflated figure of 84% is likely due to the difference in the criteria for cognitive decline used by the study authors, defined as impairments in two out of five cognitive tests post-STN-DBS. Similarly, the two controlled studies with long-term cognitive follow-up reported respective rates of 16.7% (STN-DBS) vs. 17.6% (PD control) 8 years after surgery [53] and 31% (STN-DBS) vs. 46% (PD control) 6–9 years post-surgery [54].
These divergent rates of long-term cognitive decline across studies probably reflect several factors. First, the criteria for defining cognitive decline differ widely across the studies with long-term cognitive follow-up listed in Table 2. While some used Diagnostic and Statistical Manual (DSM) or Movement Disorder Society Parkinson’s Disease Dementia (MDS PDD) criteria [59, 60] for dementia [23, 54], others relied on decline on composite scores [58] or simply failed to specify or provide the criteria for diagnosis of cognitive decline or dementia [41, 61]. Second, the rate of cognitive decline and dementia in unoperated PD is variable and depends on several demographic and clinical features. One relevant factor is age of onset of PD, with late-onset cases more likely to develop dementia than those with young-onset PD [60]. Age is another key correlate of dementia in PD [62] and age differences across the samples of studies in Table 2 is likely to be important in this regard. Male gender is another factor associated with development of dementia in PD [63]. PD-D is also associated more with the akinetic-rigid type of PD than the tremor-dominant disease [60, 64]. Thus, differences across the samples of the studies listed in Table 2, in current age, in the proportion of old-onset vs. young-onset PD, akinetic-rigid vs. tremor-dominant PD, and male vs. female patients are likely to have influenced the percentage of cognitive decline following surgery. Third, the studies listed in Table 2 differed in the cognitive selection criteria applied to their patients. Furthermore, patients in some studies had comprehensive cognitive screening and neuropsychological assessment prior to surgery, which would have detected cases of dementia or those with MCI on the cusp of further cognitive decline. Fourth, the duration of follow-up is likely to be a key relevant factor, since longer post-operative follow-ups would be associated with longer durations of illness and greater progression of PD, which are in turn associated with a larger proportion of patients developing dementia [6–8]. Finally, ‘drop out’ through death or removal of DBS devices which are likely in a small proportion of cases are likely to influence the long-term cognitive effects of STN-DBS.
Bearing these methodological and sampling variations in mind, the highest prevalence of dementia across the controlled long-term follow-up studies after surgery based on DSM or MDS PDD criteria was 32%, which does not exceed the prevalence of dementia in the general PD population over similar periods of time [5, 8]. This suggests that the cognitive decline or dementia observed in a proportion of the operated cases is similar or in some cases lower [54] than in matched PD control groups and therefore may be attributable to the progression of PD rather than specifically related to the surgery. Future controlled long-term follow-up studies of cognition after STN-DBS will provide further confirmation of this conclusion.
The controlled studies reporting long-term cognitive effects, 2 years or longer after STN-DBS which included a PD control group, listed in the order of increasing length of follow-up
PD, Parkinson’s disease; STN, subthalamic nucleus; GPi, internal segment of the globus pallidus; DBS, deep brain stimulation; B, bilateral; DRS, Dementia Rating Scale; DSM, Diagnostic and Statistical Manual of Mental Disorders; MCI, mild cognitive impairment; VF, verbal fluency; WCST, Wisconsin card sorting test; PD-D, Parkinson’s disease dementia; MMSE, Mini-Mental Status Examination; CVLT, California verbal learning test; WAIS, Wechsler intelligence scale; HVLT, Hopkins verbal learning test; BNT, Boston naming test; BVMT, Brief visuospatial memory test; RAVLT, Rey auditory verbal learning test; SDMT, Symbol digit modalities test; TMT, trail making test; JLO, Judgement of line orientation; WMS, Wechsler memory scale; PASAT, Paced auditory serial addition test; LMT, Logical memory test; CBTT, Corsi’s block tapping test; FAB, Frontal assessment battery; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment.
A further point is worth noting. In previous studies both deficits in verbal fluency and the Stroop have been identified as predictors of later dementia in PD [65–67]. As noted from the above meta-analysis, verbal fluency decline is the most consistent cognitive deficit after STN-DBS. Such a decline in seen in up to 50% of operated patients and persists and worsens over time [68–70]. Similarly, post-surgical decline on the Stroop has also been documented in the majority of the controlled studies which included this test (see Table 1), as well as after acute STN stimulation [71]. Despite the associations between deficits on verbal fluency and Stroop and dementia in PD, the current literature suggests that deficits on these tests of executive function do not necessarily or inevitably lead to later cognitive decline or dementia after STN-DBS surgery [72]. Borden et al. [72] followed up their cohort of 24 patients with STN-DBS up to eight years after surgery. Dementia diagnosed by the MDS and DSM-IV-TR criteria was identified in nine (37.5%) cases. While pre-operative values of verbal fluency were predictive of verbal fluency performance six months after surgery, they did not predict dementia and age was the only independent predictive factor for dementia, suggesting that dementia is disease-related rather than surgery-related. This is an important clinical issue and whether or not the decline in verbal fluency and Stroop which are commonly documented following STN-DBS are predictive of development of post-surgical dementia requires further investigation.
While the current evidence suggests that STN-DBS does not accelerate or increase the rate of cognitive decline or dementia in PD, this conclusion nevertheless requires further confirmation by future controlled studies comparing the long-term rates of dementia in operated and unoperated patients. Although the MMSE is not considered an ideal cognitive screening measure for PD [73], longitudinal data shows an average rate of decline on the MMSE of 2.3 points in PD-D and 1 point per year in non-demented PD [74, 75], which could be compared to similar data from PD patients who have had STN-DBS. The MDS Task Force on cognitive screening measures [73] alternatively recommends the use of the Mattis Dementia Rating Scale-2, the Montreal Cognitive Assessment and the Parkinson’s Disease-Cognitive Rating Scale to monitor cognition and development of PD-D; any of these cognitive measures could be used for the initial screening and long-term cognitive follow-up of DBS operated and unoperated patients in future studies.
A literature review of the factors that influence cognitive outcome of STN-DBS in PD
Empirical evidence suggests that depression, hallucinations and rapid eye movement (REM) sleep behaviour disorder are among the non-motor symptoms that predict dementia in PD [7, 76–78]. There is also evidence that among the neuropsychological tests of cognitive function, decline in verbal fluency and the Stroop Colour Word Interference Test, impaired performance on memory tests, as well as the pentagon copying on the Mini-Mental State Examination (MMSE), also predict later development of dementia in PD [65–67, 79]. Nevertheless, as noted above, the further decline on verbal fluency and the Stroop tests documented after STN-DBS do not seem to necessarily herald the onset of dementia in PD following surgery either in the short or long-term follow-up [72]. To our knowledge, with the exception of depression, the potential associations between hallucinations and REM sleep behaviour disorder and cognitive decline following STN-DBS have not been investigated.
Given that decline in verbal fluency is the most consistent and persistent cognitive impairment found in a proportion of operated patients following STN-DBS, a number of studies have investigated the correlates of this decline. A host of demographic and clinical variables have been shown not to significantly correlate with the decline in verbal fluency after STN-DBS; these include age, duration of illness, change in levodopa equivalent dose or depression after surgery, number of microelectrode passes during surgery, stimulation parameters, dysarthria, baseline severity of PD as rated on the Unified Parkinson’s Disease Rating Scale (UPDRS) [20, 80–86]. Conversely, decline in verbal fluency has been shown to be significantly associated with pre-operative scores on the Beck Depression Inventory and the MMSE [24]. Significant associations between verbal fluency and speed of processing [86, 87] and with apathy [68, 86–88] have also been reported, although other studies have failed to reproduce these findings [82, 89]. The evidence relating to the association of verbal fluency decline and change in UPDRS and side of DBS (with left hemisphere DBS being associated with greater decline) is also inconsistent [20, 84].
Technical aspects of DBS surgery can also influence the cognitive side-effects after STN-DBS. For instance, the angle of electrode entry, its trajectory and final position in the STN on verbal fluency has been examined. Electrode angles that were more superior and more posterior-lateral in the frontal quadrant of the brain were related to verbal fluency decline [90]. The length of the surgical trajectory was positively correlated with pre vs. post- surgical difference score of semantic verbal fluency [90], which may be related to microlesioning effects of the electrode trajectory [91, 92]. However, microlesions of white matter tracts such as the arcuate fasciculus, the superior and inferior longitudinal fasciculus, the frontal aslant and the fronto-striatal tracts were not associated with the verbal fluency decline six months after STN-DBS [93]. Electrode trajectories that intersected the ventricles were also associated with cognitive decline [94]. One multi-centre study reported that electrode trajectories that intersected the caudate were associated with verbal fluency decline [39], although other studies have not found this [95–97]. Electrodes positioned in or adjacent to the motor STN itself [90] or in the ventral STN [21] have been associated with verbal fluency decline after STN-DBS.
The literature is also suggestive of a number of factors that may be relevant to the post-STN-DBS cognitive outcome. However, for most of these the evidence is not consistent. Pre- operative cognitive status of the patients is clearly important, and presence of significant pre- operative cognitive deficits is associated with post-operative cognitive decline and dependence [12]. In a similar vein, greater executive dysfunction on tests such as Trail-Making B or the Stroop Colour-Word Interference task before surgery have been reported to be associated with poorer cognitive outcome after STN-DBS [32], and IQ and list learning were predictive of post-operative cognitive outcome in another study [98]. Age seems to be a critical factor with older patients more likely to show cognitive decline following surgery in a number of studies [19, 51]. Poorer response to and higher pre-operative doses of levodopa have also been associated with poorer cognitive outcome following STN-DBS [19, 99], although this could be confounded by greater disease severity associated with poorer responses and higher doses. Consistent with the association between cognition and axial motor symptoms such as freezing of gait in PD [100], greater axial involvement has been found to be associated with greater cognitive decline after STN-DBS [19, 101]. While most studies have generally failed to identify an association between stimulation parameters and cognitive test performance [44], tests of attention and immediate and delayed memory were found to correlate mainly negatively with the voltage and frequency of stimulation in one study [102]. Another notable exception is by Reich and colleagues, who showed that functional connectivity between DBS sites and other regions, such as the subiculum, are associated with post-DBS cognitive decline. They define a cognitive decline circuit that can be used as a clinical tool to predict which patients might develop cognitive side-effects after DBS or even use this tool to prevent cognitive decline in future patients by changing stimulation parameters [103]. DBS sites connected to the anterior cingulate, caudate nucleus, hippocampus, and cognitive regions of the cerebellum were more likely to be associated with cognitive decline [102].
In summary, our review of current evidence suggests that several variables predict worse cognition after STN-DBS. These include pre-operative cognitive status (particularly greater executive dysfunction and poorer memory), older age, higher pre-operative doses of levodopa, and greater axial involvement. Knowledge of these factors may aid neurologists and neuropsychologists in selection of suitable candidates for STN-DBS surgery.
DISCUSSION
There has been much interest in elucidating the predictors of cognitive decline following STN-DBS. To that end, a common approach is to complete pre- and post-surgical neuropsychological evaluation to firstly establish which elements of cognition may be altered by STN-DBS. These are then used to suggest predictors of cognitive decline following surgery [98, 104]. While valuable, this approach suffers from two main flaws: given that cognitive decline is a feature of PD progression, the absence of a non-surgical PD control group makes it difficult to dissociate post-surgical cognitive change from cognitive decline that occurs due to PD progression. Secondly, there exists a stun-effect following DBS, where patients experience substantial motor therapeutic benefit secondary to local trauma. The duration of the stun-effect is uncertain but may last anywhere between a few weeks to six months [105]. This issue becomes important as neuropsychological evaluation during this period may not be representative of the longer-term effects of DBS. The current meta-analysis addresses both of these issues by reviewing controlled studies and assessing both the short- and long-term effects of DBS.
We show that the short-term cognitive effects of STN-DBS included changes in general cognition, attention and concentration, psychomotor speed, and language, the latter of which included specific tests of phonemic and semantic verbal fluency. With the exception of verbal fluency, these effects were small. Acute on vs. off manipulation of STN-DBS may improve elements of cognition such as working memory and speed of processing (e.g. [70]), although such positive cognitive change was not a prominent feature of the group results of this meta-analysis of the short-term and review of long-term effects of pre- vs. post-surgery cognitive effects. Nevertheless, improvement of speed of processing and delayed memory recall from before to after surgery has been documented in a small percent of the operated patients in some studies (e.g., [26]). We also performed a review of the long-term effects of STN-DBS on cognitive decline and dementia. Our analysis of controlled studies up to nine years after surgery showed similar rates of cognitive decline in patients who had undergone STN-DBS and medicated-only PD patients. Hence our review shows that in carefully screened and selected patients STN-DBS is unlikely to increase the risk of cognitive decline and dementia in PD. Including studies with non-surgical PD control groups and measuring neuropsychological measures over time controlled for both practice effects and the effect of natural disease progression. However, it should be noted that cognitive decline is still seen in a proportion of patients with PD who undergo STN-DBS. Indeed, the study by Rothlind and colleagues [40] shows cognitive decline and poorer quality of life in patients undergoing surgery despite them not meeting the criteria for dementia. This study highlights that thresholds for dementia can seemingly mask an effect on cognitive decline and that there may be a subgroup of patients that may be susceptible to cognitive decline after surgery. Future studies should aim to use techniques such as meta-regression to delineate the effect of these factors on cognitive symptoms after STN-DBS.
The pathophysiology of STN-DBS side effects is multifactorial. For example, current spread to nonmotor areas of the STN is believed to underlie some of the cognitive side-effects [106]. In fact, model-based stimulation parameters, that serve to minimise this current spread but maintain activation of motor-STN, are shown to mitigate effects on verbal fluency while sustaining motor benefits [107]. Features of DBS surgery may also underlie STN-DBS side-effects: passage of the electrode through the prefrontal cortex may be related to executive dysfunction [108] whereas impaired verbal fluency is likely related to anterior lead placement in the STN [39] and stimulation frequency. For example, stimulation in the theta frequency range may improve verbal fluency compared to when stimulation is turned off or optimised for motor symptoms using gamma frequency stimulation [109, 110]. Frequency- and site-specific modulation of PD symptomatology represents a future avenue of research that is patient-centred. We noted in this meta-analysis, as have others [52, 111], that impairments in verbal fluency were the most notable side effect after STN-DBS and persisted as a long-term side effect. For example, in one study the decline in verbal fluency at 8 years follow-up was greater than at 5 years after STN-DBS; although even at 8 years follow-up, a major proportion of the patients had verbal fluency scores in the normal range in this sample [70]. However, it is notable that studies which examined the impact of the further decline in verbal fluency on QoL of STN-DBS operated patients, did not find any significant detrimental effect [14, 112]. The clinical significance of the smaller magnitude decline on the other cognitive domains remains to be investigated. Furthermore, it is reassuring that STN-DBS does not appear to exacerbate cognitive decline or acceleration to dementia that may naturally unfold as part of PD. Indeed, the rate of conversion to dementia in STN-DBS patients followed-up for up to ten years is no higher than that of the general PD population [113].
In the meta-analysis, the high I2 heterogeneity metric for some cognitive domains probably partly reflects the diverse measures used across studies as currently there is no standardized battery of neuropsychological tests for pre- vs. post-DBS cognitive testing. While it is reassuring that different tests assessing the same cognitive domain reveal the same direction of effect; nevertheless, going forward, achieving consensus about a neuropsychological battery to be employed across future studies would greatly benefit clinical practice and research and the interpretation of the results of different studies.
The current article provides clear evidence, from controlled studies, of the cognitive side-effects that may ensue following STN-DBS, both in the short- and long-term. However, the effect size of most short-term side effects (other than verbal fluency) is smaller than that considered to be clinically significant, and there is no convincing or consistent evidence of long-term cognitive decline or dementia induced by STN-DBS in appropriately screened and carefully selected patients. Greater executive dysfunction and poorer memory pre-operatively, older age, higher pre-operative doses of levodopa, and greater axial involvement are some of the factors associated with worse cognition after STN-DBS in PD. We hope the results of our meta-analysis and review will aid neuropsychologists, neurologists, and neurosurgeons in selection of suitable candidates for surgery and help them inform patients and their families about the risk of cognitive decline following surgery.
CONFLICT OF INTEREST
The authors have no conflict of interest to report.
