Deep Brain Stimulation for Multiple Sclerosis Tremor: A Meta-Analysis

Abstract

Objectives:

To examine the effect of deep brain stimulation (DBS) on multiple sclerosis (MS)-tremor, as measured by a normalized scale of tremor severity, with a meta-analysis of the published literature.

Methods:

Medline and EBSCO Host (January, 1998 to June, 2018) were systematically reviewed with librarian guidance, using the keywords “Deep brain stimulation” and “multiple sclerosis.” Bibliographies and experts in the field were also consulted to identify missed articles. All therapeutic studies on DBS for MS-tremor, reported in the English language, within the study period were included. Papers that reported outcomes without a measure of central tendency and/or distribution were excluded. The papers were read in their entirety and graded for risk of bias according to the American Academy of Neurology (AAN) standards. To maximize statistical power, papers using different stimulation targets were grouped together. Outcomes were reported with the Fahn-Tolosa-Marin scale (FTM), the Bain-Finchley scale (CRS) and 3- and 4-point tremor severity scales and normalized with a Hedges g.

Results:

The search produced 13 studies suitable for meta-analysis. The random-effects meta-analysis showed that DBS improved the Hedges standardized mean tremor score by 2.86 (95%CI 2.03–3.70, p < .00001). Heterogeneity was high, with an I² of 84%, suggesting that random effects model is more appropriate. Adverse event rates varied from 8% to 50%.

Conclusions:

This meta-analysis provides level III evidence that DBS may improve MS-related tremor as measured by standardized tremor severity scales.

INTRODUCTION

Tremor is a major component of multiple sclerosis (MS), effecting between 25% and 58% of patients with MS (1–3). Severe, disabling tremor effects 3–15% of all MS patients (1–3). Tremor in MS is can be varied but is most often a postural or kinetic type of tremor (1). Pure resting tremor and Holmes’s tremor are rarely seen in MS (1).

The pathogenesis of tremor in MS is poorly understood. This is related to the multifocal nature and varied neurologic presentation of the disease. The best evidence points to a primary role of cerebellar efferents from the dentato-rubro pathway via the superior cerebellar peduncle. The leading theory is that disruption of this white matter tract by demyelinating lesions disinhibits the thalamus, leading to the development of tremors. There is considerable ongoing research to more fully identify the pathophysiologic causes of tremor in MS (4).

Medications have had limited benefit for MS tremor in randomized clinical trials (RCTs) (4,5). Drug classes that have been evaluated in clinical trials for MS tremor include benzodiazepines, anti-convulsants, beta blockers, ethanol, cannabinoids, mood stabilizers, anti-emetics, antidementia, and antispasmodics (4,5). Of these, anticonvulsants and benzodiazepines are the most effective in improving symptoms but often leave patients with residual disabling tremor despite maximized medical treatment (4,5).

The prevalence of disabling tremor among patients with MS and the lack of effective medical options lead to the development of surgical options to treat this symptom. Deep brain stimulation (DBS) was first used in 1980 to successfully treat MS tremor (6). Since that time multiple studies have been completed to evaluate the effect of DBS on MS tremor (7–15). Many of these studies showed relatively small clinical effects and were composed of very small patient series (N = 4–16 per arm). Many of these studies also used various measurements of tremor severity (Including the Fahn-Tolosa-Marin (FTM) scale, non-validated ordinal scales, or the Bain-Finchley (CRS)), making direct comparison of outcomes difficult.

METHODS

Primary Research Question

The objective of this report was to examine the effect of DBS on MS tremor, as measured by a normalized scale of tremor severity, with a meta-analysis of the published literature that has examined this topic.

Literature Search and Article Selection

The key words “Deep brain stimulation” and “multiple sclerosis” were used to search the Medline database (https://www.ncbi.nlm.nih.gov/pubmed) and EBSCOHost (https://www.ebsco.com). Searches were limited to interventional studies. To ensure completeness, bibliographies, and experts in the field were also consulted to look for any missed articles. Titles and abstracts were reviewed to eliminate animal studies, obviously unrelated studies, and duplicates. Papers that remained were read in their entirety, and unrelated studies were removed. Papers that reported outcomes without a measure of central tendency and/or distribution were excluded because they are not able to be included in meta-analyses. After this final round of review, the remaining papers were included in the meta-analysis (Fig. 1, Table 1).

Figure 1.

Study flow diagram.

Table 1.

Summary of Studies Included in the Meta-Analysis.

Author	Year	N	Length of Follow Up (Months)	Outcome Measure	Target	Level of Evidence	Adverse Events, Number of Patients, (%)	Conclusion
Berk	2002	12	12	12 point scale	VIM	II	1 (8)	DBS improved tremor and self-feeding ability in MS tremor
Bittar	2005	10	15	CRS	VO/ZI	IV	3 (30)	DBS may be preferred for tremor control in MS because it has less morbidity than thalamotomy
Geny	1996	13	8–24	4 point scale	VIM	III	2 (15)	DBS may be useful postural tremor in MS
Herzog	2007	11	>4	FTM	VIM	IV	NR	Profound effect of DBS for MS tremor
Hooper	2002	15	12	FTM of contralateral limb	VIM/VO	IV	3 (20)	DBS significantly reduced tremor and improved function in MS tremor
Hosseini	2010	9	6	FTM	VIM	III	3 (33)	Confirms safety and efficacy of VIM DBS for MS tremor
Montgomery	1999	15	3–12	4 point scale	VC	III	2 (13)	VIM DBS provides tremor reduction in MS and is relatively safe
Oliveria	2017	11	6	FTM	VIM/VO	II	6 (55)	Dual lead DBS may be effective in reducing MS tremor
Plaha	2008	4	12	FTM	ZI	II	1 (25)	ZI may be an effective target for tremor control
Schuurman	2008	5	6	4 point scale	VIM	II	3 (50)	DBS must be further studied for relief of tremor in MS
Thevathasan	2010	16	12	CRS	VO/ZI	III	NR	MS tremor permanently improves with DBS even when limb strength is stable
Wishart	2003	4	18	3 point scale	VIM	IV	0 (0)	DBS often produces improved MS tremor
Zakaria	2013	7	12	FTM	VIM	III	2 (29)	VIM DBS may reduce severe MS tremor

LOE was based on AAN criteria. Conclusions are paraphrased from the conclusion section of the original papers’ discussion or abstract.

Abbreviations: VIM, ventral intermediate nucleus; VO, ventral oralis nucleus; ZI, zona incerta; MS, multiple sclerosis; DBS, deep brain stimulation; FTM, Fahn-Tolosa-Marin; CRS, Bain-Finchley Scale.

Data Abstraction

The papers that were included in the study were read in their entirety. Each paper was graded for risk of bias according to the American Academy of Neurology (AAN) standards and given a grade of I-IV, with I representing the lowest risk of bias and IV representing the highest (16). Level I studies are high-quality RCTs with masked allocation and blinded assessments. Level II studies are RCTs with excessive crossovers, non-masked allocation or assessments, or other risks of bias. Level II studies can also include high-quality cohort or cross-sectional studies. Level III would be RCTs, cohort, or cross-sectional studies that have some risk of bias (nonblinded assessments, crossovers, etc.) Level IV studies are interventional studies that have major sources of bias (i.e. a nonblinded treating physician responsible for assessments) or case reports/series. Tremor severity can be measured with several different validated scales and many separate scales were used across the literature (17).

To maximize statistical power, papers using different stimulation targets (ventral intermediate nucleus of the thalamus [VIM], ventral oralis nucleus of the thalamus [VO], ventral caudal nucleus of the thalamus [VC], zona incerta [ZI] were grouped together. All studies used targets within the ventral thalamus/VIM. Bilateral and unilateral implantation studies were also combined. In the case of unilateral DBS studies, only the normalized tremor severity score for the contralateral limb was included as an outcome.

Statistical Methods and Variables

Meta-analyses were done using both the fixed effects and random effects models. A Hedges g correction was used to standardize the outcomes of different scales of tremor severity (18). Review Manager 5 software was used to complete the meta-analyses (RevMan 5, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, DK). When ranges were reported but not standard deviations, the range was divided by 4 to estimate the standard deviation.

The only dependent variable examined was the normalized tremor severity after DBS compared to tremor severity at baseline. This outcome was available in all the included studies. Time points for comparison varied across studies. The time point closest to 12 months post-surgery was chosen for inclusion in the meta-analysis. Twelve-month outcomes were chosen since this outcome was most the most widely reported and was chosen to maximize studies that could be included.

Standard Protocol Approvals, Registrations, and Patient Consents

This study is a meta-analysis that did not require patient consent or IRB review.

RESULTS

The results of the search and numbers of articles eliminated at each step are summarized in Figure 1. Ninety-eight papers were initially identified in the database searches and an additional 10 found by searching bibliographies and with expert consultation. After removing papers that were unrelated to the study question, duplicate results, papers without comparisons for outcomes, and multiple papers based on the same patient cohort, the final number of studies included in the meta-analysis was 13 (7,8,10,12,19–27). The total number of patients included across all manuscripts was 261, 129 received DBS, and 132 received medical management. Seven studies used exclusively the VIM as a target, 2 studies used both the VIM and VO, 3 studies used either the ZI or a combination of VO and ZI, and 1 study used the VC. The descriptive information for the final set of included articles is summarized in Table 1.

The fixed-effects meta-analysis showed that DBS improved the Hedges standardized mean tremor score by 2.15 (95% Confidence Interval [CI] 1.84–2.46, p < 0.00001) (Fig. 2). Similar effects were seen in the random-effects model with an improvement of 2.86 (95%CI 2.03–3.70, p < 0.00001) (Fig. 3). Heterogeneity was high, with an I² of 84%, suggesting that random effects model is more appropriate.

Figure 2.

Fixed-effects model meta-analysis of the effect of DBS on MS tremor.

Figure 3.

Random-effects model meta-analysis of the effect of DBS on MS tremor.

Of the 13 studies included, the level of evidence (LOE) of the studies as measured by AAN criteria was varied. Four were level II, 5 were level III, and 4 were level IV. The major factor in the lower levels of evidence was the fact that no studies had randomized allocation and many studies had unmasked evaluation of outcomes.

The number of subjects per study was relatively low, ranging from 4 to 15 patients.

Adverse event rates varied from 8% to 50%, though definitions of adverse events, follow-up time and number of subjects varied considerably across the studies. The most common adverse event was device infection requiring reoperation (Table 1).

Outcome measures varied considerably across studies. Outcome measures used included variants of the Fahn-Tolosa-Marin scale (FTM), the Bain-Finchley scale (CRS) and 3- and 4-point tremor severity scales. Even when similar scales were used across studies, there was no standardized way of reporting them and often a subset of the scale was used rather than applying the entire scale (Table 1).

Publication bias was assessed by visual inspection of a funnel plot (Fig. 4). It demonstrated publication bias with a lack of small, negative studies in the literature. This is indicated by the lack of studies falling on the lower-left portion of the plot.

Figure 4.

Funnel plot of study data. The funnel plot shows a dearth of studies in the bottom, left-hand side of the graph. This indicates a publication bias against small, negative studies.

DISCUSSION

Meta-Analysis

Despite multiple studies completed on this topic at busy DBS centers, patient cohorts remain small. This is true even for case series and retrospective trials (7), indicating that MS tremor patients seeking DBS are relatively rare, especially compared to patients with essential tremor or Parkinson disease. This held true even for prospective studies, where patient enrollment was low. (10) This indicates that a large RCT addressing DBS for MS tremor is unlikely to yield sufficient patient numbers to definitively answer the question at hand. This fact makes the meta-analysis approach appropriate for this clinical question, since level I data is unlikely to be forthcoming.

This meta-analysis provides the best evidence to date that DBS is a safe and effective treatment of MS tremor. Because of the high heterogeneity, the random-effects model is preferred. It demonstrated that DBS will lead to an improvement of the standardized tremor severity score of 2.86. The use of the Hedges g correction (18) allows the combination of outcomes based on multiple different scales into one outcome, but the standardized mean difference can be difficult to interpret. Although this analysis shows that DBS improves tremor in MS, the exact magnitude to improvement that can be expected on a given tremor scale is difficult to predict. Going forward, it would improve the utility of studies on this topic to use an agreed upon, standardized measure of tremor severity. This would allow easier generalizability of results.

The LOE of these studies ranged from II-IV by AAN criteria. There were no RCTs of DBS versus medical management for MS tremor, so there were no level I studies. Most studies that were level III or IV, reached that level because of nonmasked assessments. This limits the LOE of the conclusion in this meta-analysis to level III. Nevertheless, pivotal studies on the effectiveness of DBS for other diseases have been non-level I evidence and are still accepted as definitive evidence of the effectiveness of DBS (28,29). Although these studies were accepted as pivotal trials, level I RCTs have been completed in certain diagnoses for DBS (30). To decrease the risk of bias and improve the LOE in future studies, efforts should be made to assess outcomes in a masked fashion.

The publication bias observed in this set of literature, based on the funnel plot (Fig. 4) shows that small, negative studies are unlikely to be published. This is a standard pattern of publication bias seen in many different areas. It is potentially worrisome, however, because the studies that have been conducted are all small, and, as mentioned above, large, prospective trials on this topic are unlikely to be produced. Thus, the positive outcome reported in this meta-analysis should be interpreted with an appropriate amount of caution.

The adverse event rate varied among the studies included in the analysis from 8 to 55%. This is similar than reported in trials for other diagnoses (28–30). This shows that patients with MS and tremor can tolerate DBS surgery and DBS with a similar safety profile to patients with other movement disorders. Further, many of the studies included patient cohorts over a long period of time, when stereotactic techniques and other surgical innovations now available, were not possible. It is possible that modern complication rates are lower.

There are several important limitations to this work. First, the meta-analysis technique provides evidence only as strong as the studies that are included in the analysis. As reported above and summarized in Table 1, the studies included in this meta-analysis were level II-IV studies with a small number of subjects. Also, based on funnel plot analysis, there is a publication bias present against small, negative studies. Since the positive studies are also small, this may indicate that the effect reported here is an overestimate. Because of the quality of the evidence that is available, the conclusions drawn must be tempered with the knowledge of the bias inherent in the studies that make up the meta-analysis. Even so, this report represents the best available evidence on the efficacy of DBS for MS tremor. Another limitation is the use of the Hedges g correction. While this makes different outcome scales that measure the same phenomenon comparable, it does so by normalizing the outcome scales to a standardized mean difference. This has the drawback of making the summary statistic of the meta-analysis difficult to interpret when it comes to predicting the magnitude of the clinical benefit offered by the intervention. Although it is the best mathematical solution to the problem and adds to our understanding, a better and more precise estimate of the effect of DBS for MS tremor could be obtained if researchers agreed upon a standardized scale for the measurement of tremor severity in this population. Finally, the effects of DBS on patient outcomes in the disease could be improved by better patient selection. MS and lesions associated with tremor are varied. Other symptoms of MS (ataxia, anxiety) can also exacerbate tremor in ways independent of the pathway treated by DBS (4). Futures studies will need to be done focused on selecting the patients with MS tremor who are most likely to benefit from surgery.

CONCLUSIONS

This meta-analysis provides level III evidence that DBS may improve MS-related tremor as measured by standardized tremor severity scales. Adverse events in this population are similar to those reported for other indications of DBS. In patients with severe, debilitating tremor related to MS, DBS should be considered as an effective treatment option based on the best available evidence.

COMMENT.

In this report, Brandmeir, et al have taken a unique approach to a difficult problem. As with other indications for deep brain stimulation, there have been many small studies looking at its utility, in this case for the severe debilitating tremors often experienced by multiple sclerosis patients. This meta-analysis demonstrates that the previous work does, in fact, provide good evidence to suggest that stimulation confers a significant clinical benefit to these patients and should be considered a reasonable treatment option for them. It also demonstrates the difficulty of gathering this sort of data. Given the relatively small number of patients that even busy stimulation centers see, a registry based study may be a better way to gather evidence for this indication than a traditional randomized trial.

James McInerney, MD

Hershey, PA, USA

Comments not included in the Early View version of this paper.