Article Text


Research paper
Comparative effectiveness of neuroablation and deep brain stimulation for treatment-resistant obsessive-compulsive disorder: a meta-analytic study
  1. Kevin K Kumar1,
  2. Geoffrey Appelboom1,
  3. Layton Lamsam1,
  4. Arthur L Caplan2,
  5. Nolan R Williams3,
  6. Mahendra T Bhati1,3,
  7. Sherman C Stein4,
  8. Casey H Halpern1
  1. 1Department of Neurosurgery, Stanford University, Stanford, California, USA
  2. 2Department of Population Health, Division of Medical Ethics, New York University, New York City, New York, USA
  3. 3Department of Psychiatry, Stanford University, Stanford, California, USA
  4. 4Department of Neurosurgery, University of Pennsylvania, Philadelphia, Pennsylvania, USA
  1. Correspondence to Dr Casey H Halpern, Department of Neurosurgery, Stanford University, Stanford, CA 94305-5327, USA; chalpern{at}


Background The safety and efficacy of neuroablation (ABL) and deep brain stimulation (DBS) for treatment refractory obsessive-compulsive disorder (OCD) has not been examined. This study sought to generate a definitive comparative effectiveness model of these therapies.

Methods A EMBASE/PubMed search of English-language, peer-reviewed articles reporting ABL and DBS for OCD was performed in January 2018. Change in quality of life (QOL) was quantified based on the Yale-Brown Obsessive Compulsive Scale (Y-BOCS) and the impact of complications on QOL was assessed. Mean response of Y-BOCS was determined using random-effects, inverse-variance weighted meta-analysis of observational data.

Findings Across 56 studies, totalling 681 cases (367 ABL; 314 DBS), ABL exhibited greater overall utility than DBS. Pooled ability to reduce Y-BOCS scores was 50.4% (±22.7%) for ABL and was 40.9% (±13.7%) for DBS. Meta-regression revealed no significant change in per cent improvement in Y-BOCS scores over the length of follow-up for either ABL or DBS. Adverse events occurred in 43.6% (±4.2%) of ABL cases and 64.6% (±4.1%) of DBS cases (p<0.001). Complications reduced ABL utility by 72.6% (±4.0%) and DBS utility by 71.7% (±4.3%). ABL utility (0.189±0.03) was superior to DBS (0.167±0.04) (p<0.001).

Interpretation Overall, ABL utility was greater than DBS, with ABL showing a greater per cent improvement in Y-BOCS than DBS. These findings help guide success thresholds in future clinical trials for treatment refractory OCD.

Statistics from


Obsessive-compulsive disorder (OCD) has an overall lifetime rate of 2%–3% in the general population. This neuropsychiatric disorder most commonly presents in late adolescence to early adulthood and has tremendous impact on an individual’s social and occupational functioning. Not surprisingly, severe OCD has a well-described mortality risk due to suicide, behooving rigorous investigation to identify the most suitable intervention for these vulnerable patients.1 2 The most effective treatment for reducing the symptoms of OCD is Exposure and Response Prevention, which may also act synergistically with pharmacological and other cognitive interventions.3 To date, pharmacological alternatives largely consist of selective serotonin reuptake inhibitors and antipsychotics.4 Other psychological modalities can be efficacious as well.5 In fact, recent meta-analyses have revealed the combination of psychotherapy and pharmacological treatment to be more effective than other combined treatments.6 However, while many respond well to these treatments, at least 10% of patients suffer from treatment-refractory disease.7–9 Thus, there has been a concerted effort towards the development of procedural interventions for OCD as treatments of last resort.

Neuroablation (ABL) techniques, such as anterior capsulotomy using radiofrequency ablation (RF) and stereotactic radiosurgery (SRS), have an extensive history in the treatment of OCD. Several studies have strongly suggested that these lesion procedures are safe and highly effective. Alternatively, deep brain stimulation (DBS) is intended to modulate frontal–basal ganglia–thalamic circuits reported to be dysfunctional in OCD.10 The advent of DBS as a therapy for advanced movement disorders has generated interest in novel indications such as OCD.11 While several DBS targets are under study and have exhibited promise, including the anterior limb of the internal capsule,12 the nucleus accumbens,10 13 14 subthalamic nucleus13 15 and the bed nucleus of the stria terminalis,16 there have been no reported, demonstrable differences between targets.17 Importantly, several established DBS targets are in close proximity anatomically and thus stimulation to one area has potential off-target effects.18–20 There is no consensus regarding the optimal target or stimulation protocol. ABL has theoretically fewer associated hardware obstacles and risks common to DBS procedures; these can include system maintenance requirements, device malfunctions, generator replacements and patient infections. Nevertheless, DBS has several significant advantages as a procedure. DBS is non-lesional, has some early reversibility and is adjustable, generating much interest in the field for this therapy.

A recent systematic review found comparable outcomes from both ABL and DBS for OCD.21 The purpose of this meta-analytic review is to provide a more definitive comparison of pooled ABL outcomes with regard to both safety and utility over time. Specifically, we sought to evaluate the relative efficacy of ABL and DBS as measured by per cent decrease in Yale-Brown Obsessive Compulsive Scale (Y-BOCS), complication rates and calculated units of utility as reported in the literature.


We created a decision analytical model to estimate and compare utility and complication rates of two procedural strategies for treatment-refractory OCD, namely ABL and DBS.22–24 The model projects the change in utility after surgery. The possible pathways and outcomes of the two procedures are illustrated (figure 1). Data for the model were derived from a critical review of published English language reports.

Figure 1

Decision tree, with possible outcomes following a choice of DBS or ABL for surgery of treatment-refractory OCD. ABL, neuroablation; DBS, deep brain stimulation; OCD, obsessive-compulsive disorder.

Our base case was a patient with treatment-refractory OCD of average age and disease duration based on pooled data, who was considered a suitable subject for ABL or DBS. Outcome parameters of interest were per cent improvement in the Y-BOCS score from baseline and the complications and side effects of treatment.

Search strategy and selection criteria

Data collection

We performed a EMBASE and PubMed search of articles reporting the results of ABL and DBS for OCD. The search, which was performed in January 2018, included the terms ‘obsessive-compulsive disorder’ as a subject heading or title, combined with one of the following: ‘surgery’ (subheading), ‘stimulation’ (title), ‘deep-brain stimulation (title or text) or any of the following in the title: lesion, radiofrequency, thermal, stereotaxic, stereotactic, radiosurgery, gamma or –otomy. Each article was reviewed by at least two authors. We excluded the following: non-English language publications, case reports, reviews without original data, republication of previously reported data, animal studies, technical reports and other studies lacking original clinical data. The search was supplemented by employing the ‘Related citations’ feature of PubMed and reviewing the bibliographies of review articles and selected publications. Two recent reviews were consulted to help eliminate duplicated data.17 25

Data extracted from each article included, where available: numbers of cases, study type (controlled trial, prospective and so on), intracranial target, side(s) operated, number of operated cases, number of cases with complete follow-up, demographics, postoperative complications and side effects (type and number), mean follow-up duration and mean preoperative, postoperative or per cent change in Y-BOCS scores at the longest follow-up period provided (online supplementary table S1). Data were extracted from each article to minimise the impact of individual study type on the meta-analysis. ABL targets (RF and SRS) fell into two groups: capsulotomy and cingulotomy. To simplify analysis, we also divided the DBS targets into two groups: striatal (caudate nucleus, internal capsule, nucleus accumbens with or without subthalamic nucleus, medial forebrain bundle, nucleus of the stria terminalis) and thalamic (various thalamic sites, subthalamic nucleus). Given the small numbers of reported cases, comparisons among more groups were not statistically robust. Many reports did not mention complications; therefore, we calculated complication incidences only from reports documenting complications or those stating none occurred.

Quality of life correlations

Almost all reports of DBS treatment employ the Y-BOCS (an OCD symptom severity scale) as their primary effectiveness parameter. As comparative effectiveness studies require utility or a similar measure of quality of life (QOL), we performed an independent EMBASE and PubMed search of English language literature (January 1999 to January 2018) for studies reporting both ‘obsessive-compulsive disorder’ and ‘quality of life’ as subject headings (online supplementary table S2). We downloaded and reviewed only publications reporting serial measurements of both Y-BOCS and a QOL metric, such as 36-Item Short Form Health Survey or WHO Quality of Life-BREF. This allowed us to quantify change in QOL as a function of change in Y-BOCS (figure 2). A separate literature search was made for the impact of each complication on QOL. We omitted all transient complications and minor sensations related to DBS leads and considered only those, which impacted patient well-being during the follow-up period.

Figure 2

Meta-regression of utility gained as a function of per cent decrease in Y-BOCS score. This was plotted from studies reporting pretreatment and post-treatment Y-BOCS scores and simultaneous utility scores. The black line represents the pooled means, and the grey area is the 95% CI. Y-BOCS, Yale-Brown Obsessive Compulsive Scale.

The per cent change in utility was calculated from the change in Y-BOCS based on a formula determined from meta-regression analysis (figure 2). From this meta-regression, we determined that per cent utility change after surgery equals 0.13 plus 0.167 times per cent change in Y-BOCS score.

Data analysis

Probabilities and effectiveness

The probability of a patient with an average OCD progressing towards either branch of the decision tree was derived from the literature (figure 1). Pooled values of complication types and rates were calculated. Outcomes were expected to be indirectly related to mean improvement in the Y-BOCS scores, which were pooled from published results. These were converted to mean improvement in QOL, using utility as the measure of QOL. In this context, ‘utility’ is understood to be a measure of relative patient preference for a given healthcare outcome.26 Utility is a parametric measure, with values ranging between 0 (dead) and 1 (perfect health). Each QOL score was converted to utility using published algorithms or linear extrapolation.24 27 Utility can be combined with measures of time in a particular health state to calculate both quality and quantity of life. The quality-adjusted life year (QALY) combines utility with duration in a particular health state. For example, one year in perfect health equals one QALY, a year in which utility equals 0.8 would result in 0.8 QALYs.

Data management

The mean response of Y-BOCS to ABL or DBS was determined using random-effects, inverse-variance weighted meta-analysis of observational data. The same method was used to analyse other variables including mean age, sex distribution, duration of symptoms and incidences of individual complications.28 Per cent improvement of Y-BOCS was converted to mean improved utility with the aid of meta-regression of studies measuring both changes in Y-BOCS and QOL.29 Meta-regression was also used to assess durability of ABL and DBS. We constructed a subtree of incidence of each reported complication and its effect on utility to calculate the utility of the average patient suffering from postoperative complications and side effects. This study adhered to the reporting guidelines set forth by the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group.30 Calculations of model outputs employed beta distributions of probabilities and utilities. Pooled values for per cent improvement in Y-BOCS scores after surgery were analysed for heterogeneity (χ²) and by funnel plots and Begg’s test for publication bias.


The primary analysis was the comparison of the expected utilities of ABL and DBS. Sensitivity analysis consisted of Monte Carlo simulations (100 simulated trials of 100 subjects each, following standard procedure).31 Secondary analyses included comparative effectiveness of different neural targets, SRS versus RF lesions and unilateral versus bilateral DBS. Pooled values of demographic features and other possible confounders were also compared between ABL and DBS. Statistical comparisons employed a two-tailed Student’s t-test. Differences whose probabilities were less than 5% were considered significant and when present are indicated in the Results section. Meta-analytical pooling, meta-regression and statistical analysis were performed using Stata software (Stata Statistical Software: Release 12, Stata, College Station, Texas, USA). The decision analysis model employed TreeAge Pro 2016 Software (TreeAge Software., Williamstown, Massachusetts, USA).

Role of the funding source

All funding organisations and sponsors had no role or influence in the design and conduct of the study or collection, management, analysis and interpretation of the data. Funding sources and sponsors were not involved in the preparation, review, approval of the manuscript or decision to submit the manuscript for publication. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.


The literature search returned 543 abstracts (figure 3). Fifty-six reports (38 DBS, 18 ABL) were determined suitable for analysis. These reports documented surgical results in 681 cases (367 ABL, 314 DBS) (online supplementary table S1). A mix of study types was observed and ranged from randomised, controlled trials to retrospective studies. It is notable that all ABL (including SRS and RF) lesions were bilateral, as were most DBS cases. Age at onset of OCD, sex distribution and preoperative Y-BOCS scores were not significantly different between the two groups (online supplementary table S3). Average age at surgery was significantly younger and the average length of postoperative follow-up was longer in ABL than DBS, respectively. Treatment complications were reported in 13 of ABL and 18 of DBS studies identified in the literature search.

Figure 3

Summary of literature search, articles reviewed and used in the analysis. Y-BOCS, Yale-Brown Obsessive Compulsive Scale.

Pooled abilities to reduce Y-BOCS scores were 50.4% (±22.7%) for ABL and 40.9% (±13.7%) for DBS. These values mapped onto 0.215 (±0.026) and 0.199 (±0.025) units of utility gained by surgery, respectively. Meta-regression of per cent improvement in Y-BOCS scores over time showed no statistically significant change over the length of follow-up for either ABL or DBS (p=0.585 and 0.848, respectively). These results are shown in online supplementary figures S1-S4. Meta-analyses of Y-BOCS changes for both ABL and DBS revealed significant heterogeneity (p<0.001 in both cases) but no evidence of publication bias (continuity corrected p=0.488 for ABL, 0.077 for DBS). Funnel plots illustrating the degree of publication bias were also generated (online supplementary figures S5-S6) to visualise this finding.

Individual surgical complications and side-effects are shown (online supplementary table S4). Listed are the incidences of each adverse effect for ABL and DBS, along with the impact of this adverse effect on utility. Adverse events, including complications and side effects, occurred in 43.6% (±4.2%) of ABL cases and 64.6% (±4.1%) of DBS cases, a difference which was statistically significant (p<0.001). The presence of complications reduced the utility of ABL by 72.6% (±4.0%) and reduced that of DBS by 71.7% (±4.3%).

The primary analysis compared the expected utility (as tempered by complications) of treatment with ABL to that of DBS, demonstrating that utility provided by ABL (0.189±0.03) is superior to that of DBS (0.167±0.04). This difference in utility is statistically significant (p<0.001). This result suggests that the patient with average ABL can expect to gain 0.022 QALYs at the end of the first postoperative year and 0.11 QALYs after 5 years.

The secondary analyses compared subgroups and possible confounds. These are summarised in online supplementary table S5. The only significant difference noted was a lower effectiveness of unilateral compared with bilateral DBS (p=0.04).


This is the first systematic meta-analytic study comparing outcomes of ABL versus DBS for OCD. Our analysis demonstrates that ABL is superior to DBS in cumulative overall effect on utility with time after surgery, and this effect may at least in part be due to ABL’s lower complication rate. This finding may be surprising as theoretical advantages of DBS includes its reversibility and adjustability. These advantages may bias surgeons towards DBS, though the follow-up required with patients for DBS programming can be a major disadvantage compared with ABL. Ethically, the adjustability of DBS supports its use because symptoms of OCD can fluctuate over time, whereas ABL is lesional. Moreover, in the case of a non-responder, DBS leads and the connected generator can be removed or the device can be turned off, whereas ABL is permanent. However, in its current form, the benefit DBS provides over a permanent procedure does not appear to positively affect its outcome profile. ABL techniques have been refined over time and can be targeted focally, mitigating the risk of this approach. ABL may also be favourable in scenarios where transportation to the clinic for programming is challenging or candidacy for surgical implantation of a device is in question. Our investigation has revealed that the complication rate of ABL is significantly lower than DBS. Our analysis of the published data suggests that patients, surgeons and referring physicians ought to be informed about the risk profile and therapeutic potential of both treatment options. Moreover, clinicians need to be aware that patient choices may shift as advances in both techniques continue to be made.

As techniques for ABL have evolved, the technology surrounding DBS has also seen improvement in targeting accuracy and programming strategies.32 33 In particular, biomarker-guided stimulation delivery (ie, adaptive DBS) may provide a more physiologically-based, targeted approach given the episodic nature of this brain disorder.7 34 There have been recent efforts to identify measurable transdiagnostic biomarkers using fMRI, electroencephalography, magnetoenceophalography or behavioural tasks to help refine patient selection for DBS.14 Moreover, standardisation of target selection and programming settings, in a fashion similar to those used for treating Parkinson’s Disease with DBS, may accelerate detection of clinical improvement in patients undergoing DBS for OCD.32 However, there remain circumstances where DBS may be an appropriate therapeutic alternative, based on patient desire for a reversible intervention or a surgeon’s technical preference. It is important to note that therapeutic strategies such as Exposure and Response Prevention in combination with medical therapy is the current preferred primary intervention for OCD. As a result, procedural management should only be offered to patients deemed refractory to these therapies, and our analyses reflect outcomes for those patients with particularly severe intractable forms of OCD. For these patients, recent evidence suggests that DBS in combination with cognitive behavioural therapy may be optimal for improving OCD symptoms.35

Our understanding of the complex interconnection between targets for ABL and DBS continues to improve, with potential implications for advancing current clinical therapies. There has been extensive work describing cortico–striato–thalamo–cortical circuitry thought to underlie the pathogenesis of OCD. Given published data demonstrating aberrant functional connectivity between prefrontal and striatal regions in OCD, further understanding of these networks through neuroimaging and optogenetic studies may help direct future neuromodulatory strategies.36 The challenge remains in discerning the impact of concomitant pharmacological, behavioural and surgical therapies when interpreting this data.3 As a result, future clinical trials directly comparing ABL and DBS with this supporting data would be of tremendous value. Another question remains as to the potential value of a combined procedural approach, which is beyond the scope of this study.

While this approach sought to provide a comprehensive meta-analysis of ABL and DBS for OCD, the analysis was inherently limited by the quality of pooled data and by the assumptions on which they relied. The data were derived from multiple independent studies, which were subject to varying inclusion criteria; we found they lacked common standards in targeting, stimulation parameters, prior treatment and population heterogeneity. Future studies should report detailed inclusion criteria, including previous medications and psychotherapy prior to surgery and provide loss to follow-up data as these factors may significantly impact outcome metrics. Compared with ABL, the average older age at surgery for patients with DBS could have limited the therapeutic efficacy of this technique. Furthermore, while subtle differences in efficacy of ABL and DBS targeting different brain regions may have been undetected by our analysis, this may be secondary to common circuit-based modulation and limited sample sizes for each target subgroup in the dataset. Indeed, a lack of differences in efficacy has been observed with currently approved neural targets for Parkinson’s disease.37 An additional limitation of this study is that decision analyses cannot be used to calculate certain metrics such as effect size and number needed to treat. As a result, this study cannot replace the findings of a well-conducted randomised controlled trial. While the comparison of ABL and DBS studies containing the same biases sought to minimise the impact of placebo effects on our analysis, differences in reporting may still exist between techniques. In order to generate our model, we made several baseline assumptions, defined previously, that may have impacted our results. For example, the relatively low complication rate reported in published ABL series, especially lesion-induced neurobehavioral changes that can be challenging to assess, may be a potential source of reporting bias. An advantage of ABL to DBS is less required neurosurgical follow-up; however, this provides less opportunity for documenting adverse effects of the therapy. Additionally, long-term clinic follow-up for DBS reprogramming and periodic impulse generator replacements may burden patients, decreasing the utility of this approach. There are efforts to limit this burden, such as more durable surgical solutions including use of rechargeable implantable pulse generators.38 The duration of non-transient complications was not reported. Although some complications (eg, urinary incontinence) may resolve after several months, our model assumes they persist for the entire follow-up period. We limited our search to the English language literature, which may have missed relevant reports published in international journals in other languages.

In summary, the burgeoning area of neurosurgical intervention for OCD is a rapidly evolving discipline, and one where a choice of ABL versus DBS offer patients differing therapeutic and complication profiles. As these techniques become more widespread in practice, subsequent clinical studies should ideally directly compare these alternative therapies to guide future treatment and strategies and inform patient decision-making.


We thank Christine Plant for her invaluable assistance and critical review in preparing the final draft of the manuscript.


View Abstract


  • Contributors KKK, GA, SCS and CHH had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. KKK, GA, MTB, NRW, SCS and CHH were responsible for study concept and design. KKK, GA, LLBS, MTB, NRW, SCS and CHH were responsible for the acquisition, analysis and interpretation of data. KKK, GA, MTB, NRW, SCS and CH drafted the manuscript. KKK, GA, ALC, MTB, NRW, SCS and CHH critically revised the manuscript for important intellectual content. KKK, GA, SCS and CHH were responsible for the statistical analysis. CHH obtained the funding and was responsible for study supervision. CHH, SCS and KKK had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

  • Funding This study was supported by funds from the John A. Blume Foundation, the William Randolph Hearst Foundation and start-up funds from Stanford University’s Department of Neurosurgery.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.