Lesion location and outcome following cingulotomy for obsessive-compulsive disorder

In Brief

In patients with obsessive-compulsive disorder (OCD), the authors evaluated the relation of the neuroanatomical location of cingulotomy treatment to symptomatic improvement to determine whether the lesion placement sites during treatment are related to responder status. Using an anatomical registration matrix, the authors found that cingulotomy lesions that were placed more superiorly and posteriorly within Brodmann area 32 conferred a higher likelihood of response, a finding that is important because the precise underlying neuroanatomical basis for the beneficial effects of this treatment in OCD patients has thus far been poorly understood.

Abstract

OBJECTIVE

Obsessive-compulsive disorder (OCD) is among the most debilitating and medically refractory psychiatric disorders. While cingulotomy is an anatomically targeted neurosurgical treatment that has shown significant promise in treating OCD-related symptoms, the precise underlying neuroanatomical basis for its beneficial effects has remained poorly understood. Therefore, the authors sought to determine whether lesion location is related to responder status following cingulotomy.

METHODS

The authors reviewed the records of 18 patients who had undergone cingulotomy. Responders were defined as patients who had at least a 35% improvement in the Yale-Brown Obsessive Compulsive Scale (YBOCS) score. The authors traced the lesion sites on T1-weighted MRI scans and used an anatomical registration matrix generated by the imaging software FreeSurfer to superimpose these lesions onto a template brain. Lesion placement was compared between responders and nonresponders. The placement of lesions relative to various anatomical regions was also compared.

RESULTS

A decrease in postoperative YBOCS score was significantly correlated with more superiorly placed lesions (decrease −0.52, p = 0.0012). While all lesions were centered within 6 mm of the cingulate sulcus, responder lesions were placed more superiorly and posteriorly along the cingulate sulcus (1-way ANOVA, p = 0.003). The proportions of the cingulum bundle, cingulate gyrus, and paracingulate cortex affected by the lesions were the same between responders and nonresponders. However, all responders had lesions covering a larger subregion of Brodmann area (BA) 32. In particular, responder lesions covered a significantly greater proportion of the posterior BA32 (1-way ANOVA, p = 0.0064).

CONCLUSIONS

Lesions in patients responsive to cingulotomy tended to be located more superiorly and posteriorly and share greater coverage of a posterior subregion of BA32 than lesions in patients not responsive to this treatment.

Obsessive-compulsive disorder (OCD) is a psychiatric disorder characterized by persistent intrusive recurring thoughts that lead to compulsive behaviors. While treatment with medications and cognitive behavioral therapy is effective in many patients, up to 10%–20% of patients are refractory to these treatments.¹ For patients with severe, persistent, medically refractory OCD, anterior cingulotomy can provide symptomatic relief.^2,3

Anterior cingulotomy is performed by stereotactically creating thermoablative lesions in the anterior dorsal cingulum bundle (CB) and cingulate cortex (CC).^2,4 Long-term response rates following cingulotomy in patients with severe refractory OCD are reported at 35%–70%.^3,5–8 The mechanism by which cingulotomy improves OCD symptoms and what differentiates patients who respond to cingulotomy from those who do not remain poorly understood.

In this study, we investigated whether lesion location is related to clinical response following cingulotomy. We also studied the relationship between the degree of involvement of different anatomical structures by cingulotomy and response. This analysis is a step toward better understanding the circuit components affected by cingulotomy and refining cingulotomy surgery to make this treatment more effective for patients with intractable OCD.

Methods

Patients and Surgery

This retrospective study was approved by the Massachusetts General Hospital (MGH) Institutional Review Board. Patients were referred to MGH and thoroughly evaluated by a multidisciplinary team consisting of psychiatrists, neurologists, and neurosurgeons. Surgical candidates were required to have met Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria for OCD (DSM-III, DSM-III-R, or DSM-IV), to have demonstrated severe functional impairment due to symptoms related to OCD, and to have experienced failure of psychopharmacological and behavioral treatments, including 10-week trials of a minimum of 3 serotonin reuptake inhibitors (including clomipramine), 2 trials with augmentation agents such as clonazepam and lithium, and a minimum of 20 hours of behavioral therapy (including exposure and response prevention).

We reviewed the records of patients who underwent triple-lesion cingulotomy for OCD at MGH between 1999 and 2019 in whom both postoperative MRI and pre- and postoperative Yale-Brown Obsessive Compulsive Scale (YBOCS) scores were available (n = 23 patients). Informed consent was obtained from all patients. Inclusion criteria included availability of pre- and postoperative YBOCS scores and postoperative MRI studies. Patients with postoperative imaging of unacceptably poor quality such that our Montreal Neurological Institute (MNI) coordinate registration pipeline failed were excluded (5 patients excluded; n = 18 remaining patients, 7 female). We chose to focus on patients with triple-lesion cingulotomy, as this became the treatment of choice at our institution in 2000⁶ and was also the surgery performed in the majority of patients with available postoperative imaging and behavioral follow-up. The median age at the time of surgery was 42.7 years (SD 18.3 years).

Procedures were performed by three neurosurgeons who used the same technique. Cingulotomy surgery consisted of stereotactic bilateral creation of thermoablative lesions made by heating a probe with a 10-mm exposed tip (Cosman Medical). Three lesions were placed on each side of the anterior cingulate gyrus (6 total lesions). An intraoperative microelectrode recording was used to verify the placement of lesions in the cingulate gyrus. The initial lesions on each side were placed 20 mm posterior to the anterior extent of the frontal horn of the lateral ventricles, 5–7 mm off the midline, and 5 mm above the corpus callosum. Two subsequent lesions were created bilaterally by incrementally advancing the target point 7 mm anterior and 1.5 mm inferior to the previous lesion.

Lesion Tracing

We manually recorded all the points contained by the cingulotomy lesions and saved them as regions of interest (ROIs). Lesions were traced on all slices using the ROI tool in FreeSurfer. We analyzed postoperative MR images taken within 3 days after surgery. We also included all T1-hyperintense points, which were surrounded by a rim of T1-weighted hypointensity. With these data we created full computerized reconstructions of each brain using the FreeSurfer software and automatically registered the lesion coordinates in standardized MNI152 space using the FreeSurfer command mni152reg. For several brains with poorer-quality scans (e.g., movement distortion, truncated images), we traced and transformed the ventricles of the brain to confirm appropriate registration. Among the 23 patient brains, inaccurate registration was found in the brains of 5 patients, who were removed from the study for a total of 18 remaining patients.

Lesion Analysis

Geometrical Variables

To compute the centroid of each lesion, we averaged all points in the lesion in each axis. To compute the volume of each lesion, we used the MATLAB boundary function.

Probability Density Maps

We first computed a 3D histogram describing the number of patients whose lesions overlapped with each point. For instance, the MNI point (−8, 22, 30) could be covered by 8 different responders’ lesions. This 3D histogram was generated based on the resolution of the lowest-resolution MRI scan (2-mm slices), such that more points were not counted for higher-quality images. For this reason, and for display purposes, scattered linear interpolation was used to compute values for the missing points (every other point, as the standardized images shown in this article have 1-mm resolution). All values were then divided by the number of patients to create a probability map. The MATLAB function imfuse with no scaling was used to produce the composite images shown here.

Percentage of Anatomical Regions Covered

We used the Harvard-Oxford atlas available in FSLeyes for the CC and paracingulate cortex (PC), the Johns Hopkins White Matter Tractography atlas for the CB, and the Brodmann atlas available in MRIcron. For the first two of these atlases, we summed the probability values assigned to each point in the atlas that overlapped with a point that corresponded to the lesion. We divided this total by the sum of probability values assigned to each point in the atlas for the entire anatomical structure.

Defining an Axis Along the Cingulate Sulcus

To define an axis along the cingulate sulcus (CS) along which to measure lesion placement, we computed a best-fit line based on the BA24/BA32 boundary (the CS), compressed along the MNI x-dimension. The distance of a centroid along this axis was calculated based on the point along the axis with the smallest euclidean distance to the centroid.

Statistics

All ANOVA results reported in the text were computed to confirm normalcy with the chi-square goodness-of-fit test. In addition, we confirmed equal variance between groups of data with a 2-sample F-test for equal variances.

Code Availability

All MATLAB codes used to conduct analyses reported in the Results section are available upon reasonable request.

Results

Patients and Clinical Response

We reviewed the records of patients who underwent triple-lesion cingulotomy for OCD at MGH between 1999 and 2019, in whom both postoperative MRI studies and pre- and postoperative YBOCS scores were available (n = 23 patients). All of these patients underwent placement of thermoablatic lesions using the same set of coordinates (see the Methods section). Patients with postoperative imaging of unacceptably poor quality such that our MNI coordinate registration pipeline failed (see Methods) were excluded (5 patients excluded; n = 18 patients remaining). All recorded YBOCS scores for the 18 patients are shown in Fig. 1A. The median YBOCS score prior to surgery was 35 (SD 4.37). The first analyzed follow-up occurred at least 6 months postoperatively (Fig. 1B; median 8.43 months after surgery). The median YBOCS score at this time point was 22.28 (SD 9.25). This score was significantly improved from the preoperative YBOCS score [1-way ANOVA, F(1,34) = 27.86, p = 7.47 × 10⁻⁶]. The median YBOCS score at the last available follow-up (median 12.08 months after surgery) was 22, which was also significantly improved from the preoperative YBOCS score [1-way ANOVA, F(1,34) = 17.11, p = 2.0 × 10⁻⁴]. The Beck Depression Inventory (BDI) score was also collected in 14 of the 18 patients. There was no significant correlation between preoperative BDI and postoperative change in YBOCS, using either the 6-month or latest possible follow-up (linear correlation, BDI with change in 6-month YBOCS score: −0.30, p = 0.23; BDI with change in latest follow-up YBOCS score: ρ = −0.23, p = 0.37). In addition, we did not detect differences in postoperative BDI [t-test, t(13) = −0.73, p = 0.48].

FIG. 1.

YBOCS scores of all study patients. A: All available YBOCS scores plotted over time. The triple cingulotomy time point is marked by a dashed black line, and the 6-month postcingulotomy time point is marked with a solid black line. For a patient to be categorized as a responder, the first available YBOCS score at least 6 months after cingulotomy had to be decreased by at least 35% from the baseline YBOCS score, and this 35% decrease also had to be maintained for the most recently recorded YBOC score. For a patient to be categorized as a nonresponder, the first available YBOCS score at least 6 months after cingulotomy had to be decreased by less than 25% from the baseline score, or the most recent follow-up YBOCS score had to have decreased less than 25% from the baseline score. Both of the patients who were categorized as nonresponders on the basis of reduced YBOCS scores at the latest follow-up underwent a second operation prior to these improved scores; therefore, these improved scores were not included in the analyses shown in the present work. B: YBOCS scores grouped by follow-up period. “Baseline” refers to the preoperative score, “Earliest >6mo” refers to the earliest follow-up that occurred at least 6 months following surgery, and “Latest” refers to the last available follow-up. Black dots with whiskers show median ± 1 SE for all patients; *p = 0.00002 at latest available follow-up. Figure is available in color online only.

Patients were categorized as responders or nonresponders (Fig. 1). Responders were defined as patients with a YBOCS score decrease of 35% or greater from baseline at the first follow-up at least 6 months postcingulotomy, and a maintained YBOCS decrease of 35% or greater from baseline at the latest follow-up assessment recorded. Nonresponders were defined as patients with a YBOCS score decrease of 25% or less from baseline at the first follow-up at least 6 months postcingulotomy, or a YBOCS score decrease of 25% or less from baseline at the latest follow-up assessment recorded. Nine patients were categorized as responders, and 8 patients were categorized as nonresponders. One patient was a partial responder, with a YBOCS score improvement of 34% from baseline at the 6-month follow-up and an improvement of 44% from baseline at the 12-month follow-up; this patient was categorized as a responder to simplify our analysis. We next sought to determine whether lesion placement differed in responders versus nonresponders.

Lesion Placement and Responder Status

We manually traced the cingulotomy lesions and then registered the lesion coordinates in standardized space (see Methods and Fig. 2). The centroids of the majority of lesions were centered within the rostral/middorsal CB and the dorsal CC (Fig. 3C). In MNI coordinates (in mm, x-axis, anterior/posterior; y-axis, medial/lateral; z-axis, superior/inferior), the average centroid for the left lesion was (−8.3 mm, 22.9 mm, 29.5 mm) and the average centroid for the right lesion was (8.5 mm, 22.0 mm, 29.2 mm). We asked whether the centroid of the cingulotomy lesions differed between responders and nonresponders. Grossly, MNI coordinates for the average left lesion centroid were (−8.1 mm, 24.5 mm, 28.5 mm) in nonresponders and (−8.5 mm, 21.6 mm, 30.2 mm) in responders. The average MNI right lesion centroid was (8.3 mm, 24.2 mm, 27.6 mm) in nonresponders and (8.7 mm, 20.2 mm, 30.5 mm) in responders. The absolute difference in euclidean distance between the centroids of responder and nonresponder lesions was 3.4 mm for the left lesion and 4.9 mm for the right lesion. We found that the centroids of lesions tended to be located in significantly more superior locations in responders versus nonresponders [1-way ANOVA, MNI z-coordinate: F(1,34) = 13.6, p = 0.0008; Fig. 3A and C]. In addition, superior placement of the lesion centroid was significantly correlated with change in YBOCS score, such that more superior lesions predicted a greater decrease in postoperative YBOCS score compared to baseline (Fig. 3B). This was true both when the correlation was computed using the earliest YBOCS score obtained at least 6 months postoperatively (linear correlation, lesion centroid z-coordinate, YBOCS score: ρ = −0.52, p = 0.0012) and when the correlation was computed using the latest available YBOCS score (linear correlation, lesion centroid z-coordinate, YBOCS: ρ = −0.55, p = 0.00058).

FIG. 2.

Cingulotomy lesion demarcation. Lesions were traced on each MRI slice to generate ROIs, and coordinates were transformed to MNI space. A: Coronal and sagittal postoperative T1-weighted MR noncontrast images for a representative triple cingulotomy lesion. B: Lesion ROIs (yellow areas) corresponding to lesion shown in panel A after transforming to MNI space. MNI152 1-mm resolution of the brain is shown. Figure is available in color online only.

FIG. 3.

Geometrical measures of lesion placement and size. A: MNI coordinates of lesion centroid for responders compared to nonresponders. The MNI z-coordinate (inferior/superior axis) was significantly greater for responders compared to nonresponders. There was no significant difference in MNI x- or y-coordinates between responder and nonresponder groups. Absolute values of MNI x-coordinates are displayed in order to show the distance from midline for both left- and right-sided lesions. Black dots with whiskers show median ± 2 SEs; **p = 0.94 for MNI x-coordinates, p = 0.59 for MNI y-coordinates, and p = 0.00008 for MNI z-coordinates B: MNI z-coordinate correlates with decrease in YBOCS score following cingulotomy. Graph depicts change in YBOCS scores at the most recent follow-up and the corresponding MNI z-coordinate of the lesion centroid. C: Probability map of lesion coverage in responders and nonresponders. Green shows areas covered by proportionally more responders, and pink shows areas covered by proportionally more nonresponders. Color intensity is scaled by probability. D: Lesion volume. There was no difference in lesion volume for responders compared to nonresponders. Black dots with whiskers show median ± 2 SEs. Figure is available in color online only.

Neither the anterior/posterior (MNI x) coordinate of the lesion centroids [1-way ANOVA, lesion centroid x-coordinate: F(1,34) = 0.01, p = 0.94] nor the medial/lateral (MNI y) coordinate of the lesion centroids [1-way ANOVA, lesion centroid y-coordinate: F(1,34) = 3.83, p = 0.059] was significantly different in responders versus nonresponders (Fig. 3A). Lesion volume did not significantly differ between responders and nonresponders [1-way ANOVA, lesion volume: F(1,34) = 0.34, p = 0.57; Fig. 3D]. Finally, in the 14 patients in whom pre- and postoperative BDI scores were available, we found that there were no significant correlations between the x-, y-, or z-coordinates of lesion centroid placement and changes in postoperative BDI score (linear correlation, lesion centroid x-coordinate, BDI: ρ = 0.01, p = 0.93; y-coordinate, BDI: ρ = −0.05, p = 0.80; z-coordinate, BDI: ρ = −0.15, p = 0.45). In summary, among geometrical measures describing centroid placement, only the MNI z-coordinate (inferior/superior axis) predicted clinical response (measured by the YBOCS score) to cingulotomy.

Anatomical Structures Incorporated by Lesions and Responder Status

Next, we related these geometrical measures to anatomical landmarks. We plotted the centroids of all patients’ lesions alongside Brodmann area (BA) 24, BA32, and the CS, which corresponds to the boundary between BA24 and BA32 (Fig. 4A). All lesions were centered within 6 mm of the CS. Grossly, we observed that responders’ lesions were positioned posteriorly and superiorly along the CS compared to nonresponders’ lesions (Fig. 4A). We tested this observation empirically by drawing a linear axis along the CS and asking where a particular lesion was located along this axis (Fig. 4B). Lesion centroids were located significantly posteriorly and superiorly along the CS in responders versus nonresponders [1-way ANOVA, lesion centroid distance along CS: F(1,34) = 10.5, p = 0.003; Fig. 4C.

FIG. 4.

Lesion position relative to anatomical landmarks. A: Lesion centroids relative to BA32 and BA24. B: Schematic of axis approximated to be angled along the CS in the sagittal view. Distance along the axis was calculated starting most superiorly and posteriorly. C: Distance (d) of responders’ and nonresponders’ centroids along the CS axis (shown in panel B), illustrating that responders’ lesions are located more posteriorly and superiorly along the CS. Black dots with whiskers show median ± 2 SEs; **p = 0.0003. D: To compute the proportion of BA32 covered by lesions along the anterior/superior axis, BA32 was binned in 5-mm segments along the MNI y-axis (anterior/superior axis). The proportion of each of these binned subdivisions of BA32 covered by the lesion was calculated and plotted ± 1 SE. The responders’ lesions cover a greater proportion of BA32 posterior to MNI y = 24 mm (dotted line); p = 0.035. E: Relationship between lesion areas and BA32. Illustration of left lesion ROIs (boundary containing lesion volumes of at least 25% of patients, plotted for each group) with posterior BA32 emphasized in dark blue (dotted line denotes MNI y = 24 mm). F: Left lesion placement relative to a connectivity-based ROI delineated by the atlas presented in the study by Yeo et al.,¹¹ termed the “ventral attention network” by the authors. G: Responders had greater coverage of the connectivity-based ROI delineated in Yeo et al. shown in panel F. p = 0.035. Figure is available in color online only.

We next asked whether BA24 and BA32 were differentially covered by lesions in responders and nonresponders. The proportion of BA24 and BA32 affected by the lesions was not significantly different between responders and nonresponders [1-way ANOVA, proportion of BA24 affected: F(1,34) = 0.20, p = 0.66; 1-way ANOVA, proportion of BA32 affected: F(1,34) = 0.71, p = 0.40]. However, there was a significant difference in coverage along the anterior-posterior length of BA32. We divided BA32 into equal segments along its anterior-posterior axis and computed the proportion of each BA32 segment covered by the lesion (Fig. 4D). The proportion of coverage of BA32 by the lesion varied over the length of BA32, with visibly greater coverage of BA32 posterior to MNI y = 24 mm ("posterior BA32," Fig. 4E) in responders. The proportion of posterior BA32 covered by the lesion was 9% in nonresponders and 17% in responders. This difference in coverage was highly significant [1-way ANOVA, proportion of posterior BA32 affected: F(1,34) = 8.43, p = 0.0064]. Highly significant differences in lesion coverage did not exist along BA24. In summary, responders’ lesions were centered more superiorly and posteriorly along the CS and covered a greater extent of posterior BA32 (region highlighted in Fig. 4E).

We next asked whether lesions in responders versus nonresponders corresponded to greater coverage of more recent, connectivity-based atlases. Recent studies have proposed a different parcellation of CC than described by BAs, including those based on diffusion tractography,^9,10 as well as functional connectivity. We asked whether responders’ lesions showed differential coverage of connectivity-based cortical parcellations that included the CC. We used an atlas based on functional connectivity presented by Yeo et al.¹¹ and found that responders showed greater coverage of a connectivity-based ROI termed the “ventral attention network” [Fig. 4F and G; 1-way ANOVA, coverage of connectivity-based ROI: F(1,34) = 2.19, p = 0.035]. This connectivity-based ROI includes a subregion of the dorsal cingulate gyrus that is centered more posteriorly than both BA24 and BA32, and was therefore covered more thoroughly by responders’ lesions.

Finally, we asked whether lesions in responders and nonresponders differed in coverage of surrounding anatomical structures (Fig. 5A) similarly to those reported by Yang et al. in 2014.¹² We focused on the proportion of the CB, CC, and PC (medial frontal cortex) covered by the lesions by using the probabilistic Harvard-Oxford brain atlas for the CC and PC, and the Johns Hopkins White Matter Tractography atlas for the CB (see Methods). The proportion of these anatomical structures of interest affected by lesions was not significantly different between responders and nonresponders [Fig. 5B; 1-way ANOVA, proportion of PC affected: F(1,16) = 0.40, p = 0.53; 1-way ANOVA, proportion of CB affected: F(1,16) = 2.72, p = 0.11; 1-way ANOVA, proportion of CC affected: F(1,34) = 1.98, p = 0.17]. Less than 2% of other nearby brain regions, including the premotor cortex, superior corona radiata, and anterior corona radiata, were covered by the lesions. In summary, lesion coverage of gross anatomical structures did not explain why more superior lesion placement was associated with better clinical outcomes.

FIG. 5.

A: Anatomical structures affected by lesions: cingulum bundle (CB), cingulate cortex (CC), and paracingulate cortex (PC) from the Harvard-Oxford atlas and the Johns Hopkins White Matter Tractography atlas. The density of color corresponds to probability. Lesion outlines for responders and nonresponders are boundaries drawn around points covered by lesions in at least 2 patients. B: Percentage of PC, CB, and CC covered by lesions in responders and nonresponders. Black dots with whiskers show median ± 2 SEs. Figure is available in color online only.

Discussion

Multiple brain regions and circuits have been implicated in the pathophysiology of OCD. Perturbations in cortico-striato-thalamo-cortical (CSTC) circuitry involving the orbitofrontal cortex, basal ganglia, limbic striatum, and thalamus are thought to play an important role in the disorder.¹³ The anterior CC, medial prefrontal cortex, and orbitofrontal cortex have been shown to play an important role in OCD circuitry.^14–16 Imaging studies have demonstrated decreased prefrontal cortex and anterior cingulate volume in OCD patients, as well as changes in the caudate, amygdala, and hippocampus.¹⁷ Studies have additionally identified altered prefrontal cortico-thalamic connectivity in OCD patients.¹⁸ Cingulotomy lesions target the rostral anterior CC, comprising BA24 and BA32, with the boundary between these regions defined by the CS. These regions have limbic and cognitive functions, and their dysfunction has been implicated in OCD.^19,20 Anterior cingulotomy additionally interrupts white matter fibers in the CB, putatively disrupting projections of regions implicated in OCD to other brain regions. Anatomical tracing and fiber imaging studies in both humans and nonhuman primates have shown that the CB has substantial differences in connectivity compared with other brain regions along its length.^21–23 Better understanding of the structures and circuits impacted by anterior cingulotomy and their relationship to clinical response will help to inform understanding of the pathophysiology of OCD and mechanisms of cingulotomy.

We found that patients with more superiorly placed lesions tend to have better clinical outcomes. While all lesions were centered within 6 mm of the CS, responders’ lesions were placed superiorly and posteriorly along the length of the CS compared to nonresponders’ lesions. Furthermore, responders’ lesions covered a greater extent of posterior regions of BA32. Whereas the lesion targets (from most posterior to most anterior) were 5, 3.5, and 2 mm superior to the corpus callosum, we would propose transposing the inferior-superior coordinate of each triple cingulotomy lesion by 2 mm superiorly.

Previous studies of bilateral anterior cingulotomy have reported response rates ranging from 35% to 70%.^3,5–8 The factors contributing to response are poorly understood. One study reported that decreased gray matter volume in a region of the right dorsal anterior cingulate and increased connectivity between the right cingulate and caudate, putamen, pallidum, thalamus, and hippocampus were correlated with better clinical outcome following cingulotomy.¹ Another study found that increased metabolism within a region of the right posterior cingulate was correlated with response.²⁴ Our study sought to identify how specific lesion location contributes to optimal clinical outcomes by using modern image analysis tools to compare lesion coverage of voxels within a standardized coordinate space.

Several previous studies have examined lesion location in bilateral cingulotomy. In a study of patients with depression who underwent cingulotomy, more rostrally placed lesions had better clinical outcomes, whereas the superior-inferior coordinate of the lesion centroid was not significantly related to clinical outcome.²⁵ These results cannot be directly compared to those in the present study given the different psychiatric diagnosis of the study population. Moreover, the reported average centroids of MNI y- and MNI z-coordinates were nearly 1 cm posterior and superior to the average centroid observed in the present study. Two previous studies correlated lesion location with clinical outcome following cingulotomy for OCD. One study reported more superior placement of the cingulotomy lesion in patients who experienced initial improvement of YBOCS score, as well as in 1 patient who experienced sustained improvement.²⁶ This result was replicated by our study, despite having some differences in method of lesion placement, such as placing the lesion slightly more posteriorly. Another study, consistently with our findings, found no significant differences between responders and nonresponders in the proportion of lesion area overlapping with anatomical areas.¹²

In our study, responders’ lesions were placed superiorly and posteriorly along the CS, which corresponded to greater coverage of the posterior BA32. Tractography analyses have suggested a different parcellation of the CC than is described by traditional Brodmann areas.^9,10 One study produced 9 distinct clusters within the CC based on a resting-state blood oxygen level–dependent (BOLD) functional MRI (fMRI).⁹ The region we designated the “posterior BA32” in this study corresponds closely to a dorsal cingulate sulcal region the other authors termed “cluster 4,” which resides just posterior to the majority of BA32 (contained in “cluster 3”). Both of these clusters, located within the dorsal anterior CC, shared common connections to several brain regions, including dorsal striatum and prefrontal cortex. Interestingly, cluster 4, although posterior to cluster 3, had more connections with the premotor cortex. It is intriguing that responders’ lesions appear to involve a greater proportion of this cluster, as OCD has been hypothesized to involve premotor circuit dysfunction.²⁷ In addition, our analysis based on the Yeo et al. resting-state connectivity atlas¹¹ showed that responders’ lesions covered a greater portion of one network termed the “ventral attention network,” which included part of the dorsal anterior CC centered more posteriorly relative to BA32 (Fig. 4F). This ventral attention network includes the dorsal CC and fronto-opercular areas just anterior to the premotor cortex and has been referred to in other studies as the cingulo-opercular network.²⁸ Intriguingly, activation in the cingulo-opercular network during a behavioral task implicating conflicts in decision-making predicted response to exposure-and-ritual prevention therapy. Therefore, our analysis of lesion placement utilizing a connectivity-based atlas implicates a region of the dorsal CC with distinct connectivity patterns in the frontal cortex, possibly the frontal operculum.

Disordered signaling in the CSTC circuit is thought to contribute to OCD pathophysiology.¹⁶ One prominent account is that the CSTC plays an important role in action selection: the cortex supplies an ensemble of possible actions, and the striatum chooses which one to pursue.²⁹ This model serves as a useful heuristic approach to explain OCD symptomatology: inappropriate potentiation of particular CSTC loops may result in the inability to screen out pathological sequences of behavior.³⁰ CSTC components, including the anterior thalamus, orbitofrontal cortex, anterior CC, dorsomedial prefrontal cortex, caudate nucleus, and putamen, have shown abnormalities in OCD patients in both structural^31,32 and functional imaging studies.^33–35 BOLD fMRI resting-state connectivity studies showed altered connectivity between the orbitofrontal cortex and basal ganglia,^36,37 and specific abnormalities in the connectivity of indirect pathway components of the CSTC in OCD patients.³⁸

A compelling mechanism of action of cingulotomy is perturbation of the CSTC circuit. By ablating part of the CC, inputs and outputs from several regions may be affected, including both dorsal and ventral striatum and dorsal prefrontal cortex.⁹ Some fibers from the thalamus use the CB to reach the anterior CC and medial prefrontal cortex;^21,23 ablating these CB fibers along with parts of the CC could alter the extent of excitatory thalamic output that reaches the cortex. Along these lines, differential targeting along the CB itself has been proposed.³⁹ Sweet et al.⁴⁰ showed that the variable input-output relationships along the length of the CB, which have been well characterized in primates by Heilbronner and Haber,²¹ are conserved in humans. Although data quality precluded analysis of individual diffusion tractography in the present study, future studies may investigate placement of lesions relative to individualized tractography. There may be a “sweet spot” that should be targeted in cingulotomy procedures in order to optimize the clinical response. One hypothesis is that effective cingulotomy lesions are centered around the CS, with superior and posterior placement sufficient to cover posterior portions of BA32.

Finally, it is important to consider our results in the context of alternative surgical options for refractory OCD. Surgical options include ablative procedures such as anterior cingulotomy, anterior capsulotomy (including the anterior limb of the internal capsule), and subcaudate tractotomy. With regard to deep brain stimulation (DBS) options, there is evidence for the efficacy of bilateral subthalamic nucleus (STN) DBS and bilateral ventral capsule/ventral striatum DBS (see review⁴¹). Of note, there are no head-to-head trials that compare ablative procedures and DBS. STN DBS for OCD was originally attempted after implantation in Parkinson’s disease patients showed reduction in obsessive behaviors.⁴² Retargeting STN DBS electrodes to the limbic STN, which receives afferents from the anterior CC, can be an effective treatment in OCD patients.⁴³ In this way, STN DBS likely affects a component of the CSTC circuit disrupted by anterior cingulotomy; instead of targeting the CC and associated areas of the CB, STN DBS targets efferents from those areas. Along similar lines, subcaudate tractotomy, which has been employed alongside anterior cingulotomy in a combined operation called limbic leukotomy,⁴⁴ targets pallidothalamic fibers that reside in the substantia inominata. Disrupting these fiber tracts provides yet another lesion to a component of the CSTC. For this reason, patients may benefit from redundant lesioning of these tracts following a partially effective first surgery.⁶ In summary, beneficial predominant techniques, such as STN DBS, anterior cingulotomy, and subcaudate tractotomy, may all center around disruption of various loci within the same CSTC circuit. Our study points to a dorsal cingulate cortical region that may be of particular importance for cortical lesion efficacy.

This retrospective study has limitations. First, the sample size of 18 patients is small and represents only a subset of patients at our institution who underwent triple cingulotomy. Second, the image quality is variable between patients, as some MR images dated back to 1999. Automated registration of patient brains to the MNI152 brain was checked by outlining the ventricles in the original image and visually confirming accurate registration to the standardized space. Nonetheless, in cases with poor resolution, registration (although grossly accurate) may have subtle differences from registration with high-quality images. Therefore, retrospective analysis, sample size, and image quality must all be taken into account when interpreting the results of this study.

Conclusions

Our study describes a relationship between lesion placement and clinical outcome following cingulotomy for OCD. More superiorly placed lesions correlated with better clinical outcomes. Responders’ lesions were located superiorly and posteriorly along the CS and showed greater coverage of posterior regions of BA32. Consideration of both the coordinates of lesion centroid placement along the CS and the regions of coverage along BA32 may be important when planning cingulotomy procedures and for future studies investigating the circuitry of OCD and mechanism of cingulotomy.

Acknowledgments

This project described was supported by award number T32GM007753 from the National Institute of General Medical Sciences (C.K.S.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health.

Disclosures

Dr. Dougherty reports being a consultant for Medtronic and receiving clinical or research support for the study described (includes equipment or material) from Medtronic.

Author Contributions

Conception and design: Williams, Starkweather, Bick. Acquisition of data: Williams. Analysis and interpretation of data: Williams, Starkweather, Bick. Drafting the article: Starkweather. Critically revising the article: Williams, Starkweather, Bick, Dougherty. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Williams. Statistical analysis: Starkweather. Administrative/technical/material support: Williams, McHugh. Study supervision: Williams.