Clinical outcomes of deep brain stimulation for obsessive‐compulsive disorder: Insight as a predictor of symptom changes

Aim Deep brain stimulation (DBS) is a safe and effective treatment option for people with refractory obsessive‐compulsive disorder (OCD). Yet our understanding of predictors of response and prognostic factors remains rudimentary, and long‐term comprehensive follow‐ups are lacking. We aim to investigate the efficacy of DBS therapy for OCD patients, and predictors of clinical response. Methods Eight OCD participants underwent DBS stimulation of the nucleus accumbens (NAc) in an open‐label longitudinal trial, duration of follow‐up varied between 9 months and 7 years. Post‐operative care involved comprehensive fine tuning of stimulation parameters and adjunct multidisciplinary therapy. Results Six participants achieved clinical response (35% improvement in obsessions and compulsions on the Yale Brown Obsessive Compulsive Scale (YBOCS)) within 6–9 weeks, response was maintained at last follow up. On average, the YBOCS improved by 45% at last follow up. Mixed linear modeling elucidated directionality of symptom changes: insight into symptoms strongly predicted (P = 0.008) changes in symptom severity during DBS therapy, likely driven by initial changes in depression and anxiety. Precise localization of DBS leads demonstrated that responders most often had their leads (and active contacts) placed dorsal compared to non‐responders, relative to the Nac. Conclusion The clinical efficacy of DBS for OCD is demonstrated, and mediators of changes in symptoms are proposed. The symptom improvements within this cohort should be seen within the context of the adjunct psychological and biopsychosocial care that implemented a shared decision‐making approach, with flexible iterative DBS programming. Further research should explore the utility of insight as a clinical correlate of response. The trial was prospectively registered with the ANZCTR (ACTRN12612001142820).

Obsessive-compulsive disorder (OCD) is characterized by recurrent, intrusive, unwanted thoughts, images, or impulses [obsessions] and repetitive behaviors or mental acts in response to these [compulsions]. 1 OCD and related disorders are some of the costliest, functionally disabling and treatment resistant brain conditions. 2 First line treatment of selective serotonin re-uptake inhibitors (SSRIs) and cognitive behavioral therapy (CBT) lead to suboptimal outcomes; 40-60% remain treatment resistant and 60-80% relapse. 3Around 10% of OCD patients continue to experience highly disabling symptoms 4 and may be eligible for deep brain stimulation (DBS).
The prevailing neurobiological model of OCD contends that hyperactivity and hyperconnectivity of cortico-striato-thalamo-cortical (CSTC) circuits underlies OCD pathophysiology. 5,6Deep brain stimulation (DBS) is an invasive neuromodulation therapy that provides a unique opportunity to modulate underlying CSTC circuitry in a personalized manner, 7 leading to significant global and symptomatic changes in many otherwise refractory patients.
DBS for OCD is approved (under a humanitarian exception) by the US Food and Drug Administration (FDA) and Conformitè Europëenne (CE), and a 60% response rate is well-established. 8,9][11] Yet, the level of evidence and efficacy has not translated into regulatory acceptance; many barriers to treatment access remain 12 and the therapy remains investigational in nature.
The purpose of this trial was to investigate determinants of DBS efficacy in TR-OCD.Existing literature suggests that certain symptom profiles influence response.Pre-operative factors associated with good response include sexual/religious symptoms, older age of onset and insight into symptoms. 13,146][17] Yet, these clinical factors cannot predict response on an individual level, nor should they be used to deny individuals' therapy.
The precise location of stimulation relative to an anatomical target, or neuronal tract, influences clinical outcomes. 18Yet, a wide variety of regions are targeted (Fig. 1), with many reports failing to delineate the precise stimulation location.][22][23] There is thus a pressing need to better understand which patients and which treatment protocols are most likely to be associated with clinical benefit for people with OCD undergoing DBS.In this context, the current study implemented a comprehensive treatment protocol to optimize DBS therapy for a cohort of people with TR-OCD.
The aims of the study were to: (1) assess the clinical efficacy of DBS therapy for TR-OCD, including primary and comorbid symptoms; (2) investigate predictors of OCD symptom improvement in response to DBS therapy (both clinical characteristics and lead placement were explored); and (3) report safety data.clomipramine hydrochloride at maximum dosage for at least 12 weeks, one augmentation trial with a second generation antipsychotic for 8 weeks in combination with a SSRI, and ≥16 CBT sessions).Exclusion criteria included a clinically significant comorbid DSM-IV diagnosis (except for major depression disorder [MDD], and mild anxiety disorders), and clinically significant and unstable neurological or medical illness.The project was approved by St Vincent's Hospital Melbourne Human Research Ethics Committee (HREC/12SVHM/64) and conformed to the provisions of the Declaration of Helsinki.All participants provided written informed consent.

Surgical procedure
Implantation of DBS leads was performed according to standard stereotactic procedures using frame-based magnetic resonance imaging (MRI) for target determination, under local anesthesia.Participants underwent bilateral implantation of quadripolar 3389 electrodes (Medtronic, Minnesota, USA), spaced 0.5 mm apart and 1.5 mm long (Fig. 2).The surgical target (of the electrode tip) was the NAc as identified on MRI.Stereotactic atlases place this 7 mm lateral to the midline, 3 mm anterior to the anterior border of the anterior commissure, and 4 mm inferior to the intercommissural line.The two ventral contacts were targeted in the center of the NAc (E0, E1, E8, E9).Intra-operative test stimulation was conducted to assess mood and behavior.The electrodes were connected via subcutaneous extensions to Activa PC, or RC stimulator (Medtronic, Minnesota, USA) placed in the infraclavicular pocket, under general anesthesia.Post-operative computed tomography (CT) fused with pre-operative MRI verified the position of the implanted electrodes.DBS was switched on 2 days following the surgery.

Programming
Anatomical guidance (fused imaging) was used to select the active electrode for initial programming.Programming was conducted in accordance with the Montgomery algorithm 24 ; initial parameters were 130 Hz, 90 μs, voltage was increased by 0.5 V increments, until sufficient benefit or a limiting side effect was reached (see Figure S1.).The programming adjustments involved a shared decision-making process between the psychiatrist (P.B.) and participant; benefit was defined by a global assessment of subjective reports and clinical impression.

Study design
1 Optimization phase: during the first 8-months of therapy, stimulation parameters were assessed and adjusted fortnightly.Once a 6-point drop in YBOCS was obtained, fortnightly CBT sessions commenced, and maintained for 24-weeks.CBT sessions focused on exposure and response prevention (ERP) as well as dealing with habits that accrued during long duration of OCD symptoms.Medications were maintained during this phase.2 Maintenance phase: programming assessments were conducted every 3 months, and thereafter reduced in frequency depending on the participants' needs.Medication was reduced where possible.Outreach psychosocial support was provided when appropriate (unavailable for participants 7 and 8) by a senior mental health clinician weekly or fortnightly for 12 months, and monthly thereafter.
The outreach therapy employed an eclectic approach, drawing from various treatment models including exposure response prevention, dialectic behavioral therapy, acceptance and commitment therapy, trauma informed care, and psychoeducation (see 25 for further discussion).
A one-month randomized controlled closed label phase was planned, but the first six participants declined due to fear of the therapy being terminated.Thus, the closed label condition was removed from the protocol.

Outcome measures
Our primary outcome was the YBOCS to evaluate obsessive and compulsive symptoms, on a 40-point scale. 26Clinical response was defined as a 35% improvement in the YBOCS total score from baseline.A range of secondary outcomes were collected.Depression was evaluated using the Hamilton Scale for Depression (HAMD). 27Anxiety was assessed using the Hamilton Anxiety Scale (HAMA). 28nsight into obsessive and compulsive symptoms was evaluated using the Brown Assessment of Beliefs Scale (BABS), a lower score indicating better insight. 29Additional neuropsychological, neuroimaging, functional and qualitative assessments were conducted and will be reported elsewhere. 30,31The protocol was registered on a clinical trial registry (ACTRN12612001142820).

Analyses
Clinical outcomes are reported using descriptive statistics as within-and between-subject change from baseline.Mixed linear modeling was used to investigate predictors of change in symptom severity: repeated measure = weeks; random factor = participant; covariates/predictors = HAMA, HAMD, BABS; primary outcome variable = YBOCS (unless otherwise specified).The modeling considered each participant as a random effect and assigned a random intercept to model individual differences.The AR (1) covariance structure accounted for autocorrelation across repeated measures.Covariates were expressed as change variables (t1À(t1 À 1)).The first model included all timepoints within phase 1, which were approximately 2 weeks apart (termed 'concurrent model').The second model regressed the YBOCS change scores (after baseline) on the covariates from the previous assessment period (termed 'lagged model').The lagged model allowed determination of whether the value of covariates at a given time point could predict the next YBOCS score, approximately 2 weeks later (predicts Y(t) from X(t À 1)).Within time series analysis, residuals are likely autocorrelated, thus the outcome variable is also composed of a change variable to account for possible autocorrelation.The advantage of mixed linear modeling is that it is implemented for modeling case series data over time, allows for repeated or related measures, uneven datasets and missing data.
Lead localization DBS electrodes were localized using Lead-DBS software (V2.5.3) (https://www.lead-dbs.org/). 32Briefly, pre-operative T1 and T2-weighted images and post-operative computed topography (CT) scans were linearly coregistered using Advances Normalization Tools (ANTs) (http://stnava.github.io/ANTs/), 33and normalized into ICBM 2009b non-linear Asymmetric (MNI) template space using the SyN approach implemented in ANTs.Coregistration and normalization were visually reviewed and refined if necessary.Some datasets required alternative coregistration and normalization processing methods.Subcortical refinement was applied as a module in Lead-DBS to post-operative acquisitions to correct for possible brain shift.DBS electrodes were then pre-reconstructed using the PaCER algorithm 34 and warped into MNI space.Lead reconstructions were modeled in a 3D subcortical atlas, the CIT-168 Reinforcement Learning Atlas. 35Using Lead-DBS Group, individual leads were reconstructed on a single MNI template to visualize the precise differences in surgical targeting of the NAc.The volume of tissue activated (VTA) for each lead was estimated using a finite element method 32 based on the individual stimulation settings at last follow up, and visualized within MNI space using the CIT-168 atlas.

Descriptive statistics
Table 1 provides baseline demographics and clinical characteristics of the participants.Figures 3-5 include individual patterns of response as a function of percentage change from baseline for YBOCS, HAMD and HAMA outcomes, within phase 1.Although clinical outcomes showed fluctuations within, and variability across participants, six individuals reached clinical response (75% response rate) for OCD symptoms; those who reached clinical response, did so within 6-9 weeks, and maintained response in the long term.
Table 2 shows YBOCS change for each participant following phase 1 and last follow up, as well as final stimulation parameters.Figure 6 shows individual YBOCS outcomes at 3-monthly intervals for the entirety of the follow up period.Last follow up varied from 9 months to 7 years, due to the time of implantation for each participant.
Mean change in the clinical outcomes following phase 1 and last follow up are presented in Table 3; at last follow up mean change in YBOCS total score was 45%, YBOCS obsessions 44%, YBOCS compulsions 40%, HAMD 42%, HAMA 41%, and BABS 35%.
Of the non-responders, P6 showed elected to have the DBS device removed, while P7 experienced hypomanic symptoms from stimulation increases which interfered with and limited programming refinements.

Mixed linear modeling
Residuals were tested for homoscedasticity and normality through visual inspection, which were met.A two-tailed Spearman's correlation showed YBOCS raw scores were significantly correlated (P < 0.001) with HAMD (r = 0.68), HAMA (r = 0.59) and BABS (r = 0.56) raw scores.YBOCS change (t1À(t1 À 1)) scores were also  Therefore, considering the associations between clinical variables with medium effect sizes, the relationship between clinical outcomes was further investigated using mixed linear modeling.The concurrent model analyzed predictors based on concurrent values (Table 4).Concurrent Model 1 showed that changes in HAMD and HAMA were significantly associated with the concurrent YBOCS change (P = 0.008, P = 0.013, respectively).Considering the previously established association with level of insight and clinical response 14 we further investigated directionality by changing the outcome variable to the BABS.Model 2 showed that changes in HAMA significantly predicted concurrent BABS changes (P = 0.003).
The lagged models further investigated directionality and lagged associations, by considering the previous change in covariates as predictors of subsequent (not current) YBOCS changes.Model 1 demonstrated that the previous HAMD and BABS were significantly associated with the ensuing YBOCS change (P = 0.011, P=0.008, respectively).Again, the outcome variable was changed to the BABS change variable.Model 2 demonstrated that the previous change in YBOCS, HAMD and HAMA were significantly associated with the consecutive BABS change (P = 0.042, P = 0.043, P = <0.001,respectively).
Lead localization DBS leads were reconstructed using the LeadDBS toolbox and visually evaluated.Group level visual inspection showed that five out of six responders had their leads placed more dorsal compared to nonresponders, relative to the NAc (Fig. 7).The chronic electrode contacts were the two most dorsal contacts for all participants, except for P8, who had experienced an insula hematoma and was receiving low level stimulation.Also, the responder's leads appear more posterior within the right hemisphere, yet this relationship is not observable in the left hemisphere.The localization of leads within 2D planes (Supporting Information, S2. and S3.), VTA estimates of final stimulation parameters (Supporting Information, S4), and 3D video of leads (Supporting Information, S5.) is provided in the Supporting Information to assist in the interpretation of leads relative to anatomical structures.

Discussion
This study of eight people with severe treatment-resistant OCD demonstrated the clinical efficacy of DBS, including patterns and predictors of symptom change.DBS therapy achieved a rapid, large, and sustained clinical effect.The mean change in YBOCS, the primary outcome, was 41% after the optimization phase (8 months), and 45% at last follow up (9 months-7 years).Clinical response was achieved in 75% of participants within 6-9 weeks and maintained in the long-term.The HAMD, HAMA and BABS also showed clinically meaningful improvements.Mixed linear modeling demonstrated depression and insight as predictors of changes in symptom severity, yet insight showed a stronger statistical lagged effect across timepoints.Lead localization of DBS electrodes demonstrated that the majority of responders had their leads placed slightly more dorsal.Surgical and stimulation related side effects were experienced, consistent with previous reports.Participants experienced improvement in primary symptoms, but also in various psychological functions.Primary (obsessions, compulsions) and secondary (anxiety, depression) symptoms, and symptomatic insight demonstrated statistical correlation.Changes in OCD symptoms showed some fluctuations, yet major improvements  occurred within the first 6 months, and generally plateaued thereafter.However, anxiety and depression symptoms showed dramatic fluctuations within and across participants; two patients showed deterioration beyond baseline, which later improved and two other participants improved by over 80%-a greater extent to OCD symptom changes.These findings highlight the importance of considering several outcomes of DBS efficacy.This patient group exhibits complex phenotypes and etiologies; thus, it is necessary to consider various determinants of recovery in a functionally relevant manner.We recently reported various psychosocial changes, high satisfaction, and valuable feedback relating to DBS care within this cohort. 30Further, our lived experience investigation of responders and carers identified profound changes to psychopathology and self-constructs; phenomenological changes are discussed within a proposed cognitive model. 31hanges in insight predicted subsequent symptom severity changes (P = 0.008).Yet, the initial symptomatic changes appeared to be driven by depression and anxiety that showed a concurrent effect on symptom severity.Previously, it has been proposed that DBS modulates anxiety within seconds to minutes, and obsessions within weeks, 36 our analysis supports and add to this notion on the time-series effect of DBS on various symptom domains.When considering the strongest predictors in each model, the analysis suggests that changes in depression and anxiety are occurring concurrently and driving subsequent changes in symptom severity and insight.More specifically, depression had a stronger effect on symptom severity, and anxiety had a stronger effect on insight across time points (lagged effect).In turn, the level of insight into symptoms is influencing subsequent changes in symptom severity (see Fig. 8 for a schematic).The analyses employed change variables, which eliminate autocorrelation that may still be present in raw scores and are more suggestive of causality.

Long term YBOCS changes
The impact of comorbid anxiety and depression is well understood within OCD, yet the role of insight in relation to psychopathology and treatment response is underappreciated.Insight related to mental illness is a multidimensional, dynamic, and continuous phenomenon. 37Insight in OCD involves the understanding that one's obsessions and compulsions are irrational.9][40][41][42][43] Improved insight of mental health can enhance treatment adherence and coping strategies. 41Therefore, insight is an important determinant of OCD psychopathology and prognosis.A recent large cohort analysis demonstrated good insight at baseline to have a positive predictive value of 84.4% for clinical response from DBS for OCD, yet level of insight did not yield individual predictive power. 14Programming is more difficult in those who lack awareness into cognitive-affective functions, as the utility of subjective observations is limited. 44Insight is also related to symptom domains; individuals with perfectionism, hoarding or symmetry symptoms describe their symptoms as ego-syntonic and do not respond as well as others. 16Our sample size does not allow for patient specific predictive analysis, yet it is worth noting that the nonresponders in our cohort had relatively poor baseline insight (BABS score of 11 and 14) although one of the responders had a comparable The mean AE standard deviation is presented for each primary and secondary outcome.For patient 7 the last outcome for HAMD, HAMA, and BABS was carried forward.baseline score (BABS score of 14).Therefore, insight at baseline may serve as a useful prognostic, but not, exclusion factor, and needs consideration with other patient characteristics.
In previous studies, long-term DBS stimulation of the NAc has led to 12-33% improvement in OCD symptoms, and response in 10-56%, 21,[45][46][47][48][49][50][51][52][53] yet other targets (ventral anterior limb of the internal capsule (ALIC), ventral capsule/ventral striatum (VC/VS)) have been associated with up to 80% response. 19The higher efficacy of NAc DBS identified in the current study might be due to the detailed and comprehensive fine tuning of stimulation as well as the adjunct behavioral and psychosocial therapy.
We implemented a multi-disciplinary model to support the OCD DBS treatment during the pre-operative and post-implantation periods.Psychiatric care involved extensive programming sessions with collaborative involvement from the participant, as well as exposure therapy, and continuing psychological support.Also, outreach psychosocial therapy with participants and families consolidated exposure therapy in naturalistic settings.Although DBS alone can achieve an initial and rapid improvement in OCD symptoms, it is contended that integrative therapy with a holistic consideration of the severity and impact of OCD symptoms on everyday life is necessary for sustained and optimized recovery.For further discussion, refer to our clinical guideline for managing people receiving DBS for TR-OCD. 25Adjunct therapeutic approaches in this patient group are rarely discussed in the scientific literature. 54et, post-operative management of psychiatric patients relies on a highly experienced and specialized multidisciplinary team working in collaboration with the patients and their families.Greater discussion in the scientific literature is required in this domain, with the aim to develop more standardized and personalized care.
The efficacy within our cohort, is also attributable to probable stimulation of regions dorsal to the surgical target.Five out of six responders in our study were receiving chronic stimulation in the two most dorsal contacts on the lead, which were placed above the NAc (Fig. 7) and likely modulating nearby structures, including the ventral ALIC.The ventral ALIC has been consistently identified as the optimized target across different cohorts 19,20,[55][56][57][58] and we add to this growing body of literature.0][61][62] Despite modern surgical techniques, the placement of the leads cannot be assumed to be uniform across individuals.Thus, evaluation and reporting of clinically effective DBS contacts is critical to progress scientific understanding of the best practices to optimize efficacy.Therefore, we present the precise localization of DBS leads for transparency and greater understanding in the field.
Participants declined the planned closed label phase, in which they would have known that their stimulation may or may not be turned off.There are several concerns with DBS sham conditions in psychiatry.Turning DBS off can lead to drastic deterioration of symptoms 63,64 ; as a result, sham conditions are often terminated early. 55,65,66Also, order effects may occur 53 and increased anxiety during blinded phases can complicate clinical evaluations.DBS efficacy relies on the optimization of programming, which can take 6-12 months, yet for effective blinding the participant should be naïve to stimulation; both cannot hold true in closed label conditions.As DBS is considered a 'last resort' treatment for extremely vulnerable patients, ethically we need to provide the best possible treatment options.Thus, sham methods were necessary in earlier investigations to establish a level of efficacy.However, we contend that a sufficient level of evidence has been reached to consider DBS an established therapy for TR-OCD, [9][10][11]67,68 and that closed label conditions do not satisfy ethical or validity requirements. The oerall goal of DBS therapy is to improve outcomes for patients.Therefore, if the patient experiences clinically significant long-term improvement, hasn't the endpoint been reached?
There are several limitations to this project, including the small number of participants, which limits extrapolation of findings.Having said that, we conducted the analyses with sample size at the forefront of our considerations, and the overall effect sizes were robust to analyses.There are a limited number of cases of DBS for OCD worldwide, and most are confined to single research groups, hence our cases expand the overall reported cohort and provide a perspective from another treatment center.Insight in OCD is difficult to operationalize and is unlikely to be captured with a single scale.The BABS assesses insight with reference to the individual's predominant symptom, which may lack sensitivity to insight of the condition.Finally, the protocol did not involve a blinded component, as no participant agreed to the potential for their device to be randomly switched off.One participant inadvertently allowed the battery to go flat and experienced an overwhelming exacerbation of OCD symptoms, with a huge surge in anxiety; there was immediate restitution once the battery was charged.Further, here we presented the primary and secondary outcomes of the clinical trial, which do not portray the profound changes the participants experienced in their everyday life.Subjective experiences, psychosocial outcomes, and a lived experience investigation are reported elsewhere. 30,31uture trials of DBS therapy for TR-OCD should implement a multi-disciplinary adjunct treatment approach targeting several aspects of functional recovery, with consistent programming adjustments.Follow up reports should also consider different aspects of functioning and the patients' perspective to gain a more comprehensive understanding of DBS mechanisms in obsessive compulsive symptomology.Insight should be further investigated within this context, by evaluation of different elements of insight relating to (1) condition, (2) symptom, and (3) treatment, as possible mediators of clinical response.Also, pooling of data in a clinical registry would allow comprehensive analysis to establish if baseline insight can predict individual clinical response.Although a specific clinical predictor of DBS response may not exist for this patient group, pooling multimodal data may identify a set of factors that provide more nuanced prognostic predictions.This approach would allow progression towards personalized precision medicine for this patient group.
Although the improvements described were clinically and personally meaningful, DBS therapy comes with risks from the surgery, device, and/or stimulation.In our cohort, two participants experienced serious transient adverse events from surgical complications, which required medical follow up.Another two participants experienced hypomanic symptoms from increases in stimulation voltage.Hypomania is common in individuals receiving DBS within the dorsal striatum for OCD 69 and may even signify effectiveness of the therapy. 20owever, the presence of hypomania complicates increases in stimulation, which are necessary to optimize outcomes.Therefore, a predisposition to hypomania could be screened pre-operatively and considered in the decision making of the surgical target.Further, tractography-guided surgical targeting has been implemented in one identified OCD cohort, and led to greater responders, and significantly fewer stimulation related side effects (including hypomania) compared to conventional targeting. 70This novel approach shows potential utility in prospective targeting to optimize outcomes and minimize side effects.

Conclusion
We demonstrated the efficacy of DBS for TR-OCD and provided further support for the prognostic utility of insight, within a field lacking predictive factors.We encourage a multidisciplinary and biopsychosocial approach in tailoring therapy and evaluating effectiveness in line with personally meaningful goals.DBS, and other neuromodulation therapies should not be seen as a standalone solution, rather a synergistic tool that can provide alleviation from severe distress, allowing individuals to engage in psychotherapy and consolidate DBS mediated changes.The effect size of DBS for TR-OCD is large, particularly when considering those of conventional therapies and the refractoriness of individuals undergoing DBS.

Participants
Eight participants (three females) were recruited via referral from their treating psychiatrist to the principal investigator (P.B.), at St Vincent's Hospital Melbourne, Australia, between July 2012 and August 2021, and received approval by the Mental Health Tribunal, Victoria.Inclusion criteria included a Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) diagnosis of OCD, aged between 18 and 65 years, YBOCS ≥24, at least a 5-year history of OCD with substantial functional impairment on DSM-IV criterion C, global assessment of functioning (GAF) <45, and treatment refractoriness (≥2 SSRIs at maximum dosage for 12 weeks, one treatment with

Fig. 6
Fig.6Individual changes in the YBOCS across the follow up period from baseline to 84 months (M).

Fig. 7 Fig. 8
Fig. 7 Individual patient leads are reconstructed on a normative atlas to demonstrate the slight differences in surgical placement within the NAc using the LeadDBS toolbox.DBS leads of responders are represented in blue, DBS leads of non-responders are represented in purple.

Table 1 .
Baseline demographics and clinical characteristics Fig. 3 Individual percentage change in YBOCS.Programming adjustments and clinical assessments were conducted at approximately 2-week intervals (T) during the optimization phase across approximately the first 9 months.The number of programming sessions during the optimization phase varied between participants based on individual needs.

Table 2 .
Within subject YBOCS outcomes

Table 3 .
Group level clinical outcomes