Efficacy of Mindfulness-Based Cognitive Training in Surgery

Key Points Question Is there preliminary evidence to support the effectiveness of mindfulness-based stress reduction for improving well-being and performance in postgraduate year 1 surgery residents? Findings In this pilot randomized clinical trial of 21 first-year surgery residents, taking a modified mindfulness-based stress reduction class was associated with higher mindfulness, lower stress, better executive function scores, faster motor skills, and unique activation of neural substrates associated with executive control and self-awareness during an emotional regulation task compared with control participants. Meaning Mindfulness-based stress reduction appeared to mitigate stress and enhance executive function in surgery residents, supporting the value of larger, more definitive trials of this promising intervention for surgeons.


Introduction
Among physicians, overwhelming stress has been linked to burnout, distress, depression, and suicidality 1 and, in the absence of adequate coping skills, has been posited to promote performance deficits, from surgical errors to poor professionalism. [2][3][4][5] While interventions to address work climate and work-life balance are growing, 6 individual-based interventions that successfully mitigate the effects of stress remain scarce. 7,8 This gap is particularly striking in light of a growing consensus that physician well-being is a cornerstone of sustainable health care that is hindered by overwhelming stress and equally dependent on institutional, systemic, and individual efforts. 9,10 Resilience is a means of adaptive coping that changes perceived stress through the development of specific cognitive habits. [11][12][13] Mindfulness-based interventions (MBIs) have been shown to enhance resilience 14 and improve affect, 15,16 executive function, [17][18][19][20] and performance 21,22 in other high-stress populations, such as Marines, 14,17 police special forces, 23 and elite athletes, 24,25 suggesting the potential of MBIs to serve as stress resilience training for physicians. Mindfulness meditation training involves the cultivation of key cognitive skills, including the moment-to-moment awareness of thoughts, emotions, and sensations (ie, interoception), 26,27 the development of nonreactivity in response to stimuli (ie, emotional regulation), 28 and the conscious awareness of cognitive control processes (ie, metacognition). 20,29,30 The most scientifically studied form of mindfulness training is the secular mindfulness-based stress reduction (MBSR), 29 which is trained through an 8-week codified curriculum. Mindfulnessbased stress reduction and other MBIs may work by strengthening an individual's ability to tolerate uncomfortable experiences through training in nonreactivity to difficult thoughts, events, and emotions. Such nonreactivity may reduce the magnitude of perceived stress, which theory and empirical work suggest is stimulating to a point, after which impairment occurs. 31,32 Working memory capacity (WMC), a primary domain of executive function and a key measure of cognitive load, 33 is worsened by overwhelming stress. 17 Thus, reducing perceived stress may reduce cognitive load, leading to improved WMC and executive function for complex tasks, such as clinical reasoning and surgical judgement-highly relevant skills for physicians and surgeons.
Despite evidence that MBIs enhance stress resilience, well-being, and performance in quantifiable ways, they have been little used in physicians. To address this gap, we first conducted a pilot randomized clinical trial of modified MBSR (modMBSR) in surgery interns at a US academic center and reported it to be feasible and acceptable. 34 Here, we report on additional analysis of data from that trial, exploring preliminary evidence of MBSR efficacy regarding improved stress resilience, well-being, and performance.

Trial Overview
The trial design, intervention, and feasibility findings have been described in detail elsewhere. 34 Briefly, in 2016, we conducted a pilot parallel-group randomized clinical trial, 34 Mindful Surgeon, with 1:1 allocation to modMBSR, consisting of 8 weekly, 2-hour classes, 34 vs an active control (different content, same structure) in 21 first-year surgery residents (Figure 1). The study protocol and statistical analysis plan are available in Supplement 1. All aspects of the intervention and assessment were approved in full by the University of California, San Francisco (UCSF) institutional review board.
Eligible participants were postgraduate year 1 (PGY-1) surgery residents at UCSF, without a current mindfulness meditation practice who provided written and oral informed consent and were blinded to assignment. They received no financial compensation. All data were collected at the UCSF Sandler Neurosciences Center, and deidentified data were analyzed by a biostatistician (K.L.D.). In line with the National Institutes of Health Stage Model for behavioral intervention development, 35 trial goals were to show intervention feasibility and acceptability (reported elsewhere) 34 and to generate preliminary evidence of MBI efficacy (ie, effect size) as well as feasibility of study methods in preparation for an adequately powered future trial. Thus, this study was not powered to detect statistically significant intergroup differences and comprises a sample of convenience. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline. Our working conceptual model (Figure 2) is that MBI training enhances the development of cognitive habits that change how interns experience and respond to discomfort (ie, stress), thereby decreasing the magnitude of perceived stress and its contribution to cognitive load. Thus, available 10 Allocated to active control 9 Received allocated intervention 1 Did not receive allocated intervention b a Two participants were enrolled but were withdrawn by their parent program before completing assessment battery or attending any study sessions owing to conflicts with specialty-specific didactic sessions and concern for compromised education.
b One participant was initially allocated to the active control but did not receive the intervention owing to inadvertently attending the modMBSR training class during week 1. She was therefore reassigned to the modMBSR intervention group. c Two participants did not have functional magnetic resonance imaging (fMRI) scans analyzed. One was never scanned owing to implanted metal, and the other was scanned but data were incomplete (protocol glitch) and could not be analyzed.  executive function resources (such as emotional regulation, WMC, and executive control) will increase and contribute to better mood, cognition, and performance.

Measuring Efficacy
The primary efficacy outcome was change in perceived stress. The secondary outcome was change in executive function. Additional prespecified outcomes were burnout, depression, mindfulness, resilience, grit, motor skill performance, and changes in functional neuroimaging during an emotion regulation task. Outcomes were assessed at baseline (T1, before the start of intern year), postintervention (T2, 3.5 months after baseline), and at the end of the intern year (T3, 12 months after baseline).

Psychological Assessment
Participants completed an online psychologic survey consisting of reliable and published questionnaires shown to be sensitive in our prior work. 1

Emotional Regulation Task
To explore changes in neural substrates associated with our intervention, we used a task of emotional regulation in the form of cognitive reappraisal 45 (ie, the reinterpretation of affective stimuli to alter emotional impact). During a blood oxygen level-dependent functional magnetic resonance imaging (fMRI) scan, participants viewed aversive images selected from the International Affective Picture System. 46 Images were displayed on a monitor in the scanner suite and viewed through a system of back-projecting mirrors mounted to the head coil unit. Participants viewed each image once (eAppendix 1 in Supplement 2).
Images were grouped into 3 conditions. For the neutral and negative (ie, aversive) conditions, participants were instructed to view the image, understand it, and experience resultant feelings naturally. For the decrease negative condition, participants viewed negative images and were asked to decrease the intensity of their emotional response. All participants saw the same 60 images, in 3 sets of 20, delivered in the same fashion ( Figure 3D).

Statistical Analysis
For tests and surveys, group mean scores and SDs were calculated at 3 points and evaluated using t tests and Pearson χ 2 tests. Consistent with recommendations for pilot trials, 47 we did not focus on statistical power but did use linear mixed-effects modeling (analysis of covariance) for multivariate analysis, with baseline scores as a covariate, calculating effect size (partial η 2 ) with 3 suggested cutoff points: small, less than 0.06; medium, 0.06 to 0.14; and large, greater than 0.14. 48,49 Of relevance to future trials, power calculations suggest that a sample size of 40 participants in a 2-group comparison will have 80% power to detect an effect size expressed as partial η 2 of 0.17. This is considered a large-sized effect and is approximately equivalent to a Cohen d of 0.91.
Imaging analysis was limited to 19 participants (10 from the modMBSR arm and 9 from the control arm). After standard preprocessing steps (eAppendix 2 in Supplement 2), fMRI blood oxygen level-dependent responses from the 8-second image-presentation period of each trial were

BOLD Signal
A, Patterns of activation unique to viewing negative images in the intervention and control groups at baseline. Right panel shows activation of inferior occipital gyrus (IOG), which is associated with the processing of emotionally salient images. B, Patterns of activation unique to the action of decreasing emotional response to negative images (ie, reappraisal) in the intervention and control groups at baseline. Right panel shows activation of ventrolateral prefrontal cortex (vlPFC), which includes the inferior frontal gyrus (IFG) and is associated with the reinterpretation of affective stimuli to alter the emotional impact. C, Patterns of activation unique to the action of decreasing emotional response to negative images, only seen in the intervention group and only seen after the intervention. Lower panel shows activation of dorsolateral prefrontal cortex (dlPFC), which includes the middle frontal gyrus (MFG) and superior frontal gyrus (SFG) and is associated with the functioning of the executive control hub of higher-order cognition. The precuneus (precun), which is anatomically within the posterior cingulate cortex (PCC), is associated with mental imagery, visuospatial motor skills, and self-awareness. Both areas showed activation in the modified mindfulness-based stress reduction arm postintervention. D, Schematic diagram of the timing and steps involved in the presentation of each image for the emotional regulation task. BOLD indicates blood oxygen level-dependent; fMRI, functional magnetic resonance imaging; L, left; and R, right. analyzed using a general linear model with an event-related design fit to the 3 pseudorandomly occurring image types: neutral, negative, and decrease negative. Two comparisons, negative vs neutral and decrease negative vs negative, were used to test for activations related to the viewing of negative images and the process of decreasing negative emotions, respectively. We used a fixedeffects model to combine data across runs within a scanning session and mixed-effects models to perform higher-level within-group and between-group analyses, producing group-averaged wholebrain z statistical maps of contrasts in MNI 152 standard space (Montreal Neurological Institute). 50 We used longitudinal analysis with a paired within-individual model to assess for training-induced changes in activation for each participant and groups between scans at T1 and T2.
We performed whole-brain and region-of-interest (ROI) analyses based on predefined networks of interest and networks showing unique activation in whole-brain analysis. Group results from whole-brain analysis are presented as raw z statistical maps to show that the patterns of task activation are robust and regionally specific ( Figure 3A-C). Regions of interest were defined anatomically based on the Automated Anatomic Labeling atlas of the standard Montreal Neurological Institute brain template, 51 with activations measured using the Featquery tool (FMRIB Software Library) and averaged across the 2 hemispheres. We used 1-sided 1-sample t tests to validate findings of activation during emotional reappraisal 52,53 and 1-sided 2-sample t tests to assess the effect of modMBSR on reappraisal-related activation in the intervention and control groups, separately.
Statistical significance was set at P < .05, and all statistical analyses were performed on R version

(The R Foundation).
To demonstrate task validity, we compared negative vs neutral image responses at T1 across all participants ( Figure 3A), focusing on the inferior occipital gyrus, which is involved in the processing of emotionally salient images. 52,53 To demonstrate task fidelity, we evaluated activation during the

Psychological Well-being and Distress
As shown in Table 2 for control participants increased more than twice as much as for modMBSR participants at T2, but at T3, scores decreased for both groups while remaining elevated above baseline.

Changes in Executive Function
As shown in Table 2 Abbreviations: modMBSR, modified mindfulness-based stress reduction; OMFS, oromaxillofacial surgery.
a Groups as randomized were modMBSR (n = 11) and control (n = 10). One participant (a female, white, categorical general surgery resident) inadvertently attended the wrong first class and was therefore transferred to the modMBSR group.
b No participants of Hispanic or other race/ethnicity were enrolled.

Motor Skills Performance
The modMBSR group was faster in both the peg-transfer and circle-cutting tasks, accounting for variance in baseline scores. These relative effects persisted throughout the year, although times for both groups slowed from T2 to T3 (

Activation Patterns on fMRI During Emotional Regulation Task
Group results from whole-brain analysis are presented as raw z statistical maps to show that the patterns of task activation were robust and regionally specific ( Figure 3A-C). Task validity was seen with ROI analysis based on an inferior occipital gyrus mask ( Figure 3A, right panel), which showed greater activation of this visual processing region in response to negative images compared with neutral images without significant intergroup difference. Task fidelity was seen with ROI analysis based on an inferior frontal gyrus mask ( Figure 3B, right panel), which showed statistically significant activation of this ventrolateral prefrontal cortex-associated region during the act of decreasing  c Analysis of covariance shows variance between mean group score changes from T1 to T2 and from baseline T1 to T3. This method of analysis accounts for the effects of differences in treatment (ie, modMBSR vs control) as well as differences in baseline scores.
d Partial η 2 is the index of effect size for ANCOVA analyses, with 3 suggested cutoff points: small, less than 0.06; medium, 0.06 to 0.14; and large, greater than 0.14. For reference, partial η 2 = 0.17 is approximately equivalent to Cohen d = 0.91, which is considered a large effect size. emotional response to negative images compared with simply viewing negative images, without significant intergroup difference.
Regarding intervention effects, the whole-brain analysis showed 2 primary areas of robust, regionally specific, and unique activation: the dorsal lateral prefrontal cortex (DL-PFC; centered on medial frontal gyrus and superior frontal gyrus) and the precuneus/posterior cingulate cortex. As expected, these did not reach voxelwise significance on whole-brain analysis in our small sample, but the more focused ROI analysis ( Figure 3C, lower panel) confirmed statistically significant higher activation in all 3 areas (medial frontal gyrus, superior frontal gyrus, precuneus/posterior cingulate cortex), unique to the act of emotional regulation and only seen in the postintervention modMBSR group.

Discussion
The clear need for enhanced stress resilience in surgery led us to conduct a pilot randomized clinical trial of modMBSR in PGY-1 surgery residents that was found to be feasible and acceptable. 34 The preliminary evidence of efficacy regarding improved well-being and performance supports our outcome measures as sensitive and relevant to this population and suggests that our study methods can be successfully translated to an adequately powered future trial, as per the National Institutes of Health Stage Model for behavioral intervention development. 35 We found increased mindfulness and less stress in participants in the modMBSR arm compared with those in the control arm, associated with medium to large effect sizes. We also found medium to large effect sizes associated with increased WMC among modMBSR participants and a marked decrement in cognitive control among control participants. Together with our finding of unique activation in brain regions associated with executive function and emotional regulation among modMBSR participants, our results suggest a sound working conceptual model and an intervention with potentially powerful effects.
Our finding that formal mindfulness training may improve aspects of well-being and distress is supported by our observation that mindfulness scores increased twice as much in the modMBSR arm as in the control arm, whereas stress and depressive symptoms scores increased twice as much in the control arm as in the modMBSR arm. These results were associated with medium to large effect sizes and echo statistically significant improvements in mood, affect, professional satisfaction, perceived stress, and protection from depressive symptoms seen with the use of MBIs in other groups. 55,56 Relatedly, Sen et al 57 showed a greater than 6-fold increase in depression during the intern year, with subsequent work demonstrating that higher perceived stress 58 and lower perceived well-being 59 predicted this evolution. Our results support the possibility that MBIs mitigate this risk through lowering perceived stress and increasing perceived well-being.
In contrast, our finding of increased burnout scores in both modMBSR and control participants, with little difference between the 2 groups and a small effect size, suggests the intervention does not influence burnout as measured and manifested in this population or simply reflects the underpowered nature of this study. While reports of the effect of MBIs on burnout are mixed, 7,8,16 subdomain analysis of burnout (ie, emotional exhaustion vs depersonalization) in a larger sample may reveal effects that our small sample size precludes.
Our finding that modMBSR may improve or protect aspects of executive function is supported by increases in the executive composite and WMC scores in the modMBSR arm and marked decrement in the cognitive control factor score among control participants. These findings were associated with large effect sizes over time and merit further study. Executive function subserves decision making, problem solving, and the execution of complex procedures-skills clearly involved in surgical judgment and obviously valuable for physicians. The striking decrement in cognitive control scores within the control group raises concern that the rigors of residency training may cause relative cognitive impairment over time, while the lack of such a decrement in the modMBSR group suggests that formal mindfulness training may be protective. Prolonged stress in medical students has been associated with diminished performance on tests of executive function, 60 whereas predeployment Marines have shown decrements in WMC that MBI training protected against. 17 The faster peg-transfer and circle-cutting times for modMBSR participants vs control participants may be explained by the purported influence of MBI on attention and emotional regulation, cognitive skills increasingly recognized as critical for motor-skill performance. 61-66 These results were associated with medium to large effect sizes and are supported by findings of a 2017 randomized clinical trial in general surgery residents, 67 where performance training that included an emotional regulation component (eg, modulating self-criticism and positive self-talk) resulted in significantly higher scores on Fundamentals of Laparoscopic Surgery modules and higher-rated surgical skills.
Finally, our finding of greater activation in brain regions associated with executive function and self-awareness in modMBSR participants during an emotional regulation task is supported by the enhanced activity in the DL-PFC and precuneus/posterior cingulate cortex. 68-70 These findings suggest that modMBSR affects changes at the level of neural substrates, which manifest in the setting of emotional stress. If increased activity correlates with enhanced capability in these parts of the brain, it could have broad implications for the benefit of MBIs. Clinical reasoning has been shown to decrease in internists with high burnout and concomitant deactivation of the DL-PFC, 71 raising the possibility that increased activation in the DL-PFC following modMBSR may translate to improved clinical performance. Similarly, spatially complex bimanual coordination has been shown to be associated with significant activation of the precuneus, which is thought to mediate the interface between cognition and action in complex motor function. 72

Limitations
Study limitations include the small sample size, use of volunteers, and execution at a single institution with its own unique issues and resources. Our findings, while theoretically exciting, should be interpreted cautiously. Likewise, the ability to translate our intervention, study methods, and related findings to larger groups or other centers is promising but remains to be seen. Nevertheless, our findings support the value of further work to fully understand and more clearly demonstrate the association of MBIs with well-being and performance suggested here.

Conclusions
The preliminary evidence presented here, supporting the potential benefits of MBIs for surgeons in the critical areas of well-being and executive function, justifies further exploration in a larger randomized clinical trial. Addressing perceived stress within surgical trainees may yield myriad downstream benefits, as seen in other populations. This characteristic of MBIs may provide a versatility otherwise lacking in well-being interventions and may affect key issues (eg, impaired mental health, cognition, and performance) that have otherwise been difficult to address. Moreover, as an internalized skill focused on perception and not just circumstance, mindfulness-based cognitive training has the potential to evolve and remain relevant across a surgeon's career trajectory and life, making MBIs a promising means to address Accreditation Council for Graduate Medical Education programming mandates for physician trainee well-being.