Abstract
Glucocorticoids—stress hormones released from the adrenal cortex—reduce phobic fear in anxiety disorders and enhance psychotherapy, possibly by reducing the retrieval of fear memories. Glucocorticoid signaling in the basolateral amygdala can influence connected fear and memory related cortical regions, but this is not fully understood. Previous studies investigated specific pathways moderated by glucocorticoids, for example visual-temporal pathways, however, these analyses are limited to a priori selected regions. Here, we performed whole-brain pattern analysis to localize phobic stimulus decoding related to the fear-reducing effect of glucocorticoids. We analyzed functional magnetic resonance imaging (fMRI) data from a randomized, double-blind, placebo-controlled study with spider-phobic patients while they were looking at spider images. The patients either received oral glucocorticoids (20 mg of hydrocortisone) or a placebo. Patients with phobia had higher decoding in the anterior cingulate cortex (ACC) and the left and right anterior insula compared to controls. Decoding in the ACC and the right insula showed the highest correlations with experienced fear and explained 40% of the variance across all participants. Patients with cortisol reported a reduction of fear by 10–13% and showed decoding of phobic images in the precuneus, the cerebellum and the opercular cortex. Patients in the placebo group with increased fear showed decoding in the insula, the ACC and the right frontal lateral pole which have been shown to be related to the fear circuitry and episodic memory. This study demonstrates phobic decoding in fear-related frontal regions and suggests that cortisol administration alters these fears-specific processing areas.
Introduction
Anxiety is a common disorder with a lifetime prevalence of 8%–16% (Magee et al., 1996; Vicente et al., 2006). Among anxiety disorders, specific phobias are the most common, with a lifetime prevalence of 12.5%, and affect both female and male individuals of all ages (Kessler et al., 2005). Such fears can be related to high altitudes, airplane travel, enclosed spaces, or animals such as snakes and spiders. Reactions can range from personal distress to panic. Individuals with phobia avoid the stimulus or the situation to reduce the fear, which entails considerable restrictions in their lives. Confrontation with the phobic stimulus or situation (or even its anticipation) almost invariably provokes the retrieval of past phobic memories, which consequently leads to a fear response (Cuthbert et al., 2003; de Quervain and Margraf, 2008). This mechanism supports the consolidation of additional fear memories and ultimately strengthens these fear memory traces (Sara, 2000). Eventually, the retrieval and consolidation of fearful memories seems to be an important factor in the maintenance of phobic disorders.
Evidence shows that cognitive-behavioural therapy (CBT), including exposure and cognitive restructuring, is efficacious and reorganizes processing in key regions including the amygdala, insula, and cingulate cortex (Hauner et al., 2012; Shin and Liberzon, 2010). However, up to one-third of the patients with anxiety disorders do not respond to CBT (Cuthbert, 2002; Heimberg, 2002; Hofmann and Smits, 2008). Drugs with the potential to enhance memory extinction processes are therefore promising candidates for enhancing exposure therapy. Over the last decade, various studies have demonstrated that glucocorticoids are involved in memory regulation (for an overview de Quervain et al., 2017). Specifically, glucocorticoids impair emotional long-term memory retrieval (de Quervain et al., 1998) while enhancing the consolidation of new memories (de Quervain et al., 2009). Previous studies demonstrated that acute administration of glucocorticoids reduces phobic fear in patients with anxiety disorders (Aerni et al., 2004; Soravia et al., 2006) and improves extinction-based psychotherapy (de Quervain et al., 2011; Soravia et al., 2014). Similarly, stress-induced cortisol elevation can reduce negative affect after stress (Het et al., 2012). Thus, glucocorticoid treatment in combination with exposure therapy is a novel and promising approach for a more effective treatment of phobia (Bentz et al., 2010).
Glucocorticoids interact with noradrenergic transmission in the basolateral part of the amygdala which also affects other brain regions, such as the hippocampus and cortical areas that project to the amygdala (de Quervain et al., 2009). Thus, anxiolytic effects from glucocorticoids and induced changes in the amygdala can influence connected cortical areas and influence cortical processing of phobic information. A recent study focused on the the visual temporal pathway, including the lingual, fusiform gyrus, and the amygdala (Nakataki et al., 2017). However, the role of additional areas that closely interact with the amygdala is not fully understood, in particular the prefrontal cortex, the insula, the cingulate gyrus and the hippocampus. Therefore, the aim of the present study was to elucidate the cortical changes related to the processing of phobic information under glucocorticoids. We analyzed the same dataset as the study by Nakataki et al. (2017). However, unlike the study by Nakataki et al. (2017) that focused three regions of interest, the amygdala, the fusiform gyrus, and the lingual gyrus, we implemented whole-brain multivoxel pattern analysis (MVPA), a highly sensitive multivariate analysis in order to investigate differences in decoding of phobic material throughout the cortex. The dataset contained behavioural and functional MRI (fMRI) data from a double-blind, placebo-controlled, randomized clinical study. Spider phobic patients received 20 mg of cortisol or placebo orally one hour before a picture task provoking phobic fear while fMRI images were acquired. During the experiment, participants viewed spider and non-phobic pictures and rated their experienced subjective fear. We hypothesized that decoding of spider images in the limbic and frontal regions, including the amygdala, the insula and the cingulate cortex, may be correlated with subjective fear and changes may be specifically modulated by the effects of glucocorticoids.
Materials and Methods
Participants
Thirty-six right-handed patients who fulfilled ICD-10 criteria for specific phobia for spiders and 29 healthy control participants were included in the study. Diagnosis of spider phobia was based on the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). We used a computer-based structured clinical interview (DIA-X), which was based on the Composite International Diagnostic Interview (CIDI; Robins, 1988). We assessed fear of spiders using the German version of the Spider Phobia Questionnaire (SPQ; Watts and Sharrock, 1984) and the Fear of Spider Questionnaire (FSQ; Szymanski and O’Donohue, 1995). Patients were excluded from the study if they met any of the following conditions: history of head injury, acute, or chronic medical conditions, a recent history of systemic or oral glucocorticoid therapy, psychiatric disorders other than specific phobia, psychotropic drug treatment, smoking more than 15 cigarettes per day, neurological diseases, current drug or alcohol abuse, pregnancy, use of hormonal contraceptives, or current behavioural therapy. Female participants were evaluated during the luteal phase of their menstrual cycle as previous studies showed that amygdala activation in response to psychological stressor depends on menstrual cycles (Chung et al., 2016) and that cortisol responses to stress are comparable between females in a luteal phase and males (Kirschbaum et al., 1999). After assessment, the included spider-phobic patients were randomly assigned to two groups according to a double-blind, placebo-controlled design. We carefully performed data quality checks and excluded three individuals in the cortisol group, three in the placebo group, and five controls from data analysis due incomplete data (12% excluded) or head movements (5% excluded); for details, see Supplementary Figure 1A. It is well-known that head movements can introduce spurious effects (Power et al., 2012). After exclusions, we analysed a final sample of 54 participants: 15 patients in the cortisol group, 15 patients in the placebo group, and 24 healthy controls (see Table 1 for demographic details). After providing a complete description of the study to the participants, written informed consent was obtained. The study was approved by the ethics committee of the Canton of Bern, Switzerland (Nr. 161/07) in accordance with the principles of the Declaration of Helsinki and the Swiss authority for pharmaceutical drugs (Swissmedic). The study was registered (ClinicalTrials.gov, NCT01574014).
Design and procedure
The experiments were conducted at the Bern University Hospital between 2 PM and 5 PM each day. Patients and healthy controls underwent the same experimental procedure except, for the diagnostic interview, substance administration, and collection of saliva samples, which only included the patients. Saliva samples were collected to control the effectiveness of the cortisol administration. Upon arrival, participants were informed about the procedure, asked to fill out the Spielberger State Anxiety Inventory (Spielberger et al., 1970), and rate their actual subjective fear, physical discomfort, and avoidance behaviour on a visual analogue scale (VAS) ranging from 0 (no symptoms) to 100 (maximal symptoms). The first saliva sample was collected in the patient group using a Salivette (Sarstedt Inc., Rommelsdorf, Germany). Participants were instructed regarding the picture task, and performed a few practice trials on the computer to become familiar with the rating procedure. After the oral administration of cortisol (20 mg; Galepharm, Küsnacht, Switzerland) or placebo to the patients, a resting period of 30 minutes followed. Sixty minutes after substance administration, the second saliva sample was collected, before the beginning of the fMRI task, to control for the cortisol level increase. Functional MRI images were acquired during the picture task (24 minutes). After the scanning session, the third saliva sample was collected from the patients; the level of cortisol was therefore measured at 3 time points. All participants completed questionnaires regarding anxiety before the scan. Additionally, participants were asked to retrospectively rate experienced fear on a visual scale from 0–100 while looking at the spider images in the scanner. A further questionnaire asked about side effects and whether the patient believed that he/she received cortisol or placebo. The saliva samples were stored at -20°C until required for biochemical analysis.
Paradigm
During the event-related experiment, participants viewed 80 randomized pictures from the International Affective Picture System (ISAP; Lang et al., 2008). We presented four categories (20 trials each) of images: phobic (spiders), other animals, negative, and neutral. The presentation time was five seconds, with inter-stimulus intervals (ISI) between 10.1–13.7 s (Supplementary Figure 1B). All participants rated their subjective fear after each trial on an analogue scale between 1 (no fear) and 4 (maximum fear).
Statistical analysis
Baseline variables and anxiety self-ratings were analysed using Kruskal-Wallis and Wilcoxon rank-sum tests. Cortisol levels were investigated using a 2 × 3 repeated measures ANOVA (group cortisol vs. placebo, and 3 time points). MVPA was performed with the “Searchmight” toolbox (Pereira and Botvinick, 2011) and a nonparametric ANOVA with SnPM (v. 13.1.05). For correlations Spearman’s rho was used. Statistical analysis was performed in R (v. 3.4.1); A P-value of < 0.05 was considered statistically significant and all test were two-tailed. All voxel-wise tests were corrected for multiple comparisons using FDR of 0.05. The analysis pipeline is illustrated in Supplementary Figure 1C.
Hormone analysis
We analysed free cortisol concentrations in saliva using commercially available chemiluminescence immunoassays (cortisol: CLIA; IBL-Hamburg, Germany). The inter- and intra-assay coefficients of variation were below 10%. The samples of all subjects were analysed in the same run to reduce error variance caused by intra-assay variability.
MRI data acquisition and pre-processing
Functional images were acquired with a 3T Siemens Magnetom Trio and a 12-channel head coil, using an interleaved EPI sequence (579 volumes; 37 slices; voxel, 3.6 × 3.6 × 3 mm3; gap thickness, 0 mm; matrix size, 64 × 64; FOV, 230 × 230 mm2; TR, 2500 ms; TE, 30 ms). For structural images, a high-resolution 3D T1-weighted imaging protocol (modified driven equilibrium Fourier transform, MDEFT) was used (176 sagittal slices; thickness, 1.0 mm; FOV, 256 × 256 mm2; matrix size, 256 × 256; voxel, 1 × 1 × 1 mm3; TR, 7.92 ms; TE, 2.48 ms). Pre-processing is illustrated in Supplementary Figure 1C. We performed standard pre-processing using SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and normalized to MNI space (2 × 2 × 2 mm3), except that data were not smoothed due to subsequent pattern analysis.
Multivoxel pattern analysis
We used whole-brain multivoxel pattern analysis (searchlight MVPA) with a classifier to investigate individual stimulus decoding on the subject level (Kriegeskorte et al., 2006). The resulting classification accuracy maps represent brain areas that decoded phobic content versus the three other categories (negative, animal, neutral) based on the BOLD response from multiple voxels. A major benefit of MVPA is its increased power to detect cognitive states (Haynes and Rees, 2006; Norman et al., 2006) compared to the standard mass-univariate approach. Multivariate approaches use the information from multiple voxels (patterns) to predict stimulus category based on the BOLD signal. Therefore, we can detect smaller regional differences compared to classical approaches. This shifts the interest of whether a single voxel responds differently in two conditions (mass-univariate) to whether a pattern of activity carries enough information to distinguish between two conditions (multivariate). Prior to MVPA, a GLM was performed for each trial, including a regressor for the single trial and a regressor coding for all remaining trials, which is best practice for the subsequent MVPA (Mumford et al., 2012). We also included a CSF, a WM, and six movement parameters and their first-order derivatives in the model. The resulting beta estimate maps of the individual trials were subjected to a whole-brain MVPA using a searchlight approach (Kriegeskorte et al., 2006) that involved a Gaussian Naive Bayes classifier with leave-one-sample-out cross-validation (Pereira and Botvinick, 2011). The searchlight involved a cube of 3 × 3 × 3 voxels (6 × 6 × 6 mm3). Classification was performed between 20 spider pictures and 20 randomly sampled pictures from the three other categories (negative, animal, and neutral). The classification step was performed 60 times (bootstrapping) using the same 20 spider images but always a different randomly sampled set of 20 negative, animal, and neutral pictures, and a whole-brain mean accuracy map was created for each subject. This accuracy map is the average percentage of correct classification of spider pictures versus that for the three other categories and can be related to the decoding of spider pictures.
First, we tested all the subjects (N = 54) with a voxel-wise one-way ANOVA with the three group labels as levels (non-parametric permutation/randomisation test with SnPM) and a FDR correction of 0.05 to identify regions that differed between the three groups; data were smoothed beforehand (FWHM = 8 mm). However, this analysis was only to generally test the hypothesis that fear relates regions are involved and does not account for the high spatial resolution of MVPA. Indeed, group analyses in MVPA can be challenging and classical approaches do not take into account individual spatial differences; hence summed binary significance maps have been suggested (Pereira and Botvinick, 2011). We performed this analysis for the cortisol and placebo group. We corrected the unsmoothed individual decoding maps with a FDR correction of 0.05 to get above chance decoding accuracies and applied a slight smoothing (FWHM = 2 mm); in these significant decoding maps, the average decoding accuracies for the 30 patients was 72% (range: 66% to 85%). We binarized the individual FDR corrected accuracy maps and summed them for each patient group, resulting in two count maps demonstrating brain areas that exhibited significant individual decoding in a specific number of subjects. To directly compare decoding in the cortisol to the placebo group, we subtracted the two maps, resulting in a maps with relative increase or decrease in the number of subjects with significant decoding (difference map). To threshold this map we performed a randomization test. We randomly sampled two groups each n = 15 from the pooled cortisol and placebo sample, with each group containing individuals from both groups. We summed the maps and created two count maps, one for each group, and subtracted the maps. We performed this k = 500 times to created a distribution under the null hypothesis of no difference between the cortisol and placebo group. To control the alpha error, we determined the upper and lower percentile (2.5% and 97.5%); thus only in 5% or less of the cases the the data are as extreme as these boundaries (alpha level). These boundaries were determined voxel-wise and used to threshold the original difference map. The average upper and lower thresholds across voxels were +3.6 subjects (SD: 0.51) and -3.4 subjects (SD: 0.54). We used the Harvard-Oxford cortical atlas to identify brain regions.
Results
Demographics, baseline fear and endocrine measures
Demographics are shown in Table 1. The three groups did not significantly differ in age between the three groups (Kruskal-Wallis test, χ2 = 0.80, df = 2, P = .67). There were a higher proportions of females in the two patient groups, however, the three group did not differ significantly regarding gender (χ2 = 2.73, df = 2, P = .26). There was also no difference regarding body mass index (Kruskal-Wallis test, χ2 = 1.03, df = 2, P = .60). The cortisol and placebo groups did not differ regarding spider phobia symptoms assessed by FSQ and SPQ at baseline before the experiment (Wilcoxon rank-sum test; FSQ: W = 99, P = .98; SPQ: W = 121.5, P = .29). At baseline the patients had significantly higher scores in spider phobic symptoms compared to healthy controls, who had no fear (Kruskal-Wallis test; FSQ: χ2 = 36.6, df = 2, P < .0001; SPQ: χ2 = 37.0, df = 2, P < .0001; Table 1). The cortisol group had 5.2 times higher cortisol levels at the beginning of the fMRI experiment, and 4.3 times higher levels at the end of the experiment compared to their baseline, while the placebo group showed no increase over time (Table 1) which was confirmed by a significant interaction group × time (repeated measures ANOVA; F4,98 = 46.6, P < .0001).
Subjective fear
During the fMRI task, patients (cortisol and placebo group) exhibited 2.1 times higher subjective fear in response to spider pictures compared to controls (Kruskal-Wallis rank sum test, χ2 = 28.8, df = 2, P < .0001), Figure 1A. Patients with cortisol treatment had a 9.8% decrease in subjective fear while looking at spider pictures compared to patients with placebo (Wilcoxon rank sum test, W = 47.5, P = .021), Figure 1A. We found a 2.4 times higher subjective fear in patients with spider-phobia compared to controls when subjects were asked to retrospectively rate their perceived fear in the scanner after the experiment (χ2 = 23.2, df = 2, P < .0001), see Figure 1B, and patients with cortisol had a subjective fear reduction of 13.3% compared to the placebo group (W = 42.5, P = .011). The three groups did not differ with respect to subjective fear while looking at the negative pictures (χ2 = 4.57, df = 2, P = .10), Figure 1C.
Cortical decoding of spider images
We first performed a voxel-wise test for group difference with the null hypothesis that decoding of phobic images is equal in all the three groups (non-parametric one-way ANOVA, FDR 0.05). We found the strongest group effects in the left anterior insula, the right anterior insula, and in the anterior cingulate gyrus (ACC) (Figure 2). These areas showed a higher median decoding in both patient groups (cortisol: insula left 57%, insula right 56%, ACC 59%; placebo: insula left 56%, insula right 55%, ACC 56%) compared to healthy controls (insula left 50%; insula right 50%, ACC 50%), see Figure 3A. The decoding magnitudes were averaged across a larger area of the anterior insula and the ACC but are locally much higher. Decoding in all three areas correlated with subjective fear rated during the scanner (Spearman’s rank correlation; insula l., ρ = .60, P < .0001; insula r., ρ = .64, P = .0001; ACC, ρ = .64, P < .0001), see Figure 3B.
We investigated significant decoding accuracies on the subjects level (FDR 0.05 corrected) and transformed these into binary maps that we summed across individuals of the placebo and the patient group (count maps, Supplementary Figure 2). We subtracted the count maps of the placebo and cortisol group to show group differences, and thresholded this map with a randomization test and alpha level of p = 0.05 (Figure 4AB, for more axial slices. see Supplementary Figure 3). In the cortisol group, an increased number of individuals showed decoding of spider images in the right precuneus cortex (+8 individuals; 12 cortisol; 4 placebo), the left central opercular cortex (+6; 8 cortisol; 2 placebo), the left cerebellum (+7; 8 cortisol; 1 placebo), the right parietal opercular cortex (+7; 9 cortisol; 2 placebo), see Table 2 for a complete list. In the placebo group, an increased number of individuals demonstrated higher decoding in the insula (+6; 10 placebo; 4 cortisol), the ACC (+5; 7 placebo; 2 cortisol), and the frontal pole (+6; 6 placebo; 0 cortisol), see Table 3 for a complete list.
Discussion
In this study, we investigated cortical decoding of spider images related to the administration of glucocorticoids in patients with spider phobia. The anxiolytic effect of glucocorticoids in phobic fear and anxiety disorders has previously been demonstrated (Aerni et al., 2004; Soravia et al., 2006) which can be beneficial for patients not responding to standard therapy. In this study, subjective fear was reduced by 10–13% in the cortisol group during viewing of phobic images compared to the placebo group. Such a reduction can be clinically relevant and, in combination, improves extinction-based psychotherapy (Bentz et al., 2010; de Quervain et al., 2011; Soravia et al., 2014).
The aim of this study was to investigate cortical changes in phobic information decoding after glucocorticoid administration. Decoding of phobic images in the anterior insula and the ACC was higher in the patients compared to the controls and was correlated with subjective fear. These areas are well-known key players in the processing of emotions and part of the fear network (Greco and Liberzon, 2016), particularly in specific phobia (Del Casale et al., 2012), and it was shown that exposure therapy reduces activity specifically in these regions (Hauner et al., 2012). In the placebo group, a larger number of individuals showed decoding in frontal regions: the insula, the ACC and the right lateral frontal pole. It has been demonstrated that these regions are involved in fear and memory processes: the insula and ACC are key player in the fear circuitry and its role has been well demonstrated in anxiety disorders (Shin and Liberzon, 2010). The frontal pole is involved in various cognitive functions, such as mentalizing, multitasking, but also episodic memory (Christoff and Gabrieli, 2000; Gilbert et al., 2006). It has been shown that the lateral frontal pole co-activates with the ACC and the anterior insula (Gilbert et al., 2010). Our result are consistent with the findings that symptom reduction achieved through CBT was associated with lower responsiveness in the bilateral insula and the ACC and a reduction in cerebral blood flow (Hauner et al., 2012; Soravia et al., 2016).
The cortisol group showed decoding in the precuneus, the anterior cerebellum and in the opercular cortex. The precuneus is a hub involved in multimodal, attentional and memory processes, and is also involved in the default mode network. Functional connectivity mapping of this hub showed negative connectivity with the amygdala (Zhang and Li, 2012). Thus, it may be possible that glucocorticoid related changes in the amygdala may moderate the precuneus cortex, but this needs to be further investigated. Anterior parts of the cerebellum, such as the vermis, are involved in memory formation and fear conditioning (Sacchetti et al., 2002). The left opercular cortex is related to auditory processing and speech.
A recent study with healthy participants showed that cortisol disrupts ventromedial prefrontal cortex functioning and its communication with other brain regions such as the cingulate cortex and parahippocampal gyrus (Kinner et al., 2016). This is a potential mechanism behind our finding that the cortisol group showed reduced decoding in the ACC. The insula is particularly involved in processing of emotionally salient information as it is strongly connected with limbic structures, the cingulate cortex, the amygdala, and also prefrontal regions. Additionally, several fMRI studies showed that successful cognitive-behavioral therapy (CBT) for spider phobia is accompanied by reduced CBF in the bilateral insula (Schienle et al., 2009, 2007). Hence, the acute administration of glucocorticoids and similarly CBT can reduce the hyperactivation in response to phobic stimuli in brain regions that are crucially involved in identifying fearful stimuli and generating fear response, such as the insula (Duval et al., 2015). Thus, the decoding areas in the placebo group are strongly consistent with previous findings of areas engaged in a fear response.
Pattern analysis is a method applied on the individual subject level. This yields individual decoding pattern that only partially overlap due to local functional decoding differences among individuals. It can therefore be challenging to find consistent results in classical voxel-wise group analyses, and such group analyses are generally not suggested with MVPA (Stelzer et al., 2013). We addressed this issue using count maps (Pereira and Botvinick, 2011) that were thresholded on the individual, and on the group level with a randomization test, providing robust statistics but also accounting more thoroughly for individual variation in decoding.
Some limitations of this study merit discussion. First, we present data from a sample with small subgroups, but this is not uncommon in neuroimaging studies involving clinical samples and drug administration. We excluded some subjects due to head movements; it is well known that head movements can confound results in neuroimaging (Deen and Pelphrey, 2012; Power et al., 2012). Correlations and functional connectivity seem especially sensitive to movement and can exhibit false effects in response. It is unknown to what extent pattern analysis can be confused by this, but it is possible in cases where the movement correlated with the experimental design. It is not surprising that some phobic patients moved their head while confronted with spider images. It is important to realize that small sample sizes may overestimate effects and have low reproducibility (Button et al., 2013), and careful statistics are required to control false positives. We addressed this problem by using a FDR 0.05 on the subject-level and a randomization test on the group level. Non-parametric test are suitable for small samples, do not make the assumption of normality, and are generally more conservative in showing significant effects. The gender imbalance is a limitation of this study: there are 46% males in the control group, but only 33% in the cortisol and 20% in the placebo group. An explanation is that spider phobia may be more common among females, and therefore, the control group should have been more carefully matched. However, our main result is the comparison between the cortisol and the placebo group, which both had a lower proportion of males that did not significantly differ.
In sum, this study elucidates cortical decoding patterns of phobic stimuli related to glucocorticoid treatment. Glucocorticoids can improve the treatment of patients because they inhibit the retrieval of fear memories while they enhance the consolidation of new corrective experiences achieved through exposure therapy (de Quervain et al., 2011, 2009; Soravia et al., 2014). We have identified frontal decoding (insula, ACC and frontal pole) in the placebo group, and posterior decoding (precuneus, cerebellum, and operculum) in the treatment group providing new insight into the information processing of the phobic brain after glucocorticoid administration.
Acknowledgments
This work was supported by the Swiss National Science Foundation (No. 124947, No. 162066) and by the Medical Faculty of the University of Bern (520.10). We thank Melanie Fisler, Joëlle Witmer, Basil Preisig, and Yvonne Renevey for their assistance. We thank all participants for their time and efforts.
Footnotes
Conflict of Interest: The authors declare no competing financial interests.