Cognitive Subtypes of Schizophrenia Characterized by Differential Brain Volumetric Reductions and Cognitive Decline

Key Points

Question  What is the extent to which brain volumetric differences further define intellectually based subgroups of people with schizophrenia?

Findings  In this study of 96 patients, cluster analytical techniques revealed 4 IQ-based subgroups in which all those assessed had significantly reduced inferior parietal volume relative to healthy controls; however, a severely deteriorated subgroup had the most extensive reduction of gray matter volume relative to all groups.

Meaning  Classifying the cognitive heterogeneity associated with schizophrenia can provide a platform to better characterize the neurobiological underpinnings of the illness and its treatment.

Importance  Cognitively distinct subgroups of schizophrenia have been defined based on premorbid and current IQ, but little is known about the neuroanatomical differences among these cognitive subgroups.

Objectives  To confirm previous findings related to IQ-based subgroups of patients with schizophrenia in an independent sample and extend those findings to determine the extent to which brain volumetric differences correspond to the IQ-based subgroups.

Design, Setting, and Participants  A total of 183 participants were assessed at the outpatient settings of Neuroscience Research Australia and Lyell McEwin Hospital from September 22, 2009, to August 1, 2012. Patients were classified using cluster analysis on the basis of current and premorbid IQ differences. Regional magnetic resonance imaging (MRI) brain volumes were compared among the IQ-based subgroups using analysis of covariance with intracranial volume and age as covariates.

Main Outcomes and Measures  Wechsler Adult Intelligence Scale, third edition, scores; Wechsler Test of Adult Reading scores; Positive and Negative Syndrome Scale scores; and MRI brain volumes.

Results  Ninety-six outpatients (mean [SD] age, 35.7 [8.4] years; age range, 18-51 years; 59 men) with schizophrenia or schizoaffective disorder and 87 healthy controls (mean [SD] age, 31.9 [8.4] years; age range, 20-50 years; 46 men) were studied. Sixty-two patients and 67 healthy controls underwent structural MRI of the brain. Cluster analyses revealed 25 putatively preserved patients (26%), 33 moderately deteriorated patients (34%), 27 severely deteriorated patients (28%), and 11 compromised patients (12%). Negative symptom scores were significantly worse in the severely deteriorated group relative to the putatively preserved group (F_2,82 = 13.8, P < .001, effect size [ES] = 1.40). Patient subgroups analyzed revealed significantly reduced inferior parietal volume relative to controls (F_3,113 = 9.7, P < .001, ES = 0.85-1.24). The severely deteriorated group had significantly reduced total hippocampal (mean [SEM], 8309.6 [175.0] vs 9024.0 [145.5]; P = .01), lingual gyrus (mean [SEM], 11 996.0 [531.5] vs 13 838.1 [441.9]; P = .05), and superior temporal sulcus (mean [SEM], 4697.8 [192.0] vs 5446.0 [159.6]; P = .05) gray matter volumes relative to the putatively preserved group (ES = 0.91-1.10).

Conclusions and Relevance  Using an independent sample, we obtained proportions in each IQ-based subgroup that were similar to our previous work. Inferior parietal volume reduction was characteristic of schizophrenia relative to controls, and the severely deteriorated IQ group had widespread volumetric reductions. Classifying cognitive heterogeneity in schizophrenia provides a platform to better characterize the neurobiological underpinnings of the illness and its treatment.

Quiz Ref IDCognitive deficits in schizophrenia, although heterogeneous in nature, are characteristic of the disease and predate illness onset.¹ Both neurocognitively impaired patients (74%) and putatively cognitively normal patients (26%) were previously identified.² Three distinct groups were identified based on the putative temporal pattern of IQ trajectories before and after illness onset: those patients who had a significant decrease from normal premorbid IQ estimates, those with normal premorbid IQ estimates who appeared to maintain or preserve their IQ, and those with low premorbid IQ estimates and global cognitive deficits.³ These 3 groups were labeled deteriorated, preserved, and compromised, respectively, and made up 51%, 25%, and 23% of the cohort, respectively. These findings were recently replicated in a large independent sample of 534 chronically ill patients with schizophrenia and 635 healthy controls.⁴ These IQ-based subgroups were also identified in similar proportions of individuals with first-episode psychosis.^5,6

Despite the identification of unique, intellectually based subgroups, the nature and extent of cognitive deficits in these subgroups remain somewhat unclear because investigators^3,7-10 have found that even those with putatively preserved IQ perform significantly worse than matched healthy individuals on a range of cognitive abilities. Conversely, some suggest that there is a subgroup of people with schizophrenia having putatively preserved IQ and minimal neuropsychological impairment who appear less symptomatic.¹¹ Thus, further clarification is needed around the extent and nature of IQ preservation and decrease in schizophrenia.

Extending the previous IQ-based classification work by establishing whether these different intellectual subgroups are associated with differences in brain structure would help determine whether these IQ-based classifications are related to underlying neurobiological differences. Although there is some evidence of regionally distinct neural abnormalities in these cognitive subgroups, previous studies^12,13 have been limited in sample size, and inconsistencies in the methods applied make direct comparisons difficult. Neuropsychologically impaired patients were different from healthy controls in relation to reduced hippocampal, amygdala, and thalamic volume and larger lateral ventricle size,¹² whereas neuropsychologically near-normal and impaired patients only differed in white matter volume in the sensorimotor and parietal-occipital regions. Others have found that a putatively preserved patient group had significantly less gray matter than controls in a region extending from the orbital prefrontal cortex to the anterior cingulate gyrus.¹³ Woodward and Heckers¹⁴ found that neuropsychologically normal patients with bipolar disorder and schizophrenia had regional changes in white matter but not gray matter compared with controls; however, neuropsychologically impaired patients (compromised and deteriorated) had reduced total brain volume and more widespread reductions in gray matter and white matter. Taken together, these studies^12-14 suggest that cortical or subcortical brain volumes may not be uniformly reduced in all people with schizophrenia, but instead they may differ with the degree of cognitive deterioration, such that a more widespread reduction in volume may be related to more extensive cognitive decline. Thus, given the paucity of studies that compare cognitively putatively preserved, deteriorated, and compromised subgroups on the basis of brain volumes, additional work is needed to confirm these preliminary findings.

The aim of the present study was to replicate previous findings³ identifying distinct IQ-based subgroups of patients in a cohort of 96 outpatients with schizophrenia or schizoaffective disorder and, more important, to extend that work by determining the extent to which brain volumetric differences further define these intellectually based subgroups. Our hypothesis was that the more severely deteriorated groups would have more extensive brain volume changes.

Outpatients with a diagnosis of schizophrenia or schizoaffective disorder were recruited into a treatment trial at 2 sites: Neuroscience Research Australia, Sydney, New South Wales (48 men and 26 women), and Lyell McEwin Hospital, Adelaide, South Australia (11 men and 11 women). All participants were between the ages of 18 and 51 years and had been receiving antipsychotics for at least 1 year before their entry into the study. Diagnostic and exclusion criteria are described in the eMaterial in the Supplement. Healthy adults between the ages of 20 and 50 years met the criteria for entry into the study and were recruited to the 2 sites (Sydney: 37 men and 31 women; Adelaide: 9 men and 10 women) as a comparison group. Participants were assessed from September 22, 2009, to August 1, 2012. All participants provided written informed consent, and the study was approved and monitored by the South Eastern Sydney and Illawarra Area Health Services, the University of New South Wales, and the Adelaide Health Service Human Research Ethics Committees. Data were deidentified and the code stored with identifying data in a separate file that was not accessible to all staff working on the study.

The eMaterial in the Supplement provides a description of the cognitive, symptoms, negative emotional state, quality of life, and functional outcome measures.

Only participants at the Sydney site underwent magnetic resonance imaging (MRI), which was performed using a 3-T Phillips Achieva magnet (Phillips Healthcare) with an 8-channel head coil at Neuroscience Research Australia, Randwick, New South Wales, from September 22, 2009, to August 1, 2012. T1-weighted, high-resolution anatomical scans were obtained with 1-mm section thickness and no gap for a total of 180 sections with a repetition time of 5.4 milliseconds, echo time of 2.4 milliseconds, and a field view of 256 mm. Scans were visually inspected for motion or other artifacts and for significant neuroanatomical abnormalities. The eMaterial in the Supplement provides the reasons and number of participants excluded from MRI. Sixty-two patients and 67 healthy controls underwent structural MRI of the brain. Scans were processed with FreeSurfer software, version 5.1.0 (Athinoula A. Martinos Center for Biomedical Imaging) on a Mac OSX 10.8 (Apple Inc). A detailed description of the FreeSurfer cortical reconstruction and volumetric segmentation method (including published references) is available at http://surfer.nmr.mgh.harvard.edu/. All processed scans were visually inspected for errors and, if necessary, manually edited and reprocessed in FreeSurfer by trained personnel. Volumetric data for each brain region were then calculated using the FreeSurfer Desikan-Killiany atlas. Volumes for each region of interest in the left and right hemisphere were summed, and the total volume was included in statistical analyses.

Both clinical IQ decline–based and empirical cluster analysis subgroups were formed,³ and the subgroups were compared on the basis of cognitive, symptomatic, and structural imaging variables. Given that the empirical clustering analysis method provides a relatively unbiased, data-driven approach and the 2 classification strategies produced similar results, the empirical classification methods (eMaterial in the Supplement) are featured with empirical results presented below, whereas clinical classification methods and results are reported in the eMaterial in the Supplement.

All data analyses were performed using SPSS statistical software, version 22, for Windows (IBM), with the α value set to .05. Antipsychotic dose was converted to mean daily chlorpromazine equivalent dose based on standard guidelines.^15,16 Demographic variables, mean current daily chlorpromazine equivalent dose, symptom severity, emotional states, quality of life, daily function measures (Positive and Negative Syndrome Scale [PANSS], Depression Anxiety and Stress Scale, Schizophrenia Quality of Life Scale total, and Medical Outcomes Study 36-Item Short-Form Health Survey, version 2), and the cognitive measures (Letter Number Sequencing, Controlled Oral Word Association Test, Logical Memory I and II, and Trail Making Test Form A) were compared as appropriate among groups using univariate analyses of variance with follow-up Scheffe pairwise comparisons. The compromised subgroup was not included in the statistical analyses because of the small numbers of patients in that group. False-positive results related to multiple comparisons were controlled using the Benjamini-Hochberg false discovery rate method.¹⁷ Effect size (ES) calculations were measured as Cohen d.

To determine whether there were regionally distinct brain volumetric differences among the cognitive subgroups and healthy controls, a series of univariate analyses of covariance were applied to total regional brain volumes with cognitive subgroup as the between-group factor and age and intracranial volume as covariates with false discovery rate corrections.¹⁷ Follow-up Tukey pairwise comparisons were conducted with Bonferroni corrections. The ES calculations were measured as Ψ root-mean-square standardized effect.

Quiz Ref IDNinety-six outpatients (mean [SD] age, 35.7 [8.4] years; age range, 18-51 years; 59 men) with schizophrenia or schizoaffective disorder and 87 healthy controls (mean [SD] age, 31.9 [8.4] years; age range, 20-50 years; 46 men) were studied. The final MRI sample for analysis consisted of 54 patients and 65 controls. The analyses yielded 4 IQ groups that defined 3 general IQ-based trajectories (ie, putatively preserved, deteriorated, and compromised). The 4 clusters included a putatively preserved group that consisted of 25 patients (26%) with a mean Wechsler Test of Adult Reading (WTAR) score (a premorbid IQ estimate) of 110.3 and a mean Wechsler Adult Intelligence Scale, third edition (WAIS-III), full-scale IQ (FSIQ) score of 107.8; 2 deteriorated groups, with one consisting of 33 patients (34%) with a mean WTAR score of 103.9 and a mean WAIS-III FSIQ score of 92.6 (moderate deterioration) and the other deteriorated group consisting of 27 patients (28%) having a mean WTAR score of 99.7 and a mean WAIS-III FSIQ score of 77.3 (severe deterioration); and a compromised group consisting of 11 patients (12%) with a mean WTAR score of 85.0 and a mean WAIS-III FSIQ score of 80.5.

Regarding demographic, cognitive, and symptom variables, there were many similarities and some differences among the IQ subgroups (Table 1). No significant differences were found in relation to sex and ethnicity ratios among the IQ subgroups. The patient subgroups did not differ significantly on age at onset, illness duration, or mean daily chlorpromazine equivalent dose. No significant differences were found in age among the groups with the exception that the moderately deteriorated group was significantly older than controls (ES = 0.65). All the groups had significant differences in education (ES = 0.67-1.73) with the exception of the controls vs the putatively preserved group and the moderately deteriorated vs severely deteriorated groups.

Quiz Ref IDThe putatively preserved subgroup was not significantly different from healthy controls on premorbid or current IQ. The 2 deteriorated groups, however, had significantly lower premorbid IQ relative to the putatively preserved group (ES = 1.14-2.03), and the severely deteriorated group had significantly lower premorbid IQ compared with healthy controls (ES = 1.19). Both deteriorated groups had significantly lower current IQ measures relative to both the controls and the putatively preserved subgroup (ES = 1.36-5.39). Although the putatively preserved group did not differ significantly from healthy controls on IQ, they had significant deficits relative to controls in relation to attention/processing speed, negative emotional states, quality of life, and functional outcome (ES = 1.03-1.70). The deteriorated groups performed significantly worse on all cognitive, negative emotional states, quality of life, and functional outcome measures relative to healthy controls (ES = 0.75-1.78) with the exception of verbal fluency, which in the moderately deteriorated group did not differ significantly from controls. The moderately and severely deteriorated groups performed significantly worse than the putatively preserved group on verbal memory and working memory (ES = 0.79-2.03). The severely deteriorated group performed significantly worse than the putatively preserved group on PANSS negative symptom severity (ES = 1.40) and verbal fluency (ES = 1.11). The severely deteriorated group performed significantly worse on verbal fluency (ES = 0.86) and had significantly higher PANSS negative symptom severity scores than the moderately deteriorated group (ES = 1.00). Few differences were found between the empirical and clinical grouping strategies, which were primarily limited to differences between the putatively preserved group and controls (premorbid IQ and working memory), deteriorated group and controls (verbal fluency), and putatively preserved and deteriorated groups (premorbid IQ and PANSS scores) (Table 1 vs eTable 1 in the Supplement).

Quiz Ref IDSignificant differences were found in brain volumes among the groups (Tables 2, 3, and 4, Figure, and eFigure in the Supplement). There were no strong (r < .19 for all), significant (P > .13 for all) correlations between any brain region and mean daily chlorpromazine equivalent dose (eTable 2 in the Supplement). The compromised subgroup was not included in the analysis of structural brain changes because of the small numbers of patients in this group (n = 5). The putatively preserved and both deteriorated groups had significantly lower inferior parietal volumes than controls (ES = 0.85-1.24). Both deteriorated groups had significantly decreased insula volume compared with controls (ES = 1.13-1.27). The severely deteriorated group had the most extensive abnormalities compared with controls, with significantly reduced volumes in many cortical regions and significant decreases in total cortex, total gray matter, cortical white matter (Table 2), hippocampus volumes, and increased lateral ventricles (ES = 0.59-1.31) (Table 4 and eFigure in the Supplement). The moderately deteriorated subgroup did not differ significantly from the putatively preserved subgroup in relation to any regional brain volumes. The severely deteriorated group had significantly smaller banks of superior temporal sulcus, lingual gyrus, and hippocampal volumes compared with the putatively preserved group (ES = 0.91-1.10) and significantly smaller lingual and supramarginal gyri and superior temporal volumes compared with the moderately deteriorated group (ES = 0.36-1.05). Some differences were found between empirical and clinical grouping strategies in relation to brain volumes, with most differences occurring between the deteriorated and control groups followed by the putatively preserved vs control groups (see the Elaborated Results portion of the eMaterial in the Supplement and compare Tables 2, 3, and 4 with eTables 3, 4, and 5, respectively, in the Supplement).

Generally consistent with previous findings,³ empirical cluster analysis of this independent sample revealed 4 distinct IQ-based subgroups: a putatively preserved subgroup of 25 patients, a compromised subgroup of 11 patients, and 2 deteriorated subgroups of 33 patients (moderately deteriorated) and 27 patients (severely deteriorated). Although 3 clusters were previously defined,³ based on the current analysis, we further subdivided the deteriorated group into 2 subsets. Compared with the previous results,^3,4 we identify smaller proportions of compromised patients and a larger proportion in the composite deteriorated group. In the clinical grouping method (eMaterial in the Supplement), the compromised group was also underrepresented compared with the previous studies of these patients^3,4 and other studies.^6,18,19

Despite having apparently stable IQ levels, the putatively preserved subgroup had deficits on attention/processing speed, a finding consistent with the previous studies of these patients.^3,5,20,21 In the present study, the 2 deteriorated groups were distinguished by a larger mean decline in IQ (11 vs 7 points), and the severely deteriorated group had additional deficits in verbal fluency, negative symptom severity, and cortical and subcortical brain volumes beyond that found in the moderately deteriorated group. These findings suggest that these deteriorated groups are distinct. Although the putatively preserved and moderately deteriorated groups differed significantly based on verbal and working memory measures, these 2 groups did not differ significantly on the basis of other cognitive measures or brain volumes. Thus, the deteriorated group, as originally presented in the previous work,³ may need to be revised on the basis of the IQ classification analysis results presented here, which are supported by the volumetric differences. Overall, our findings confirm that schizophrenia may be classified on the basis of distinct subgroups related to IQ deterioration from premorbid levels, which may account for the neurobiological heterogeneity present in the illness, as shown in other disorders²² in which temperament-based attention-deficit/hyperactivity disorder subgroups were related to resting-state functional connectivity MRI.

Lending further support to the existence of a distinct, severely deteriorated subset of patients with schizophrenia is our finding that this group had the largest and most widespread reductions based on cognitive, functional, emotional, and brain volumetric measures relative to healthy controls, which is generally consistent with the findings of previous studies.^12,14 The severely deteriorated group had a reduction in 22 of the 48 whole-brain, cortical, or subcortical regions examined and increased lateral ventricle size. These regions are among those most commonly reported to be decreased in schizophrenia, including the superior temporal gyrus, insula, hippocampus, and total gray matter.^23,24 In contrast, the putatively preserved group differed from the healthy controls only in inferior parietal volume, whereas the moderately deteriorated group had only significantly reduced inferior parietal and insula volumes relative to healthy controls. The inferior parietal cortex was the only region with significantly reduced volume across all 3 schizophrenia subgroups relative to the controls. The inferior parietal cortex is considered a heteromodal association area, which is essential to sensory integration.²⁵ Sensory integration dysfunction is one of the earliest and most ubiquitous symptoms of schizophrenia, and a review²⁶ of many studies found disruptions in connectivity, circuitry, and white matter integrity of the tracts within the inferior parietal region in schizophrenia.

Our present results are somewhat consistent with one of the few studies assessing structural differences on the basis of cognitive subtypes of schizophrenia,¹² revealing significant enlargement of the lateral ventricles and significant reductions in total hippocampal volume in the neuropsychologically impaired group but not the putatively preserved group relative to healthy controls. In a departure from previous work,^12-14 we did not find specific differences between the neuropsychologically near-normal subgroup vs the impaired on white matter volume, and we did not find significant volumetric differences between the putatively preserved subgroup and controls in the medial orbital frontal volumes.^12,13 Previous studies^12,13 used only memory or executive function to define their subgroups, which may have contributed to the different findings. Our present study found significant differences between putatively preserved and severely deteriorated patients in the lingual gyrus, banks of the superior temporal sulcus, and hippocampal volumes. The superior temporal sulcus has roles in social cognition and language processing,²⁷ whereas the lingual gyrus is associated with visual processing and word recognition.²⁸ Hippocampal and lingual gyrus volume are significantly reduced in schizophrenia^24,29; however, the present results are the first report, to our knowledge, of a significant volumetric difference between putatively preserved and deteriorated subgroups.

Quiz Ref IDOne limitation of our study pertains to the use of a retrospective premorbid intelligence measure. The WTAR is less accurate as a premorbid IQ estimate for people who have superior or highly impaired intellect. Some patients in the putatively preserved subgroup may have deteriorated from relatively high premorbid IQ levels, but their decline was undetected because of the inability of the WTAR to accurately assess their premorbid IQ level. Although antipsychotics may influence brain volumes in schizophrenia, our series of correlational analyses among brain regions and chlorpromazine equivalent dose suggested no confounding effect of antipsychotics on brain volumes. Underrepresentation of the compromised group may also be a limitation; however, at least 2 other studies^5,9 reported 10% to 15% of patients in the compromised IQ group. Nevertheless, inclusion of a larger compromised group would be informative in future studies.

The present study confirms and extends previous findings of distinct, intellectually putatively preserved, deteriorated, and compromised subgroups in schizophrenia in a relatively large independent sample. These IQ-based subgroups were differentially associated with the extent of gray matter volume loss; thus, classifying the cognitive heterogeneity associated with schizophrenia can provide a platform to better characterize the neurobiological underpinnings of the illness and its treatment.

Corresponding Author: Thomas W. Weickert, PhD, Neuroscience Research Australia, Barker Street, Randwick, New South Wales, Australia 2031 (t.weickert@unsw.edu.au).

Accepted for Publication: September 18, 2016.

Published Online: November 9, 2016. doi:10.1001/jamapsychiatry.2016.2925

Author Contributions: Ms Weinberg had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: C. S. Weickert, T. W. Weickert.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Weinberg, C. S. Weickert, T. W. Weickert.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Weinberg, Lenroot, Jacomb, Allen, Bruggemann, C. S. Weickert, T. W. Weickert.

Obtained funding: C. W. Weickert, T. W. Weickert.

Administrative, technical, or material support: All authors.

Study supervision: Galletly, C. S. Weickert, T. W. Weickert.

Conflict of Interest Disclosures: Dr Liu has conducted clinical trials for Lundbeck, Sunovion, Bristol-Myers, Janssen-Cilag, AstraZeneca, Reviva Pharmaceuticals, Envivo, and Bionomics; Dr Galletly has conducted clinical trials for Bristol-Myers Squibb, Janssen, ICON Clinical Research, Envivo Pharmaceuticals Inc, and she has received professional fees from Lundbeck and Janssen-Cilag; Dr Catts has been an advisory board member a sponsored educational speaker, and/or has conducted research projects for Janssen-Cilag, Eli Lilly, Lundbeck, Novartis, Pfizer, Bristol-Myers Squibb, Sanofi, Hospira, and AstraZeneca, he is a trustee for Psychosis Australia Trust and the Queensland Schizophrenia Research Foundation, and he also is a member of Clear Thinking Queensland; and Dr C. S. Weickert (partner to Dr T. W. Weickert) is on the advisory board of Lundbeck, Australia Pty Ltd, and she has received funds for research performed in collaboration with Astellas Pharma Inc, Japan. No other disclosures were reported.

Funding/Support: This work was supported by grant 568807 from the National Health and Medical Research Council (NHMRC) of Australia, the University of New South Wales, Neuroscience Research Australia, the Schizophrenia Research Institute (using infrastructure funding from the New South Wales Ministry of Health, the Macquarie Group Foundation), and the Australian Schizophrenia Research Bank supported by NHMRC, the Pratt Foundation, Ramsay Health Care, the Viertel Charitable Foundation, and the Schizophrenia Research Institute. Dr C. S. Weickert is a recipient of grant 1021970 from the National Health and Medical Research Council (Australia) Senior Research Fellowship.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Additional Contributions: We thank everyone involved in the Cognitive and Affective Symptoms of Schizophrenia Intervention study for their assistance, in particular, Michelle Hill, BSc(Hons), Julia Hill, BSc(Hons), and Bronwyn Overs, BPsych(Hons), Neuroscience Research Australia, for their work using FreeSurfer to process and/or present the structural MRI data. In addition, Maryanne O’Donnell, MBBS, FRANCZP, University of New South Wales School of Psychiatry, and Daniel Pellen, MBBS, FRANZCP, Neuroscience Research Australia, contributed to patient recruitment and patient management; Julia Langton, PhD, and Loretta Moore, BSc, Neuroscience Research Australia, administered cognitive and/or symptom assessments; Ans Vercammen, PhD, University of New South Wales School of Psychiatry, conducted some cognitive, symptom, and psychiatric assessments; and Ashley Skilleter, PhD, University of New South Wales School of Psychiatry, collected the imaging data. All were compensated for their work except for Drs O’Donnell and Pellen. We also sincerely thank the participants in this study.