An Introduction to the Human Connectome Project for Early Psychosis

Abstract

Background

The time following a recent onset of psychosis is a critical period during which intervention may be maximally effective. Studying individuals in this period also offers an opportunity to investigate putative brain biomarkers of illness prior to the long-term effects of chronicity and medication. The Human Connectome Project for Early Psychosis (HCP-EP) was funded by the National Institutes of Mental Health (NIMH) as an extension of the original Human Connectome Project’s approach to understanding the human brain and its structural and functional connections.

Design

The HCP-EP data were collected at 3 sites in Massachusetts (Beth Israel Deaconess Medical Center, McLean Hospital, and Massachusetts General Hospital), and one site in Indiana (Indiana University). Brigham and Women’s Hospital served as the data coordination center and as an imaging site.

Results

The HCP-EP dataset includes high-quality clinical, cognitive, functional, neuroimaging, and blood specimen data acquired from 303 individuals between the ages of 16–35 years old with affective psychosis (n = 75), non-affective psychosis (n = 148), and healthy controls (n = 80). Participants with early psychosis were within 5 years of illness onset (mean duration = 1.9 years, standard deviation = 1.4 years). All data and novel or modified analytic tools developed as part of the study are publicly available to the research community through the NIMH Data Archive (NDA) or GitHub (https://github.com/pnlbwh).

Conclusions

This paper provides an overview of the specific HCP-EP procedures, assessments, and protocols, as well as a brief characterization of the study participants to make it easier for researchers to use this rich dataset. Although we focus here on discussing and comparing affective and non-affective psychosis groups, the HCP-EP dataset also provides sufficient information for investigators to group participants differently.

Introduction

The Human Connectome Project (HCP) was initiated in 2009 by the National Institutes of Health (NIH) Blueprint for Neuroscience Research collaborative framework to accelerate progress in understanding the organization of the human brain by mapping neural pathways and connections that subserve human brain function.¹ The original HCP developed advanced protocols used to collect high-quality behavioral, cognitive, and multimodal neuroimaging data on a large sample of healthy young adults (n = 1206; ages 22–35). The HCP data and methods are shared publicly, and more than 1500 journal articles using HCP data or methods have been published to date.² In a later wave of initiatives, the NIH sought to build upon the original HCP by extending this approach to human brain diseases with the aim of acquiring the same high-quality data as in the original HCP and sharing it with the larger scientific community. There are now 14 Connectome Disease initiatives funded across NIH institutes with the goal of understanding brain structural and functional connectivity, and how this connectivity differs between healthy individuals and individuals with neuropsychiatric illnesses (https://www.humanconnectome.org/disease-studies). Here we describe the Human Connectome Project for Early Psychosis (HCP-EP), funded through this initiative.

Psychotic disorders, which include schizophrenia spectrum disorders and affective disorders with psychosis, are severe psychiatric illnesses often accompanied by pronounced functional decline and poor quality of life.³ These disorders also frequently lead to disability globally⁴ and represent a substantial economic and public health burden.^5,6 Psychosis is also associated with alterations in neurocircuitry and brain structure and function. These brain alterations are further associated with clinical symptoms, cognitive impairments, and functional decline, although findings in the early course of illness are relatively sparse.⁷

Importantly, the time closely following the onset of psychosis is a critical period during which interventions are more effective at improving outcomes related to symptom severity, treatment response, quality of life, and social and occupational functioning.^8–11 It is also noteworthy that within a few years of psychosis onset, there is already evidence of progressive brain structural alterations that worsen as patients become chronically ill, and are moderated by pharmacological treatments.^12–15 Thus, focusing on the early period following psychosis onset offers an opportunity to study putative brain biomarkers of illness prior to the long-term effects of both illness chronicity and medication use.

Accordingly, the HCP-EP aimed to acquire high-quality data to characterize the pathological substrates across the psychosis spectrum within the first 5 years following illness onset, defined as the time of initial psychosis symptom presentation determined by medical records and by raters during an evaluation of psychiatric history. In addition to focusing on the early stages of the illness course, it is critical to improve our understanding of how co-morbid and overlapping symptoms are related to different neurobiological alterations. The HCP-EP is a transdiagnostic sample that makes possible the study of potential biomarkers and neural mechanisms related to both early affective psychosis (mood disorders with psychosis) and non-affective psychosis (schizophrenia spectrum and other psychotic disorders). Affective and non-affective psychoses have several similarities, including common risk factors such as overlapping genetic variants, prenatal exposures, and childhood psychopathology.^16,17 Further, illnesses such as schizophrenia and bipolar disorder tend to co-aggregate in families.¹⁸ However, both common and distinct findings are reported in brain imaging studies across affective and non-affective disorders,^19–23 and the interpretation of these findings is complicated by substantial heterogeneity within diagnostic groups.^24,25

The main goals of the HCP-EP study were thus to create a dataset of multimodal clinical, cognitive, behavioral, functional, and structural neuroimaging, and blood specimen data to be used to investigate both individuals with affective and non-affective psychosis at the early stages of illness onset, and to make these data available to the scientific community. Although the current paper focuses primarily on discussing and comparing affective and non-affective psychosis groups, the HCP-EP dataset also provides the opportunity for investigators using this dataset to group participants differently (eg, by specific diagnosis) or to take a more dimensional approach in their analyses.

Methods

Recruitment Sites

Participants were recruited at 3 sites in Massachusetts (Beth Israel Deaconess Medical Center [BIDMC], McLean Hospital [McLean], and Massachusetts General Hospital [MGH]), and one site in Indiana (Indiana University [IU]) (figure 1). Study procedures and data collection and processing were matched across sites. Imaging data were initially collected at Brigham and Women’s Hospital (BWH) for all Massachusetts sites. McLean was added as an additional magnetic resonance imaging (MRI) scan site in April 2018. The Psychiatry Neuroimaging Laboratory at BWH, led by Dr Martha Shenton (contact Principal Investigator [PI]), served as the data coordination center. Recruitment and informed consent procedures were approved by the Partners Institutional Review Board Committee, which served as the single Institutional Review Board (IRB) of record for the Boston area sites. The IRB of record for IU was their IRB.

Fig. 1.

Overview of HCP-EP’s recruitment sites, data collection, data management, and quality assurance/quality control of the data, as well as data sharing with the National Institute of Mental Health Data Archive (NDA).

All sites recruited healthy controls using locally approved flyers and internet research recruitment methods, along with local referrals. The following clinical programs were used for the recruitment of early psychosis participants. In addition to the clinical programs described below, 2 early psychosis participants from the Boston sites were recruited through the Mass General Brigham Rally Recruitment website.

The Prevention and Recovery for Early Psychosis (PARC) Program is in Indianapolis, Indiana, and is an affiliate of IU. PARC is the only comprehensive early psychosis specialty program in central Indiana and serves a population of over 1 million people. Dr Alan Breier (Site-PI and MPI) founded PARC in 2009.

The BIDMC-Massachusetts Mental Health Center (MMHC) is an early psychosis program in Boston, Massachusetts. Dr Matcheri Keshavan (Site-PI; formerly Dr Larry Seidman) is the director of this program, which includes the Prevention of and Recovery from Early Psychosis (PREP) clinical outpatient program, the Center for Early Detection, Assessment and Response to Risk (CEDAR) high risk (“prodromal”) outpatient program, and Deaconess 4 (DEAC-4), an inpatient psychiatry unit.

The McLean OnTrack program in Belmont, Massachusetts, is a specialty first-episode psychosis program within the Psychotic Disorders Division at McLean Hospital. Dr Dost Öngür (Site-PI) is the Chief of the Division, and Dr Kathryn Lewandowski (Co-I) serves as Director of Clinical Programming for McLean OnTrack. Participants in the study from McLean were also referred by treating physicians from the McLean Psychotic Disorders inpatient, residential, and partial hospital programs.

The Massachusetts General Hospital (MGH) First Episode and Early Psychosis Program (FEPP) in Boston, Massachusetts, is a specialty program within the Psychosis Clinical and Research Program, led by Dr Daphne Holt (Site-PI) and Dr Oliver Freudenreich, in the outpatient department of MGH. Individuals meeting the study criteria for psychosis were also referred by treating physicians in the MGH inpatient unit.

Inclusion and Exclusion Criteria

Eligible participants were between the ages of 16 and 35 and able to provide informed consent or had a legally authorized representative or guardian provide informed consent. They were also fluent in English and willing to share de-identified data with the National Institute of Mental Health (NIMH) Data Archive (NDA). Those participants who were included in the study in the affective psychosis group met the criteria for having a DSM-5²⁶ mood disorder with psychosis diagnosis, including major depression with psychosis (single or recurrent episodes), or a bipolar disorder with psychosis (including most recent episodes of depressed and manic types). Participants with bipolar disorder without psychosis were not included in the study. The participants included in the non-affective psychosis group met the criteria for a DSM-5 Schizophrenia Spectrum and Other Psychotic Disorders diagnosis, including schizophrenia, schizophreniform, schizoaffective, psychosis not otherwise specified, delusional disorder, or brief psychotic disorder. For both early psychosis groups, onset was required to be within the last 5 years prior to study entry. The diagnosis was determined by trained raters, supervised by licensed psychiatrists or clinical psychologists through the administration of the Structured Clinical Interview for DSM-5 Research Version (SCID-5-RV) in conjunction with available medical records. Diagnoses were finalized using the SCID and other data (eg, chart reviews, collateral history) reviewed in the research team meetings, which included experienced clinicians. All participants were outpatients at the time of study entry. Participants were excluded if the above diagnoses were determined to be substance-induced or due to a medical condition.

Participants were excluded based on the following: met SCID-5-RV criteria for current severe substance use disorder within the last 90 days; had a positive pregnancy test; known IQ less than 70 based on medical history or an estimate from the Wechsler Abbreviated Scale of Intellegence-2nd Edition (WASI-II), 2-subtest version; insufficient English language ability; known medical history of Human Immunodeficiency Virus positive (HIV+) status; an active medical condition that affects the brain or cognitive functioning (eg, seizure disorder, epilepsy, stroke, traumatic brain injury, significant loss of consciousness, or other neurological disorder); received electroconvulsive therapy (ECT) in the last 12 months; or had MRI contraindications (ie, implantation of a pacemaker, medication pump, vagal stimulator, deep brain stimulator, transcutaneous electrical nerve stimulation unit, ventriculoperitoneal shunt, etc.). Additional exclusion criteria for the healthy control group included meeting criteria for any of the DSM-5 diagnoses listed above, current anxiety disorder, or a lifetime history of anxiety if the duration of illness was more than 12 months, and/or had been present in the last 12 months or required the use of medication. Healthy controls could be included if they had a major depressive disorder with only a single past episode (but not multiple episodes), a past alcohol use disorder, or a specific phobia. Healthy controls could not have bipolar disorder.

Other exclusion criteria for the control group included a first-degree family member diagnosed with a schizophrenia spectrum disorder, use of psychiatric medications at the time of study entry, or a lifetime history of psychiatric hospitalization. Healthy control recruitment was carefully matched to the psychosis group participants based on sex, age, handedness, and parental socioeconomic status. Healthy control ages were matched within 1–2 years of the affective and non-affective psychosis groups. [Note: differences were observed when the psychosis group was separated into affective and non-affective groups, see “Results” section.]

Materials and Measures

Participants completed screening, clinical interviews, self-report questionnaires, cognitive assessments, a blood draw, and multimodal neuroimaging scans. Protocols used for HCP-EP were similar to the HCP Lifespan Pilot Project (https://www.humanconnectome.org/study-hcp-lifespan-pilot) with certain modifications and additions for an early psychosis sample. Table 1 provides a comparison of the assessments in the HCP-EP and the HCP Lifespan Pilot Project. Clinical assessments and procedures, including select NIH Toolbox measures and additional cognitive assessments, took place for Boston area participants at BIDMC-MMHC, McLean, and MGH. MRI scans for the Boston area participants were conducted at BWH and McLean. Clinical and cognitive assessments as well as MRI scans for IU participants were conducted at the IU School of Medicine.

Table 1.

All Screening, Cognitive, and Clinical Assessments Collected as Part of the HCP-EP, and How Assessments Differ From the HCP Lifespan Pilot Project (+ Indicates an Assessment was Included)

Assessments	Included in the HCP-EP Protocol	Included in HCP Lifespan Pilot Project Protocol
Screening assessments
Consent	+	+
Demographics	+	+
Family interview for genetic studies (family psychiatric history)	+	−
Traumatic brain injury screen	+	+
Magnetic resonance imaging screen	+	+
Chlorpromazine (CPZ) equivalence	+	−
Handedness	+	+
Parental socioeconomic status	+	−
Pregnancy test	+	+
Structured Clinical Interview for DSM-5, Research Version (SCID-5-RV)	+	−
Olfactory Questionnaire	+	−
Cognitive Assessments
NIH Toolbox
Cognition
Picture sequence memory	+	+
Dimensional change card sort	+	+
Flanker inhibitory control and attention	+	+
Picture vocabulary	+	+
Pattern comparison processing speed	+	+
List sorting working memory	+	+
Oral reading recognition	+	+
Emotion
Self-report emotion questionnaires	+	+
Sensation
Words in noise	+	+
Odor identification	+	+
Taste	−	+
Dynamic visual acuity	+	−
Motor function
9-Hole Pegboard Dexterity	+	+
Grip strength	+	+
2 minute walk	−	+
4-meter walk	−	+
Other cognitive measures
Visual/cognition
Delay discounting	+	+
Penn emotion recognition	+	+
Penn progressive matrices	−	+
Electronic visual acuity	−	+
Auditory
Seidman Auditory Continuous Performance Test	+	−
IQ
Wechsler Abbreviated Scale of Intelligence-II, 2-subtest version	+	−
Function
Adult Self Report 18-59	−	+
Personality
NEO Five-Factor Inventory-60	−	+
Clinical and global functioning assessments
Positive and Negative Syndrome Scale (PANSS)	+	−
Clinical Assessment Interview for Negative Symptoms (CAINS)	+	−
Brief Negative Symptom Scale (BNSS)	+	−
Young Mania Rating Scale (YMRS)	+	−
Montgomery-Asberg Depression Rating Scale (MADRS)	+	−
MIRECC Global Assessment of Functioning (GAF)	+	−

Demographic, Psychiatric History, and Clinical Assessments.

Screening involved ascertainment of demographic information, family psychiatric history using the Family Interview for Genetic Studies,²⁷ history of traumatic brain injury, and contraindication for MRI scans. The Hollingshead Four-factor Index of Social Status²⁸ was administered to assess parental socioeconomic status. Additionally, all participants underwent an assessment of psychiatric diagnoses (see “Inclusion and Exclusion Criteria” above), lifetime medical history, lifetime and current substance use, lifetime and current antipsychotic medication history, and other medication use. To enhance the comparability of doses across different antipsychotic medications, chlorpromazine (CPZ) equivalent dosages were calculated. The literature in this field is complex and the conversion values differ, attributed to the use of varying methodologies. These include clinical consensus approaches,²⁹ the Minimum Effective Dose Method,³⁰ the Mean Dose Method,³¹ and the Defined Daily Dose Method.³² The conversion rates utilized in our study are derived from these established methods and detailed in Supplementary table 1. Recognizing the approximate nature of CPZ equivalence in reflecting medication dosages, we decided to also incorporate CPZ equivalence categories as proposed by.³³ This categorization divides participants into 4 groups: CPZ = 0 mg/day; CPZ < 300 mg/day; CPZ 300–1000 mg/day; and CPZ > 1000 mg/day.

Symptom severity was assessed using several interview-based measures. The Positive and Negative Syndrome Scale (PANSS³⁴) was used to assess symptoms of psychotic disorders including positive, negative, and general symptoms. Negative symptoms were further assessed using the Clinical Assessment Interview for Negative Symptoms (CAINS^35–37), which maps more closely to current formulations of the structure of negative symptom dimensions and is especially valuable for assessing affect and avolition/motivation. We additionally used the Brief Negative Symptom Scale (BNSS^38,39), because it assesses all 5 negative symptom domains (anhedonia, asociality, avolition, blunted affect, and alogia). The Young Mania Rating Scale (YMRS⁴⁰) and the Montgomery-Asberg Depression Rating Scale (MADRS⁴¹) were used to assess mania and depression symptoms, respectively. The Mental Illness Research, Education, and Clinical Center (MIRECC) version of the Global Assessment of Functioning (GAF) was included to provide global assessments of symptom severity, occupational, and social functioning, and is commonly used in early-stage psychosis studies.⁴² Note that the MIRECC GAF is a distinct version of the GAF,⁴³ which has the advantage that it separates social and occupational functioning from symptom severity, instead of combining these three measures into a single score.

Cognitive Assessments.

A subset of the NIH Toolbox measures,^44–46 which were used in the HCP Lifespan Pilot Project, were administered to test performance across multiple domains including Cognition (Picture Sequence Memory, Dimensional Card Sort, Flanker Inhibitory Control and Attention, Picture Vocabulary, Pattern Comparison Processing Speed, List Sorting Working Memory, and Oral Reading Recognition), Emotion (self-report emotion questionnaires), Sensation (Words-In-Noise, Odor Identification, and Dynamic Visual Acuity), and Motor function (9-Hole Pegboard Dexterity and Grip Strength). Additional assessments from the HCP Lifespan protocol included the Delay Discounting Task and the Penn Emotion Recognition Test.^47,48 Cognitive tests added to the HCP-EP protocol that were not used in the HCP Lifespan protocol included measures specific to the assessment of psychosis. These measures included the Seidman Auditory Continuous Performance Test^49,50 and the WASI-II, 2-subtest version (Vocabulary and Matrix Reasoning subtests), as an estimate of IQ.⁵¹ The Penn Progressive Matrices from the HCP Lifespan Pilot Project were not included, as this measure was redundant with the WASI-II. Electronic Visual Acuity was also not assessed due to redundancy with the NIH Toolbox Visual Acuity test and the requirement of a specific testing room without windows or windows with shades, which was not always available for a given testing session.

Blood Specimen Collection.

A minimum of 10 mL of blood was collected from participants and processed at the Center for Clinical Investigation at BWH or by trained study staff at McLean or IU. Blood samples were then shipped to the Rutgers University Cell and DNA Repository (RUCDR) Infinite Biologics Biorepository. Data are available to qualified researchers for future analyses in adherence with the NIH Genomic Data Sharing Policy through the NIH-funded data repository, NIMH Repository and Genomics Resource (NRGR).

Imaging Acquisition.

The imaging protocol used for HCP-EP included structural MRI, diffusion MRI, and resting-state functional MRI (rs-fMRI) collected during one imaging session. All participants were scanned using the same sequence on 3 Siemens MAGNETOM Prisma 3T scanners at BWH, McLean, or IU. BWH and IU used a 32-channel head coil; McLean used a 64-channel head and neck coil, with the neck channels turned off.

The imaging protocol used was the 2016 Connectome Coordination Facility (CCF) template protocol (https://www.humanconnectome.org/hcp-protocols-ccf-template). This protocol is based on the HCP Lifespan Pilot protocol, which has several modifications from the original HCP to shorten it for participants who may be at increased risk of fatigue and movement during the participants’ scans. Modifications included the exclusion of task-evoked fMRI and a shortened time interval for the diffusion-weighted MRI and the rs-fMRI scans. By selecting the imaging parameters of the CCF protocol, it is possible to harmonize and to compare the HCP-EP imaging data across HCP studies that are using the same or very similar imaging protocols.

Prior to scanning, participants completed a 20-minute MRI safety screening in which procedures were described. Participants were instructed to remain still during scanning and a deformable foam cushion was used to minimize head motion. Noise-attenuating headphones and ear stopples were used, which provided excellent noise reduction while permitting adequate auditory perception. Trained study staff reviewed all images in real-time on the scanning console for quality assurance. If there was a detectable problem, the scan was repeated. Blood was drawn during the MRI visit prior to the scan.

The specific imaging protocol took approximately 65 minutes to complete and included a localizer and auto-align scout, and structural T1w (MPRAGE; 0.8 mm isotropic; Repetition time [TR] 2400 ms; Inversion time 1000 ms; Flip angle 8°) and T2w (SPACE; 0.8 mm isotropic; TR 3200 ms; Echo time [TE] 563 ms) scans. Participants underwent about 23 minutes of rs-fMRI (2 mm isotropic; multiband (MB) acceleration × 8; TR 800 ms) acquired across four 5-minute 46 seconds scans (420 measurements in each), and in 2 blocks, each block with one scan in Anterior-Posterior (AP) and 1 in Posterior Anterior (PA) phase encoding. During the rs-fMRI scans, participants were asked to keep their eyes open and focus their gaze on a black cross against a light gray background. Participants also underwent 2 blocks of diffusion-weighted MRI that included a total of 4 scans (1.5 mm isotropic; TR: 3230 ms; TE: 89.2 ms; MB acceleration × 4). The diffusion scans included 92 directions with b = 3000 s/mm², 93 directions with b = 1500 s/mm², and 6 b = 0 volumes. These 197 volumes were identical to the CCF protocol. To sensitize the CCF protocol to faster diffusion components we included 9 additional volumes (3 directions in b = 200 and 6 directions in b = 500). In total, the diffusion scans included 206 volumes collected twice, once in AP and once in PA phase encoding, and split into 2 blocks. Field maps were acquired to correct for intensity and geometric distortions. Detailed imaging protocols can be found in Supplementary appendix A.

Imaging Quality Control and Pre-processing.

Quality assurance/quality control (QA/QC) procedures for MRI included the use of documented standard operating procedures (SOPs), including pre-acquisition scanner monitoring, automated verification of scan acquisition parameters, post-acquisition visual review by trained study staff, and semi-automated quality control procedures developed to detect signal drops in the diffusion-weighted MRI scan.

As multiple scanning sites were used to acquire the imaging data, several procedures to harmonize the data were implemented, including the use of Siemens-specific QA tools, Functional Biomedical Informatics Research Network (fBIRN), and the National Institute of Standards Technology (NIST) phantoms. To monitor the variability between acquisition sites, benchmark measurements of scanner performance were established using the fBIRN phantom and the NIST diffusion phantom, along with measurements of baseline intra- and inter-site variability. Additionally, 2 individuals were scanned at IU, BWH, and McLean Hospital, and a third person was scanned at both IU and BWH. Scanner stability was then assessed weekly by running Siemens-specific QA tools and scanning the fBIRN and NIST phantoms. Phantom QA scripts were automatically run within the database system and reports were generated in near real-time within the web interface of the database system. In the event of a scanner hardware and/or software change, these QA procedures were implemented immediately before and after the upgrade. If biases or variance increases were noted, sites worked with the relevant vendor field-service engineers to identify and correct the source of change.

Protocols further included automated verification of scan acquisition parameters and subsequent manual review, which included inspection for image contrast, blurriness, motion, and other artifacts. Each scan was assigned an overall quality rating on a 4-point scale (1 = poor, 2 = fair, 3 = good, 4 = excellent). If the reviewed session did not include at least one T1w scan and at least one T2w scan with ratings of 3 (“good”), or higher, repeat imaging scan sequences were requested. As this QC was done in almost real-time, this increased the likelihood that data of poor quality could be repeated within a short period of time (often on the same day or the next day). A further novel semi-automated quality control procedure developed at PNL/BWH, Slicer Diffusion QC, was used to detect signal drops in dMRI data. QA/QC reports were stored in the internal database and visible through the web interface to all of the project researchers.

Following the transfer of the imaging data to the Connectome Coordinating Facility (CCF) at Washington University, T1w and T2w data were preprocessed with the HCP Structural pipeline,⁵² including MSM-Sulc-registered versions of PreFreeSurfer, FreeSurfer, and PostFreeSurfer pipeline outputs, and intermediate files in the FreeSurfer v6 output directory. Details on each of the HCP pipelines can be found at: https://github.com/Washington-University/HCPpipelines.

Data Acquisition.

Data collection began in August 2016 and was completed in April 2022 (the last 2 years of collection occurred during the coronavirus disease 2019 [COVID-19] pandemic). Study procedures took place over 2–3 days based on scheduling and participant preference. The first visit included consent procedures, as well as lifetime psychiatric, family, and medical history using the FIGS and SCID-5-RV to determine eligibility for all participants. Administration of SCID-5-RV and clinical scales took approximately 2.5–3 hours to complete. Cognitive assessments took approximately 1.5–2 hours to complete. Assessment staff were extensively trained in the administration of all measures by clinical psychologists or psychiatrists with expertise in clinical and cognitive assessments. Due to the COVID-19 pandemic, most of the screening, clinical, and cognitive assessments were conducted remotely between March 2020 and the completion of the study in April 2022. Neuroimaging and blood collection continued in person. The average time between consent and screening assessments and MRI acquisition was less than 1 month (26 days), and 91% of participants had their MRI within 3 months of consent and screening assessments. Prior to the pandemic, all measures were conducted in person. Any screening measures that could not be done remotely occurred during the MRI study visit.

Data Management.

Clinical and behavioral data were stored centrally on the Harvard Catalyst Research Electronic Data Capture (REDCap) system at BWH, a secure web-based application designed to support data capture for research studies. All data shared across the project team were de-identified and did not contain personal health information (PHI), which was captured through a special REDCap form with stringent access privileges.

Imaging data were stored via a central Extensible Neuroimaging Archive Toolkit (XNAT⁵³) database system at BWH. This system was customized to host the project data, manage daily operations, and perform quality control procedures, as well as was automatically synched with the CCF XNAT database at Washington University. The MRI data were automatically stripped of PHI locally at BWH, McLean, or IU prior to transfer to the CCF.

Results

Sample

In total, 390 participants (289 early psychosis and 101 healthy controls) were consented, and 347 participants (256 early psychosis and 91 healthy controls) were enrolled in the study. Participants were included in the final sample if they had completed screening measures and at least 1 major clinical or cognitive measure. The final sample is comprised of 303 participants in total, made up of 75 individuals with affective psychosis, 148 individuals with non-affective psychosis, and 80 healthy controls (Supplementary figure S1). Healthy controls did not differ from the psychosis group as a whole on sex, age, handedness, or parental socioeconomic status, although there were differences when the psychosis group was separated into affective and non-affective (see table 2). All available scans (no pre-selection based on QC) will be shared via the NDA, and 288 participants have structural T1-weighted and T2-weighted MRI data (95%), 281 participants have diffusion MRI data (93%), and 287 participants have rs-fMRI data (95%). A subset of 205 participants have blood specimen data (68%). In addition, 167 early psychosis participants were taking antipsychotic medication at the time of study recruitment (75%). Compared to the affective psychosis participants, the non-affective psychosis participants were slightly younger (t(145) = 2.3, P = .02), had fewer females (X² = 10.7, df = 1, P = .001), a greater number of Black or African-American individuals (X² = 15.9, df = 3, P = .001), a lower estimated IQ (t(177.6) = 4.9, P < .001) (see figure 2), lower education (t(167.6) = 6.1, P < .001), and lower parental socioeconomic status (t(140.2) = −3.0, P = .003).

Table 2.

Characteristics of HCP-EP Participants.

	Affective Psychosis (n = 75)	Non-affective Psychosis (n = 148)	Healthy Controls (n = 80)	Statistic
	Number (%)
Recruitment site
BIDMC	6 (8%)	22 (15%)	20 (25%)	X 2 = 68.4, df = 6, P < .001
Indiana University	19 (25%)	94 (64%)	26 (32%)
McLean Hospital	49 (65%)	25 (17%)	26 (32%)
MGH	1 (1%)	7 (5%)	8 (10%)
Sex (% females)	41 (55%)	46 (31%)	37 (46%)	X 2 = 12.7, df = 2, P < .001 Aff, HC > NonAff
Race
Asian	4 (5%)	5 (3%)	10 (12%)	X 2 = 40.2, df = 6, P < .001
Black or African-American	8 (11%)	50 (34%)	7 (9%)
White	11 (15%)	25 (17%)	5 (6%)
Other	52 (69%)	68 (46%)	58 (72%)
Hispanic or Latino	6 (8%)	13 (9%)	9 (11%)	X 2 = 0.6, df = 2, n.s
Diagnosis
Schizophrenia		98 (44%)
Schizophreniform		12 (5%)
Schizoaffective disorder		26 (18%)
Psychosis not otherwise specified		6 (3%)
Delusional disorder		3 (1%)
Brief psychotic disorder		3 (1%)
Major depression with psychosis	12 (5%)
Bipolar disorder with psychosis	63 (28%)
Alcohol use disorder			2 (3%)
Major depression (single episode)			3 (4%)
Specific phobia			1 (1%)
Handedness
Right	65 (87%)	129 (87%)	66 (82%)	X 2 = 8.0, df = 6, n.s
Left	9 (12%)	11 (7%)	13 (16%)
Ambidextrous	1 (1%)	7 (5%)	1 (1%)
Missing		1 (1%)

	Mean (SD)
Age (y, 16–35)	24 (3.8)	22.8 (3.7)	24 (3.8)	F = 4, df = 2, P = 0.02 HC, Aff > NonAff
Duration of illness (y, 0–5)	1.7 (1.4)	2 (1.4)		F = 1.9, df = 1, n.s
Education level (1–11)	5.8 (1.5)	4.4 (1.8)	6.8 (2)	F = 48.6, df = 2, P < .001 HC > Aff > NonAff
Parental socioeconomic level	1.9 (1.1)	2.5 (1.3)	2 (1)	F = 6.2, df = 2, P = .002 NonAff > Aff, HC
WASI-II Full Scale IQ Score Estimate	111.8 (12)	102.3 (16.1)	116.5 (11.2)	F = 29.2, df = 2, P < .001 HC > Aff > NonAff
Positive PANSS Score	9.7 (3.9)	12.5 (4.6)		t = −4.7, df = 164.4, P < .001
Negative PANSS Score	10.9 (4.3)	13.9 (5.4)		t = −4.3, df = 171.5, P < .001
General PANSS Score	22.7 (5.8)	25.5 (5.6)		t = −3.4, df = 134.6, P < .001
CAINS Score	14.4 (8.8)	18.2 (11)		t = −2.6, df = 162.2, P = .009
BNSS Score	10.7 (12.3)	14.1 (14.2)		t = −1.5, df = 136.3, n.s
YMRS Score	5.4 (6.9)	6 (5.8)		t = −0.5, df = 117.1, n.s
MADRS Score	9.6 (10)	7.2 (7.4)		t = 1.8, df = 107.8, n.s
MIRECC GAF Symptom Severity	69.1 (18.6)	61.8 (17.8)	89.1 (9.7)	F = 73, df = 2, P < .001 HC > Aff > NonAff
MIRECC GAF Occupational Functioning	71.7 (22.7)	66.9 (22.7)	91.5 (7.2)	F = 40.1, df = 2, P < .001 HC > Aff, NonAff
MIRECC GAF Social Functioning	77.5 (14.7)	70.3 (14.8)	89.6 (10.2)	F = 51.1, df = 2, P < .001 HC > Aff > NonAff

Note: Two participants with early psychosis were recruited using the Mass General Brigham Rally Recruitment website and not through hospital psychosis programs. In addition to their main diagnoses, a subset of psychosis participants also had current or past attention-deficit/hyperactivity disorder (n = 7), obsessive-compulsive disorder (n = 9), alcohol use disorder (n = 33), cannabis use disorder (n = 52), other substance use (n = 8), panic disorder (n = 9), anxiety disorder (n = 18), specific phobia (n = 3), eating disorder or body dysmorphia (n = 4), post-traumatic stress disorder or other trauma/stress disorders (n = 18), and depressive or bipolar disorders (n = 12). Groups were compared in the Statistic column using ANOVAs or t-tests for continuous variables and chi-squared tests for discrete variables.

Aff, affective psychosis; BIDMC, Beth Israel Deaconess Medical Center; BNSS, Brief Negative Symptom Scale; CAINS, clinical assessment interview for negative symptoms; HC, healthy control; MADRS, Montgomery-Asberg Depression Rating Scale; MGH, Massachusetts General Hospital; MIRECC GAF, Mental Illness Research, Education, and Clinical Center version of the Global Assessment of Functioning; NonAff, non-affective psychosis; PANSS, Positive and Negative Syndrome Scale; WASI-II, Wechsler Abbreviated Scale of Intelligence-II; YMRS, Young Mania Rating Scale; education level is the highest education level obtained by the participant: 1 = Middle/Junior High School, 2 = High School, no degree, 3 = GED, 4 = High School, degree, 5 = Some University Courses, 6 = Associate’s Degree, 7 = Bachelor’s Degree, 8 = Some Graduate Level Courses, 9 = Master’s Degree, 10 = Doctorate Level Courses, and 11 = Doctorate Degree. Parental socioeconomic level was evaluated using the Hollingshead Four-Factor Parental Socioeconomic Status Scale,²⁸ which weighted parental/guardian education and occupation scores to obtain a single score (1 = highest SES classification, 5 = lowest SES classification).

Fig. 2.

Demographic characteristics of affective psychosis, non-affective psychosis, and healthy control groups, including (A) Wechsler Abbreviated Scale of Intelligence-II (WASI-II) estimated IQ score, (B) age, (C) race, and (D) sex. T-tests were used to compare differences between groups. Significance level indicators for (A) and (B): *<.05, **<.01, ***<.001, ****<.0001.

There were no differences in the age, sex, ethnicity, or handedness of participants between recruitment sites, although there were site differences in the number of affective psychosis, non-affective psychosis, and healthy control participants (X² = 492.2, df = 6, P < .001), as well as race (X² = 54.0, df = 9, P < .001), education (F = 10.4, df = 3, P < .001), parental socioeconomic status (F = 9.2, df = 3, P < .001), and estimated IQ of participants (F = 5.7, df = 3, P = .002) (see Supplementary table S2).

Initial comparisons of differences between the psychosis groups on clinical and global functioning assessments using t-tests showed that compared to the affective psychosis group, the non-affective psychosis group had higher (more severe) PANSS positive symptom scores (t(164.4) = −4.7, P < .001), PANSS negative symptom scores (t(171.5) = −4.3, P < .001), PANSS general symptom scores (t(134.6) = −3.4, P < .001), and CAINS negative symptom total scores (t(162.2) = −2.6, P = .009) (see figure 3). Further, the MIRECC GAF social functioning (t(139.6) = 3.4, P < .001) and the MIRECC GAF symptom severity score (t(133.2) = 2.8, P = .007) were higher (better) in the affective group compared to the non-affective group. The MADRS depression scores, YMRS mania scores, BNSS negative symptom scores, and MIRECC GAF occupational functioning scores did not differ significantly between the 2 early psychosis groups. As the affective and non-affective psychosis groups did not significantly differ in depression scores, we sought to investigate whether this was due to the inclusion of individuals with schizoaffective disorder in the non-affective psychosis group. Of the 148 participants included in the non-affective psychosis group, 26 participants (17.6%) had a DSM-5 diagnosis of schizoaffective disorder (see “Inclusion and Exclusion Criteria” section). The schizoaffective subgroup had significantly higher (more severe) MADRS depression scores (t(29.4) = −3.1, P = .005) compared to the rest of the non-affective psychosis group, although no significant differences were observed in the YMRS mania scores or BNSS negative symptom scores. When the schizoaffective subgroup was excluded, those remaining in the non-affective psychosis group (n = 122) showed significantly lower (less severe) MADRS depression scores compared to the affective psychosis group (t(105.7) = 2.6, P = .01). There were no differences between non-affective and affective groups in YMRS mania scores or BNSS negative symptom scores when the schizoaffective subgroup was excluded.

Fig. 3.

Clinical scores for affective and non-affective psychosis groups. T-tests were used to evaluate clinical score differences between affective and non-affective psychosis for each of the clinical assessments collected as part of HCP-EP including the PANSS, BNSS, CAINS, MADRS, and YMRS. Significance level indicators: *<.05, **<.01, ***<.001, ****<.0001.

Data Sharing

All data will be uploaded to the NDA and made available to the research community. These include diagnostic, clinical, cognitive, and imaging data. In addition to sharing the raw imaging data, minimally preprocessed structural T1w and T2w data will also be made available.

For the HCP-EP project, imaging data are processed and uploaded to the NDA by the CCF, while all other data are directly uploaded to the NDA by the HCP-EP data coordination center at BWH. Detailed guides and resources about accessing and downloading the HCP-EP data from the NDA can be found at: https://www.humanconnectome.org/study/human-connectome-project-for-early-psychosis/document/hcp-ep. Information pertaining to the HCP-EP file names and directory structure can be found in Supplementary appendix B. Upload of the final release is currently in progress and is scheduled to be completed by the end of 2024. Additionally, longitudinal data for a subset of HCP-EP participants will also be shared through the NDA in a separate collection entitled “Neuroprogression Across the Psychosis Spectrum in the Early Course of Illness” (NDA Collection ID: 3179; NIMH R01MH117012).⁷

Novel and Modified Neuroimaging Tools

Several novel and/or modified tools for neuroimaging were developed in parallel to data collection as part of this project, including tools for harmonization, analysis, and quality control. These tools can be downloaded from a public GitHub repository (https://github.com/pnlbwh). Specific tools are a harmonization algorithm for diffusion MRI data acquired across multiple sites (dMRIharmonization⁵⁴). Another tool developed is the Slicer Diffusion QC, a dMRI signal drop detection algorithm to facilitate quality control (https://github.com/pnlbwh/SlicerDiffusionQC). A workflow pipeline also developed by our group is available for automated MRI processing pipelines that join together individual processing modules (https://github.com/pnlbwh/luigi-pnlpipe). We also include tools developed prior to the HCP-EP project, which are available to the general research community. One such tool is UKF Tractography, which is a multi-tensor white matter tractography algorithm (ukftractography⁵⁵).

Discussion

The HCP-EP provides a rich resource of clinical, cognitive, behavioral, blood specimen, and multimodal MRI data for a large transdiagnostic sample of individuals within the first 5 years of psychosis onset. This valuable dataset extends the original HCP by applying similar protocols for the acquisition and QC of high-quality data with the goal of accelerating knowledge in a manner not previously possible. Progress has been slow in translating knowledge of the brain to new and more effective treatments for human brain diseases. Severe mental disorders, including psychotic disorders, are often devastating, resulting in significant disruptions during adolescence and early adulthood, and too often leading to chronic disability and early mortality.⁵⁶ There is thus a critical need to improve our knowledge of dysfunctions in structural and functional brain connectivity in these disorders and to translate this knowledge to treatment.

The HCP-EP provides a sample that is clinically comparable to other early psychosis samples with neuroimaging data, such as the Personalised Prognostic Tools for Early Psychosis Management (PRONIA) recent-onset psychosis sample.^57,58 However, HCP-EP participants are on the lower end of psychotic symptom severity and have had their first onset of psychosis within the past 5 years (mean illness duration = 1.9 years, standard deviation = 1.4 years). In contrast, PRONIA recent-onset psychosis participants have more severe psychotic symptoms on average compared with the HCP-EP study sample and had their first onset of psychosis within the past 2 years.

Similar to other transdiagnostic psychosis samples with more chronic symptoms, such as the Bipolar-Schizophrenia Network for Intermediate Phenotypes (B-SNIP), the HCP-EP non-affective psychosis group had more severe positive, negative, and general psychotic symptoms, as well as lower estimated IQ, lower global functioning, and less education as compared to the affective psychosis group.²⁴ Also consistent with the B-SNIP sample, in the current HCP-EP study, MADRS depression scores were elevated in the HCP-EP schizoaffective participants compared to the other non-affective psychosis participants.²⁴

The HCP-EP has several notable strengths. First, we collected multimodal data from a relatively large sample of people in the early stages of psychosis, when confounds such as medication and chronicity of illness are minimal. Additionally, this project involves careful clinical characterization of participants and the acquisition of high-quality, carefully QC’d data, including imaging data. Finally, the HCP-EP expands upon the innovation of the original HCP with the development and sharing of novel neuroimaging tools that advance our abilities to harmonize diffusion MRI data across different scanners, assess structural MRI scan quality, and analyze white matter tracts and microstructure.

While there are many strengths of the HCP-EP, several limitations should be noted. First, although there is an ongoing longitudinal study involving many of the HCP-EP participants, the current data are cross-sectional. This is important for certain psychosis diagnoses such as schizophreniform and brief psychotic disorder, as these may change for individuals over time. Second, by shortening the imaging protocol to adapt to the early psychosis population, we were unable to collect the breadth of imaging data acquired in the original HCP Young Adult Study. Third, imaging data were not collected at 1 site, but 3. Acquiring data across multiple sites is a technical challenge for harmonizing the data for further analyses. However, several procedures were used to address site effects. For example, a diffusion MRI harmonization method was developed and is one of the neuroimaging tools that is publicly available (https://github.com/pnlbwh/dMRIharmonization). Fourth, while participants were early in the course of their illness and treatment, many were already taking medication at the time of study entry. To address this, we collected detailed information about medication history (ie, type, dose, frequency, start and stop date), which can be included in planned analyses. Fifth, while participants in the psychosis and healthy control group were matched based on sex, age, handedness, and parental socioeconomic status, they were not matched on race, which will limit the investigation of racial differences in this early psychosis sample. Lastly, data collection was interrupted by COVID-19, which resulted in a slightly reduced sample size compared to the initial target size of 320 participants with psychosis and 80 healthy controls.

In summary, the HCP-EP is positioned to contribute to important discoveries about brain structure and connectivity underlying early affective and non-affective psychosis that can be used to inform the development of early and more personalized interventions. Future opportunities include integrating the HCP-EP data with other HCP datasets, such as the Young Adult and Lifespan datasets. Moreover, longitudinal data collected as part of a separate Neuroprogression Study (NDA Collection ID: 3179)⁷ will extend the scope of the HCP-EP to study how trajectories in neural structure, function, and connectivity are related to cognition and affective and non-affective psychosis symptomatology over time.

Supplementary Material

Supplementary material is available at https://academic.oup.com/schizophreniabulletin/.

Appendix A. Imaging protocols for the three sites from https://www.humanconnectome.org/study/human-connectome-project-for-early-psychosis/document/hcp-ep.

Appendix B. Information about the file structure from https://www.humanconnectome.org/study/human-connectome-project-for-early-psychosis/document/hcp-ep.

Funding

This work was supported by funding from the following National Institutes of Health (NIH) grant: U01MH109977 (Martha E. Shenton, Alan Breier, Sylvain Bouix, Daphne Holt, Matcheri Keshavan, Dost Öngür, Larry J. Seidman).