Associations Between Resting State Functional Brain Connectivity and Childhood Anhedonia: A Reproduction and Replication Study

Page 1 of 34 1 Title Page: 2 Associations Between Resting State Functional Brain Connectivity and 3 Childhood Anhedonia: A Reproduction and Replication Study 4 Authors: 5 Yi Zhou1 MSc 6 Narun Pat2 PhD 7 Michael C. Neale1 PhD 8 Affiliations: 9 1Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth 10 University 11 2 Department of Psychology, University of Otago, New Zealand 12 Corresponding Author Email: 13 [email protected] 14 15 Short/Running Title: 16 Anhedonia and Brain Connectivity in Children 17 18 6 Keywords: 19 Childhood Anhedonia, rsfMRI Brain Connectivity, Depressed Mood, Reproducibility, 20 Replicability, Development. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. Page 2 of 34 21 Abstract: 22 Background: Previously, a study using a sample of the Adolescent Brain Cognitive 23 Development (ABCD)® study from the earlier 1.0 release found differences in several 24 resting state functional MRI (rsfMRI) brain connectivity measures associated with 25 children reporting anhedonia. Here, we aim to reproduce, replicate, and extend the 26 previous findings using data from the later ABCD study 4.0 release, which includes a 27 significantly larger sample. 28 Methods: To reproduce and replicate the previous authors’ findings, we analyzed data 29 from the ABCD 1.0 release (n = 2437), in an independent subsample from the newer 30 ABCD 4.0 release (n = 6456), and in the full ABCD 4.0 release sample (n = 8866). 31 Additionally, we assessed whether using a multiple linear regression approach could 32 improve replicability by controlling for the effects of comorbid psychiatric conditions and 33 socio-demographic covariates. 34 Results: We could only replicate the significant association between anhedonia and the 35 Within Cingulo-Opercular network connectivity measure in an independent subsample 36 of the ABCD 4.0 data release. When using the larger full ABCD 4.0 sample, six out of 37 the eleven previously reported associations remained significant. Accounting for socio- 38 demographic covariates and comorbid conditions using multiple linear regression did 39 not improve replicability but allowed for the identification of specific and independent 40 effects of anhedonia on 16 rsfMRI connectivity measures in the full ABCD 4.0 release 41 sample. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 3 of 34 42 Conclusion: Replication of previous findings were limited. A multiple linear regression 43 approach helped resolve the specificity of rsfMRI connectivity associations with 44 anhedonia. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 4 of 34 62 Main Article Text 63 Introduction 64 Anhedonia is defined as a markedly diminished interest or pleasure in previously 65 enjoyable activities and is a transdiagnostic symptom that is a core component of major 66 depressive disorders (MDD) (1) and schizophrenia (SZN) (2). Symptoms of anhedonia 67 are also present in substance use disorders (3), PTSD (4), bipolar depression 68 (Gałuszko-Węgielnik et al., 2019), and ADHD (5). Anhedonia in children and 69 adolescents is a significant prognostic predictor of greater depression severity (6), 70 treatment resistant depression (7), and suicidal behaviors (8). 71 Functional neuroimaging approaches have been widely used to explore the 72 neurocircuitry of anhedonia (9). Functional brain connectivity is a measure of the 73 coactivation of different brain regions, which measures the degree of synchrony 74 between the blood oxygen level dependent (BOLD) signals across time between 75 regions in the brain (Lv et al., 2018). In other words, functional connectivity allows for 76 the characterization of networks of brain activity rather than activity in single brain 77 regions. Importantly, functional connectivity can be measured at rest. 78 While there have been many studies of brain activity and connectivity in 79 anhedonic adults, fewer have been conducted in children and adolescents. However, 80 findings from these studies generally converge on the significance of disruptions in the 81 reward, default mode, and salience networks (10–12). A recent study using the early 1.0 82 release of the Adolescent Brain Cognitive Development (ABCD) study data found 83 several resting state functional MRI (rsfMRI) brain network connectivity measures 84 associated with anhedonia in children aged 9-10 years old (13). Importantly, it was one . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 5 of 34 85 of the largest studies of anhedonia in children with a sample size of ~2,500 participants 86 including 215 children reporting past and/or present anhedonia. Presently, the latest 4.0 87 release of the ABCD study, which includes a significantly larger sample of participants 88 with neuroimaging and behavioral data (n = 11,878), has been made available. Given 89 that prior releases of the ABCD data are archived and publicly available, there is a 90 significant opportunity to both reproduce and replicate these findings. 91 We define reproducibility as the ability to achieve exactly the same results as a 92 previous study by using the same data and analytical approach, and replicability as the 93 ability to achieve the same (or similar) results as a previous study in a different dataset 94 (14). By our definition, reproducibility is better able to assess the consistency of results 95 while replicability is better able to assess the generalizability of those results. The aims 96 of this present study are to reproduce the previously reported associations between 97 rsfMRI connectivity and childhood anhedonia using the ABCD 1.0 release sample, and 98 to replicate those findings using an independent subset of the larger ABCD 4.0 release 99 sample, excluding participants from the ABCD 1.0 release sample. 100 Importantly, in depressive disorders, anhedonia is characterized as the loss of 101 pleasure and interest that is distinct from feelings of sadness or other dysphoric 102 moods(15). Thus, there is great need to elucidate the specific neurobiological 103 underpinnings associated with anhedonia, distinct from other comorbid symptoms, to 104 better understand the underlying brain dysfunction. Thus, we also aim to extend our 105 analyses and evaluate the specificity of rsfMRI connectivity associations with anhedonia 106 by evaluating the effects of significantly comorbid psychiatric symptoms and diagnoses. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 6 of 34 107 Methods and Materials 108 All of our analyses were performed in R (version 4.0.3) and Rstudio . We used 109 several scripts, including the utils.R and combat.R (16) scripts for data harmonization, 110 which were provided by the previous authors immediately upon request. The code and 111 data structures used for our study can be accessed from the associated NDA study. 112 The code we used can also be found at our Open Science Framework repository 113 (https://osf.io/vy85h/?view_only=0497385708874a6a9cce2bbfc5c30600). 114 ABCD Study Data 115 The ABCD® study is the largest longitudinal study of brain development in 116 children in the United States (https://abcdstudy.org/). The study has collected structural 117 and functional brain imaging measures as well as detailed psychiatric and behavioral 118 data from almost 12,000 children starting from when they were 9-10 years old. Notably, 119 data is released on a continuous basis. For this study, we used baseline data from the 120 ABCD 1.0 and ABCD 4.0 releases. 121 rsfMRI Connectivity Measures and Quality Control (QC) 122 Neuroimaging processing pipelines and analyses for the ABCD study are 123 reviewed by (17). Briefly, the functional scans include twenty minutes of resting-state 124 data acquired with eyes open and passive viewing of a crosshair (18). From the ABCD 125 Data Repository, we obtained rsfMRI connectivity measures which were constructed 126 using a seed-based correlational approach where regions of interest (ROIs) within 127 Gordon parcellations (19) were grouped together into predefined cortical networks. 128 Briefly, correlations between unique pairs of ROI’s were obtained and Fisher 129 transformed into z-statistics. Connectivity measures represent the averaged Fisher- . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 7 of 34 130 transformed correlations of all the unique pairs of ROIs either within a cortical network, 131 between cortical networks, or between cortical networks and subcortical regions. From 132 the ABCD Data Repository, we obtained rsfMRI connectivity measures between 19 133 subcortical regions and 12 cortical networks (data structure: mrirscor02), as well as 134 rsfMRI connectivity measures from within and between the 12 cortical networks (data 135 structure: abcd_betnet02). Thus, there were 228 (12 x 19) subcortical ROI vs. cortical 136 network rsfMRI connectivity measures and 78 (12C2 network pairs + 12 within network) 137 within/between cortical network rsfMRI variables, for a total of 306 rsfMRI connectivity 138 measures. 139 For QC, we used the IQC_RSFMRI_GOOD_SER variable, which represents the 140 number of rsfMRI runs that were complete, passed protocol compliance and QC, and 141 had field maps acquired within 2 scans prior to the run that were complete and passed 142 QC and protocol compliance. Like the previous authors’, we retained subjects who had 143 IQC_RSFMRI_GOOD_SER values greater than or equal to four. 144 For analyses using only ABCD 1.0 release sample, we also removed individuals 145 who were scanned by “Philips Medical Systems” MRI machines because of a post- 146 processing issue in the ABCD 1.0 release, which was resolved in later releases. When 147 working with ABCD 4.0 release sample, we retained all the subjects who were scanned 148 by Philips Medical Systems scanners because the post-processing errors identified in 149 the ABCD 1.0 release had been fixed for the ABCD 4.0 release. 150 Data Harmonization 151 Different MRI scanners were used across the 21 sites in the ABCD study. The 152 original authors harmonized the data across MRI scanners by using the ComBat tool . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 8 of 34 153 (combat.R) to adjust for batch effects due to the different scanners used. Here, we do 154 the same and harmonize the data separately for the subcortical ROI vs. cortical network 155 and within/between cortical network rsfMRI measures as it is possible these two 156 variable types may be affected by scanners differently(20). Note, before the 157 harmonization step, listwise deletion of subjects with any missing rsfMRI data was done 158 as data harmonization requires complete data. There were 23 different scanners used 159 in the study for the ABCD 1.0 release and 29 different scanners for the ABCD 4.0 160 release. Thus 23 and 29 batch effects were used to adjust the ABCD 1.0 and 4.0 161 releases, respectively. 162 Psychiatric Symptoms and Diagnoses 163 The focus of this study were past/present symptoms of anhedonia. However, we 164 were also interested in other psychiatric conditions that may be comorbid with 165 anhedonia. Psychiatric data were obtained from the youth (data structure: 166 abcd_ksad501) and parent (data structure: abcd_ksad01) Kiddie Schedule for Affective 167 Disorders and Schizophrenia (KSADS) data structures from the ABCD study. Using 168 both youth and parent KSADS items, we combined past and present items for the same 169 symptom or diagnosis and consolidated some items into a single variable. For example, 170 we consolidated 18 past and present suicide related diagnosis items into one single 171 suicide thoughts and behavior variable, which was similarly done in another study (21). 172 For psychiatric conditions besides anhedonia, we first selected the KSADS items 173 representing psychiatric diagnoses and not individual symptoms. However, in both the 174 Youth and Parents KSADS data, no diagnosis variables for major depressive disorder 175 (MDD) were available. Thus, we selected two MDD related symptoms (besides . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 9 of 34 176 anhedonia), irritability and depressed mood, from both the Youth and Parent data to be 177 used in our analyses. Similarly, no diagnostic variable was available for ADHD. Thus, 178 we created a representative variable, inattention_distracted_p, which is a combination 179 of two prevalent ADHD related symptom items (Symptom - Difficulty sustaining attention 180 since elementary school and/or Symptom - Easily distracted since elementary school) 181 from the parent KSADS data. 182 Statistical Analyses 183 Student’s and Bayes Factor T-Tests 184 Prior to statistical analyses, participants with missing data for psychiatric 185 symptoms/diagnoses were removed. We then proceeded to identify and remove outliers 186 for each rsfMRI connectivity measure as values 1.5 times greater than the interquartile 187 range (IQR) of values. 188 We performed Student’s T-Tests for each of the 306 rsfMRI measures between 189 controls and individuals with past and/or present psychiatric symptoms/diagnoses of 190 interest (the reference group consisted of individuals endorsing psychiatric 191 symptoms/diagnoses, such as anhedonia). We applied the Benjamini-Hochberg 192 adjustment for multiple testing corrections. Finally, we performed Bayes Factor T-Tests 193 for each of the rsfMRI measures and reported natural logarithms of the Bayes Factors 194 (lnBF). As with the previous authors’, lnBF values greater than 1.1 were considered 195 significant. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 10 of 34 196 Tetrachoric Correlations 197 Tetrachoric correlations are suitable for use with binary or categorical variables 198 as it assumes responses arise from an underlying normal distribution with thresholds 199 that delineate response categories. The tetrachoric() function from the psych R package 200 was used in our analyses. Tetrachoric correlations were used to estimate the 201 correlations between pairs of 34 binary psychiatric variables. 202 Multiple Linear Regressions 203 For multiple linear regression analyses, the ABCD rsfMRI data (from both 1.0 204 and 4.0 releases) were harmonized for MRI scanner using the ComBat tool as 205 previously described, except we also adjusted for batch effects with covariates (20). The 206 covariates included during data harmonization were: age, sex, race/ethnicity, 207 anhedonia, bipolar II, irritability, and depressed mood. 208 We performed linear mixed effects modeling using the lmer4 package in R, a 209 form of multiple linear regression, on the harmonized data. Individual rsfMRI 210 connectivity measures were modeled as outcome variables while age, sex, 211 race/ethnicity, anhedonia (reference group was the control group), depressed mood, 212 irritability, and bipolar II disorder were modeled as independent explanatory variables. 213 Age, sex, and race/ethnicity were considered potential confounding variables. The 214 independent effects of anhedonia, bipolar II, irritability, and depressed mood symptoms 215 on rsfMRI connectivity measures were assessed by identifying corresponding 216 statistically significant partial regression coefficients, after multiple testing corrections. 217 Family ID was included as a random effect to control for the non-independence of 218 values from participants who belonged to the same family. We adjusted for multiple . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 11 of 34 219 comparisons with the Benjamini-Hochberg method to control for a False Discovery Rate 220 of 0.05. 221 Results 222 Reproduction of previous findings 223 To reproduce the previous authors’ results, we used the ABCD 1.0 release 224 sample. Socio-demographic characteristics for this sample can be found in Table 1. 225 Note that although the two groups appear to exhibit differences in a few of these 226 characteristics, they were not controlled for statistically when we performed in our t-tests 227 in order to remain consistent with the previous authors’ approach. 228 Table 1. Comparison of Socio-demographic Measures Between Controls and 229 Those with Anhedonia in the ABCD 1.0 Release Sample. Measures Control (N = 2209) Anhedonia (N = 215) Statistic p-value % Male 51.8 54 0.286 0.593 Mean Age (months) 120.6 119.9 1.183a 0.238 % Asian 1.8 0.9 0.45 0.502 % Black 8.2 18.1 22.296 <0.001 % Hispanic 20 26.5 4.753 0.029 % Other 9.3 9.8 0.012 0.911 % White 60.8 44.7 20.386 <0.001 % Bipolar II 0.8 7.9 64.358 <0.001 % Depressed Mood 8.1 28.4 87.975 <0.001 % Irritability 4.7 26.5 148.192 <0.001 230 The proportion (%) of participants in each group for each measure are shown. Student’s 231 t-test was done to compare age (in months) between control and anhedonia groups. 232 Chi-square tests of independence were done for all other measures. Note there is a 233 slightly lower number of controls here than reported below due to the exclusion of . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 12 of 34 234 participants with missing sociodemographic and/or comorbid psychiatric syndrome and 235 diagnoses measures. 236 a Student’s t-statistic. 237 We were able to successfully reproduce their findings (13). To assess how well we 238 reproduced the previous authors’ results, we calculated Pearson correlations of the t- 239 statistics and lnBF values between our values and those reported by the previous 240 authors when comparing individuals with or without anhedonia (Fig. 1A and 1B). 241 Figure 1 – Reproduction of previous statistical analyses. Pearson correlations 242 between A) Student’s t-statistics and B) lnBF statistics derived from the previous 243 author’s analyses and those derived from our replication analyses. LnBF – natural 244 logarithm of Bayes Factors. 245 Like the previous authors, we identified 215 individuals who endorsed past 246 and/or present anhedonia and 2,222 controls who reported neither past nor present 247 anhedonia at the baseline timepoint. In line with the previous authors findings, 11 248 rsfMRI connectivity measures were found to be associated with anhedonia using the 249 lnBF(10) statistic (Table. 2; full table of results found in Supplementary Table. 1). To be 250 consistent with the previous authors, individuals with anhedonia were the reference 251 group. 252 Table. 2 Reproducing Results of rsfMRI Network Connectivity Measures 253 Associated with Anhedonia. rsfMRI Measure t-stat p-value p.adj lnBF Meancontrol SDcontrol Meananhedonia SDanhedonia DorsalAttentionLeftHippocampus -3.931 0 0.027 5.06 -0.129 0.095 -0.101 0.096 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 13 of 34 WithinRetrosplenialTemporal 3.644 0 0.042 3.989 0.525 0.152 0.485 0.143 CinguloParietalBrainStem -3.46 0.001 0.045 3.366 -0.001 0.091 0.023 0.095 DefaultDorsalAttention -3.441 0.001 0.045 3.29 -0.129 0.063 -0.113 0.063 SensorimotorHandBrainStem 3.304 0.001 0.051 2.854 0.11 0.106 0.084 0.096 SalienceLeftVentraldc -3.294 0.001 0.051 2.809 -0.12 0.092 -0.097 0.094 WithinCinguloOpercular 3.142 0.002 0.074 2.323 0.291 0.078 0.273 0.08 RetrosplenialTemporalRightCerebellumCortex 3.084 0.002 0.079 2.151 0.127 0.136 0.097 0.127 CinguloParietalRightPallidum 3.024 0.003 0.086 1.974 0.127 0.119 0.101 0.132 SensorimotorHandRightHippocampus 2.965 0.003 0.094 1.809 0.099 0.099 0.078 0.097 CinguloOpercularBrainStem 2.75 0.006 0.167 1.204 0.026 0.101 0.005 0.102 254 The results of statistically significant univariate Student’s T-tests and Bayes Factor T- 255 tests comparing differences in rsfMRI connectivity measures between controls and 256 individuals reporting anhedonia at baseline using the ABCD 1.0 release sample (n = 257 2437). Student’s t-statistics (t-stat), p-values, Benjamini-Hochberg adjusted p-values 258 (p.adj), natural logarithms of Bayes Factors (lnBF), means (Mean) and standard 259 deviations (SD) for the control and anhedonia groups are reported. lnBF values greater 260 than 1.1 were considered statistically significant. 261 Replication of previous findings 262 At the time of writing, the ABCD 4.0 release has been made available. We 263 wanted to replicate the previous findings by using the full cohort, excluding the subjects 264 used in the previous analyses, which is analogous to replication in an independent 265 sample. In this sub-sample of the ABCD 4.0 release (which excludes participants from 266 ABCD 1.0 release), we found 591 participants who endorsed past and/or present 267 anhedonia and 5,865 controls who did not. Socio-demographic characteristics for this 268 sample can be found in Table 3. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 14 of 34 269 Table. 3 Comparison of Sociodemographic Measures Between Controls and 270 Those with Anhedonia in the ABCD 4.0 Release, Excluding ABCD 1.0 Release, 271 Sub-sample. 272 Measures Control (N = 5863) Anhedonia (N = 591) Statistic p-value % Male 51.2 56.7 6.202 0.013 Mean Age (months)a 118.5 119.1 -1.705 0.089 % Asian 2.3 0.8 4.703 0.03 % Black 15.2 20.8 12.449 <0.001 % Hispanic 20.5 26.4 11.087 0.001 % Other 10.4 12.9 3.055 0.08 % White 51.6 39.1 33.287 <0.001 % Bipolar II 0.3 8.3 313.34 <0.001 % Depressed Mood 6.4 31.8 431.052 <0.001 % Irritability 4.4 27.9 484.563 <0.001 273 The proportion (%) of participants in each group for each measure are shown. Student’s 274 t-test was done to compare age (in months) between control and anhedonia groups. 275 Chi-square tests of independence were done for all other measures. Note there is a 276 slightly lower number of controls here than reported above due to the exclusion of 277 participants with missing sociodemographic and/or comorbid psychiatric syndrome and 278 diagnoses measures. 279 a Student’s t-statistic. 280 When comparing the groups using Student’s and Bayes Factor t-tests, 18 rsfMRI 281 connectivity measures were found to be significantly associated with anhedonia 282 according to the lnBF statistic (Table. 4; full table of results found in Supplementary 283 Table 2). However, only the w ithinCinguloOpercular network rsfMRI connectivity 284 identified by the previous authors, was also found to be significant. 285 Table. 4 Replicating Results of rsfMRI Connectivity Measures Associated with 286 Anhedonia. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 15 of 34 rsfMRI Connectivity Measure t-stat p-value p.adj lnBF Meancontrol SDcontrol Meananhedonia SDanhedonia VisualLeftHippocampus 4.328 0 0.005 6.273 0.056 0.089 0.039 0.09 CinguloOpercularLeftAmygdala 4.02 0 0.009 5.002 0.07 0.094 0.053 0.096 RetrosplenialTemporalRightThalamusProper 3.852 0 0.012 4.338 0.158 0.083 0.144 0.087 AuditoryRightPutamen 3.574 0 0.022 3.321 0.124 0.107 0.107 0.107 WithinSensorimotorHand -3.507 0 0.022 3.092 0.271 0.071 0.282 0.072 CinguloParietalLeftThalamusProper 3.498 0 0.022 3.053 0.224 0.093 0.21 0.091 SensorimotorHandRightPutamen 3.477 0.001 0.022 2.98 -0.011 0.105 -0.027 0.112 RetrosplenialTemporalRightVentraldc 3.367 0.001 0.028 2.611 0.018 0.093 0.004 0.094 VentralAttentionLeftCerebellumCortex -3.332 0.001 0.028 2.491 -0.004 0.076 0.007 0.08 SensorimotorHandLeftPallidum 3.315 0.001 0.028 2.436 0.074 0.098 0.06 0.101 CinguloOpercularLeftCaudate 3.274 0.001 0.029 2.306 0.018 0.092 0.004 0.094 DorsalAttentionRightAmygdala -3.252 0.001 0.029 2.226 -0.009 0.078 0.002 0.076 VisualRightPallidum 3.236 0.001 0.029 2.183 -0.032 0.076 -0.043 0.075 CinguloOpercularRightAmygdala 3.213 0.001 0.029 2.111 -0.014 0.105 -0.029 0.11 CinguloParietalRightCerebellumCortex 3.087 0.002 0.041 1.708 0.163 0.108 0.148 0.11 SensorimotorHandRightAccumbensArea 3.068 0.002 0.041 1.661 0.224 0.093 0.211 0.091 AuditorySensorimotorHand -3.048 0.002 0.042 1.603 0.12 0.057 0.128 0.058 CinguloParietalLeftCaudate 2.987 0.003 0.048 1.407 0.259 0.079 0.249 0.077 WithinCinguloOperculara 2.968 0.003 0.048 1.35 0.299 0.067 0.29 0.066 AuditorySalience 2.931 0.003 0.052 1.244 0.042 0.067 0.034 0.068 287 The results of statistically significant univariate Student’s T-tests and Bayes Factor T- 288 tests comparing differences in rsfMRI connectivity measures between controls and 289 individuals reporting anhedonia at baseline using an independent subsample of the 290 ABCD 4.0 release, excluding ABCD 1.0 release, data (n = 6456). Student’s t-statistics 291 (t-stat), p-values, Benjamini-Hochberg adjusted p-values (p.adj), natural logarithms of 292 Bayes Factors (lnBF), means (Mean) and standard deviations (SD) for the control and 293 anhedonia groups are reported. lnBF values greater than 1.1 were considered 294 statistically significant. 295 a replicated association between rsfMRI connectivity measure and anhedonia. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 16 of 34 296 To increase our power to detect genuine associations, we next performed our 297 analyses on the full ABCD 4.0 release sample, including participants from the ABCD 1.0 298 release. In the full ABCD 4.0 release sample, there were 800 participants who endorsed 299 past and/or present anhedonia and 8066 controls who did not. Socio-demographic 300 characteristics for this sample can be found in Table 5. 301 Table 5. Comparison of Sociodemographic Measures Between Controls and 302 Those with Anhedonia in the Full ABCD 4.0 Release Sample. Measures Control (N = 8064) Anhedonia (N = 800) Statistic p-value % Malea 51.3 55.8 5.589 0.018 Mean Age (months) 119.1 119.3 -0.845 0.398 % Asiana 2.2 0.9 5.363 0.021 % Blacka 13.3 19.8 25.19 <0.001 % Hispanica 20.4 26.6 16.579 <0.001 % Other 10.1 12.1 2.916 0.088 % Whitea 54 40.6 52.043 <0.001 % Bipolar IIa 0.5 7.9 347.62 <0.001 % Depressed Mooda 6.9 30.9 507.29 <0.001 % Irritabilitya 4.4 27.4 627.02 <0.001 303 The proportion (%) of participants in each group for each measure are shown. Student’s 304 t-test was done to compare age (in weeks) between control and anhedonia groups. Chi- 305 square tests of independence were done for all other measures. Note there is a slightly 306 lower number of controls here than reported below due to the exclusion of participants 307 with missing sociodemographic and/or comorbid psychiatric syndrome and diagnoses 308 measures. 309 a Student’s t-statistic. 310 Notably, 6 out of the 11 rsfMRI connectivity measures identified by the previous 311 authors were also significantly associated with anhedonia in this analysis (Table. 6; full . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 17 of 34 312 table of results found in Supplementary Table 3). As well, 17 out of the 20 significant 313 rsfMRI connectivity measures identified in the independent sub-sample of the ABCD 4.0 314 release (from Table. 4) were also found to be significant in the analyses using the full 315 ABCD 4.0 release sample, including the within CinguloOpercular network connectivity 316 measure (Table. 6). 317 Table. 6 Significant Associations Between rsfMRI Connectivity Measures and 318 Anhedonia in the Full ABCD 4.0 Sample. rsfMRI Connectivity Measure t-stat p-value p.adj lnBF Meancontrol SDcontrol Meananhedonia SDanhedonia RetrosplenialTemporalRightThalamusProperb 4.962 0 0 9.057 0.161 0.082 0.146 0.086 CinguloOpercularLeftAmygdalab 4.252 0 0.003 5.816 0.072 0.093 0.057 0.094 VentralAttentionLeftPutamen -4.207 0 0.003 5.626 -0.055 0.065 -0.045 0.067 AuditoryRightPutamenb 4.085 0 0.003 5.125 0.125 0.106 0.109 0.105 SensorimotorHandRightPutamenb 4.029 0 0.003 4.902 -0.01 0.104 -0.026 0.108 SensorimotorHandLeftPallidumb 4.027 0 0.003 4.893 0.076 0.098 0.061 0.099 WithinCinguloOpercularab 3.865 0 0.004 4.254 0.303 0.066 0.293 0.066 VentralAttentionLeftCaudate 3.839 0 0.004 4.153 0.036 0.066 0.026 0.07 CinguloParietalLeftThalamusProperb 3.836 0 0.004 4.142 0.224 0.092 0.211 0.092 SensorimotorHandRightAccumbensAreab 3.787 0 0.005 3.961 0.225 0.093 0.212 0.092 VentralAttentionLeftCerebellumCortexb -3.762 0 0.005 3.865 -0.007 0.075 0.003 0.08 VisualLeftHippocampusb 3.691 0 0.006 3.605 0.056 0.088 0.044 0.089 DefaultDorsalAttentiona -3.604 0 0.007 3.284 -0.127 0.051 -0.121 0.051 CinguloOpercularRightAmygdalab 3.601 0 0.007 3.281 -0.013 0.103 -0.027 0.106 RetrosplenialTemporalRightCerebellumCortexa 3.52 0 0.009 2.993 0.128 0.106 0.114 0.106 WithinSensorimotorHandb -3.513 0 0.009 2.975 0.271 0.07 0.28 0.072 WithinVisual 3.411 0.001 0.012 2.61 0.409 0.089 0.398 0.09 CinguloOpercularBrainStema 3.358 0.001 0.013 2.436 0.025 0.074 0.016 0.074 CinguloOpercularLeftCaudateb 3.315 0.001 0.013 2.3 0.02 0.09 0.008 0.093 VisualRightPallidumb 3.31 0.001 0.013 2.283 -0.03 0.076 -0.039 0.076 DorsalAttentionRightVentraldcb 3.312 0.001 0.013 2.282 0.046 0.067 0.038 0.069 FrontoParietalVisual -3.307 0.001 0.013 2.266 -0.107 0.044 -0.102 0.044 SensorimotorHandBrainStema 3.268 0.001 0.015 2.136 0.129 0.08 0.119 0.079 SalienceLeftVentraldca -3.246 0.001 0.015 2.07 -0.121 0.071 -0.113 0.072 FrontoParietalLeftCerebellumCortex -3.228 0.001 0.015 2.011 -0.063 0.065 -0.055 0.068 DorsalAttentionLeftVentraldc 3.189 0.001 0.017 1.885 0.27 0.116 0.256 0.119 VentralAttentionLeftThalamusProper 3.154 0.002 0.017 1.773 -0.096 0.11 -0.109 0.113 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 18 of 34 CinguloParietalLeftCaudateb 3.149 0.002 0.017 1.76 0.26 0.078 0.251 0.076 FrontoParietalRightPallidum -3.148 0.002 0.017 1.755 -0.035 0.058 -0.029 0.058 RetrosplenialTemporalRightVentraldcb 3.114 0.002 0.018 1.657 0.021 0.092 0.01 0.095 FrontoParietalLeftCaudate 3.111 0.002 0.018 1.645 0.084 0.044 0.079 0.044 DorsalAttentionRightAmygdala -3.108 0.002 0.018 1.629 -0.012 0.077 -0.003 0.076 CinguloOpercularLeftAccumbensArea 3.091 0.002 0.019 1.581 0.111 0.077 0.102 0.076 SalienceVentralAttention 3.04 0.002 0.021 1.423 0.089 0.061 0.082 0.062 DorsalAttentionLeftAmygdala -3.035 0.002 0.021 1.408 -0.076 0.064 -0.068 0.063 319 The results of statistically significant univariate Student’s T-tests and Bayes Factor T- 320 tests comparing differences in rsfMRI connectivity measures between controls and 321 individuals reporting anhedonia at baseline using the full ABCD 4.0 release sample (n = 322 8866). Student’s t-statistics (t-stat), p-values, Benjamini-Hochberg adjusted p-values 323 (p.adj), natural logarithms of Bayes Factors (lnBF), means (Mean) and standard 324 deviations (SD) for the control and anhedonia groups are reported. lnBF values greater 325 than 1.1 were considered statistically significant. 326 a rsfMRI connectivity measures also reported by the previous authors as significantly 327 associated with anhedonia. 328 b rsfMRI connectivity measures that were also found to be significant in the replication 329 analyses. 330 It is important to note that individuals reporting anhedonia may also report other 331 symptoms or psychiatric diagnoses. Thus, it is important to evaluate the specificity of 332 the associations between rsfMRI connectivity and anhedonia. In order to identify 333 psychiatric conditions significantly comorbid in individuals reporting anhedonia, we 334 performed tetrachoric correlations between anhedonia and 33 additional psychiatric 335 diagnoses and symptoms collected at baseline (Supplementary Figure. 1). Three 336 psychiatric conditions exhibited correlation coefficients greater than or equal to 0.5 with 337 anhedonia: irritability, depressed mood, and bipolar II disorder. In the full ABCD 4.0 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 19 of 34 338 release sample, 32%, 28%, and 8% of the participants reporting anhedonia also 339 reported depressed mood, irritability, and bipolar II disorder (Supplementary Table. 4). 340 We next performed student’s and Bayes Factor t-tests to compare rsfMRI 341 connectivity between individuals with and without depressed mood, irritability, and 342 bipolar II, respectively (Supplementary Tables. 5-7). We found that several rsfMRI 343 connectivity measures associated with anhedonia were also associated with these other 344 psychiatric conditions. 345 Multiple Linear Regression Approach 346 While we were able to characterize which rsfMRI connectivity measures were 347 specifically associated with anhedonia vs. shared with other psychiatric conditions, 348 simple t-tests were not able to estimate the independent effects of each psychiatric 349 condition on rsfMRI connectivity. Furthermore, t-tests are not able to control for 350 potentially confounding sociodemographic variables such age, sex, and race/ethnicity. 351 Chi-square tests of independence showed significant differences in race/ethnicity and 352 sex between individuals reporting anhedonia and those who do not, across the different 353 samples used in our analyses (Tables 1, 3, and 5). 354 By using a multiple linear regression approach where a rsfMRI connectivity 355 measure is modeled as the response (or outcome) variable, comorbid psychiatric 356 conditions as well as confounding factors can be included as explanatory (or predictor) 357 variables. Thus, multiple linear regression allows for the estimation of the main effects 358 of anhedonia on rsfMRI connectivity, independent of the effects of depressed mood, 359 irritability, and bipolar II disorder (and vice versa), and the effects of confounding 360 covariates. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 20 of 34 361 Since the ABCD 1.0 release sample and the subsample of ABCD 4.0 release, 362 excluding the ABCD 1.0 sample, were not matched on age, sex, or race/ethnicity, we 363 hypothesized that controlling for these potential confounders may improve replicability. 364 Thus, we first performed multiple linear regression in the ABCD 1.0 release sample and 365 then in the ABCD 4.0 release sample, excluding the ABCD 1.0 release sample. The 366 reference group was the control group without symptoms with anhedonia. In the ABCD 367 1.0 release sample, only two rsfMRI connectivity measures exhibited statistically 368 significant partial regression coefficients for anhedonia, after multiple testing corrections 369 (Table 4; full multiple regression results found in Supplementary Table 8). 370 Table. 4 rsfMRI Connectivity Measures with Significant Partial Regression 371 Coefficients for Anhedonia in the ABCD 1.0 Sample. rsfMRI Connectivity Measure Effect of Anhedonia Std.Err t-value p-value p.adj R2 DorsalAttentionLeftHippocampus 0.024 0.007 3.208 0.001 0.019 0.012 CinguloParietalBrainStem 0.021 0.007 2.968 0.003 0.036 0.006 372 Statistically significant effects of anhedonia as a predictor of rsfMRI connectivity 373 measures from multiple linear regression tests are shown. The standard errors 374 (Std.Err), t-values, p-values, Benjamini-Hochberg adjusted p-values (p.adj) for the 375 effects of anhedonia, and overall coefficients of determination (R2) for each regression 376 are reported. 377 It is likely that including the additional explanatory variables into the model 378 reduced the statistical power to detect significant associations in the relatively small 379 sample. Accordingly, while more rsfMRI connectivity measures were significantly 380 associated with anhedonia in the ABCD 4.0 release, excluding ABCD 1.0 release, sub- 381 sample, we were unable to replicate the associations found using the ABCD 1.0 release . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 21 of 34 382 sample (Table. 5). rsfMRI connectivity measures with significant partial regression 383 coefficients for bipolar II and depressed mood were also identified for analyses in the 384 ABCD 1.0 release sample (Supplementary Table. 8) and for depressed mood, irritability, 385 and bipolar II in the ABCD 4.0 release, excluding ABCD 1.0 release, sub-sample 386 (Supplementary Table. 9). No significant partial regression coefficients for irritability 387 were detected in the ABCD 1.0 release sample. 388 Table. 5 rsfMRI Connectivity Measures with Significant Partial Regression 389 Coefficients for Anhedonia in the ABCD 4.0 Release, Excluding ABCD 1.0 390 Release, Sub-sample. rsfMRI Connectivity Measure Effect of Anhedonia Std.Err t-value p- value p.adj R2 AuditoryRightPutamen -0.013 0.005 -2.563 0.01 0.03 6 0.04 8 DorsalAttentionRightHippocampus 0.009 0.004 2.524 0.012 0.04 0 DorsalAttentionRightAmygdala 0.012 0.004 3.226 0.001 0.00 6 0.00 6 RetrosplenialTemporalRightThalamusPrope r -0.01 0.004 -2.661 0.008 0.02 8 0.02 9 VentralAttentionLeftCerebellumCortex 0.01 0.004 2.715 0.007 0.02 5 0.00 6 VisualLeftHippocampus -0.013 0.004 -3.067 0.002 0.00 9 0.04 5 AuditoryCinguloParietal -0.008 0.003 -2.465 0.014 0.04 6 0.01 8 AuditorySensorimotorHand 0.007 0.003 2.614 0.009 0.03 2 0.03 3 CinguloOpercularSalience -0.009 0.003 -2.903 0.004 0.01 5 0.00 2 CinguloParietalSensorimotorHand -0.008 0.003 -2.626 0.009 0.03 1 0.00 6 FrontoParietalVisual 0.006 0.002 2.732 0.006 0.02 3 0.01 1 RetrosplenialTemporalSensorimotorHand -0.007 0.002 -2.88 0.004 0.01 5 0.02 1 WithinSensorimotorHand 0.009 0.003 2.837 0.005 0.01 7 0.03 8 WithinSalience -0.015 0.005 -2.802 0.005 0.01 9 0.00 3 391 Statistically significant effects of anhedonia as a predictor of rsfMRI connectivity 392 measures from multiple linear regression tests are shown. The standard errors . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 22 of 34 393 (Std.Err), t-values, p-values, Benjamini-Hochberg adjusted p-values (p.adj) for the 394 effects of anhedonia, and overall coefficients of determination (R2) for each regression 395 are reported. 396 In order to maximize our power to detect associations between rsfMRI 397 connectivity measures and anhedonia, we performed multiple linear regression 398 analyses in the full ABCD 4.0 release sample. We found that 16 rsfMRI connectivity 399 measures exhibited significant partial regression coefficients for anhedonia (Table. 6; 400 full table of results found in Supplementary Table 10). Of note, 9 out of these 16 rsfMRI 401 connectivity measures were previously identified to be uniquely associated with 402 anhedonia from our t-tests. Furthermore, the Retrosplenial Temporal vs. Right 403 Cerebellum Cortex and CinguloOpercular vs. Brainstem connectivity measures were 404 reported to be significantly associated with anhedonia by the previous authors. 405 Table. 6 rsfMRI Connectivity Measures with Significant Partial Regression 406 Coefficients for Anhedonia in the Full ABCD 4.0 Release Sample. rsfMRI Connectivity Measure Effect of Anhedonia Std.Err t-value p- value p.adj R2 RetrosplenialTemporalRightThalamusPropera -0.012 0.003 -3.583 0 0.00 1 0.02 9 VentralAttentionLeftCerebellumCortexa 0.009 0.003 3.161 0.002 0.00 6 0.00 8 AuditoryRightPutamen -0.013 0.004 -3.034 0.002 0.00 9 0.04 5 DorsalAttentionRightAmygdalaa 0.008 0.003 2.751 0.006 0.02 0.00 7 VentralAttentionLeftPutamen 0.007 0.003 2.745 0.006 0.02 0.01 9 CinguloOpercularSalience -0.007 0.003 -2.667 0.008 0.02 5 0.00 2 WithinSensorimotorHanda 0.007 0.003 2.649 0.008 0.02 6 0.03 7 VisualRightPalliduma -0.008 0.003 -2.615 0.009 0.02 9 0.03 1 FrontoParietalVisuala 0.005 0.002 2.609 0.009 0.02 9 0.01 2 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 23 of 34 WithinSalience -0.011 0.005 -2.449 0.014 0.04 4 0.00 1 CinguloOpercularBrainStema -0.007 0.003 -2.437 0.015 0.04 5 0.03 9 RetrosplenialTemporalSensorimotorHand -0.005 0.002 -2.422 0.015 0.04 7 0.02 VisualRightVentraldc 0.008 0.003 2.412 0.016 0.04 8 0.00 7 RetrosplenialTemporalRightCerebellumCortex a -0.01 0.004 -2.407 0.016 0.04 9 0.04 1 VentralAttentionLeftCaudatea -0.006 0.003 -2.404 0.016 0.04 9 0.02 5 SensorimotorHandLeftPallidum -0.009 0.004 -2.398 0.017 0.05 0.05 6 407 Statistically significant effects of anhedonia as a predictor of rsfMRI connectivity 408 measures from multiple linear regression tests are shown. The standard errors 409 (Std.Err), t-values, p-values, Benjamini-Hochberg adjusted p-values (p.adj) for the 410 effects of anhedonia, and overall coefficients of determination (R2) for each regression 411 are reported. 412 a rsfMRI connectivity measures also found to be specifically associated with anhedonia 413 from previously performed t-tests in the full ABCD 4.0 release sample. 414 Two rsfMRI connectivity measures, the Auditory vs. Right Putamen and Ventral 415 Attention vs. Left Putamen were previously found to be associated with both depressed 416 mood and anhedonia from our t-tests but now only exhibit significant partial regression 417 coefficients for anhedonia. Interestingly, the Sensorimotor-Hand vs. Left Pallidum 418 connectivity measure exhibited significant partial regression coefficients for both 419 anhedonia and depressed mood (Table. 6, Supplementary Table. 10), consistent with 420 previous t-test results, suggesting the presence of significant independent effects of 421 both symptoms on the same rsfMRI connectivity measure. 422 The CinguloOpercular vs. Left Amygdala connectivity measure was also 423 previously associated with both depressed mood and anhedonia, based on their t-tests, . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 24 of 34 424 but now only exhibit significant partial regression coefficients for depressed mood 425 (Supplementary Table. 10). Similarly, the Ventral Attention vs. Left Thalamus Proper 426 connectivity measure was previously associated with both anhedonia and irritability 427 based on t-tests, but now only exhibit significant partial regression coefficients for 428 irritability (Supplementary Table. 10). Only one rsfMRI connectivity measure, Within 429 Ventral Attention, exhibited a significant partial regression coefficient for bipolar II 430 disorder (Supplementary Table. 10). 431 Altogether, while multiple linear regression analyses did not improve replicability 432 between the ABCD study sub-samples, it did allow us to estimate the independent 433 effects of anhedonia on rsfMRI connectivity measures in the full ABCD 4.0 release 434 sample and helped resolve the specificity of the effects of connectivity measures that 435 were previously found to be associated with anhedonia and other psychiatric symptoms. 436 Discussion 437 Reproduction and replication of previous findings 438 While we were able to successfully reproduce the previous authors’ findings, we 439 were mostly unable to replicate them using a larger independent subset of the full 440 ABCD 4.0 release sample. Interestingly, a recent study exploring the replicability of 441 brain-behavior association studies using simulations and parametric bootstrapping 442 methods found that relatively small sample sizes (n<500) produced results with 443 significantly inflated effect sizes, low precision, and low replicability and it was only 444 when the sample sizes were increased to the high hundreds or thousands were they 445 able to produce stable effects that were significantly more replicable (22). Thus, it is . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 25 of 34 446 possible that associations found by the original authors and in our replication analyses 447 using relatively small samples are at risk of being driven by random sample variation. 448 To maximize statistical power and to reduce inflated associations between 449 rsfMRI connectivity associations with anhedonia, future analyses should be conducted 450 using larger samples, such as in the full ABCD 4.0 release sample. Furthermore, Marek 451 et al., 2022 found that controlling for socio-demographic covariates further reduced 452 effect size inflation. Thus, the results from our multiple linear regression analyses using 453 the full ABCD 4.0 release sample, where we control for socio-demographic covariates, 454 are more likely to represent less-inflated and more replicable findings. 455 Specificity of associations 456 We found depressed mood, irritability, and bipolar II disorder to be significantly 457 comorbid with anhedonia. Using a multiple linear regression approach in the full ABCD 458 4.0 dataset, we were able to estimate the effects of anhedonia on rsfMRI connectivity 459 measures independent of those comorbid conditions, allowing us to disentangle rsfMRI 460 connectivity measures associated with more than one condition based on t-test results. 461 In doing so, however, the interpretation of anhedonia requires careful consideration and 462 reflection. 463 As mentioned previously, while anhedonia and depressed mood are core 464 symptoms of major depressive disorders, they are considered distinct processes (15). 465 Alternatively, the hierarchical Taxonomy of Psychopathology (HiTOP) (23), a recently 466 developed dimensional framework for psychopathology, has classified anhedonia as a 467 symptom belonging to two high level sepctra of psychopathology; the internalizing and 468 detachment spectra. Interestingly, low/depressed mood and irritability also fall under the . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 26 of 34 469 internalizing spectrum whereas bipolar II falls under the thought disorder spectrum. 470 Thus, when estimating the effects of anhedonia independent of other internalizing and 471 thought disorder related symptoms, we could interpret the remaining effect to 472 emphasize detachment processes. As detachment is a component of the psychosis 473 super-spectrum (24) our findings may represent anhedonic neurocircuitry that may, in 474 part, be related to schizophrenia, schizotypal personality, or other psychotic disorder 475 processes. Further work is required to assess the extent to which differences in rsfMRI 476 connectivity specific to anhedonia better associates with or even predicts dimensional 477 measures of internalizing or detachment related psychopathology. 478 We included race/ethnicity as a covariate in our multiple linear regression 479 analyses and note they exhibited significant partial regression coefficients for many of 480 the rsfMRI connectivity measures we analyzed. Furthermore, we found there were 481 significantly higher proportions of Black and Hispanic participants in the anhedonia 482 group compared to non-anhedonic controls in the full ABCD 4.0 release sample. Race 483 and ethnicity are social constructs representing complex social and cultural factors (25) 484 deserving careful consideration. Several previous studies have reported significantly 485 higher risk of anhedonia in Black and Hispanic compared to non-Hispanic White adults 486 (26,27) and that these associations may, in part, be accounted for by socioeconomic 487 factors, such as household income and education, as well as other social determinants 488 of health (28), such as disparities in access to healthcare (26). In the ABCD sample, 489 racial discrimination may be an important factor contributing to risk of anhedonia as well 490 as differences in brain-based measures. For example, several recent studies have 491 found that racial discrimination is associated with lower total brain volume (29) and . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 27 of 34 492 alterations in pre-frontal white matter tracts in adults (30,31). While out of the scope of 493 this study, it will be very important to investigate how social determinants of health and 494 other related factors contribute to differences in health and brain-based outcomes 495 between different racial and ethnic groups during child and adolescent development. 496 Limitations 497 One major limitation of our study was the use of seed-based correlational 498 methods to compute network connectivity measures in functionally-defined networks. As 499 such, these network connectivity measures are averages over large and distributed 500 networks where signals from sub-regions potentially highly associated with anhedonia 501 may be drowned out by signals from sub-regions with low levels of association. Another 502 concern is that the Gordon brain parcellations were produced using a boundary- 503 mapping approach in adult brains (19) so whether they are generalizable to the brains 504 of developing children is important to consider. For example, one study found that the 505 functional topography of connectivity networks does change with age which was 506 predictive of individual differences in executive function (32). One alternative method is 507 to use a decomposition-based method, such as independent components analysis 508 (ICA), to define functional connectivity measures (33). ICA is a data driven approach 509 that extracts components that maximally explain the data and thus, may enhance 510 predictive performance. One study took such an approach and found that using a 511 decomposition-based, compared to a seed-based, extraction of functional networks 512 during a social cognition task achieved significantly greater performance in predicting 513 the degree of social anhedonia in around 70 adolescents/young adults with varying . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 28 of 34 514 levels of schizotypy (34). Since the raw neuroimaging data from the ABCD data are 515 publicly available, this may be a feasible approach to implement in a future study. 516 Another limitation in our study is the use of a binary classifier for childhood 517 anhedonia. Several studies have reported greater predictive performance of 518 neuroimaging measures on anhedonia symptom scores (34,35) which suggests that 519 functional neuroimaging measures may be more useful for predicting symptom severity 520 rather than for disorder classification. Thus, it may be more reliable to investigate the 521 associations between functional neuroimaging measures and clinical scales for 522 assessing behavioral problems (such as the Child Behavior Checklist) or neurocognitive 523 performance in individuals with anhedonia. 524 Finally, the DSM-V definition of anhedonia conflates two distinct reward 525 processes: motivational (interest/wanting) and consummatory (pleasurable/liking) 526 behaviors. These behaviors have been shown to have distinct neurobiological and 527 behavioral components (9,36). We are limited in our study because we do not 528 distinguish between these processes. However, the ABCD study data does include 529 task-based functional neuroimaging of participants completing the monetary incentive 530 delay task, which is able to assess the anticipatory, consummatory, and learning 531 aspects of reward (37). These processes were studied previously by 532 Pornpattananangkul et al., but exceeded the scope of this study. Nevertheless, 533 exploration of brain connectivity specific to each of these reward-based components in 534 subjects with anhedonia may help elucidate the underlying circuitry underlying this 535 complex psychiatric condition. 536 Acknowledgements . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 29 of 34 537 Data used in the preparation of this article were obtained from the Adolescent Brain 538 Cognitive Development SM (ABCD) Study (https://abcdstudy.org), held in the NIMH Data 539 Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 540 10,000 children age 9-10 and follow them over 10 years into early adulthood. The 541 ABCD Study® is supported by the National Institutes of Health and additional federal 542 partners under award numbers U01DA041048, U01DA050989, U01DA051016, 543 U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, 544 U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, 545 U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, 546 U01DA041148, U01DA041093, U01DA041089, U24DA041123, U24DA041147. A full 547 list of supporters is available at https://abcdstudy.org/federal-partners.html. A listing of 548 participating sites and a complete listing of the study investigators can be found at 549 https://abcdstudy.org/consortium_members/. ABCD consortium investigators designed 550 and implemented the study and/or provided data but did not necessarily participate in 551 the analysis or writing of this report. This manuscript reflects the views of the authors 552 and may not reflect the opinions or views of the NIH or ABCD consortium investigators. 553 The ABCD data repository grows and changes over time. The ABCD data used in this 554 report came from http://dx.doi.org/10.15154/1523041. 555 This manuscript reflects the views of the authors and may not reflect the opinions or 556 views of the NIH or ABCD consortium investigators. 557 This study received no external funding. Yi Zhou was supported by the department of 558 Psychiatry at Virginia Commonwealth University. 559 Disclosures . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 30 of 34 560 The authors report no financial interests or potential conflicts of interest 561 References 562 1. American Psychiatric Association, American Psychiatric Association, editors. 563 Diagnostic and statistical manual of mental disorders: DSM-5. 5th ed. Washington, 564 D.C: American Psychiatric Association; 2013. 947 p. 565 2. Gutkovich Z, Morrissey RF, Espaillat RK, Dicker R. Anhedonia and pessimism in 566 hospitalized depressed adolescents. Depress Res Treat. 2011;2011:795173. 567 3. Garfield JBB, Lubman DI, Yücel M. Anhedonia in substance use disorders: a 568 systematic review of its nature, course and clinical correlates. Aust N Z J Psychiatry. 569 2014 Jan;48(1):36–51. 570 4. Frewen PA, Dozois DJA, Lanius RA. Assessment of anhedonia in psychological 571 trauma: psychometric and neuroimaging perspectives. Eur J Psychotraumatology. 572 2012 Jan 11;3:10.3402/ejpt.v3i0.8587. 573 5. Sternat T, Fotinos K, Fine A, Epstein I, Katzman MA. Low hedonic tone and 574 attention-deficit hyperactivity disorder: risk factors for treatment resistance in 575 depressed adults. Neuropsychiatr Dis Treat. 2018 Sep 17;14:2379–87. 576 6. Luby JL, Agrawal A, Belden A, Whalen D, Tillman R, Barch DM. Developmental 577 Trajectories of the Orbitofrontal Cortex and Anhedonia in Middle Childhood and Risk 578 for Substance Use in Adolescence in a Longitudinal Sample of Depressed and 579 Healthy Preschoolers. Am J Psychiatry. 2018 Oct;175(10):1010–21. 580 7. McMakin DL, Olino TM, Porta G, Dietz LJ, Emslie G, Clarke G, et al. Anhedonia 581 predicts poorer recovery among youth with selective serotonin reuptake inhibitor 582 treatment-resistant depression. J Am Acad Child Adolesc Psychiatry. 2012 583 Apr;51(4):404–11. 584 8. Robbins DR, Alessi NE. Depressive symptoms and suicidal behavior in adolescents. 585 Am J Psychiatry. 1985 May;142(5):588–92. 586 9. Treadway MT, Zald DH. Reconsidering anhedonia in depression: lessons from 587 translational neuroscience. Neurosci Biobehav Rev. 2011 Jan;35(3):537–55. 588 10. Gabbay V, Ely BA, Li Q, Bangaru SD, Panzer AM, Alonso CM, et al. Striatum-based 589 circuitry of adolescent depression and anhedonia. J Am Acad Child Adolesc 590 Psychiatry. 2013 Jun;52(6):628-641.e13. 591 11. Rzepa E, McCabe C. Decreased anticipated pleasure correlates with increased 592 salience network resting state functional connectivity in adolescents with depressive 593 symptomatology. J Psychiatr Res. 2016 Nov;82:40–7. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 31 of 34 594 12. Rzepa E, McCabe C. Anhedonia and depression severity dissociated by dmPFC 595 resting-state functional connectivity in adolescents. J Psychopharmacol Oxf Engl. 596 2018 Oct;32(10):1067–74. 597 13. Pornpattananangkul N, Leibenluft E, Pine DS, Stringaris A. Notice of Retraction and 598 Replacement. Pornpattananangkul et al. Association between childhood anhedonia 599 and alterations in large-scale resting-state networks and task-evoked activation. 600 JAMA Psychiatry. 2019;76(6):624-633. JAMA Psychiatry. 2020 Oct 1;77(10):1085– 601 6. 602 14. Stevens JR. Replicability and Reproducibility in Comparative Psychology. Front 603 Psychol. 2017;8:862. 604 15. De Fruyt J, Sabbe B, Demyttenaere K. Anhedonia in Depressive Disorder: A 605 Narrative Review. Psychopathology. 2020;53(5–6):274–81. 606 16. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression 607 data using empirical Bayes methods. Biostatistics. 2007 Jan 1;8(1):118–27. 608 17. Hagler DJ, Hatton S, Cornejo MD, Makowski C, Fair DA, Dick AS, et al. Image 609 processing and analysis methods for the Adolescent Brain Cognitive Development 610 Study. NeuroImage. 2019 Nov 15;202:116091. 611 18. Casey BJ, Cannonier T, Conley MI, Cohen AO, Barch DM, Heitzeg MM, et al. The 612 Adolescent Brain Cognitive Development (ABCD) study: Imaging acquisition across 613 21 sites. Dev Cogn Neurosci. 2018 Aug 1;32:43–54. 614 19. Gordon EM, Laumann TO, Adeyemo B, Huckins JF, Kelley WM, Petersen SE. 615 Generation and Evaluation of a Cortical Area Parcellation from Resting-State 616 Correlations. Cereb Cortex N Y NY. 2016 Jan;26(1):288–303. 617 20. Orlhac F, Eertink JJ, Cottereau AS, Zijlstra JM, Thieblemont C, Meignan M, et al. A 618 Guide to ComBat Harmonization of Imaging Biomarkers in Multicenter Studies. J 619 Nucl Med Off Publ Soc Nucl Med. 2022 Feb;63(2):172–9. 620 21. Vidal-Ribas P, Janiri D, Doucet GE, Pornpattananangkul N, Nielson DM, Frangou S, 621 et al. Multimodal neuroimaging of suicidal thoughts and behaviors in a US 622 population-based sample of school-aged children. Am J Psychiatry. 2021 Apr 623 1;178(4):321–32. 624 22. Marek S, Tervo-Clemmens B, Calabro FJ, Montez DF, Kay BP, Hatoum AS, et al. 625 Reproducible brain-wide association studies require thousands of individuals. 626 Nature. 2022 Mar;603(7902):654–60. 627 23. Conway CC, Forbes MK, South SC. A Hierarchical Taxonomy of Psychopathology 628 (HiTOP) Primer for Mental Health Researchers. Clin Psychol Sci. 2022 Mar 629 1;10(2):236–58. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 32 of 34 630 24. Kotov R, Jonas KG, Carpenter WT, Dretsch MN, Eaton NR, Forbes MK, et al. 631 Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): I. 632 Psychosis superspectrum. World Psychiatry. 2020 Jun;19(2):151–72. 633 25. LaVeist TA. Beyond dummy variables and sample selection: what health services 634 researchers ought to know about race as a variable. Health Serv Res. 1994 635 Apr;29(1):1–16. 636 26. Vyas CM, Donneyong M, Mischoulon D, Chang G, Gibson H, Cook NR, et al. 637 Association of Race and Ethnicity With Late-Life Depression Severity, Symptom 638 Burden, and Care. JAMA Netw Open. 2020 Mar 2;3(3):e201606. 639 27. Liu CH, Tronick E. Prevalence and predictors of maternal postpartum depressed 640 mood and anhedonia by race and ethnicity. Epidemiol Psychiatr Sci. 2014 641 Jun;23(2):201–9. 642 28. Braveman PA, Cubbin C, Egerter S, Williams DR, Pamuk E. Socioeconomic 643 disparities in health in the United States: what the patterns tell us. Am J Public 644 Health. 2010 Apr 1;100 Suppl 1:S186-196. 645 29. Meyer CS, Schreiner PJ, Lim K, Battapady H, Launer LJ. Depressive 646 Symptomatology, Racial Discrimination Experience, and Brain Tissue Volumes 647 Observed on Magnetic Resonance Imaging: The CARDIA Study. Am J Epidemiol. 648 2019 Apr 1;188(4):656–63. 649 30. Okeke O, Elbasheir A, Carter SE, Powers A, Mekawi Y, Gillespie CF, et al. Indirect 650 Effects of Racial Discrimination on Health Outcomes Through Prefrontal Cortical 651 White Matter Integrity. Biol Psychiatry Cogn Neurosci Neuroimaging [Internet]. 2022 652 May 18 [cited 2022 Oct 19]; Available from: 653 https://www.sciencedirect.com/science/article/pii/S2451902222001203 654 31. Fani N, Harnett NG, Bradley B, Mekawi Y, Powers A, Stevens JS, et al. Racial 655 Discrimination and White Matter Microstructure in Trauma-Exposed Black Women. 656 Biol Psychiatry. 2022 Feb 1;91(3):254–61. 657 32. Cui Z, Li H, Xia CH, Larsen B, Adebimpe A, Baum GL, et al. Individual Variation in 658 Functional Topography of Association Networks in Youth. Neuron. 2020 Apr 659 22;106(2):340-353.e8. 660 33. McKeown MJ, Hansen LK, Sejnowsk TJ. Independent component analysis of 661 functional MRI: what is signal and what is noise? Curr Opin Neurobiol. 2003 Oct 662 1;13(5):620–9. 663 34. Krohne LG, Wang Y, Hinrich JL, Moerup M, Chan RCK, Madsen KH. Classification 664 of social anhedonia using temporal and spatial network features from a social 665 cognition fMRI task. Hum Brain Mapp. 2019 Dec 1;40(17):4965–81. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 33 of 34 666 35. Mellem MS, Liu Y, Gonzalez H, Kollada M, Martin WJ, Ahammad P. Machine 667 Learning Models Identify Multimodal Measurements Highly Predictive of 668 Transdiagnostic Symptom Severity for Mood, Anhedonia, and Anxiety. Biol 669 Psychiatry Cogn Neurosci Neuroimaging. 2020 Jan 1;5(1):56–67. 670 36. Der-Avakian A, Markou A. The Neurobiology of Anhedonia and Other Reward- 671 Related Deficits. Trends Neurosci. 2012 Jan;35(1):68–77. 672 37. Wilson RP, Colizzi M, Bossong MG, Allen P, Kempton M, Abe N, et al. The Neural 673 Substrate of Reward Anticipation in Health: A Meta-Analysis of fMRI Findings in the 674 Monetary Incentive Delay Task. Neuropsychol Rev. 2018 Dec 1;28(4):496–506. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint Page 33 of 34 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 26, 2022. ; https://doi.org/10.1101/2022.10.24.22281441doi: medRxiv preprint