Adolescents were asked to maintain abstinence from alcohol and other substances from the time they completed their CDDR; the initial appointment was scheduled to coincide with 4-10 days of abstinence. This ensured that any cognitive impairment observed in the heavy drinkers could not be indicative of acute alcohol withdrawal and detoxification. Adolescents completed a TLFB at their first NP appointment to gather information about alcohol and substance use that may have occurred in the time after completion of the CDDR. Although adolescents typically provide valid self-reports of alcohol and other drug use , self-report of abstinence was verified with breathalyzer and urine toxicology screens . If an adolescent came up with a positive result for any alcohol or drug use, a trained research assistant discussed the finding with the participant and sought verification of use from the adolescent. A TLFB was used to assist the adolescent in recall. If adolescents acknowledged use but reported continued interest in the study, they were allowed to continue, provided they remained abstinent for the number of days that passed since their first NP assessment plus the number of days that they reported being abstinent prior to their first NP assessment. A Motivational Interviewing protocol was employed when necessary,grow supplies for weed which assisted participants in maintaining their abstinence throughout the duration of the protocol. All imaging data were collected from the 3.0 Tesla General Electric short bore Excite-2 MR system with an eight-channel phase-array head coil located at the UCSD Keck fMRI Center. Eight high bandwidth receivers for ultra-short repetition time reduced signal distortions and signal dropout.
Scan sessions collected: 1) localizer scans to assure good head placement and slice selection to cover the Whole brain ; 2) a high resolution 3d T1-weighted sequence obtained using a sagittally acquired spoiled gradient recalled sequence to allow volumetric analyses of brain structures and the defining and outlining of regions of interest ; 3) a T2-weighted axially acquired echo-planar imaging sequence measured BOLD signal during the BART task; 4) field map scans employed two different echo times to assess field in homogeneities and signal distortions under the same parameters used in the echo-planar sequence ; these were applied to the fMRI acquisitions to minimize warping and signal dropout. fMRI task stimuli were back projected from a laptop to a screen at the foot of the scanner bed visible via an angled mirror attached to the head coil. Responses were recorded using a fiber-optic trackball mouse . All participants were administered the same BART task during fMRI acquisition. The task utilized a rapid event-related design in order to measure changes in the hemodynamic response corresponding to the three stages of decision making . Twenty balloons trials were presented one at a time on the screen, each with a predetermined, random “explosion point”. Participants were asked to “pump up” each balloon by responding to a series of six phases; a status bar was present throughout the task, which indicated which phase of the task was currently active. In the “Think” phase, participants viewed a question mark on the screen and were asked to think about the number of pumps they would like to enter. In the “Pump” phase, participants manually entered their chosen number of pumps, between 1-128. In the “Wait” phase, participants viewed the uninflated balloon on the screen and waited for the task to continue. In the “Inflate” phase, participants watched the balloon slowly expand as they waited to receive the outcome of their decision. In the “Pop or Win” phase, participants found out whether or not they earned money for the trial.
If the chosen number of pumps was greater than the predetermined explosion point, the balloon on the screen “popped” at the end of the inflation period, and no money was earned. If the chosen number was less than the explosion point, participants “won” that balloon, and the inflated balloon remained on the screen for a moment, as a counter box in the lower right-hand corner displayed the amount of money earned for that trial . Last, in the “Rest” phase, balloons were removed from the screen, and participants viewed a blank screen and waited for the next balloon trial to begin. The task consisted of 241 repetitions in all, with a total run time of 8 minutes and 2 seconds . This work is being prepared for submission for publication as “fMRI Correlates of Risky Decision-Making in Adolescent Alcohol Users: The Role of Abstinence.” The dissertation author will be the primary author of this material along with co-authors Alan Simmons, Ph.D., Carmen Pulido, Ph.D., Susan Tapert, Ph.D., and Sandra Brown, Ph.D. All neuroimaging data were processed and analyzed using Analysis of Functional NeuroImages . Each repetition of each slice was examined for artifact and atypical signal levels, using an automated program developed by the UCSD Laboratory of Cognitive Neuroimaging. Motion in the time series data were corrected by registering each acquisition to the maximally stable base volume with an iterated least squares algorithm to estimate 3 rotational and 3 displacement parameters for each participant, which were used in the model to control for spin history effects . To evaluate task related motion, the reference vector was correlated with the six motion parameters generated for each individual dataset. Datasets with significant task-correlated or bulk motion were excluded from analyses. Trained raters then manually examined the time series data, and omitted any repetitions containing excessive head movements. If more than 20% of repetitions in a task were discarded, the participant was excluded. Two participants were excluded from each time point due to excess motion . For the remaining participants in the sample, 89% or greater repetitions were retained. Deconvolution was conducted on time series data with a reference function which coded the alternating task conditions and convolved the behavioral stimuli with a hemodynamic response model , while covarying for linear trends and the degree of motion correction applied.
As the specific task stimuli are unique each time the task is run, according to each participant’s performance , separate customized reference functions were created for each run and each participant. This resulted in fit coefficients for each voxel representing the change in BOLD signal across the task conditions , as well as percentage signal change and threshold statistics. Standardization transformations were made for each high-resolution anatomical image , and functional datasets were warped in accordance to these images, to manage individual anatomical variability. Functional data were resampled into isotropic voxels ,equipment for weed growing and a spatial smoothing Gaussian filter minimized the influence of individual anatomic variability. Co-registration of structural to functional images was performed with a mutual information registration program that robustly handles images with different signal characteristics and spatial resolutions. ROI masks were created for each hypothesized ROI using the TT_Daemon brain atlases available in AFNI. As there were no pre-existing DLPFC and VMPFC masks, these were created by adding together masks from relevant Brodmann areas . Six planned contrasts allowed for the examination of BOLD response unique to the three phases of decision-making modeled in the task , and to the effect of “winning” vs. “losing” . Active task conditions were contrasted against control task conditions . The six contrasts included: 1) “Think” trials where the previous balloon popped vs. “Rest”, 2) “Think” trials where the previous balloon did not pop vs. “Rest”, 3) “Inflate” trials where the previous balloon popped vs. “Wait”, 4) “Inflate” trials where the previous balloon did not pop vs. “Wait”, 5) “Pop or Win” trials where the balloon popped vs. “Wait”, 6) “Pop or Win” trials where the balloon did not pop vs. “Wait”. ROI masks were applied separately to these six planned contrasts, which enabled the extraction of fit coefficients averaged across each ROI for each participant; these were imported into SPSS for use in further analyses. Hypothesis 1 was tested using the AFNI program 3dttest++, with group as the between-subjects variable, to determine the effect of group on baseline BOLD response during active vs. control task conditions for each of the six planned contrasts, averaged across each ROI. There were a total of 6 ROIs identified using AFNI-based atlases, with 2 relevant to the pre-response assessment phase of decision-making , 3 relevant to the anticipation phase , and 3 relevant to the outcome evaluation phase . For each ROI, AFNI program 3dClustSim determined the number of contiguous voxels necessary within a given area of activation , to keep the family-wise alpha level at 0.05 within the ROI. Cluster sizes were identified as 5 voxels for the amygdala, 9 voxels for the ventral striatum, 10 voxels for the anterior cingulate, 11 voxels for the DLPFC and insula, and 12 voxels for the VMPFC. In addition to testing the six contrasts of interest originally identified a priori , two additional contrasts were tested: 1) “Inflate” trials where the previous balloon popped vs. “Rest”, 4) “Inflate” trials where the previous balloon did not pop vs. “Rest”; these contrasts were added because of the hypothesized overlap between the active “Inflate” condition and the control “Wait” condition. Specifically, as the “Wait” condition takes place immediately after participants input their chosen pump number, it is likely that participants are “anticipating” the outcome of their decision during “Wait” trials before viewing the visual cue of an inflating balloon during the “Inflate” trials. Thus, to more effectively separate out changes in BOLD response related to the anticipation phase of decision-making, the “Inflate” trials were contrasted with the other control condition, “Rest.”
These two contrasts were also examined in Hypothesis 2 and the exploratory whole-brain analysis. There were a total of 22 tests run for Hypothesis 1 . Hypothesis 2 was tested using AFNI program 3dLME to run a mixed-model 2 x 3 ANOVA with group as the between-subjects factor, time as the within-subjects factor, and subjects as a random factor, to determine main effects of group and time, as well as interaction effects between group and time on BOLD response during active conditions vs. control task conditions in the 8 planned contrasts in specified ROIs. Contrasts and ROIs with between-group differences at baseline were first tested to assess changes in BOLD activation patterns over time in these regions as specified in Hypothesis 2. To do this, masks were created representing the specific clusters within an ROI that showed significant between-group differences in BOLD activation. In addition, all other contrasts and ROIs were tested for interaction and main effects in order to assess any other patterns of change in BOLD response across time that may exist other than the expected pattern of change . Significant interaction effects were followed up with pairwise comparisons as well as visual inspection of the data, in order to characterize the direction of between and within-group differences. Cluster sizes determined by 3dClustSim were again used , to keep the family-wise alpha level at 0.05 within each ROI. There were a total of 25 tests run for Hypothesis 2 . Hypothesis 3 was tested using hierarchical regressions to examine if baseline BOLD responding in specified ROI regions predicted baseline NP performance on executive functioning and risk-taking tasks or self-reported risk-taking behavior. Only ROIs with significant group differences at baseline were used as predictors in these analyses. As the DLPFC and ACC ROIs during the pre-response assessment phase of decision-making at baseline did not evidence significant group differences, the impact of baseline BOLD response in these regions on baseline NP performance during executive functioning tasks was not examined as originally planned in Hypothesis 3. However, baseline BOLD response in the insula and VMPFC ROIs during the anticipation and outcome evaluation stages of decision-making, respectively, were examined in relation to dependent measures of risktaking . Dependent measures of BART performance included: total number of “wins” and “pops”, mean number of pumps , and number of pumps on trial 1, on the first “pop” trial, and on the trial immediately after the first “pop” trial. Dependent measures of self reported risk-taking behavior included: the Rule-Breaking sub-scale of the YSR, the Impulsivity/Sensation Seeking sub-scale of the ZKPQ, and the “Drive”, “Reward Responsiveness”, “Approach total”, and “Inhibition total” scales from the BIS/BAS.