3.1. Participant characteristics
describes the sample. The three groups did not differ significantly on any of these socio-demographic or HIV disease characteristics. Participants were 69% male, aged 27-58 years (48.1 ± 7.4, M ± SD), ethnically diverse (41% African-American, 35% Caucasian, and 23% Hispanic), and socioeconomically disadvantaged (41% living under the Federal poverty threshold). They had been diagnosed with HIV for 3-26 years (15.2 ± 6.0), with a mean CD4 cell count of 617.41 (SD= 354.05), and 64% had an AIDS diagnosis. The active and recovered groups had used cocaine regularly for an average of 20.3 ± 8.2 and 14.4 ± 5.6 years, respectively. Active participants had used cocaine on an average of 8.1 ± 7.5 days in the past month, and recovered participants had been in sustained full remission for an average of 10.9 ± 6.4 years. Co-occurring substance dependence was rare; only two participants in the active group had current alcohol dependence, and none had other drug dependencies.
Participant characteristics by cocaine group
3.2. Behavioral performance on delay discounting tasks
The mean discount rates on the MCQ for the active, recovered, and naïve groups were 0.059 ± 0.071, 0.025 ± 0.031, and 0.017 ± 0.027, respectively. To illustrate this, an active cocaine user with a discount rate of 0.059 was approximately indifferent in choosing between $31 immediately and $85 in 30 days. In contrast, a drug naïve participant with a discount rate of 0.017 was approximately indifferent in choosing between $56 immediately and $85 in 30 days. As in prior studies (Heil et al., 2006
; Monterosso et al., 2007
), the MCQ discount rates were skewed, so the data were normalized using a natural log transformation. The group difference in discount rates was not statistically significant [F(2,36)= 2.27, P
= 0.118], but the effect sizes were strong (d = 0.620 for active versus recovered and d = 0.782 for active versus naïve).
As expected, the three groups performed similarly on the ICT administered during fMRI scanning. Because participants' choices were individualized based on their MCQ discount rates, we were able to predict their responses for the no and easy choice trials. For no trials, 98.2%, 97.1%, and 98.7% of active, recovered, and naïve participants, respectively, chose responses in the predicted direction [F(2,36)= 0.324, P= 0.725]. For easy trials, 80.9%, 87.4%, and 89.8% of active, recovered, and naïve participants chose responses in the predicted direction [F(2,36)= 1.509, P= 0.235]. For hard choices, the now and later choices were of equivalent value, so it was not possible to predict individuals' choices. All of these behavioral data were rerun controlling for age, gender, and HIV disease characteristics (years since HIV diagnosis, current CD4 count, and AIDS diagnosis), and results were unaffected.
3.3. Brain activation during ICT
Initially, data from all groups were combined to examine task-related activation during the ICT. As expected, there was increased activation during easy versus no choice trials (easy-no), and even greater increases during hard versus no choice trials (hard-no). shows the five clusters of activation during easy-no trials. Activation was observed bilaterally in the prefrontal cortices, primary motor cortex (M1), ACC, SMA, IPS and parietal lobules, lateral occipital cortex, primary visual cortex, and cerebellum. During hard-no trials, activation was observed in similar regions captured as a single large cluster of 426.51 ml, with peak activation within the occipital pole (32, -96, 0; z-max= 7.93, cluster P< 0.00001). As shown in , there were significantly greater increases in activation during hard versus easy choice trials (hard-easy) bilaterally in the prefrontal cortices, ACC and SMA, IPS and parietal lobules, thalamus, and cerebellum.
Task-related activation during delay discounting choices (all participants combined).
Next, we compared the groups on the three contrasts. In clusters with significant group differences, active cocaine users consistently had smaller increases in brain activation relative to drug naïve participants (). In the easy-no contrast, the active group had smaller increases bilaterally in the IPS, parietal lobules, precuneus, ACC, and postcentral gyrus. In the hard-no contrast, they had smaller increases bilaterally in the VLPFC, IPS, parietal lobules, in the right DLPFC, thalamus, striatum, and insula, and in the left M1 and cuneal cortices. In the hard-easy contrast, they had smaller increases bilaterally in the M1 and ACC and in the right frontal pole (see ).
Clusters showing smaller increases in activation during delay discounting choices among active cocaine versus drug naïve participants.
Fig. 1 Differences in brain activation between naïve and active groups during hard versus easy choice trials. Colored regions represent clusters in which the naïve group had significantly greater increases in activation compared to the active (more ...)
The pattern of activation was more complex for the recovered group (). In the easy-no contrast, recovered users had greater increases relative to active users bilaterally in the lateral occipital cortex and in the left frontal pole, and they had similar increases (i.e., no differences) relative to drug naïve participants. In the hard-no contrast, recovered users had larger increases relative to active users in the left inferior parietal lobule and IPS, the right cerebellum, and the brain stem, but they had smaller increases relative to drug naïve participants in the left inferior parietal lobule and bilaterally in the prefrontal cortices, postcentral gyrus, IPS, precuneus, insula, among other regions. Conversely, in the hard-easy contrast, recovered users had similar increases in brain activation relative to active cocaine users, and they had smaller increases relative to drug naïve participants bilaterally in the frontal pole, including prefrontal cortices. These results suggest that recovered cocaine users have similar patterns of activation relative to drug naïve participants during easy choices, but demonstrate impairments during hard choices.
Clusters showing smaller or larger increases in activation during delay discounting choices among recovered versus active and naïve participants.
3.4. ROI analyses
shows the percent BOLD signal change during easy-no (2A) and hard-no (2B) contrasts in the nine ROIs (listed in as clusters activated in hard-easy contrast). Both types of choices were associated with increased activation relative to baseline in the plotted ROIs, with hard choices requiring substantially more activation than easy choices. In general, active cocaine users had smaller increases in activation during both easy and hard choices. In the easy-no contrast, the active group had minimal increases in BOLD signal change relative to the recovered and naïve groups, with significant differences in the left frontal pole [t(37)= -2.27, P= 0.029, d= -0.717], left IPS [t(37)= -3.11, P= 0.004, d= -0.956), and right IPS [t(37)= -3.01, P= 0.005, d= -0.920). Active users had greater increases in activation in the hard-no contrast than the easy-no contrast. For example, the mean change in the right IFG cluster was 0.21% for hard choices versus 0.03% for easy choices. However, once again, active users had smaller increases relative to naïve participants, with significant differences in the right frontal pole [t(24)= 2.60, P= 0.016, d= -1.037], left ACC/SMA [t(24)= 2.58, P= 0.016, d= -0.717], right DLPFC [t(24)= -2.40, P= 0.025, d= -0.985], and right IPS [t(24)= 2.40, P= 0.025, d= -0.980]. In the hard-no contrast, recovered users had increases that were generally intermediate relative to active and naïve users; they differed significantly only relative to active cocaine users in the right IPS [t(26)= -2.05, P= 0.050, d= -0.785].
Fig. 2 Percent BOLD signal change (mean ± standard error) for each group for: (A) easy relative to no choice trials, and (B) hard relative to no choice trials. In the easy-no contrast, the active group had significantly reduced increases in activation (more ...)