Scores on the Gambling Task did not significantly differ between the cocaine-dependent subjects (mean=1.1, SD=29.9) and the comparison subjects (mean=13.7, SD=25.4) (t=0.85, df=26, n.s.). As expected, however, resting rCBF in the right (Talairach coordinates: x=38, y=40, z=–6) and left (x=–38, y=20, z=–8) orbitofrontal cortex was lower in the cocaine-dependent subjects than in the comparison subjects (unpaired t-image, p<0.005). There was no relationship between the counts per voxel in these two regions of interest and the Gambling Task score (right orbitofrontal cortex: r=0.13, df=27, n.s.; left orbitofrontal cortex: r=0.09, df=27, n.s.).
To examine whether rCBF in other regions was correlated with Gambling Task performance, we performed exploratory, post hoc t-image analysis using data for subjects who scored more than 0 (N=13) and less than 0 (N=13) on the Gambling Task (two subjects scored 0). At p<0.001, the anterior cingulate (Brodmann’s area 32; x=4, y=14, z=36) and left dorsolateral prefrontal cortex (Brodmann’s areas 6, 9, and 44; x=–40,y=–2, z=34) showed higher resting rCBF in the subjects with high Gambling Task scores than in those with low scores. Multiple regression analysis revealed that rCBF counts per voxel in these two regions of interest accounted for a significant amount of the variability in Gambling Task scores across groups (R2=0.44, F=6.19, df=3, 24, p=0.003) (F1). The interaction of group and region of interest was not significant (R2=0.45, F=0.26, df=2, 22, n.s.), suggesting that Gambling Task performance is related to anterior cingulate and left dorsolateral prefrontal cortex rCBF irrespective of drug history. Neither region alone offered additional predictive power beyond that contributed by the two areas combined. This result may be partly accounted for by the high correlation between the counts per voxel for these two regions of interest (r=0.69, df=27, p<0.01). Of interest was a region of the left dorsolateral prefrontal cortex (Brodmann’s areas 9 and 44; x=–36, y=8, z=36) that was identified in the rCBF comparison (p<0.01) between the patients and the comparison subjects and that was in close proximity to the dorsolateral prefrontal cortex area identified earlier. The counts per voxel of the two left dorsolateral prefrontal cortex regions were highly correlated (r=0.86, df=27, p<0.5×10–10), and the counts per voxel of the latter dorsolateral prefrontal cortex region correlated with Gambling Task performance (r=0.51, df=27, p<0.005).