Volumes at First Scan
Initial scan volumes were computed (see Supplemental Table 1
for descriptive statistics) and graphed with both hemispheres summed (). The three-factor ANOVA model showed a tissue by group interaction (F6,96
= 5.40, p
< .0001) without a main effect of group or a tissue by hemisphere by group interaction. Subsequent two-factor ANOVAs for each tissue class revealed that NCGM volume differed between groups (F2,96
= 7.65, p
= .001), without a hemisphere by group interaction. Post hoc pairwise comparisons revealed that, compared with HC, both FESZ (5.8% less, t63
= 3.66, p
= .001) and FEAFF (3.9% less, t68
= 3.00, p
= .004) had smaller (left + right) NCGM but did not differ from each other (t61
= 1.07, p
= .29, Cohen's d
= .27). Groups did not differ in CWM (F2,96
= .36, p
= .70) without a hemisphere by group interaction. Groups differed in SCSF volume (F2,96
= 5.89, p
= .004), without a hemisphere by group interaction. Both FESZ (18.6% more, t63
= 3.24, p
= .002) and FEAFF (14.8% more, t68
= 2.78, p
= .007) showed greater (left + right) SCSF volume than HC but did not differ from each other (t61
= .65, p
= .52, Cohen's d
= .16). Groups showed a marginally significant difference in LV volumes (F2,96
= 3.10, p
= 050), without a hemisphere by group interaction. Both FESZ (24.5% more, t63
= 2.11, p
= .039) and FEAFF (20.0% more, t68
= 2.08, p
= .041) showed greater (left + right) LV volumes than HC. Age, which was group-matched, showed significant negative associations with total NCGM relative volumes in each group (FESZ: β = −.52, t28
= 3.12, p
= .004, FEAFF: β = −.60, t33
= 4.18, p
= .0002, HC: β = −.40, t35
= 2.54, p
= .016), with an age by group interaction in NCGM volume (F3,95
= 17.32, p
Figure 2 Cross-sectional relative volume comparisons. Each symbol represents a subject with first-episode schizophrenia (FESZ, red square), a subject with first-episode affective psychosis (FEAFF, blue circle), or a healthy control subject (HC, black inverted (more ...)
Within each group, there was a negative correlation between NCGM and SCSF (FESZ: r = −.75, p < .0001, n = 29, FEAFF: r = −.48, p = .004, n = 34; HC: r = −.523, p = .001, n = 36). No group showed a correlation between SCSF and LV.
Longitudinal Volume Comparison ()
Groups did not differ in initial ICC, did not show ICC change over time, and did not differ in ICC percent change (F2,61
= 2.17, p
= .12), indicating ICC did not indirectly affect relative volume comparisons. Age was not correlated with any longitudinal volume changes. Descriptive statistics are summarized in Supplemental Table 2
The three-factor ANOVA model showed a main effect of group (F2,61 = 6.39, p = .003), a tissue by group interaction (F6,61 = 5.04, p = .0003) and a tissue by hemisphere by group interaction (F6,61 = 3.57, p = .002). Subsequent two-factor ANOVAs revealed that groups showed different amounts of NCGM volume change (F2,61 = 13.87, P < .0001), without a hemisphere by group interaction. Post hoc pairwise comparisons on total (left + right) NCGM percent change revealed significant volume reduction in FESZ (−1.7%) compared with HC (+.05%, t41 = 2.58, p = .014) and FEAFF (+3.6%, t36 = 4.18, p = .0002). In contrast, FEAFF showed a significant volume increase (+3.6%,) compared to HC (+.05%, t45 = 3.59, p = .001) and FESZ (−1.7%, t36 = 4.18, p = .0002). Although initially smaller, the relative NCGM volume in FEAFF group was not different from HC at second scan (t45 = .09, p = .93). Groups did not differ in CWM volume change (F2,61 = 1.47, p = .24) without the hemisphere by group interaction. Total SCSF volume change showed a trend-level main effect of group (F2,61 = 2.58, p = .084), with a hemisphere by group interaction (F2,61 = 3.68, p = .031). Follow-up comparisons on SCSF percent change were performed in each hemisphere separately, revealing that only FESZ group showed SCSF enlargement in the left hemisphere (+9.1%) compared with HC (−.3%, t41 = 3.13, p = .003). When total (left + right) SCSF enlargement (+7.2%) in FESZ was compared with HC (−.05%), the same group difference was still significant (t41 = 2.54, p = .015). Additionally, this SCSF enlargement in FESZ group was positively associated with the interscan interval (β = .67, t16=3.47, p = .003), the longer the interscan interval, the greater the SCSF enlargement. No other tissue volume change showed a significant association with interscan interval.
Groups differed in LV volume change (F2,61= 12.66, p < .0001), without a hemisphere by group interaction. Post hoc comparisons of percent change in total (left + right) LV volume revealed that FESZ showed greater enlargement (+10.4%) compared with HC (+1.0%, t41 = 5.54, P < .0001) and FEAFF (−2.1%, t36 = 3.70, p = .001). There was no statistical difference between FEAFF and HC groups (t45 = 1.25, p = .22, Cohen's d = .35).
Figure 3 Percentage longitudinal volume changes over 1.5 years. Color symbols are as described in . The number of subjects in each group showing a volume increase and decrease are indicated. Horizontal lines are group means; the numeric value of the statistically (more ...)
Lobar Parcellation of Neocortical Gray Matter ()
In the cross-sectional volume comparison, there was no group difference at the initial scan. At the follow-up scan, the FESZ group showed smaller volume in all three lobar regions as compared to the HC group (frontal lobe: t41 = 3.15, p = .003, temporal lobe: t41 = 2.65, p = .011, parieto-occipital lobes: t41 = 2.62, p = .012), and in frontal and temporal lobes compared with the FEAFF group (frontal lobe: t36 = 2.84, p = .007, temporal lobe: t36 = 2.63, p = .013, parieto-occipital lobes: t36 = 2.07, p = .046), without any difference in any lobar region between FEAFF and HC groups.
In the longitudinal volume comparisons, the three-factor ANOVA model showed a main effect of group (F2,61 = 12.38, p < .0001). Subsequent two-factor ANOVAs revealed that each lobar region showed a main effect of group (frontal lobe: F2,61 = 11.36, p < .0001, temporal lobe: F2,61 = 8.96, p = .0004, parietooccipital lobes: F2,61 = 9.73, p = .0002) without a hemisphere by group interaction. Post hoc pairwise comparisons revealed that the NCGM volume reduction observed in the FESZ group was localized to frontal and temporal lobes. The frontal lobe percent decrement in FESZ (−2.4%) was significantly more than HC (−.3%, t41 = 3.19, p = .003) and FEAFF (+2.8%, t36 = 3.85, p = .0005) and the temporal lobe percent decrement in FESZ (−2.6%) was significantly more than HC (+.5%, t41 = 2.79, p = .008) and FEAFF (+3.3%, t36 = 3.56, p = .001), whereas the parietooccipital lobe percent change in FESZ (−.7%) was not different from that in HC (+.2%, t41 = 1.01, p = .32). In contrast, the NCGM volume increase in the FEAFF group was more diffuse. Percent volume changes in the FEAFF group (frontal lobe: +2.8%, temporal lobe: +3.3%, parieto-occipital lobe: +4.6%) were significantly larger than FESZ (frontal lobe: t36 = 3.85, p = .0005, temporal lobe: t36 = 3.56, p = .001, parieto-occipital lobe: t36 = 3.32, p = .002) and HC (frontal lobe: t45 = 2.86, p = .006, temporal lobe: t45 = 2.28, p = .028, parieto-occipital lobe: t45 = 3.47, p = .001), although the temporal lobe volume increase compared with HC did not quite reach the Bonferroni-corrected p value.
Figure 4 Relative volumes and percent volume changes in the neocortical gray matter (NCGM) lobar parcellation. Because hemisphere by group interaction was not observed in any lobar region, bilateral volumes (left + right hemispheres) are used in this figure. In (more ...)
Patient Group Subdivision by Medication History
There were no demographic or clinical differences between the medication-based patient subgroups. There were no volume differences between lithium-treated and valproate-treated patients. The 12 FEAFF on antipsychotics showed significantly more NCGM volume increase than the 9 FEAFF not on antipsychotics (5.4% vs. 1.2%, t19 = 2.25, p = .036), and there was no significant difference in NCGM change between FESZ subgroups on and not on antipsychotics (). Mood stabilizer medication was not statistically significantly associated with NCGM percent volume change within each group, although effect sizes were moderate to high (). Change in SCSF in FESZ with MS (n = 7) was nearly zero (−.07%), whereas FESZ without MS (n = 10) showed +12.3% SCSF enlargement (t15 = 2.63, p = .019).
Figure 5 Percentage volume changes of NCGM and medication in the interscan interval. Left panel: Purple symbols indicate patients who received only atypical antipsychotics (AP). Pink symbols indicate patients who took both typical and atypical AP. Note the large (more ...)
The FEAFF patients (n = 4) on neither MS nor antipsychotics showed a slight NCGM reduction (−.2%), whereas the FEAFF patients (n = 8) on both MS and antipsychotics showed a 6.1% NCGM increase, a significant difference (t10 = 2.51, p = .031, Cohen's d = 1.73). In this medication contrast, FESZ showed no statistical significance, although the effect size was large (Cohen's d = 1.09).
Combining FESZ and FEAFF groups, the 20 patients on MS showed a significant NCGM increase compared with the 18 patients not on MS (+2.8% vs. −.6%, t36 = 2.35, p = .024), whereas the combined FESZ/FEAFF groups on and not on antipsychotics were not statistically different (moderate effect size, Cohen's d = .60).
Clinical Correlations with Volume Change over Time
In FESZ, longitudinal NCGM change was negatively correlated with changes in the BPRS score (). Whereas most patients tended to improve in their clinical status, the more the NCGM volume loss, the less the symptom improvement (or the greater the symptom deterioration) in BPRS total score (r = −.57, p = .018), thought disturbance factor (r = −.67, p = .003), and anxiety-depression factor (r = −.73, p = .001). The BPRS thought disturbance and anxiety-depression factors correlations remained significant with a Bonferroni correction for the n = 5 BPRS measures. Within the FESZ group, there was no lobar-specific clinical correlation with lobar NCGM volume changes. Also within the FESZ group, the more the bilateral LV enlargement, the higher the change in the BPRS withdrawal-retardation factor (r = .61, p = .010, ). In contrast, the FEAFF group did not show any clinical correlations with their NCGM increase (), even in the lobar parcellation, and these correlation coefficients differed significantly (Fisher r-to-z transformation) from those of the FESZ. Changes in CWM and SCSF did not correlate with change in total BPRS score or factors in either patient group.
Figure 6 Correlation between percentage volume change and symptom change in Brief Psychiatric Rating Scale (BPRS). In Y axes, plus values mean symptom deterioration over time and minus values mean symptom improvement. Note the association between the degree of (more ...)