Caudate nucleus volume in medicated and unmedicated patients with early- and adult-onset schizophrenia

Main analysis

Adult patient analysis

In the bivariate analysis, patients with EOS were on average seven years younger (p < 0.001), had three years longer DOI (p < 0.001) and 3 ½ months longer DUP (p = 0.009) compared to patients with AOS, assessed with t-tests (Table 1). Among both patients with EOS and AOS, women had smaller caudate than men (assessed with point-biserial correlations, r_pb = -0.319, p = 0.003 and r_pb = -0.326, p < 0.001 for EOS and AOS, respectively). In EOS only, medication use, r_pb=0.322, p = 0.003, and CPZ, assessed with Spearman’s correlation, r_s = 0.248, p = 0.033, were both positively correlated with caudate volume (Table 1 & Suppl. Figure 3).

In the ANCOVA of diagnostic status (EOS/AOS) on caudate module (model 1; Fig. 1 & Suppl. Table 1), there was a statistically significant main effect of EOS/AOS, F(1,322) = 7.200, p = 0.008, η² = 0.022. In the main ANCOVA (model 2; Fig. 1; Table 2), there was a significant EOS/AOS-by-antipsychotic medication interaction (p = 0.004) on caudate: among patients with EOS, there was a statistically significant effect of antipsychotic medication use on caudate, F(1,320) = 13.019, p < 0.001, η² = 0.039, whereas among patients with AOS, there was no such effect, F(1,320) = 0.202, p = 0.654, η² = 0.001. Specifically, adjusted mean caudate volumes in medicated patients with EOS were significantly larger than in non-medicated patients with EOS, a difference of 967 mm³ (95% CI, 440 to 1494), whereas adjusted mean caudate volumes in medicated and non-medicated patients with AOS did not differ (Fig. 1; Table 2). Conversely, among medicated patients, there was a statistically significant effect of EOS/AOS status on caudate volume, F(1,320) = 11.882, p < 0.001, η² = 0.036, whereas among unmedicated patients, there was no such effect, F(1,320) = 3.185, p = 0.075, η² = 0.010. Specifically, among medicated patients, adjusted mean caudate volume was significantly larger in EOS than in AOS, a difference of 366 mm³ (95% CI, 157 to 575), whereas among unmedicated patients, EOS and AOS patients did not significantly differ (Fig. 1; Table 2). Inserting DOI and DUP into the model (model 3; Suppl. Table 2), there was still a significant EOS/AOS-by-antipsychotic medication interaction (p = 0.008). Finally, EOS exhibited a significantly higher frequency of lifetime cannabis use compared to AOS (Suppl. Table 5). When the cannabis use variable was included in model 3, the interaction term remained significant (p = 0.012).

Caudate nucleus volume (in mm³) differences between adult patients with schizophrenia (SZ), early-onset SZ (EOS), adult-onset schizophrenia (AOS) and healthy controls (HC) (up), and between currently medicated with antipsychotics (AP+) and currently non-medicated with antipsychotics (AP-) adult patients with EOS and AOS (down). Adjusted means from age-, sex, scanner- and estimated total intracranial volume-adjusted analyses of covariance are shown. Error bars represent the 95% confidence intervals calculated using standard errors. The green bars represent the HC group, while the blue bars represent the patient groups. ***p < 0.001. **p < 0.01.

In the ANCOVA of the continuous age of onset of psychosis variable, age of onset was not associated with the caudate volume, F(1,322) = 0.647, p = 0.422, whilst controlling for sex, age, ICV and scanner (model 1). However, adding the antipsychotic medication variable and the age of onset-by-antipsychotic medication interaction term (model 2) we found a significant interaction (p = 0.004): among patients on antipsychotics, the age of onset was inversely (non-significantly) associated with caudate, F(1,287) = 3.061, p = 0.081, η² = 0.011, whereas among patients not on antipsychotics, the age of onset was positively (non-significantly) associated with caudate, F(1,28) = 3.060, p = 0.091, η² = 0.099.

We finally conducted two MANCOVAs on accumbens, amygdala, caudate, hippocampus, pallidum, putamen and thalamus. In the first MANCOVA we inserted age, sex, scanner, ICV, EOS/AOS, antipsychotic use and the interaction term EOS/AOS-by-antipsychotic use. In the second MANCOVA, we inserted age, sex, scanner, ICV, age of onset of psychosis (continuous), antipsychotic use and the interaction term age of onset of psychosis-by-antipsychotic use. In the first MANCOVA, there was no significant main effect of EOS/AOS (p = 0.948) or EOS/AOS-by-antipsychotic use interaction effect (p = 0.120). Similarly, in the second MANCOVA, there was no significant main effect of age of onset (p = 0.647) or age of onset-by-antipsychotic use interaction effect (p = 0.097). Follow-up analysis of the second MANCOVA showed that there was a significant age of onset-by-antipsychotic use interaction effect on the caudate (p = 0.004, also shown in the previous paragraph), but not on the accumbens (p = 0.086) amygdala (p = 0.584), hippocampus (p = 0.851), pallidum (p = 0.501), putamen (p = 0.078) or thalamus volumes (p = 0.401).

Adult patient/control analysis

In the bivariate analysis of the adult sample, patients (n = 329) were three years younger than HC (n = 774); patients and HC did not significantly differ in sex distribution or ICV (Table 3). In the sex-, age-, ICV-, and scanner-adjusted ANCOVA, patients had significantly larger caudate volume than HC, F(1,1096) = 19.372, p < 0.001, η² = 0.017, (Fig. 1 & Suppl. Table 3). Stratifying by age of onset, both EOS patients, F(1,850) = 14.580, p < 0.001, η² = 0.017, and AOS patients, F(1,1013) = 10.100, p = 0.002, η² = 0.010, had significantly larger caudate than HC.

Adolescent patient/control analysis

In the bivariate analysis of the adolescent samples, patients and HC did not differ in sex distribution, age or ICV (Table 3). In the sex-, age-, ICV-, and scanner-adjusted ANCOVA, patients had non-significantly larger caudate volume than HC, F(1,146) = 3.864, p = 0.051, η² = 0.026 (Fig. 2 & Suppl. Table 4).

Caudate nucleus volume (in mm³) differences between adolescent patients with non-affective psychosis and healthy controls (HC) (up), between medicated (currently on antipsychotics; AP+) adolescent patients and HC as well as unmedicated (AP-) adolescent patients and HC (middle), and between AP + adolescent patients and AP + adult patients with early-onset schizophrenia (EOS) (down). Adjusted means from age-, sex, scanner- and estimated total intracranial volume-adjusted analyses of covariance are shown. Error bars represent the 95% confidence intervals calculated using standard errors. The green bars represent the HC group, while the blue bars represent the patient groups.

Post-hoc analysis

Caudate specificity and laterality in EOS vs. AOS medicated adult patients

The results of the post-hoc analysis on all left and right subcortical structures are shown in Table 4. Both left (p < 0.001) and right (p = 0.003) caudate volumes were significantly larger in medicated patients with EOS relative to medicated patients with AOS, whereas none of the other analyzed subcortical structures differed between medicated patients with EOS and AOS.

Medicated and non-medicated adolescent patients vs. adolescent HC

Analyzing medicated patients only and HC, patients had significantly larger caudate volume than HC, F(1,126) = 5.220, p = 0.024, η² = 0.040, whereas analyzing non-medicated patients only and HC, patients and HC did not differ in caudate volume, F(1,105) = 0.083, p = 0.774, η² = 0.001 (Fig. 2 & Suppl. Tables 6 and 7).

Medicated adult patients with EOS vs. medicated adolescent patients

There was no significant difference in caudate volume between medicated adult patients with EOS and medicated adolescent patients, F(1,102) = 1.401, p = 0.239, η² = 0.014 (Fig. 2 & suppl. material). When we excluded adolescent patients with psychosis not otherwise specified, there was still no difference in caudate volume between adolescent patients with EOS and adult patients with EOS, F(1,74) = 0.237, p = 0.628, η² = 0.003.