We included a total of 312 patients; 187 with a DSM-IV diagnosis of schizophrenia spectrum disorder (schizophrenia, n = 143; schizophreniform, n = 11; schizoaffective disorder, n = 33) and 125 with a diagnosis of bipolar spectrum disorder (bipolar I disorder, n = 72; bipolar II disorder, n = 45; bipolar not otherwise specified, n = 8). The patients with bipolar disorder had a mean GAF score of 59 for symptoms and 57 for function. The patients with schizophrenia had GAF scores of 41 for symptoms and 42 for function. We included 239 consecutively recruited healthy controls.
There were no differences in age and sex between the patient group and the control group, but the controls had a significantly higher rate of European ancestry (). The schizophrenia group had higher proportion of men than the bipolar and control groups. At the time that blood samples were collected, 13% of patients in the bipolar disorder group and 7% of the schizophrenia group were not taking medication. None of the participants in the control group were taking any medications.
In total, 70% of the patients (n = 216) were taking antipsychotics: olanzapine (n = 105, 34%), quetiapine (n = 24, 8%), ziprasidone (n = 18, 6%), risperidone (n = 16, 5%), aripiprazole (n = 13, 4%), perfenazine (n = 15, 4%), clozapine (n = 11, 4%), amisulpride (n = 3, 1%), zuclopenthixol (n = 4, 1%), haloperidol (n = 3, 1%) and chlorprothixene (n = 2, 1%). A total of 31% were taking mood-stabilizing agents: lamotrigine (n = 44, 14.1%), valproate (n = 26, 8.3%), lithium (n = 23, 7.4%), carbamazepine (n = 5, 1.3%) and topiramate (n = 1, 0.3%). A total of 32% were taking antidepressive medications: selective serotonin reuptake inhibitors (n = 65, 21%), venlafaxine (n = 21, 6.7%) and mirtazapine (n = 14, 4.5%).
Of the patients, 2% were regularly taking medications for somatic disorders (insulin, cetirizin, omeprazol, metoprolol and diclofenac). None used anticoagulants. Fifty-five percent of the patients were current smokers (56% of the schizophrenia patients, 54% of the bipolar disorder patients). We did not have information about the smoking habits of the controls, but among patients, there was no correlation between smoking and OPG levels. Of the patients, 7% had used illegal substances in the previous 2 weeks, 1% had diabetes, 4% had cardiovascular disease and 4% had an autoimmune disorder (autoimmune thyroiditis n = 8, psoriasis n = 2, ankylosing spondylitis n = 1, celiac disease n = 1, regional enterocolitis n = 1).
The mean OPG level was 10% higher in the patient group (mean 2.78 ng/mL, standard deviation [SD] 1.47 ng/mL) than in the control group (mean 2.52 [SD 0.92] ng/mL; t531 = 2.6, p = 0.011; ). The comparison of OPG levels across the 3 groups (schizophrenia, bipolar disorders and controls) indicated a significant between-group difference (F550 = 3.1, p = 0.046); however, none of these group differences was statistically significant after correction for multiple testing (bipolar disorder v. controls, p = 0.08; schizophrenia v. controls, p = 0.20; schizophrenia v. bipolar disorder, p = 0.94).
Osteoprotegerin levels in study participants
We performed further analyses of the mean OPG levels according to diagnosis subtype: schizophrenia (n = 143), mean 2.72 (SD 1.36) ng/mL; schizophreniform (n = 11), mean 2.25 (SD 0.67) ng/mL; schizoaffective disorder (n = 33), mean 3.04 (SD 2.13) ng/mL; bipolar 1 (n = 72), mean 2.76 (SD 0.05) ng/mL; bipolar not otherwise specified (n = 8), mean 2.22 (SD 0.34) ng/mL; and bipolar 2 (n = 45), mean 3.06 (SD 0.92) ng/mL. Although there were some differences between the diagnostic subgroups, they were not statistically significant.
The regression analyses indicated that OPG levels were significantly elevated in the patient group compared with the control group after adjustment for possible confounders (t485,12, 473 = −1.97, p = 0.049; ). There were slightly more women in the control group (56%) than in the patient group (50%). Women had higher OPG levels (mean 2.84 [SD 1.35] ng/mL) than men (mean 2.48 [SD 1.14] ng/mL; p = 0.001), but the group-by-sex interaction term was not significant in the whole sample (p = 0.35).
Multiple linear regression analysis of the relation between severe mental disorders and osteoprotegerin levels after adjustment for potential confounding variables*
The known risk factors for cardiovascular disease, cholesterol and hsCRP, were weakly, but significantly, correlated with OPG level (cholesterol, r = 0.14; hsCRP, r = 0.17; p < 0.001). A diagnosis of cardiovascular disease or diabetes was not correlated with OPG level. Osteoprotegerin level was weakly negatively correlated with creatinine level (r = −0.09, p = 0.04). The variation in OPG levels was higher in patients than in controls (F = 9.5, p = 0.002; ).
Distribution of the osteoprotegerin levels in individual patients, showing a higher variation among patients than among controls (p = 0.002).
The use of antipsychotic drugs may be associated with inflammatory response.32
We therefore performed subgroup analyses by medication. Patients not taking antipsychotics (n
= 80) had significantly higher OPG levels than did controls (mean 2.93 [SD 1.41] ng/mL v. mean 2.52 [SD 0.92] ng/mL; t102
= 2.4, p
= 0.019). There was no difference in mean OPG level between patients using first- and second-generation antipsychotics (first generation, n
= 25, mean 2.57 [SD 0.80] ng/mL; second generation, n
= 192, mean 2.78 [SD 1.57] ng/mL; t259
= −0.65, p
= 0.65). This suggests that medication use did not confound the results.
Patients who had used illegal drugs within the 2 weeks before testing did not have different OPG levels than the patients who had not used illegal drugs. Within the patient group as a whole, high OPG levels were positively correlated with age (r = 0.15, p < 0.001) and female sex (r = 0.15, p < 0.001).
The bipolar group included more women than men. Because both the women and bipolar patients tended to have higher OPG levels, we tested whether sex differences were confounding the differences in OPG levels between the schizophrenia and bipolar groups. Multiple linear regression analyses by OPG level in patients revealed that the group-by-sex interaction term was significant (t284 = 2.0, p = 0.049). However, there were no significant differences in OPG levels between the schizophrenia and bipolar groups after adjustment for group, sex and other potential confounders. Adding group-by-sex as a factor did not affect the result.
The levels of OPG were not significantly correlated with smoking status in patients. To further investigate whether the elevated OPG levels were confounded by smoking, we compared OPG levels in nonsmoking patients and controls. The OPG levels in nonsmoking patients (n = 136) were significantly higher than in the controls (mean 2.84 [SD 1.51] ng/mL v. mean 2.52 [SD 0.92] ng/mL; t194 = 2.2, p = 0.026). The non-smoking patients had similar OPG levels as patients who smoked (mean 2.74 [SD 1.47] ng/mL; t312 = 0.57, p = 0.58).
To test whether elevated body mass index (BMI) in the patient group was a confounding variable, we analyzed the subgroup of individuals for whom BMI was available (297 patients and 29 controls). There were no significant correlations between OPG levels and BMI (r = 0.03, p = 0.6). We compared OPG levels in nonobese participants (BMI < 30; n = 239 patients with mean BMI of 24.1 [SD 3.26]) with the control group. The patients with normal BMIs had significantly higher OPG levels than the controls (mean 2.79 [SD 1.5] v. mean 2.52 [SD 0.92]; t391 = 2.3, p = 0.021). We also controlled for BMI-related metabolic factors (i.e., glucose and cholesterol) in the analysis ().