Study population
The NFBC1966 cohort population and validation of the characteristics related to PCOS diagnoses have been described previously (Karjula et al. 2017; Ollila et al. 2016; Taponen et al. 2003, 2004). In brief, the cohort includes individuals expected to be born in 1966 in Northern Finland (n total = 12,058; n female = 5889), altogether comprising 96.3% of total births in the region (Rantakallio 1988, University of Oulu 1966). This study utilized the cohort data collected at the age of 31 years as well as the national register data of psychiatric diagnoses and medications. The characteristics of the study groups are presented in Table 1. Data were gathered via a questionnaire sent to all cohort subjects and from clinical examinations.
Table 1 Clinical and socioeconomic characteristics of the control women and women with PCOS at age 31 in the Northern Finland Birth Cohort, 1966 The questionnaire was sent to 5608 women of the cohort at age 31 (1996–1997), 4523 (81%) of whom responded. The questionnaire included questions concerning clinical factors such as weight and height. PCOS was screened with the following questions: (1) Is your menstrual cycle longer than 35 days more than twice a year? (considered as having oligo-amenorrhea, OA); and (2) Do you have excessive, bothersome body hair? (considered as having hirsutism, H). In total, 463 women reported OA, 471 reported H, and 153 reported having both symptoms (the latter category was considered to be women with PCOS). Women who were pregnant or using hormonal contraceptives or who did not permit the use of their personal register data were excluded from the final study population. The final analysis group at age 31 consisted of 2145 asymptomatic women (the control group), 325 women with isolated OA, 322 with isolated H, and 124 with PCOS. A flow chart of the study is presented in Fig. 1. Clinical examinations, including measurements of weight, height, serum testosterone (T), and free androgen index (FAI), were performed for 3127 women (76%).
All participants provided informed consent, and the study was approved by the Ethics Committee of the Northern Ostrobothnia District in Finland (EETTMK 94/2011).
Identification of psychoses
For the identification of psychoses, several registers were utilized:
(1) Care Register for Health Care (CRHC) (1972–2016)
(2) Finnish outpatient registers: special health care (1998–2016) and primary health care (2011–2016)
(3) Social Insurance Institution (SII) registers: sick days (1974–1999), disability pensions (1974–2000), and reimbursable medicines (1974–2005)
(4) Finnish Centre for Pensions: disability pensions (1974–2016)
The identification of psychoses is described in more detail by Filatova et al. (2017). Diagnoses of non-organic psychosis were included up to 2016 (i.e., ICD‐8 295–299; ICD‐9 295, 2961E, 2962E, 2963E, 2964E, 2967, 297–299, 2988A, 2989X; ICD‐10 F20, F22‐F25, F28-F29, F302, F312, F315, F323, F333), representing the any psychosis group (Supplementary Table 1). For subgroup analysis, we separated the psychosis cases into schizophrenia and other psychosis as shown in Supplementary Table 1. For estimating the risk of psychosis between ages 31 and 50, we used the first onset of diagnoses of psychoses from age 31 onward.
Parental psychosis
Parental psychosis was defined as a parent (mother and/or father) having non‐organic psychosis (i.e., ICD‐8 295–299; ICD‐9 295, 2961E, 2962E, 2963E, 2964E, 2967, 297–299; ICD‐10 F20, F22‐F29) at any time between 1964 and 2005. Information about parental psychosis was available from the disability pension register of the Finnish Centre for Pensions (1964–2005) and the CRHC (1972–2005), including outpatient registers from special health care (1998–2005). The proportions of parental psychosis in different study groups are shown in Table 1. There was no difference in the proportions of parental history of psychoses between the PCOS group (6.5%) and the controls (6.2%; p = 0.921).
Psychopathology scales
As part of the clinical visit, the participants filled out a questionnaire consisting of mental health-related true/false (scored 0/1) questions collected from several psychological scales. The scales are used for the identification of psychopathological symptoms in individuals (Miettunen et al. 2011). The psychopathology scales used were the Social Anhedonia Scale (SAS), Physical Anhedonia Scale (PHAS), Perceptual Aberration Scale (PAS), Hypomanic Personality Scale (HPS), Bipolar II scale (BIP2), and Schizoidia Scale (SCHD). A more detailed description of the scales utilized in the same birth cohort has been published (Miettunen et al. 2011; Miettunen & Jääskeläinen 2010). In cases where more than 10% of the response items were missing, the scale was excluded.
Confounding variables
Body mass index
In the clinical examinations, weight (kg) was measured on a digital scale which was calibrated regularly. Height (cm) was measured twice, using a standard and calibrated stadiometer. Body mass index (BMI) was calculated in kg/m2. If a measurement was missing, the self-reported values were used. No statistical difference was observed between the measured and the self-reported BMIs (Ollila et al. 2016).
Testosterone and FAI
Serum T and the sex hormone binding globulin (SHBG) were assayed at age 31, as previously described (Karjula et al. 2017). Testosterone levels were assayed using Agilent triple quadrupole 6410 LC/MS equipment with an electrospray ionization source, operating in positive-ion mode (Agilent Technologies). At age 31, SHBG was assayed using fluoroimmunoassay (Wallac, Inc. Ltd., Turku, Finland). FAI was calculated as follows: 100xT (nmol/L)/SHBG (nmol/L).
Socioeconomic status/education
Socioeconomic status was based on years of education. The variable was classified into three groups according to the number of education years: ≤ 9, 9–12, and > 12 years.
Statistical analysis
Statistical analyses were performed using SPSS version 25 for Windows. The analysis of categorical variables was carried out using the Chi-square test and Fisher’s exact test when appropriate. Cox regression analysis (using hazard ratios, HR) was used to estimate the association between PCOS and psychotic disorders. Kaplan-Meier survival analysis (Mantel-Cox estimation) was used to estimate the incidence of psychotic disorders. Parental history of psychosis, BMI, education, testosterone, or FAI was used as a covariate. The analyses of psychometric scales were done using Student’s t-test for variables with normal distributions and the Mann-Whitney U-test as a non-parametric test. The data was also analyzed after removing women with a diagnosis of psychosis before age 31 to assess the incidence after the establishment of PCOS symptoms. The results are reported as means with standard deviations (SD), medians (25% and 75% quartiles), prevalence (%), and HR with a 95% confidence interval (CI). p Values < 0.05 were considered statistically significant.