TIPP (Treatment and Early Intervention in Psychosis Program), a specialized early psychosis program, was launched in 2004 at the Department of Psychiatry, Lausanne University Hospital Switzerland (40). Entry criteria to the program are: [I] aged between 18 and 35; [II] residing in the catchment area [Lausanne and surroundings; population about 300 000]; [III] meeting threshold criteria for psychosis, as defined by the “Psychosis threshold” subscale of the Comprehensive Assessment of At Risk Mental States [CAARMS Scale (40, 41). In the TIPP program, each patient is followed up by a psychiatrist and a Case manager. The program involves multiple meetings over 3 years between patients and their case managers, as well as scheduled regular appointments with psychiatrists. The Ethics Committee of the Canton of Vaud (CER-VD) granted access to TIPP clinical data (demographic, assessment of trauma history, levels of symptoms, and other clinical variables). More details on the TIPP program can be found elsewhere (21).

In the Department of Psychiatry at Lausanne University Hospital, patients aged 18 years old or older starting a psychotropic treatment that is known to have a potential risk of inducing metabolic disturbances (clozapine, olanzapine, risperidone, quetiapine, aripiprazole, amisulpride, lithium, valproate and/or mirtazapine) are prospectively followed up for metabolic parameter evolution (see the Biological measures section below). Baseline clinical data were obtained during hospitalization and follow‐up data were obtained in the hospital or in out‐patient centers during a medical examination based on the department guideline for metabolic follow‐up performed on a routine basis (42, 43). Since 2007, a study (PsyMetab) has been ongoing to determine the clinical, environmental, and genetic determinants of metabolic adverse effects of psychotropic drugs (44).

This observational prospective study has been approved by the ethics committee of the Canton de Vaud (CER-VD). Informed written consent was obtained for all participants. Data collected between 01.01.2007 and 06.08.2021 were extracted for the present study. In addition, the CER-VD granted access to the data of patients with a metabolic follow-up in the Department of Psychiatry, Lausanne University Hospital, collected before 01.01.2016 (PsyClin). The present analysis involves participants included in the TIPP program and who were also part of the PsyMetab or PsyClin study between 08.05.2007 until 31.12.2015 (PsyClin) and until 06.08.2021 (PsyMetab). A detailed description of PsyMetab and PsyClin can be found elsewhere (45).

The assessments are described in detail elsewhere (21). In summary, CT is defined if at least one of the three abuses (sexual, physical, or emotional) has been experienced by the patient before the age of 16. History of childhood trauma is assessed by the case manager during the 3-year follow-up of the TIPP program. Details about assessment can be found in Supplementary Informations . Patients are asked at baseline about the quality of their relationship with each parent. This categorical variable ranges from 0 to 5 (0: absent parent; 1 to 5: very bad, bad, suitable, good, very good relationship, respectively). The number of subjects in each categorical variable was insufficient to run statistical analyses separately; therefore, this variable was dichotomized as good relationship with the father or mother (GRF or GRM, respectively; scores 3 to 5) and lack of good relationship with the father or mother (LGRF or LGRM respectively; scores 0 to 2).

PsyMetab and PsyClin collect metabolic (i.e., body mass index (BMI), fasting plasma glucose (FPG), triglycerides, total cholesterol (TC), and low-density lipoprotein (LDL) and high-density lipoprotein (HDL) cholesterol) and clinical (i.e., diagnosis, age, sex, and smoking status) data from patients’ medical records at the beginning of weight-gain-inducing psychotropic treatments, and at 2, 3 and 12 months. Additional metabolic observations can be obtained during the hospital stay. Because we aimed to focus on blood parameters, clinical parameters were not analysed but, rather, used as covariates in statistical models.

All results from all available blood samples were taken into account except for FPG and triglycerides, which were kept only in patients who were fasting (46). Non-HDL cholesterol was calculated (TC minus HDL cholesterol). The analysis was defined using the subsequent timepoints in relation to the initiation of a psychotropic treatment: 0 (-30 to +7 days), 2 (45 to 75 days), 3 (75 to 105 days), and 12 (290 to 430 days) months in order to reach a sufficient number of patients at each timepoint to run statistical analysis. For prolactin determinations, samples from PsyMetab with no antipsychotic use at least 15 days before the introduction of an antipsychotic were used. Details about determinations of prolactin concentrations are described elsewhere (47). Hyperprolactinaemia was defined for men and women as ≥16.5 ng/ml and ≥28.3 ng/ml, respectively (48). Prolactinaemia, triglycerides, and FPG were analysed at baseline. Other time points were not analysed due to insufficient data availability. Baseline was defined up to 7 and up to 3 days after the onset of antipsychotic treatment for metabolic variables and for prolactin, respectively. In the case of an interruption of the psychotropic medication for more than two weeks or in the case of replacement by another drug, the follow-up is restarted from baseline. In the case of an introduction of a second drug from the list (see list in Supplementary Table 1 ), the follow-up is restarted. In such case, only the earliest follow-up, which has less missing information, was considered.

Gender refers to biological sex as recorded in official civil status documents. Participants were categorized as male or female based on this information. Ethnicity was dichotomized into “Westerners” and “Non Westerners” according to clinical data form Psymetab/PsyClin. When possible, missing values were filled in with data from the Genome-Wide Association Studies (see Supplementary Information for more details). Psychotropic medications were categorized into low, medium, and high risk of weight gain described in Supplementary Table 1 . In order to adjust for the sum of the doses of antipsychotic intake per day, a variable was created named “weighting of the total daily dose of antipsychotic”. This variable weights the daily dose of an antipsychotic to its upper limit of the average dose per day adapted from Swissmedic ( Supplementary Table 2 ). A sum of weightings of each antipsychotic taken in the same day is calculated. Medications were also categorized by their effect on prolactin levels based on a large meta-analysis (49) (see Supplementary Table 3 ). Socio-economic status was calculated according to an area-based index of Swiss socio-economic position (SSEP), ranging between 0 (most disadvantaged) and 100 (most privileged). The characterization of the SSEP index by place of residency was based on 2000 census data including income, education, occupation, and housing conditions. To estimate patients’ SES, postal addresses were obtained and geocoded using Google API via the ggmap R package. More details can be found elsewhere (50).

Descriptive data are shown as numbers and percentages for categorical variables or as median and interquartile range for continuous variables. For comparisons between groups, t-tests, Wilcoxon-Mann-Whitney rank-sum tests, or the Chi-squared test were used, depending on the variable type. For multivariate analyses, generalized linear models and linear mixed-effects models were used to examine the associations between plasma levels of prolactin and metabolic parameters with CT or the relationship with the mother or the father. Backward selection was performed on each model to keep only the significant covariates, resulting in different covariate adjustments between models (see Supplementary List 1 for full models).

Data preparation was conducted using Stata 17.0 (StataCorp; College Station, Texas). Univariate and multivariate analyses were performed in R environment (version 4.1.2; Rstudio, Inc; Boston, Massachusetts). Statistical significance was determined at α = 0.05.

A total of 226 subjects were included in both PsyMetab and TIPP cohorts. The demographic and clinical parameters are shown in Table 1 , while metabolic parameters are displayed in Table 2 . Of note, no significant associations between biological variables and the outcomes were found at 3 months and were therefore not included in the tables and figures. Considering the percentage of missing values for biological parameters in Table 2 , the number of dropouts after 1 year was greater in the groups reporting no CT, GRM, and GRF. The number of dropouts was also reported after 2 months in the group reporting GRM.

Median levels of TC were above the physiological range at 12 months in the CT, LGRF and LGRM groups (5.2 mmol/L, 5.2 mmol/L, and 5.0 mmol/L,respectively; Table 2 ), but not in patients with no CT and in those with GRF and GRM (4.4 mmol/L, 4.4 mmol/L and 4.7 mmol/L, respectively). Median levels of LDL were above the physiological range at 12 months in the CT, LGRF and LGRM groups (3.2 mmol/L, 3.1 mmol/L and 3.1 mmol/L, respectively) but not in patients with no CT, with GRF and with GRM (2.7 mmol/L, 2.7 mmol/L, 2.8 mmol/L, respectively).

CT in the forms of sexual, physical, and emotional abuse occurred in 27% of the cohort. Patients with CT were older at the beginning of symptoms (25 y vs 22 y, p=0.02; Table 1 ) and at entry into the cohort (27 y vs 23 y, p=0.01). CT was associated with a decreased risk of having non-HDL cholesterol level above physiological values (2.8 mmol/L) at baseline (OR=0.39, 95%CI:[0.2; 0.95], Figure 1A ). CT was associated with higher levels of non-HDL at 12 months (0.48 mmol/L, 95%CI:[0.03; 0.93]; Figure 1B ) with a 17% higher increase (95%CI:[0.4;34]; Figure 1C ) compared to the non-CT group after multivariate analysis. TC was significantly higher in patients with CT (5.2 mmol/L vs 4.4 mmol/L, p<0.05; Table 2 ), which increased by 11% (95%CI:[0.4;21]; Figure 1C ) at 12 months in the CT group as compared to the non-CT group. No associations were found between CT and LDL, HDL, FPG, or triglycerides levels.

LGRF occurred in 33% of the included patients ( Table 1 ). LDL levels were significantly higher in patients reporting LGRF (2.5 mmol/L vs 2.3 mmol/L, p<0.05; Table 2 ). At 2 months, LGRF patients presented elevated LDL (3.0 mmol/L vs 2.4 mmol/L, p<0.05), and in multivariate analysis, LGRF was significantly associated with approximately a 5-fold increased risk of having LDL levels above the median after 2 months (median=2.58 mmol/L, OR=5.5, 95%CI:[1.3;28]; Figure 1A ) and approximately a 4-fold increased risk after one year of follow-up (median=2.8 mmol/L, OR=3.7, 95%CI:[1.3;12]). After one year of treatment, LGRF was significantly associated with a higher level of HDL (0.19, 95%CI:[0.02;0.35] mmol/L; Figure 1B ). Non-HDL at baseline was significantly higher in patients reporting LGRF (3.1 mmol/L vs 2.8 mmol/L, p=0.04; Table 2 ). Similar to results with LDL, LGRF was significantly associated in multivariate analysis with approximately a 5-fold increased risk of having non-HDL levels above the median after 2 months (median=3.1 mmol/L, OR=4.7, 95%CI:[1.3; 19]; Figure 1A ) and approximately a 3-fold increased risk after one year (median=3.4 mmol/L, OR=2.8, 95%CI:[1.03; 8.0]).

TC at baseline was significantly higher in patients reporting LGRF (4.5 mmol/L vs 4.1 mmol/L, p=0.03; Table 2 ). After one year of treatment, LGRF was significantly associated with a higher level of TC (0.42, 95%CI:[0.02; 0.82] mmol/L; Figure 1B ).

Finally, no association was found between LGRF with prolactin, FPG, or triglycerides.

In univariate analysis, no significant differences in terms of demographic or biological results were found between patients with and without GRM ( Tables 1 , 2 ). In multivariate analysis, LGRM was associated with an increase of LDL by 34% after 2 months (95%CI[6.1; 62]; Figure 1C ) and with an increase of TC by 21% (95%CI=[5.9; 36]) and 12% (95%CI:[1.1; 24]) after 2 months and one year, respectively. No association was found for HDL, FPG, or triglycerides.

Levels of prolactin at baseline were above the physiological threshold in men regardless of history of CT and quality of relationship with both parents ( Table 3 ). In women, levels of prolactin were also above threshold except for LGRF and LGRM (28 ng/ml and 21 ng/ml, respectively). In multivariate analysis, levels of prolactin at baseline were elevated in women reporting LGRM (124, 95%CI:[91, 157]ng/ml, p<0.001; Figure 2A ) but not in men (7.2, 95%CI:[-3; 17] ng/mL; Figure 2B ).

Finally, multivariate analyses for lipidaemia, prolactin, and FPG were conducted combining CT with LGRF or LGRM to examine whether or not lack of good relationships with parents would worsen those values in patients with CT, but no significant results were found (data not shown).

This study has several limitations and strengths. First, this is an observational, longitudinal study, which does not allow for establishing a causal relationship. Second, patients with GRF or GRM were more likely to drop out of the study, an unexpected finding as secure attachment has been associated with greater compliance with treatment and follow-up (61). This could have introduced a selection bias with metabolic alterations observed in patients with LGRF or LGRM because of the long follow-up period. Third, because of the moderate sample size, not all relevant covariates were included in the analysis (i.e., Positive and Negative Syndrome Scale (PANSS) scores for prolactin or waist circumference for metabolic parameters). Fourth, prolactin analyses were made in non-fasting patients regardless of their menstrual cycle. Analyses of prolactin levels were also performed without excluding macroprolactin, which could have increased the rate of elevated levels of prolactin. Nonetheless, a recent study showed that hyperprolactinaemia in women with psychotic disorders was not significantly related to the presence of macroprolactin (62). Fifth, the definition of quality of relationship with parents was not assessed on a standardized scale, thus making it difficult to compare with previous studies. Given this, the results should therefore be interpreted with caution. Finally, the results presented here considered only relationships with parents and did not take other potential relationships into account.

Strengths of the present study include the longitudinal follow-up after antipsychotic initiation, enabling the prospective assessment of CT, LGRF, and LGRM effects on prolactin and metabolic parameters in real-life conditions.