Clinical records were used for psychiatric diagnoses according to DSM IV 10 criteria. Patients' socio-demographic, clinical and treatment information (current antipsychotic, doses and duration of treatment; previous antipsychotic treatment; psychotropic drugs with potential metabolic side effects and any other drug that might affect glucose or lipid homeostasis) was collected using an ad hoc schedule.

Assessments for the present analyses included fasting blood samples for glucose, cholesterol and triglycerides. The metabolic fasting sample was drawn between 8 and 10 a.m., after at least 8 hours' fasting, before medication administration. Plasma glucose, cholesterol and triglyceride levels were determined by enzymatic procedure applying the Roche/Hitachi Modular D-P automated chemistry analyzer and using the standard analytical system packs Glucose/God-pap, Cholesterol/CHOD cod-pap and Triglycerides/GPO-pap.

Weight, height and BMI were determined. Body mass index was computed as body weight (kg) divided by the square of height (m²). Height and weight for each patient in the study group were measured by CMHC nurses at the time of blood examination.

A reference group of adult men and women without any psychiatric history or antipsychotic treatment were randomly selected from outpatients suffering from hyperhydrosis and referred to the Department of Neurological Sciences in Bologna. Fasting plasma glucose, cholesterol and triglyceride levels were derived anonymously by chart review of these reference subjects. BMI was not available in the reference group, but the clinical charts reported if patients were overweight or obese or not. Thus it was known from the clinical charts that none of the patients in the reference group was overweight or obese.

We used analysis of variance (ANOVA) to examine differences among the treatment groups and reference group in mean glycaemia and cholesterolaemia levels and to compare mean BMI levels among treatment groups. Since triglyceride levels were not normally distributed, median trigliceridaemia levels were compared among groups, using non parametrical analysis (Kruskal Wallis test).

At a second stage, by way of co-variants in ANCOVA analysis we included the sociodemographic variables which differed among groups. Alpha levels were considered significant when below 0.01 (Bonferroni corrected: 0.05/5 groups).

To compare the prevalence of hyperglycaemia, hypercholesterolaemia and hypertriglyceridaemia among groups we used chi square analysis and the Fisher exact text. We calculated the crude and adjusted Odds Ratio (OR) to evaluate the relative risk of hyperglycaemia, hypercholesterolaemia and hypertriglyceridaemia for SGA-treated patients compared to the reference group. At a later stage we calculated the crude and adjusted Odds Ratio (OR) of metabolic disorders for SGA-treated patients without schizophrenia, obesity or over-protracted SGA treatment, since these conditions were described as risk factors for metabolic disorders and could thus be confounding factors in our analysis 11121314.

Abnormal metabolic levels and BMI were defined in our study on the basis of the National Cholesterol Education Program (NCEP) 15 and World Health Organisation (WHO) 16 criteria as follows: 1) an abnormal BMI was defined as between 25 and 29.9 for overweight and equal to or greater than 30 for obesity; 2) an abnormal blood glucose level was defined as equal to or greater than 110 mg/dl (equal to or greater than 126 mg/dl for diabetes); 3) an abnormal blood cholesterol level was defined as equal to or greater than 200 mg/dl; 4) an abnormal blood triglyceride level was defined as equal to or greater than 150 mg/dl.

Finally we evaluated the Pearson bi-variate correlations between metabolic parameters and drug doses. For the whole analysis we used Statistica for Windows (Statsoft, Kernel Release 5.5).

Table 1 shows the sociodemographic, clinical and treatment features of the study sample and of the reference group. Seventy-six patients were on maintenance treatment with SGA in the index period and all of them consented to receiving fasting blood glucose, cholesterol and triglyceride evaluations. This study sample was composed of 76 patients, 38 males and 38 females. The reference group was composed of 36 subjects, 19 females and 17 males. The psychiatric patients had a mean age of 44.6 ± 16, while the control group had a mean age of 37.3 ± 8.6 (post hoc treatment groups vs reference group: df 1,110, F 6.39, p = 0.0129). As expected, a large majority of the patients treated with SGA suffered from schizophrenia or other psychotic disorders.

Table 2 reports the treatment characteristics of the study sample. Patients in the 4 treatment groups had a similar duration of SGA exposure: 77% were treated for less than one year (on average for six months), and the remaining 23% for more than one year. Treatment duration longer than one year showed a a differing distribution trend among treatment groups (risperidone n = 8, 42%; clozapine n = 2, 33%; olanzapine n = 5, 16%; quetiapine n = 2, 10.5%; chi-square = 6.8, df = 3, p = 0.077); post hoc analysis showed that significantly more patients treated with risperidone had a treatment duration longer than one year compared with other treatment groups (42% of patients treated with risperidone vs 16% of patients treated with other SGA, chi-sq = 5.5, df = 1, p = 0.019). Dosages were at the lower end of the normal range. The distribution of co-therapies with possible metabolic side effects did not differ among the four treatment groups. A significantly higher rate of patients with quetiapine were treated with SGA in the past and a lower proportion were drug naïve (table 2) compared with patients in other treatment groups. None of the patients in the treatment or reference groups was treated with hypoglycaemic or hypolipidaemic agents or other drugs that might affect glucose or lipid homeostasis.

Table 3 shows the mean values of glycaemia, cholesterolaemia, triglyceridaemia and BMI. Differences among groups were evaluated including as co-variants age, previous treatment and treatment duration longer than one year, which differed among groups. A trend towards a significant difference in glucose mean levels was observed (p = 0.026) and post hoc analysis showed that the reference group had significantly lower glycaemia (df 1,107, F = 7.95, p = 0.006). Cholesterol mean levels were pathological in all groups and did not differ among groups. The triglyceride levels were non-normally distributed; thus we used non parametric analysis (median and Kruskal Wallis test). Median triglyceride levels were pathologic in the clozapine and quetiapine groups. We observed statistically significant differences in median triglyceridaemia among groups (p = 0.001); post hoc analysis highlighted that the reference group had significantly lower triglyceridaemia (df 1,112, H = 7.47, p = 0.006). We observed that mean BMI values were over the clinical threshold of overweight in all groups and over the clinical threshold of obesity in the clozapine group. BMI mean values differed among groups (p = 0.005); post hoc analysis revealed that patients with clozapine had a significantly higher BMI (df 1,69, F = 8.18, p = 0.006).

Table 4 shows pathological cases in each group. Hyperglycaemia prevalence showed a trend towards significant difference among groups; post hoc analysis showed that as a whole patients treated with SGA had a trend towards higher prevalence of hyperglycaemia than the references group (chi-sq = 5.15, df = 1, p = 0.02; Fisher exact text p = 0.076). Of the patients with hyperglycaemia, none of the reference subjects and 4 SGA-treated patients (2 patients with olanzapine, 1 with risperidone and 1 with quetiapine) reached the glycaemia threshold of diabetes (126 mg/dl). The prevalence of hypercholesterolaemia did not differ among groups. The hypertriglyceridaemia rate differed significantly among groups; post hoc analysis showed that the prevalence of hypertriglyceridaemia was significantly greater in SGA-treated patients than in the reference group (chi-sq = 7.26, df = 1, p = 0.007; Fisher exact text p = 0.003). Treatments groups had a significantly different rate of obesity (p = 0.015); post hoc analysis showed that patients on clozapine had a significantly higher rate of obesity (clozapine n 5, 83%, others n 19, 27%; chi-sq 8.09, df = 2, p = 0.013; Fisher exact text p = 0.012). The rate of overweight did not differ among treatment groups (p = 0.2).

The Odds Ratio (OR) of having a diagnosis of hyperglycaemia was 8 times significantly greater for patients treated with SGA than for the reference group (OR 7.90, 95%CI 1.00–62.70, p = 0.023); this risk, when adjusted for age, was still higher in SGA-treated patients (OR 6.12, 95%CI 0.75–49.73, p = 0.09). Moreover the odds of receiving a diagnosis of hypertriglyceridaemia was 4 times significantly higher for patients treated with SGA than for the reference group (OR 4.00, 95%CI 1.39–11.46, p = 0.007); this risk was not affected by adjusting for age (OR 4.00, 95%CI 1.39–11.46, p = 0.007). The OR of receiving a diagnosis of hypercholesterolaemia was not greater for SGA-treated patients than for the reference group (OR 1.09, 95%CI 0.48–2.49, p = 0.834). The higher risk of receiving a diagnosis of hyperglycaemia and hypertriglyceridaemia for SGA-treated patients compared to reference subjects decreased slightly when considering only the subgroup of patients without obesity or overweight and did not decrease for the subgroup of patients without a diagnosis of schizophrenia and treatment duration shorter than one year (data available on request).

Finally Pearson's correlation between SGA doses and study variables (BMI, glycaemia, triglyceridaemia and cholesterolaemia) was not significant (data available on request).

The most important limitations of our study are due to the small sample size and the cross-sectional design. The small sample size may not have allowed us to clarify all possible differences among treatment groups. The cross-sectional study design did preclude determination of the temporal relationship between the prescription of SGA and the development of obesity and metabolic disorders. Thus the limitation due to the cross-sectional nature of this study makes it difficult to understand if these disorders are due to SGA itself or to other risk factors for diabetes mellitus and other metabolic disorders which are frequently observed in psychiatric patients, such as poor overall physical health, unhealthy life style, inadequate health care, familial and genetic vulnerability 404142.

The lack of matching with the control group in terms of body composition, and the dearth of background variables available, limit the information that can be gleaned from such a comparison. We cannot assume that our control group is representative of the general population, since this sample was drawn from hyperhydrotic patients. Nevertheless, in our experience there is no evidence that patients suffering from idiopathic hyperhydrosis are at higher risk of diabetes or dyslipidaemia than the general population. In our sample SGA-treated patients were significantly older than the reference group. However, co-varying for age we found that age did not influence our results. This result is consistent with several studies reporting that age did not influence the risk of SGA metabolic disorders, which also occur in young patients 2043. Finally our results are limited by lack of information about some indexes of metabolic disorders, such as fat distribution and blood pressure, and other important clinical variables, such as the severity of psychiatric illness.

The relative level of risk associated with each type of SGA cannot be clearly defined by our cross-sectional study design. Only 25 (33%) patients were drug naïve before starting the studied treatment, while 23 patients (30%) had been treated in the past with different SGA and 28 (37%) with FGA. We should bear in mind that patients may have been switched from a previous antipsychotic following the development of obesity or metabolic disorders. Although we cannot rule out this potentially confounding variable, we have attempted to adjust for it by co-varying the effect of previous treatment. This did not change the results.

Finally we could not calculate the Relative Risk of developing metabolic disorders of SGA-treated patients as compared to FGA-treated patients, in view of the lack of FGA-treated patients in our sample.