The Institutional Review Board at the University of Iowa reviewed and approved the procedures in this retrospective chart review study. Billing data was utilized to identify 217 outpatients seen between January 1, 2006 and June 30, 2006 for a primary diagnosis bipolar disorder from an adult, outpatient psychiatry clinic at a tertiary care hospital. The first visit within the specified time frame served as the index clinic visit. Primary diagnoses of Bipolar I, Bipolar II, and Bipolar NOS were identified based on billing codes and verified from psychiatric records. Records were reviewed from January 1, 2002 to the earlier of one year after index visit or April 20, 2007.
Age at index visit was obtained from the billing data and verified on chart review. Race and ethnicity were obtained from electronic hospital records when reported. The following diagnoses were systematically abstracted from chart records: hypertension, hyperlipidemia, diabetes mellitus, and thyroid disease. These diagnoses were recorded if identified from electronic record problem lists or physician records indicating diagnosis of and treatment for the condition. These diagnoses were not inferred from vital signs or laboratory values. Psychotropic medications and dosages from index visit were consistently reported and recorded. Non-psychiatric medications were not individually recorded though were considered to assess treatment for the aforementioned diagnoses.
Metabolic monitoring was not obtained consistently enough to allow distillation of simple point or one year prevalence. Measures of weight, fasting glucose, and fasting lipids were systematically recorded. The index visit was utilized as reference point for the following time periods: prior to index visit or baseline, within six months of following index visit, and six to 12 months following index visit. When multiple measures were present for any of these periods, the closest to index visit was selected for baseline, the closest to three months following index visit for the first six months of follow-up, and the closest to 12 months for the six to twelve month period following index visit. For the primary and secondary hypotheses, the value associated with maximum risk was chosen from among these three time points if recordings were present for multiple time periods.
Statistical analyses were performed using SAS software (36
). We utilized the more popular National Cholesterol Education Program (NCEP) definition of metabolic syndrome as published in its third Adult Treatment Protocol (ATP III), which requires at least three of the following: abdominal obesity (waist circumference > 40 inches in males or 35 inches in females), elevated triglycerides (≥ 150 mg/dL), low HDL (< 40 mg/dL in men or < 50 mg/dL in women), elevated blood pressure (≥ 130/85 or on antihypertensive medication), or elevated fasting glucose (≥ 110 mg/dL or on medication for diabetes) (37
). Descriptive statistics were compiled for body mass index, triglycerides, low density lipoprotein (LDL), HDL, and fasting glucose. A body mass index ≥ 25 and < 30 was defined as overweight with ≥30 defined as obese. We further substituted a body-mass index (BMI) ≥30 for visceral obesity as measured by waist circumference to estimate the prevalence of metabolic syndrome in this sample. BMI and waist circumference are highly correlated (r
= 0.86) with a BMI of 30 corresponding to a waist circumference of 37.2 inches (38
). The National Health and Nutrition Examination Survey (NHANES) 1999–2000 estimates the correlation between BMI and waist circumference as 0.91–0.94 for men and 0.88–0.94 for women (39
). With 96% of those with a BMI ≥30 and 43% of those with a BMI between 25 and 30 meeting NCEP waist circumference criteria (39
), our use of obesity by BMI ≥30 as a surrogate for gender-specific waist circumference would be expected to yield a conservative underestimate of the prevalence of this metabolic syndrome criterion. Consistent with the NCEP guidelines for metabolic syndrome, elevated triglycerides was delineated by values ≥150 mg/dL, low HDL by values < 40 mg/dL in men or < 50 mg/dL in women, and elevated fasting glucose by values ≥110 mg/dL. Because of missing data compromising estimates of metabolic syndrome, we utilized two sub-samples in a sensitivity analysis: a restricted sample of those with all of these measurements available and an extended sample of any participants with at least three of BMI, fasting glucose, fasting triglycerides, fasting HDL or blood pressure available. A 95% confidence interval for the prevalence estimates of cardiovascular risk factors was calculated from the binomial distribution for gross comparison to national data. Chi-square tests compared differences by categorical variables between those currently prescribed second-generation antipsychotics associated with weight gain or valproic acid derivatives and those not prescribed these medications, selected by strength of association with weight gain and metabolic complications.