Impact of Patients' Preexisting Metabolic Risk Factors on the Choice of Antipsychotics by Office-Based Physicians

From 2005 through 2007, an estimated 43.8 million (14.6 million annually) visits were associated with antipsychotic medications and accounted for 1.5% of all office-based physician visits during these three years. A majority (80%) of the visits mentioned one or more antipsychotics with high or medium risk of causing metabolic abnormalities. The six most commonly mentioned antipsychotics were quetiapine (34%), risperidone (23%), olanzapine/olanzapine-fluoxetine combination (17%), aripiprazole (14%), ziprasidone (7%), and haloperidol (3%). Nearly four million visits (9%) resulted in mentions of multiple antipsychotic agents. The most common combinations were quetiapine with aripiprazole (2%), quetiapine with risperidone (1%), and quetiapine with ziprasidone (1%) (Table 1).

Table 2 reports the patient and provider characteristics of antipsychotic visits compared with other visits. Patients with antipsychotic visits were more likely to be nonelderly adults (ages 18–64), less likely to be Hispanic, and significantly more likely to have schizophrenia (13% versus <.1%), bipolar disorder (20% versus .3%), and depression (53% versus 8%). Over half of the antipsychotic visits (54%) were to psychiatrists, whereas only 2% of other visits were to psychiatrists. Patients' BMI or weight measurements were more likely to be missing from data for antipsychotic visits than for other visits (73% versus 63% for BMI and 54% versus 32% for weight).

Table 3 compares patients' preexisting metabolic risk and mental health conditions across the metabolic risk categories of the antipsychotics mentioned during the visits. Among those with available BMIs, the proportion of visits with mentions of high-risk antipsychotic drugs was lower among more obese patients, and the reverse was true for low-risk agents (p=.007). Patients with schizophrenia were more likely than other patients to receive antipsychotic polypharmacy (p<.001).

Table 4 reports the adjusted relative risk ratios (RRRs) from the MNL regressions, where risk refers to the propensity of a prescribed antipsychotic to cause metabolic abnormalities. Compared with patients of normal weight or who were underweight according to BMI, patients who were obese or overweight were significantly less likely to receive high-risk antipsychotics than low-risk agents. Obese patients were also less likely to receive medium-risk antipsychotic agents than low-risk agents. Separate analysis using an indicator for obesity defined by physicians' diagnoses or via the checklist also found that obesity was associated with lower likelihood of receiving high-risk antipsychotics versus low-risk agents (adjusted RRR=.49; 95% confidence interval=.25–.96, p=.04).

The primary finding of this study is that in a representative national sample of physicians, obese patients were more likely to receive or continue antipsychotics with lower risk of causing metabolic abnormalities during an office visit. This finding suggests that the patient's weight may be a key consideration when providers make decisions to order antipsychotic medication with lower metabolic risk or to continue an antipsychotic medication. However, having preexisting metabolic conditions, such as diabetes, hyperlipidemia, or hypertension, appeared to have little effect on physicians' choice of antipsychotics with regard to their metabolic risk properties, even after we controlled for patient and physician characteristics, including patients' psychiatric diagnoses. One possible explanation is that these conditions often occur together and with obesity among patients receiving antipsychotics (16–18); thus including these metabolic conditions in the same model with BMI categories or the obesity indicator may mask their effects. However, separate regression analyses with each of these three conditions as a predictor variable without the BMI or obesity indicator variables did not show statistically significant effects for these variables.

Another plausible explanation for lack of association between these metabolic conditions and antipsychotic choice may be that psychiatrists, who accounted for 54% of antipsychotic visits, may be less aware than other providers of patients' metabolic comorbidities. A significantly higher number of antipsychotic visits to psychiatrists were missing BMIs (89% versus 54% in other visits). In addition, metabolic comorbidities were reported less frequently during these visits compared with visits to other providers, even though these conditions co-occur more frequently among patients with schizophrenia than in the general population (16–18). To check this possible explanation, we ran separate regressions stratified by specialties and found that the association between BMI levels and antipsychotic risk category was statistically significant for only primary care physician visits. Regression analysis including only primary care physician visits also found significantly fewer mentions of high-risk antipsychotics compared with low-risk agents for patients with diabetes. However, this result was sensitive to model specification and should be further explored with other databases.

Our findings are contrary to the report by Dassori and colleagues (17), who found that preexisting diabetes was associated with an increase in second-generation antipsychotic switching from an agent with higher metabolic risk to one with lower risk among older veterans with schizophrenia but obesity, hypertension, and dyslipidemia were not (17). That study focused on a patient population (older veterans with schizophrenia), study periods (2002–2003 and 2004–2005), and study design (prescription changes over two time periods) different from ours, which may account for the different findings. However, even in their study, switching to antipsychotics with lower metabolic risk occurred among only 1.5%–6.0% of patients despite the high prevalence of comorbid metabolic conditions in this cohort (15%, 24%, 34%, and 45% of patients with obesity, diabetes, dyslipidemia, and hypertension, respectively), suggesting that switching to antipsychotics with lower metabolic risk in the presence of these comorbidities is rare. Such switching patterns may arise from concerns of clinicians, patients, and sometimes caregivers regarding worsening psychiatric symptoms or tolerability (19). This consideration may be particularly important for patients with schizophrenia because changes in treatment have been found to be associated with increased acute-care service use and higher annual health care costs (20). Nonetheless, among those switching from one second-generation antipsychotic agent to another due to adverse events, weight gain was more often the reason than other adverse events (21).

Metabolic risk categories of antipsychotics used in this study were derived largely from findings on risk of weight gain (6,8), which may partially explain the sensitivity to weight abnormality but not other metabolic risk parameters examined in this study. Weight gain is the most widely studied adverse effect of antipsychotics; findings in the literature regarding the risk of weight gain for various antipsychotic agents were generally consistent (6,7). However, the risk of dyslipidemia and diabetes mellitus has been less studied and is more controversial (6,7). Growing evidence shows that medications for diabetes or hyperlipidemia, such as metformin or statins, were effective in treating and preventing antipsychotic-induced weight gain and other metabolic abnormalities (22–24). If patients with diabetes, dyslipidemia, or both conditions experienced no worsening of their glycemic or lipid control with appropriate treatments, continuing an antipsychotic in the higher-risk category may be clinically justifiable.

Several limitations of this study should be recognized. First, this study was cross-sectional. For continuous users of antipsychotics, we were not able to observe previous clinical response to these treatments or medication tolerance and therefore could not determine whether not switching to antipsychotics with lower metabolic risk was due to positive treatment response to a higher-risk agent. Also, providers may weigh the importance of a patient's preexisting metabolic risk differently depending on whether a new antipsychotic regimen is needed. As a sensitivity analysis, we stratified visits by mentions of any newly prescribed antipsychotics; the metabolic risk categories of antipsychotics mentioned during visits with a change in antipsychotics and visits with only continued use of antipsychotics were not statistically significantly different. Therefore, no further distinction was made in this study. However, only 10% of antipsychotic visits had mentions of any newly prescribed antipsychotics, which may explain the lack of statistical significance. Larger databases are needed to further explore this issue.

Also, the NAMCS is nationally representative of visits rather than patients. It is possible that a sample patient may be counted more than once in the data if he or she visited different physicians (for example, a primary care physician and a psychiatrist) included in the sample. However, the chance of this occurring is low given that physicians were randomly assigned to one of the 52 weeks in a year and reported visits during only that week.

A large number of visits were missing BMI or weight measurements, which was more common for visits to psychiatrists. Providers for antipsychotic visits with available BMI or other weight measures may be more aware and concerned about metabolic risk of antipsychotics. Thus our results may overstate the importance of weight when choosing among antipsychotic medications with different metabolic risk. However, using an indicator for obesity based on physician diagnoses and on reported preexisting conditions or body weight over 200 pounds resulted in findings consistent with the analysis using BMI.

Moreover, our classification on the basis of the associated metabolic risk of antipsychotics did not consider the actual doses and serum concentration of these agents, neither of which is reported in the database. A recent study showed that compared with higher doses of quetiapine for patients with major depression, lower doses may have lower risk of metabolic abnormalities (25), suggesting that our current categorization without considering doses could have misclassified the metabolic risk of some antipsychotics taken at lower doses.

Only up to eight medications per visit are recorded in the NAMCS. Antipsychotic polypharmacy was defined as prescriptions of multiple antipsychotic agents during the same visit. This definition may have misclassified the cross-titration during antipsychotic switching as polypharmacy. As a sensitivity analysis, we reanalyzed our data by assigning polypharmacy visits to the highest-risk category of all antipsychotics mentioned during the visit and found results consistent with our previous results. Long-term antipsychotic polypharmacy (≥60 days) may have higher metabolic risk (26), but that could not be assessed in this data set because of its cross-sectional nature.

At the time of this study, 2007 was the latest year with available data. The years reviewed (2005–2007) may have been too soon after the 2004 ADA/APA recommendations to have observed significant changes in prescribing patterns. Future studies using more recent data are needed to track the changes.

Patients' weight appeared to be the key consideration in providers' decision to order antipsychotics with different metabolic risk or to continue an antipsychotic treatment. Further studies are warranted to understand factors that determine the choice of antipsychotic for patients with preexisting metabolic conditions, such as diabetes, hyperlipidemia, or hypertension.

Dr. Li has received research funds from Novartis Pharmaceuticals Corporation for an unrelated study on gout. The other authors report no competing interests.