There has been a dramatic increase in antipsychotic prescribing since the introduction of second-generation antipsychotics in the 1990s (1). Second-generation antipsychotics were thought initially to represent an improvement over first-generation antipsychotics because of a decreased risk of movement disorders and apparent superior efficacy (2,3). These perceived advantages led to a substantial expansion in both U.S. Food and Drug Administration (FDA) approved and off-label uses. Off-label use now accounts for a majority of second-generation antipsychotic prescriptions (4).
However, many second-generation antipsychotics are now known to be associated with weight gain as well as elevations in glucose and lipids, potentially increasing the risk of hyperlipidemia, diabetes, and cardiovascular disease. Also their superior efficacy and lower risk of neurological side effects compared with first-generation antipsychotics have been called into question by recent large, long-term comparative effectiveness trials (5–8).
The potential cardiometabolic risk associated with certain antipsychotics is an important factor in the decision to prescribe these agents, especially medications with the greatest metabolic risk—clozapine, olanzapine, and, to a lesser extent, quetiapine and risperidone. Olanzapine and clozapine have been associated with dose-related weight gains of almost 9 pounds over ten weeks (9,10) and 13–27 pounds over one year (11,12). Risperidone and quetiapine have been associated with more moderate weight gains of 4 pounds over ten weeks and 4–11 pounds over one year (13,14), and aripiprazole and ziprasidone have been associated with little or no weight gain (5,12). In addition, glucose and lipid metabolism are similarly affected, although this research is less well developed (15).
These findings have spawned several consensus recommendations for the monitoring, treatment, and differential use of these agents (15,16). However, the response by providers to these recommendations has not been studied, and an important question remains, Do clinicians avoid prescribing medications with more metabolic risk in favor of medications that are deemed less risky if there is evidence that a patient already has a metabolic condition? To address this question, we used data from a survey of prescribers administered at the time they initiated treatment with a second-generation antipsychotic to evaluate the relationship of obesity, hypertension, hyperlipidemia, diabetes mellitus, and cardiovascular disease among patients and the level of cardiometabolic risk posed by the specific agent selected. Furthermore, we evaluated the association between antipsychotic choices and the reasons prescribers gave for these selections as evidence for the justification of using high-risk medications.
Data were gathered as part of an intervention at a single Veterans Health Administration (VHA) medical center conducted between October 2007 and May 2009. The facility provided outpatient and inpatient treatment to approximately 10,000 veterans with mental illness annually during this period (17). The intervention consisted of an educational session detailing the comparative efficacy and risks associated with second-generation antipsychotics and a survey of providers regarding decision making. The survey involved all clinicians, including psychiatric residents and fellows, attending psychiatrists, psychiatric advanced practice registered nurses, and psychiatric physician assistants, at the medical center. The prescription of on-patent second-generation antipsychotics was limited to these providers during this time period as a cost control measure.
A copy of the survey was electronically delivered to a provider whenever he or she ordered an on-patent second-generation antipsychotic as a new, nonrefill prescription. Completion was required before the prescription could be electronically sent to the pharmacy. Therefore, 100% of providers writing prescriptions for a second-generation antipsychotic were assumed to have been surveyed about 100% of patients receiving those agents. The institutional review boards of the Veterans Affairs Connecticut Healthcare System and Yale University approved this study.
The survey consisted of 20 questions that documented the agent prescribed, sociodemographic data (age and race), psychiatric diagnoses under treatment, comorbid general medical diagnoses, the professional background of the provider, and health characteristics of the patient, such as weight and height. A list of forced-choice selections and room for write-in answers were given for each question. Providers identified the presence of seven comorbid conditions: tardive dyskinesia, extrapyramidal symptoms, akathisia, hyperlipidemia, obesity, hypertension, and atherosclerotic cardiovascular disease. In addition, providers identified reasons for prescribing the medication from the following list: intolerance to current drug; greater efficacy; lower risk of extrapyramidal symptoms; lower risk of tardive dyskinesia; lower risk of akathisia; less sedation; increased sleep or sedation; treatment of tardive dyskinesia; and patient preference.
Body mass index (BMI) was calculated by dividing the individual's weight in pounds by the square of his or her height in inches multiplied by 703. According to the World Health Organization, BMI ≤18.5 indicates underweight; >18.5 and ≤25, normal weight; >25 and ≤30, overweight; and >30, obesity (18). Individuals who were underweight, normal weight, and overweight were grouped together and compared with the obese group. The presence of hyperlipidemia, hypertension, diabetes, or atherosclerotic cardiovascular disease was determined by positive responses to dichotomous survey questions.
The second-generation antipsychotics for which surveys were completed were ranked according to the risk of weight gain and cardiometabolic disorders, starting with olanzapine, with the highest risk, followed by quetiapine, oral risperidone, long-acting injectable risperidone, aripiprazole, and ziprasidone (12,15). Clozapine was not included because it was available as a generic.
We hypothesized that there would be relatively less use of high-risk agents and increased use of agents with less risk when obesity, hyperlipidemia, diabetes, hypertension, and cardiovascular disease were present. Chi square tests were used to evaluate the statistical significance of these associations. To evaluate the magnitude of these potential differences, we compared the proportions of patients with and without each cardiometabolic disorder who were prescribed each agent. We also analyzed differences in the proportion of agents of varying risk prescribed to patients with each cardiometabolic disorder.
To further investigate the selection of high-risk second-generation antipsychotics in the presence of cardiometabolic disorders, we used chi square tests to evaluate the reasons given for selection of each medication by cardiometabolic disorder.
The analytic sample included 2,643 completed surveys. After excluding surveys (N=30) in which information on the agent prescribed was missing, the final sample included 2,613 surveys from 259 providers who prescribed on-patent second-generation antipsychotics. The patients’ mean±SD age was 54.6±14.6 years, and 1,991 (82%) patients were white, 427 (18%) black, and 12 (1%) classified as other race-ethnicity (Table 1). One-third (33%, N=901) of patients were obese, 306 (12%) were underweight, 521 (20%) were normal weight, and 885 (34%) were overweight. In addition, 35% were reported to have hypertension, 30% hyperlipidemia, 16% diabetes mellitus, and 11% cardiovascular disease. As expected, significant associations were observed between obesity and hyperlipidemia, diabetes, and hypertension, although the trend in the association of obesity and cardiovascular disease did not reach statistical significance (Table 1).
Medications with relatively higher metabolic risk, such as quetiapine (47%), oral risperidone (17%), and olanzapine (16%), represented over three-quarters (79%) of all prescriptions for second-generation antipsychotics (Table 1).
There was a statistically significant association between the agent selected and obesity (p≤.001) (Table 1). Among obese versus nonobese patients, the proportion of patients receiving olanzapine and quetiapine was smaller by 4 and 2 percentage points, respectively. The proportion of patients receiving aripiprazole and ziprasidone was larger by 5 and 2 percentage points, respectively. Among obese patients, 75% received quetiapine (45%), olanzapine (13%), or oral risperidone (17%), but only 18% and 6%, respectively, received aripiprazole and ziprasidone, a pattern that suggested medication selection was unrelated to obesity.
There was also a statistically significant association between the agent prescribed and hyperlipidemia (p<.001) and diabetes (p<.001) (Table 2). The proportions of patients who received olanzapine and quetiapine were only 4% and 5% smaller, respectively, among patients with hyperlipidemia compared with patients without hyperlipidemia. The proportions of patients who received oral risperidone, aripiprazole, and ziprasidone were 5, 3, and 2 percentage points larger, respectively, among patients with hyperlipidemia versus those without it (Table 2). Evaluated differently, 43%, 13%, and 20%, respectively, of patients with hyperlipidemia received quetiapine, olanzapine, and oral risperidone, a total of 76%, but only 17% and 6%, respectively, received aripiprazole and ziprasidone.
The proportions of patients who received olanzapine and ziprasidone were 8 and 1 percentage points, respectively, smaller among patients with diabetes than those without diabetes, and the proportions who received risperidone, aripiprazole and quetiapine were 6, 1, and 1 percentage points, respectively, larger among diabetic patients than among patients without diabetes (Table 2). A total of 78% of patients with diabetes received quetiapine (47%), olanzapine (9%), and oral risperidone (22%), but only 16% and 4%, respectively, received aripiprazole and ziprasidone. Although the association between the agent prescribed and the hyperlipidemia or diabetes was statistically significant, the pattern indicated no systematic response to metabolic risk.
There was no statistically significant association between the agent prescribed and the presence of either hypertension or cardiovascular disease (Table 3). There was a statistically significant association between the agent prescribed and any cardiometabolic disorder (χ2=13.29, df=5, p=.02), but it did not remain significant after correction for multiple comparisons. [The results of these analyses are available online as a data supplement to this article.]
Across the entire sample, the reason given mostly frequently for choosing a medication was efficacy (43% of surveys) followed by patient preference (27%), and sleep or sedation (27%) (Table 1). In the analysis of the specific medication by the reason for its prescription, several significant trends emerged (Table 4). Intolerance of the prior medication (p<.001) and less sedation (p<.001) were more commonly given as reasons for prescription of aripiprazole and ziprasidone compared with other medications. Efficacy was chosen more commonly as a reason for prescription of olanzapine compared with aripiprazole and ziprasidone (63% versus 51% and 45%, p<.001). Increase of sleep or sedation was chosen more commonly as a reason for prescription of quetiapine compared with aripiprazole and ziprasidone (50% versus 5% and 7%, p<.001). Intolerance was the only reason for choosing a medication that had a statistically significant association with all categories of cardiometabolic disease except obesity (Tables 1–3), chosen, on average, 5% more often for individuals with those disorders compared with those without them.
Using data from a survey conducted at the time any second-generation antipsychotic was prescribed, this analysis sought to assess if the provider’s choice of antipsychotic displayed sensitivity to the presence of cardiometabolic disorders. Medications previously shown to increase the risk of these disorders, such as olanzapine and quetiapine, were prescribed to obese patients and those with hyperlipidemia and diabetes in significantly but only slightly lower proportions. Medications with far less risk, such as aripiprazole, were prescribed in only slightly higher proportions. In contrast, the proportion of patients prescribed risperidone, which has a moderate cardiometabolic liability (12,15), was higher among patients with cardiometabolic disorders than among patients without cardiometabolic disorders. More important, among patients with known cardiovascular disease, for many an “end stage” condition, there was no statistically significant difference in the choice of antipsychotic.
In none of these analyses did provider sensitivity to comorbid disorders affect the overall proportion of second-generation antipsychotics prescribed. Agents with higher cardiometabolic risk were prescribed to over 75% of individuals with cardiometabolic disorders. Although prescribers of these agents showed some responsiveness to the metabolic profiles of their patients, the response was small and was not consistent across all cardiometabolic diseases.
The rates of obesity and other cardiometabolic disorders among patients in this sample were not dissimilar from those of the general VHA population. Das and others (19), using height and weight data from a national VHA database, rather than ICD-9 codes (20,21), found that approximately 70% of individuals were at least overweight, and 35% were obese. Other studies of VHA populations with serious mental illness have found rates of hyperlipidemia (30%) (22), diabetes (15%) (21), and hypertension (35%) (23) similar to those found in this study.
The proportion of individuals in this study with cardiovascular disease (11%) was similar to the proportion of a 1999 national VHA sample with ischemic heart disease (23). It is possible that a “cat’s out of the bag” prescribing heuristic is at work, whereby providers do not differentially prescribe second-generation antipsychotics to those who already have cardiovascular disease because they judge that the risk of developing the disease cannot be decreased if it is already present. However, one would also expect this trend to be at work among those with diabetes or hyperlipidemia, especially because these conditions are more easily manageable with pharmacotherapy. The finding that providers considered the cardiometabolic effects of second-generation antipsychotics among patients with antecedent conditions but not among patients with known cardiovascular disease is unexpected and of concern.
Risperidone was used relatively more frequently by overweight and obese individuals and by those with diabetes and hypertension. The drug has a modest weight gain profile, similar to that of quetiapine (24). However, its effects on diabetes risk and lipids are less clear (12,25). Complex and sometimes contradictory reports may make it difficult for prescribers to stay accurately informed about the comparative risk of this drug. Second to clozapine, risperidone is the oldest and most researched second-generation antipsychotic. Providers may be more comfortable with the choice of this drug instead of newer drugs, such as aripiprazole or ziprasidone, that have less cardiometabolic risk but may be less familiar.
Few studies have evaluated prescribing behavior with respect to second-generation antipsychotics and cardiometabolic disorders. A recent VHA study showed that a comorbid diagnosis of diabetes but not hyperlipidemia or obesity was related to switches to lower-risk agents (20). Larsen and others (26) found that a change in antipsychotic was recommended for only 10% of patients with evidence of cardiometabolic disorders. A study of Japanese inpatients with schizophrenia found that significantly more individuals with diabetes were prescribed first-generation antipsychotics compared with second-generation antipsychotics (27). The Japanese study underscores a potential limitation of this study, which did not identify prescriptions for first-generation antipsychotics. However, fewer than 15% of VHA patients with schizophrenia are prescribed these drugs (28).
A study interviewing psychiatrists providing inpatient care found that clinical decisions regarding which antipsychotic to prescribe were subject to complex rules, with each decision involving a mean of eight factors (29). Some of the most common factors cited were specific target symptoms, assumed differences in drug efficacy, drug side effects, and general medical comorbidities. In addition, a nationwide survey of psychiatrists about their prescribing practices found that 85% claimed they would switch drugs, lower current dosing, or avoid prescribing a second-generation antipsychotic to lower the risk of metabolic disorders (30). These studies confirm our finding that providers are aware of cardiometabolic risks when prescribing second-generation antipsychotics. Unfortunately, the data presented here demonstrated only weak evidence that such considerations affect provider prescribing behavior.
The Mount Sinai Conference on Medical Monitoring guidelines (16), the Consensus Development Conference on Antipsychotic Drugs and Obesity and Diabetes guidelines (15), guidelines on the treatment of schizophrenia (31), and multiple other works, including seminal textbooks of psychiatry (32,33), are designed to guide clinicians on the optimal use of antipsychotics and the consideration of cardiometabolic risk. The thrust of these reviews is that either the presence of these disorders at the onset of treatment or the development of these disorders after treatment has begun should be a consideration in the selection of agents. However, these guidelines do not give specific direction about the degree to which these disorders should influence drug selection, nor do they address the use of second-generation antipsychotics specifically by patients with known cardiovascular disease, the leading contributor to mortality of adults with serious mental illness (34). The absence of specific recommendations about these topics mirrors the inconsistent behavior observed in this study.
An extensive body of work supports the selection of second-generation antipsychotics other than clozapine on the basis of side-effect profile and cardiometabolic risk (15,16,31–33). However, there is also evidence, although limited, that some higher-risk agents, namely olanzapine, are more effective in treating refractory psychosis and have better efficacy compared with other second-generation antipsychotics (35–37). Moreover, in the Clinical Antipsychotic Trials of Intervention Effectiveness, patients on olanzapine were less likely to discontinue their medication for any reason and particularly for lack of efficacy or because of patient decision (5). Data from this study showing olanzapine was chosen for efficacy slightly more frequently than other drugs supports this contention. Providers may think that the potential for increased efficacy of drugs like olanzapine outweighs the risk of cardiometabolic side effects among patients who already have a cardiometabolic disorder.
Several methodological limitations deserve comment. This study represented a secondary analysis of a study with a different primary objective than to evaluate the selection of antipsychotics in the context of cardiometabolic disorders. Direct questions about the concern for cardiometabolic risk were not asked. The study also was not able to evaluate the concurrent use of other psychotropic or metabolically active medications, potentially weight reducing therapies, or psychosocial treatments, which may represent a more robust response to cardiometabolic risk. In addition, it is not known with certainty how many medications were attempted by patients prior to the study nor the medication that preceded switches of medication. Moreover, neither the duration nor the dose of prescriptions—which may be important in determining side effect burden—was recorded.
The link between cardiometabolic disorders and mortality is well established, as is the greater mortality of persons with serious mental illness. In light of this information and the recommendations of several task forces and textbooks, one might expect stronger evidence of the differential selection of second-generation antipsychotics among vulnerable patients. Although this study found that providers showed some sensitivity with respect to cardiometabolic disorders, these responses were neither robust nor consistent. This study indicated that on average, the presence of cardiovascular disease has limited influence on providers’ selection of specific agents. Further investigation into the decision-making process of providers with respect to prescribing second-generation antipsychotics to patients with existing cardiometabolic conditions is needed. Peer-driven educational efforts, such as academic detailing aimed at psychopharmacologic decision making, may lower the risk for some patients.
This analysis was supported by the New England Mental Illness, Research, Education and Clinical Center (MIRECC). The authors thank Elina Stefanovics, Ph.D., for help in data management and analysis. The New England MIRECC had no role in the design, analysis, or interpretation of data or in the preparation of the report or decision to publish.
Dr. Sernyak has received honoraria from Pfizer. Dr. Rosenheck has received research support from Janssen Pharmaceutica, Wyeth Pharmaceuticals, AstraZeneca, Bristol-Myers Squibb, Eli Lilly and Company, and Roche Pharmaceuticals. He has testified as an expert in Jones ex rel the State of Texas v Janssen Pharmaceutica. Dr. Hermes reports no competing interests.