An Assessment of the Independent Effects of Olanzapine and Risperidone Exposure on the Risk of Hyperlipidemia in Schizophrenic Patients FREE

IN RECENT decades, several new antipsychotic agents exhibiting improved mechanisms of action compared with conventional antipsychotic agents have been widely adopted in the clinical treatment of schizophrenia. The clearest advantage of the newer antipsychotic agents is the reduction of extrapyramidal adverse effects.¹ However, the new-generation antipsychotic agents are associated with a different spectrum of adverse effects, including weight gain,² alterations in glucose metabolism,^3- 5 and elevations of blood cholesterol and lipid levels.^2,6 Most of these observations are based primarily on case reports^7- 9 and1 small comparative study.⁶ If these associations were confirmed in larger studies, one might conclude that the nutritional and metabolic effects of the new-generation antipsychotic agents are as serious as the extrapyramidal effects seen with conventional antipsychotic agents.¹⁰ For example, lipid elevations have been unequivocally associated with an increased risk for cardiovascular events, such as myocardial infarction and stroke, in general populations.^11- 14 More recently, the US Adult Treatment Panel III guidelines¹⁵ recommend aggressive modification of lipid levels to reduce cardiovascular risk.

Previous studies have suggested that olanzapine use is associated with the development of hyperlipidemia. A report by Sheitman et al⁷ found that olanzapine treatment resulted in marked increases in triglyceride levels for some patients. Another cohort study⁸ of25 olanzapine-treated patients demonstrated an increase in mean fasting triglyceride levels of 37% over baseline at 12 weeks. Melkersson et al¹⁶ found increased rates of hypertriglyceridemia and hypercholesterolemia among olanzapine-treated schizophrenic patients at 5 months. One comparative study by Meyer⁶ demonstrated increases in total cholesterol levels by 24 mg/dL (0.62 mmol/L) and fasting triglyceride levels by 88 mg/dL (0.99 mmol/L) in 47 patients after 1 year of olanzapine treatment. Among a nongeriatric subset, the increases were 31 mg/dL (0.80 mmol/L) and 105 mg/dL (1.19 mmol/L) for cholesterol and triglyceride levels, respectively. Meyer¹⁷ also reported 14 cases of severe hypertriglyceridemia at levels exceeding 600 mg/dL(7.74 mmol/L) in 12 olanzapine- and 2 quetiapine fumarate–treated patients.

In contrast, there is no evidence of an association between risperidone use and hyperlipidemia in the literature. Ghaeli and Dufresne⁹ reported decreases in triglyceride levels after patients switched from clozapine to risperidone. The Meyer study⁶ demonstrated no clinically significant increase in total cholesterol level and only modest increases in fasting triglyceride level (30 mg/dL [0.34 mmol/L], P = .03) compared with baseline for the 47 risperidone-treated patients.

Last, there is no evidence that schizophrenia itself is associated with an increased risk of hyperlipidemia. In fact, low cholesterol levels have been found in some schizophrenic patients.^18- 19 Modai et al¹⁸ showed that patients with schizophrenia had significantly lower cholesterol levels than patients with unipolar depression. Boston et al²⁰ demonstrated that cholesterol levels were significantly lower among treatment-resistant patients vs responsive patients. However, lifestyle factors may increase the risk of lipid abnormalities in patients with schizophrenia.^21- 23

The association between treatment with newer antipsychotic agents, particularly olanzapine, and incident hyperlipidemia warrants a more scientifically comprehensive examination. Given that other factors, such as concomitant medications, disease conditions, and the unhealthy lifestyle of schizophrenic patients, may contribute to the development of hyperlipidemia, it is important to explore these associations in data sources that contain information on these potential confounders. For this study, we used the General Practice Research Database (GPRD) to quantify the risk of hyperlipidemia associated with exposure to various antipsychotic agents, including conventional and atypical agents (olanzapine and risperidone).

The GPRD is a computerized medical database, containing data collected from approximately 400 general medical practices in England and Wales. In1996, the GPRD included more than 6% of the total population.²⁴ Continuous information has been collected from most of the practices for more than 10 years, giving a total experience of more than 30 million patient-years of observation.²⁵ The present study population was defined as individuals with a record of a physician diagnosis of schizophrenia registered with medical practices that submitted data to the GPRD between June 1, 1987, and September 24, 2000. The information for the present study was derived from the patient registration record, the medical record, the prevention record, and the detailed prescription record of the GPRD.

Patient eligibility begins from the time that "up-to-standard" data are provided to the GPRD. Up to standard refers to data recording that reaches an appropriate standard for use in research. However, all previous diagnoses and treatments are recorded in the medical record. Eligibility criteria for this study included a diagnosis of and treatment for schizophrenia recorded at any time in the medical and treatment records. For patients identified as having schizophrenia before up to standard medical records, the start of the study period is defined as the date when the practice began submitting up to standard data for the patient up to the date when the patient transferred out, died, or provided the last data submission to the GPRD. For all other patients, the start of the study period is defined as the earliest date of diagnosis with schizophrenia. To be eligible for the present study, patients must have contributed at least 3 months of up to standard data.

For the cohort analysis, the incidence rates were calculated from the number of patients who developed hyperlipidemia within 3 months of exposure to the drug of interest, divided by the person-time exposure for the drug of interest. The person-time exposure is calculated from the number of prescriptions of the drug of interest divided by 12 (the typical prescription in the GPRD is written for 1 month).

Incident cases of hyperlipidemia were defined as the earliest date of a diagnosis of (by medical code) or treatment for (by drug code) hyperlipidemia, occurring at least 3 months after the beginning of the patient's study period. Issued prescriptions were used as a proxy for drug exposure. The date of hyperlipidemia diagnosis is defined as the index date. A diagnosis of hyperlipidemia was defined as an Oxford Medical Information System diagnostic code or a Read medical code for an increased cholesterol or triglyceride level in the medical record. Treatment of hyperlipidemia was defined as a prescription written for any of the following medications: acipimox, atorvastatin calcium, bezafibrate, cerivastatin sodium, cholestyramine resin, ciprofibrate, colestipol hydrochloride, fenofibrate, fluvastatin sodium, gemfibrozil, inositol niacinate, ispaghula husk, niacin, pravastatin sodium, simvastatin, or Ω-3 marine triglycerides(fish oils). To ensure that the hyperlipidemia cases were incident, the medical and prescription records were checked for any record of a diagnosis of or treatment for hyperlipidemia before the start of the subject study period. If there was a record of hyperlipidemia before the start of the study period, then these patients were excluded. All remaining patients must have contributed at least 3 months of up to standard data between the start of the study period and the index date. Patients identified as cases must not have obtained a prescription for lipid-lowering agents within these 3 months, otherwise they are excluded for being prevalent cases. Prescriptions in England and Wales are usually dispensed with a 1-month supply of medication and, therefore, 3 months is sufficient time to assess treatment before the index date. The use of a 3-month window period as opposed to a longer period avoids the exclusion of hyperlipidemia cases who contributed no more than 3 months of data.

For each patient defined as a case, 6 controls with study periods at least as long as the study period of the case were matched by age at index date (±5 years), sex, and index date. All controls who met the matching criteria were assigned a random number using SAS statistical software (SAS Institute Inc, Cary, NC) procedures. Then, the 6 controls for each case were selected at random from the pool of eligible controls. Controls were selected from patients with a diagnosis and treatment of schizophrenia who did not have a diagnosis or treatment of hyperlipidemia during or before the study period. Control patients were assigned the same index date as the case patients to whom they were matched. Furthermore, the start of the study period of the controls was truncated to match the start of the study period of the cases. Therefore, by definition, the study periods were the same for cases and controls.

Antipsychotic drugs were classified as conventional agents (depot and/or nondepot), olanzapine, risperidone, and other newer drugs, including amisulpride, remoxipride, and sertindole. There were no patients undergoing quetiapine and clozapine therapy because the latter must be initiated during hospitalization.²⁶ There are no other restrictions on the use of atypical antipsychotic agents in England and Wales.

All prescriptions written by the general practitioner for the treatment of schizophrenia between the start of the study period and the index date were abstracted. The dates of all antipsychotic prescriptions within the study period were compared with the index date. Exposure was defined as the receipt of at least 1 prescription for an antipsychotic agent within the 3 months before the index date. Nonexposed subjects are those who did not have a prescription for an antipsychotic agent within the 3 months before the index date.

All analyses were conducted using SAS statistical software for personal computers, version 7.0 (SAS Institute Inc). The primary study design was a case-control analysis nested within a cohort of schizophrenic patients. To account for the matched study design, the effect of exposure to the different antipsychotic agents on the odds of hyperlipidemia development was modeled using conditional logistic regression.²⁷ Unadjusted incidence rates for hyperlipidemia were also calculated from the cohort experience.

Two reference groups were used to estimate the odds of hyperlipidemia development among those exposed to olanzapine, risperidone, and conventional antipsychotic agents. The first referent group included all individuals prescribed no antipsychotic medications within the 3 months before the index date. The second referent group included individuals prescribed conventional antipsychotic medications within the 3 months before the index date. In both comparisons, individuals were divided into mutually exclusive categories of antipsychotic exposure. In addition, patients exposed to 2 or more antipsychotic agents from 2 or more different classes within the 3-month period before the index date were grouped into a combination antipsychotic exposure category.

In addition to the matching variables, the analyses were adjusted for prescription of other drugs known to affect the risk of hyperlipidemia, including β-adrenergic blockers, thiazide diuretics, combination β-blockers and thiazide diuretics, corticosteroids, female sex hormones (estrogens and progestins), and valproate sodium.²⁸ Disorders associated with secondary hyperlipidemia, including hypothyroidism and diabetes mellitus, were also adjusted for in the analysis.²⁸

Between June 1, 1987, and September 24, 2000, 20 865 subjects were diagnosed as having schizophrenia and received treatment for schizophrenia. Of these subjects, 1429 (7%) contributed less than 3 months of follow-up and were excluded. Another 1127 (5%) who were diagnosed as having or were treated for hyperlipidemia before the start of their study period were also excluded, leaving 18 309 subjects (88%) available for study. The study population was almost equally divided between men and women, with a mean ± SD age of 51 ± 20 years (Table 1). Of the 18 309 subjects in the cohort, 85% received at least 1 prescription for nondepot conventional antipsychotic medications during the follow-up period;22%, for depot conventional antipsychotic medications; 5%, for olanzapine;8%, for risperidone; and 3%, for other newer antipsychotic agents.

During a mean ± SD follow-up of 4.07 ± 2.80 years, a total of 1269 hyperlipidemia cases (7%) were identified; 205 case identifications were based on a diagnosis of hyperlipidemia only, 1171 on treatment for hyperlipidemia, and 107 on a diagnosis and on treatment for hyperlipidemia. Among all antipsychotic-treated schizophrenic patients, the incidence rate of hyperlipidemia was 17.04 per1000 person-years. Women had a higher incidence rate (21.6 per 1000 person-years;95% confidence interval [CI], 20.14-23.10) of hyperlipidemia compared with men (12.3 per 1000 person-years; 95% CI, 11.21-13.49). The incidence rate of hyperlipidemia within 3 months of a prescription was 26.6 per 1000 person-years for olanzapine (95% CI, 17.15-41.19), 11.5 per 1000 person-years for risperidone(95% CI, 7.41-17.81), and 18.5 per 1000 person-years for conventional antipsychotic medications (95% CI, 17.30-19.82). Olanzapine-exposed subjects exhibited a40% increase in the risk of hyperlipidemia compared with those exposed to conventional antipsychotic medications (risk ratio, 1.4; 95% CI, 0.92-2.22). The increase in the risk of hyperlipidemia in subjects prescribed olanzapine compared with those prescribed conventional antipsychotic medications was8.1 per 1000 person-years.

A total of 1268 hyperlipidemia cases were matched to 7598 controls. One case could not be matched and was excluded. Table 1 details the demographic characteristics for the matched cases and controls. The mean ± SD age for the cases and controls was56 ± 16 years.

The prevalences of conventional antipsychotic prescriptions and risperidone were similar in the cases and controls, whereas olanzapine prescribing was2.1 times higher among the cases compared with the controls. Table 2 summarizes the odds ratios and 95% CIs for the development of hyperlipidemia in individuals exposed to the different antipsychotic agents compared with patients not receiving antipsychotic agents. Compared with no antipsychotic exposure, olanzapine significantly increased the odds of hyperlipidemia, followed by conventional antipsychotic medication exposure, but not risperidone exposure. As shown in Table 3, exposure to olanzapine was associated with a significant increase in the odds of hyperlipidemia compared with conventional antipsychotic medications, whereas risperidone was again not associated with an increased odds of hyperlipidemia compared with conventional agents. Figure1 summarizes the odds ratios for olanzapine and risperidone exposure relative to the 2 comparison groups.

Our analyses demonstrate an increased risk of hyperlipidemia among olanzapine-treated patients, adjusted for demographic risk factors and concomitant medications and disorders. Subjects prescribed olanzapine exhibited more than a 3-fold increase in the odds of developing hyperlipidemia compared with subjects prescribed conventional antipsychotic medications, and olanzapine exposure was associated with a nearly 5-fold increase in the odds of developing hyperlipidemia compared with schizophrenic patients not prescribed antipsychotic agents. Conventional antipsychotic exposure was also associated with an increased odds of developing hyperlipidemia, although to a lesser extent than olanzapine. In contrast, exposure to risperidone was not associated with an increased odds of hyperlipidemia in any of the models. These results corroborate the association of olanzapine with elevated lipid levels seen in several case studies^7- 8,16 and1 comparative study.⁶ However, to our knowledge, the magnitude of effect seen in this study has not been observed previously. Similarly, the observation of no association between risperidone use and hyperlipidemia is consistent with other reports^6,9 from the literature.

The clinical relevance of cholesterol and triglyceride elevations has been repeatedly demonstrated for general populations,^11- 14 and therapies focus on aggressive interventions to modify lipid levels.^15,29 The British, the European, and, more recently, the US Adult Treatment Panel III guidelines have quantified the risk for cardiovascular events associated with an increased cholesterol level; these guidelines recommend lifestyle and pharmacological therapies to reduce risk. Hypertriglyceridemia was also shown to be a strong risk factor for ischemic heart disease independent of other major risk factors, including high-density lipoprotein cholesterol level.³⁰ In addition, the spectrum of cardiovascular risk factors defined as the metabolic syndrome or syndrome X has also received considerable attention among clinicians in recent years.³¹ The clinical relevance of the syndrome is related to the combination of morbid factors that results in diabetes mellitus and subsequent cardiovascular disease. Furthermore, many of the risk factors associated with the metabolic syndrome, such as obesity, impaired glucose homeostasis, and dyslipidemia, including hypertriglyceridemia, may be exacerbated by exposure to olanzapine.

The increased risk of hyperlipidemia among subjects prescribed olanzapine in the present analysis raises serious questions about the risk-benefit ratio for olanzapine use among schizophrenic patients. This is particularly true because schizophrenic patients also tend to exhibit other major risk factors for cardiovascular events, such as smoking,²³ a sedentary lifestyle,³² diabetes mellitus,^33- 36 and weight gain.²

There are several limitations to the present study. Antipsychotic drug exposure was inferred from automated prescription data, and it is not possible to determine whether patients actually obtained or consumed the prescribed medication. Patient-specific data are limited to those recorded in the automated database. However, one study³⁷ showed that95% of all known GPRD prescriptions and 74% of all consultations were recorded in the computer records compared with 42% and 75% in written records, respectively. There is no direct information on the severity of schizophrenia, race, social class, or weight gain. We are, thus, unable to adjust for the potential confounding effect of these variables. It was not possible to study the association between clozapine and hyperlipidemia, because clozapine therapy must be initiated during hospitalization²⁶; this information is not available.

Surveillance bias is a possibility if olanzapine-treated patients were more likely to have lipid tests ordered than patients receiving other newer or conventional antipsychotic agents. If this were so, then the proportion of patients undergoing laboratory tests would be higher in olanzapine-treated patients vs others. We examined this issue by measuring the proportion of patients undergoing lipid tests within 3 months before the index date using the GPRD prevention and medical records. We found no differences in the percentage of patients treated with olanzapine, risperidone, and conventional agents who had a plasma lipid or cholesterol test ordered. Furthermore, we calculated the number of patients for whom a test for cholesterol or lipid levels was ordered within 6 months of their first prescription for antipsychotic agents. Again, there were no differences in the proportion of tests ordered according to specific agent. These data indicate that physicians were not more likely to monitor the lipid levels of patients taking the newer antipsychotic agents.

The diagnosis of hyperlipidemia included a diagnosis of increased cholesterol or triglyceride levels. Because of lack of laboratory data on high- or low-density lipoprotein cholesterol, total cholesterol, and triglyceride levels, it was not possible in this study to determine which specific lipoprotein fraction is elevated.

A recent study by Primatesta and Poulter³⁸ found that at least 25% of English adults have adverse lipid profiles; of these adults, only 2% use a lipid-lowering agent. Among those treated, only 10% are effectively controlled to target low-density lipoprotein cholesterol levels.³⁸ The prevalence of obesity among schizophrenic patients suggests that they are at considerable risk for hypercholesterolemia.³¹ Brown et al²³ found that the schizophrenic patient's diet is typically higher in fat and lower in fiber compared with the general population. In addition, compliance with lipid-lowering therapy may prove more problematic in these patients. Therefore, the effect of antipsychotic drug therapy on the odds of hyperlipidemia observed in this study may be underestimated.

It can be concluded from this study that exposure to olanzapine is associated with clinically important increased odds of hyperlipidemia, after adjusting for other factors, in schizophrenic patients. The potential cardiovascular consequences of olanzapine therapy, and its association with the metabolic syndrome, warrant serious consideration of its risk-benefit ratio by treating physicians.

Submitted for publication January 22, 2002; final revision received March 19, 2002; accepted March 19, 2002.

Corresponding author: Carol E. Koro, PhD, School of Pharmacy, University of Maryland, 506 W Fayette St, Suite 201, Baltimore, MD 21201 (e-mail: ckoro001@umaryland.edu)