Atypical Antipsychotic Drugs and Diabetes Mellitus in the US Food and Drug Administration Adverse Event Database: A Systematic Bayesian Signal Detection Analysis

Ross A. Baker; Andrei Pikalov; Quynh-Van Tran; Tatyana Kremenets; Ramin B. Arani; P. Murali Doraiswamy


Psychopharmacol Bull. 2009;42(1):1-21. 

In This Article


Pharmacovigilance studies provided an early signal, and subsequent epidemiological or clinical studies have mostly suggested a higher incidence of DRAEs and/or increased diabetes risk with clozapine and olanzapine, and, to a lesser extent, risperidone and quetiapine. The results from this study suggest that ziprasidone and aripiprazole are associated with the lowest risk of DRAEs among atypicals, although these products were introduced relatively more recently than the other atypical antipsychotics, which may have resulted in less data available for analysis. However, in agreement with the results from this study, clinical evidence to date has shown a lower risk of metabolic AEs with these agents than with older atypicals.[14,22] Of particular interest is the emerging evidence that both agents can reverse metabolic AEs related to treatment with other antipsychotics.[25,26,24] Furthermore, cumulative AE reporting showed that diabetes associations for clozap ine and olanzapine were evident within 2 years of introduction, where as the time from introduction to association is longer with the other agents. This might also explain why an earlier pharmacoepidemiologi- cal study failed to show a significantly greater risk of diabetes with olan zapine than with haloperidol.[37] Interestingly, the reports of diabetes increased sharply around 2003-2004, following warnings issued by Japan and the UK and influential case series published by Drs. Szarfman and Doraiswamy;[19,11] subsequently, the FDA warning language was introduced to the label of antipsychotics and the American Diabetes Association (ADA) published their consensus on metabolic monitoring in psychiatry. This is of relevance because aware ness of these important AEs is crucial to encourage physicians to mon itor and manage them appropriately. The impact of more recent events, such as changes to the label of olanzapine[16] and publicity around litigation, are yet to be determined.

This analysis of the extensive FDA AERS database revealed that all seven antipsychotics were associated with at least one DRAE (as demonstrated by EB05 values ≥2). However, there were noticeable differences in the number of DRAEs reported for each agent, and the degree of association observed. Overall, olanzapine demonstrated the highest frequencies of DRAEs (13 DRAEs) and five of these associated drug-event combinations were well above the threshold (EB05 values >10) of a potential signal (>2) defined in this study. Clozapine, queti-apine and risperidone were also more frequently associated with DRAEs, whereas drug-event associations with DRAEs were generally low for aripiprazole, haloperidol and ziprasidone. Adjusted reporting ratios and odds ratios for diabetes mellitus support this grouping and suggest three clusters of drugs associations: a) olanzapine, b) clozapine, quetiapine and risperidone, and c) aripiprazole, haloperidol and ziprasidone. Although the mechanism of action of antipsychotic-related DRAEs requires further elucidation, antipsychotic-induced weight gain with consequent insulin resistance has been implicated as a potential mechanism. Thus, the high frequency of DRAEs with clozapine and olanzapine observed in this study may reflect the increased risk of weight gain associated with both agents.[12] Conversely, lower drug-event associations between aripiprazole or ziprasidone and DRAEs may correlate with their lower weight-gain risk profiles.[38,39] There may also be other mechanisms at play and, as such, AERS analyses cannot determine causal mechanisms.

Of the 24 DRAEs investigated in this study, diabetes mellitus was the most frequently reported DRAE, with olanzapine, clozapine, quetiap-ine, risperidone and ziprasidone showing drug-event associations. Comparison of these findings to the entire AERS database shows that olanzapine and clozapine have some of the highest number of reports of diabetes among all 4,000 plus drugs, giving these findings some context. Hyperglycemia was the second most frequently reported DRAE; however, only olanzapine showed a drug-event association.

The present findings are broadly consistent with recent epidemiological[22] and pharmacovigilance[19,20,11,21] studies, as well as monitoring guidance[12] and an earlier analysis of data from the AERS database.[36] The present study extends this analysis to include more information on the recently introduced atypicals aripiprazole and ziprasidone, and allows for the impact of 2004 updated warning language. For example, in the present analysis, a total of 2,784 cases of diabetes mellitus were reported, versus 955 in the earlier analysis.[36]

Despite the smaller relative number of DRAEs reported in juvenile patients (<17 years) and older patients (>65 years), drug-event associations were still evident in these populations. Olanzapine showed drug-event associations with three DRAEs in both juvenile and older patients (diabetic coma; diabetes mellitus; and non-insulin-dependent diabetes mellitus). This is an important finding given that there is increasing use of atypical antipsychotics in juvenile patients[40,41,42,43,44] and evidence suggests that this population is less likely to be monitored for metabolic AEs.[9] Ultimately, clinicians need to be aware of the risk of diabetes in juveniles receiving atypical antipsychotic agents and take precautions, including using these agents judiciously and routinely monitoring patients for metabolic AEs in line with current guidelines.[12]

The notable advantage of safety databases such as the AERS is their ability to provide warning signals of drug-event combinations, which may not necessarily become apparent until the post-marketing stage when therapeutic agents become available to a larger population than those participating in clinical studies. Analyses of these databases are strengthened with the use of systematic disproportionality methods (for example, the MGPS algorithm and logistic regression of the odds ratio), which are able to provide adjusted estimates of association between drugs and AEs in the absence of large, long-term, controlled, comparative trials of multiple agents. However, there are limitations in the quality and/or scope of these databases as a result of uncertainties in establishing causality between the suspected drug and the AE, incomplete drug and medical histories, and inconsistencies in the use of diagnostic terms between reporters. Reporting and detection biases can also arise based on differences in dates of drug launch, prescription volume and publicity. However, our analyses of the data by year suggests that diabetes signals were evident with olanzapine in the first 2 years of its launch. Furthermore, prescription volume for olanzapine is lower than that for risperidone (and also quetiapine in recent years). Differences in the rates of total adverse events between individual antipsychotic drugs can dilute or amplify the relative reporting of diabetes signals. For example, clozapine can be used only within the context of a registry -its overall higher rate of all adverse events probably dilutes individual diabetes signals (since the denominator is larger) and result in an underestimate of diabetes reporting ratios. This should be borne in mind when comparing drugs with overall higher rates of events to drugs with lower rates of events. Haloperidol was launched prior to the introduction of modern safety reporting practices and hence is underestimated. Thus, our results must be interpreted in this context. Reporting patterns of metabolic as well as other types of adverse effects of drugs can change over time due to various factors and hence periodic updates of such data may prove useful to inform clinicians.