Lipscombe LL, Lévesque L, Gruneir A, Fischer HD, Juurlink DN, Gill SS, Herrmann N, Hux JE, Anderson GM, Rochon PA. Antipsychotic Drugs and Hyperglycemia in Older Patients With Diabetes. Arch Intern Med. 2009;169(14):1282-1289. doi:10.1001/archinternmed.2009.207
Evidence suggests that there is an association between antipsychotic drugs and new-onset diabetes, but little is known about the risk of hyperglycemia among persons with preexisting diabetes.
Using a nested case-control design and population-based health databases in Ontario, Canada, persons aged 66 years or older with diabetes who started treatment with an antipsychotic drug from April 1, 2002, to March 31, 2006, were followed up from treatment start until March 31, 2007. The cohort was subdivided into 3 groups: insulin-treated, oral hypoglycemic agent only–treated, and no diabetes treatment. We defined cases as patients hospitalized (emergency department visit or hospital admissions) for hyperglycemia. Each case was matched with up to 10 controls. We compared the likelihood of hyperglycemia among current users of atypical and typical antipsychotic agents with that among remote antipsychotic users (discontinued >180 days), based on prescriptions closest to event date.
Of 13 817 patients studied, 1515 (11.0%) were hospitalized for hyperglycemia. Current antipsychotic treatment was associated with a higher risk of hyperglycemia compared with remote antipsychotic use in all diabetes treatment groups (overall adjusted rate ratio, 1.50; 95% confidence interval, 1.29-1.74). The risk was increased among patients who were treated with atypical and typical antipsychotic agents and was extremely high among patients who were just starting treatment (only 1 prescription before event).
Among older patients with diabetes, the initiation of treatment with antipsychotic drugs was associated with a significantly increased risk of hospitalization for hyperglycemia (P < .001). The risk was particularly high during the initial course of treatment and was increased with the use of all antipsychotic agents.
Antipsychotic drug use among seniors has increased dramatically over the last decade to treat a variety of conditions, including behavioral symptoms of dementia.1,2 Recent warnings have been issued regarding the use of these drugs in older persons owing to a number of adverse effects, including parkinsonism, stroke, and diabetes.3- 6 The risk of diabetes may be partly related to chronic effects of the weight gain associated with antipsychotic agents.7 However, case reports of acute hyperglycemia after the initiation of therapy with these drugs8- 10 suggest that they may also be associated with acute glycemic changes. This effect may be of greater concern for patients with diabetes, who may be more vulnerable to hyperglycemia with antipsychotic treatment, particularly in the setting of psychotic symptoms. However, the risk of hyperglycemia during antipsychotic drug treatment in patients with diabetes is not well characterized. This risk may be especially important among persons aged 65 years or older, who represent more than 40% of patients with diabetes.11
Most studies exploring the association between hyperglycemia and antipsychotic therapy have focused on younger patients with schizophrenia.12- 14 The few studies that have looked at this risk in older adults have not shown a consistent association.15- 19 However, those studies primarily assessed the risk of new-onset diabetes with the use of these drugs and did not consider the potential risk of metabolic decompensation in patients with preexisting diabetes. Furthermore, most existing research has focused on atypical antipsychotic agents or olanzapine only,12- 14 although hyperglycemia has been described with the use of typical antipsychotic agents as well.20,21
We sought to quantify the association between the initiation of therapy with antipsychotic drugs and hospitalization for hyperglycemia among older patients with diabetes. We examined this association for both typical and atypical agents.
We used linked population-based health databases from Ontario, Canada, which have been described in detail elsewhere.6 We identified prescription records using the Ontario Drug Benefit database, which contains data on publicly funded medications dispensed to Ontarians aged 65 years and older.22 Hospitalizations were identified using the National Ambulatory Care Reporting System database and the Canadian Institute for Health Information Discharge Abstract Database, which provide information regarding emergency department visits and hospital admissions, respectively. The Ontario Health Insurance Plan database was used to identify physician service claims. We obtained demographic information and deaths from the Registered Persons Database, and diabetes status from the Ontario Diabetes Database, a validated registry of Ontarians with diagnosed diabetes.23 These databases were linked anonymously using encrypted health card numbers.
We conducted a case-control study nested within a cohort of older patients with diabetes who were newly treated with antipsychotic agents. The study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre in Toronto. The cohort consisted of all individuals 66 years and older with preexisting diabetes, defined as persons with diagnosed diabetes before cohort entry date, who began therapy with an antipsychotic drug between April 1, 2002, and March 31, 2006. Patients with diabetes were identified from the validated Ontario Diabetes Database, in which diabetes is defined as 1 hospitalization record with a diagnosis of diabetes or 2 physician service claims for diabetes within 2 years.23 Cohort entry was defined as the date of the first prescription for any antipsychotic drug. Individuals were not eligible for inclusion in the cohort if they had received a previous antipsychotic prescription within the preceding year. We excluded patients who were younger than 65 years to avoid incomplete medication records, as well as those receiving dialysis or palliative care and those residing in long-term care facilities. Because of the potential modifying effects of diabetes severity on the risk of hyperglycemia, we stratified patients by diabetes treatment based on prescription records in the 180 days before cohort entry. Patients with at least 1 prescription for insulin were categorized as insulin treated; patients with at least 1 prescription for an oral hypoglycemic agent but no insulin prescription were categorized as oral hypoglycemic only treated; and the no-treatment group comprised patients with no diabetes drug prescription during that time. Once in the cohort, individuals remained there until the first occurrence of the outcome of interest (defined below), death, or the end of the study period (March 31, 2007).
We identified cases as cohort members who were hospitalized with hyperglycemia during the observation period. Hyperglycemia was defined as an emergency department visit or a hospital admission with a discharge diagnosis of hyperglycemia, diabetic ketoacidosis, or hyperosmolar nonketotic coma or with diabetes as the most responsible discharge diagnosis or emergency department main diagnosis (International Statistical Classification of Diseases, 10th Revision, codes: R739, E1010, E1012, E1110, E1112, E1310, E1312, E1410, E1412, E1100, E1101, E1300, E1301, E1400, E1401, E11, E13, and E14). The date of hospitalization served as the index date.
For each case, we selected up to 10 potential control patients from the cohort who were still at risk for an event on the index date. Controls were matched with cases on the basis of age (±1 year), sex, year of cohort entry, and duration of follow-up (±30 days). Controls were assigned the same index date as that of their respective case. Patients could serve as a control more than once and were eligible to become a case at a later time.
Exposure status was defined according to both antipsychotic type and timing of use relative to the index (event) date. All antipsychotic drugs funded by the Ontario Drug Benefit program during the study period were identified and classified as either (1) atypical agents (olanzapine, quetiapine fumerate, and risperidone) or (2) typical agents. We did not include clozapine because it is not funded by the provincial program.
Timing of use relative to the index date was classified as current use, recent past use, or remote use. These categories were defined using the date of the most recent prescription before the index date and the number of days supplied to estimate treatment duration. Treatment duration was calculated as the number of days supplied plus an additional 14 days to allow for nonadherence. An individual was classified (1) as a current user if treatment duration overlapped with the index date, (2) as a recent past user if treatment was discontinued 15 to 180 days before the index date, or (3) as a remote user if treatment was discontinued more than 180 days before the index date.
We used conditional logistic regression to estimate associations between the current use of each antipsychotic agent class with the occurrence of hyperglycemia, while controlling for potential confounders, including age, sex, neighborhood income quintile, diabetes duration, Charlson comorbidity index,24 comorbid conditions, other prescription drug use before cohort entry, and health service use history (Table 1). Baseline variables were determined based on records at or before cohort entry date: a 180-day look-back window was used for drug prescriptions; a 1-year look-back window was used for health care use and number of unique prescriptions; a 2-year window was used for hyperglycemia hospitalizations, liver disease, renal disease, and Charlson comorbidity scores; and a 5-year window was used to capture records of schizophrenia, other psychoses, and dementia. Associations are expressed as adjusted rate ratios (aRRs) with 95% confidence intervals (CIs).25 Remote users were the reference group. This approach has been shown to provide unbiased estimates of the rate ratios that would be obtained from a traditional time-to-event analysis of the full cohort, with little or no loss of precision but with substantial gains in computational efficiency.26 Analyses were conducted for the entire cohort and stratified by diabetes treatment at cohort entry.
To isolate a potential acute effect of these drugs on hyperglycemia, we also examined the risks of hyperglycemia separately among patients who were first-time current antipsychotic users (only 1 antipsychotic prescription before index date) vs prevalent users (more than 1 prescription before index date from the same antipsychotic category). Also, a third separate exposure group was created to account for “switchers” from one antipsychotic category to another, and this category was included in all the models. Switchers were defined as current users who had their 2 most recent prescriptions from different antipsychotic categories before the index date.
We conducted 3 secondary analyses to assess the robustness of our findings. First, we tested the specificity of our findings by examining prescriptions for antibiotic eye drops, which we expected to have no association with either hyperglycemia or antipsychotic drug therapy. Second, to assess the reliability of our outcome definition, we examined the association between hyperglycemia and the use of oral corticosteroid drugs, which are known to be associated with hyperglycemia.27 Analogous to our primary analysis of antipsychotic use, for each of these drugs we used conditional logistic regression to compare current exposure to the drug of interest between the cases of hyperglycemia and their matched controls. Third, nonadherence to the use of diabetes medication may be more common among patients requiring an antipsychotic agent, which might influence an association between new antipsychotic use and hyperglycemia. To partly address this issue, we performed subgroup analyses between first-time antipsychotic users who did and did not have a current diabetes drug prescription at the time of their hospitalization but who were originally treated with insulin and/or an oral hypoglycemic agent at cohort entry. If disruption of diabetes treatment is a major contributor to this association, then we would expect to find an increase in hyperglycemia risk predominantly among antipsychotic-treated patients without current diabetes prescriptions. In contrast, while we cannot be certain of drug adherence among persons with recently dispensed prescriptions, a similar increase in that group would argue against this type of confounding. All analyses were done using SAS version 9.1 (SAS Institute Inc, Cary, North Carolina), and a 2-tailed P value of .05 was used as the threshold for statistical significance.
We estimated the absolute risk difference of hyperglycemia and the number needed to harm associated with antipsychotic agents based on the baseline event rate for each diabetes treatment group and their aRR estimates for antipsychotic treatment. Although these aRRs were calculated based on a comparison with remote users of antipsychotic agents, the event rate of remote users was likely the same or lower than that of the entire cohort.
The study cohort consisted of 13 817 individuals with diabetes (mean age, 78 years) who were newly treated with an antipsychotic drug and followed up for an average of 2 years. Among these, 5549 (40.2%) received no medications for diabetes, 6284 (45.5%) received 1 or more oral hypoglycemic agents, and 1984 (14.4%) had treatment that included insulin at cohort entry. During follow-up, 1515 patients (11.0%) were hospitalized for hyperglycemia and 5288 (38.3%) died. There were 479 cases (24.1%) of hyperglycemia among insulin-treated patients, 823 (13.1%) among patients taking oral hypoglycemic agents only, and 213 (3.8%) among patients not taking diabetes medications. There were 11 cases for whom no matching controls were found (9 insulin-treated patients and 2 oral hypoglycemic agent–treated patients). The characteristics of the remaining cases and their matched controls are described in Table 1.
Compared with no use of antipsychotic drugs in the preceding 180 days, current antipsychotic use was associated with a significantly higher likelihood of hyperglycemia among all diabetes treatment groups (insulin-treated: aRR, 1.40; 95% CI, 1.06-1.84 [P = .02]; oral hypoglycemic agent-treated: aRR, 1.36; 95% CI, 1.12-1.66 [P = .002]; and no diabetes treatment: aRR, 2.43; 95% CI, 1.61-3.66 [P < .001]) (Table 2). This finding was similar when we looked at the complete cohort without stratification by diabetes treatment (aRR, 1.50; 95% CI, 1.29-1.74 [P < .001]). The risk of hyperglycemia was increased among patients who used atypical or typical antipsychotic agents. The risk of hyperglycemia was not increased among recent past users compared with remote users, suggesting that it is unlikely that treatment with these agents has a residual effect on hyperglycemia risk once the treatment is discontinued.
When we examined cases of hyperglycemia separately among patients newly treated with antipsychotic drugs, defined as only 1 prescription before the index date, first-time recipients in all 3 groups had a markedly higher likelihood of hospitalization than untreated patients, with aRR estimates ranging from 8 to 15 times higher than those of untreated patients (Table 2). This big effect was seen with both atypical and typical agents, although the number of first-time typical antipsychotic users who were not taking diabetes medication was too small to permit meaningful statistical inference. Among first-time antipsychotic users, nearly 70% of events occurred within 14 days of treatment initiation, consistent with an immediate metabolic effect. Ongoing (prevalent) use of any antipsychotic drug was associated with a small but significantly increased risk of hyperglycemia in all diabetes treatment groups (insulin-treated: aRR, 1.36; 95% CI, 1.03-1.79 [P = .02]; oral hypoglycemic agent-treated: aRR, 1.31; 95% CI, 1.08-1.60 [P = .007]; and untreated: aRR, 2.23; 95% CI, 1.48-3.37 [P < .001]). However, in contrast to first-time users, a significant association was found only for atypical antipsychotic drugs in prevalent users (Table 2). There were only 29 patients overall who switched from one category of antipsychotic agent to another before the index date (23 typical to atypical and 6 atypical to typical); therefore, the numbers were too small for analysis. Estimates for this group are thus not shown in Table 2.
As expected, we found no association between antibiotic eye drops and hospitalization for hyperglycemia (aRR, 0.67; 95% CI, 0.32-1.41), but we found a significant association between corticosteroid use and hyperglycemia (aRR, 2.13; 95% CI, 1.67-2.71 [P < .001]). When we analyzed first-time antipsychotic users who did have a current diabetes drug prescription at the index date vs those who did not, we found similar significant associations between first-time antipsychotic use and hyperglycemia (P < .001). These findings suggest that the increase in the risk of hyperglycemia among new antipsychotic users was not primarily related to nonadherence to diabetes treatment during that time.
The annual estimated number needed to harm was 21 (95% CI, 10-138) for insulin-treated patients, 42 (95% CI, 23-127) for oral agent–treated patients, and 37 (95% CI, 20-86) for untreated patients (Table 3).
Our study indicates that the initiation of antipsychotic therapy represents a critical period during which seniors with diabetes are particularly vulnerable to metabolic decompensation. The new use of both atypical and typical antipsychotic drugs was associated with a significant increase in hospitalizations for hyperglycemia, which appeared independent of baseline diabetes treatment and was strikingly high during the initial period of antipsychotic therapy. These findings highlight the importance of enhanced glycemic monitoring when antipsychotic therapy is initiated in patients with diabetes.
While a link between antipsychotic treatment and hyperglycemia has been documented,12- 14 the majority of previous studies have focused on younger patients with schizophrenia, who may be at higher baseline risk of diabetes.28 This association has made it difficult to disentangle the causal effects of the use of antipsychotic agents from the underlying effects of schizophrenia.29 Previous studies among older persons have not focused specifically on those with diabetes and may have had insufficient power to demonstrate an association between antipsychotic treatment and hyperglycemia.16- 18 Moreover, although case reports of severe hyperglycemia and ketoacidosis after the initiation antipsychotic therapy have suggested an acute effect,8- 10 most prior studies were not designed to quantify the short-term risks of these drugs. By restricting our analysis to new users and examining first-time users separately, our study more accurately represents events that happen shortly after treatment initiation. Such events are often underestimated by including long-term users of a drug (ie, survivors), who may have a lower susceptibility to an outcome.30 Our population-based study advances the argument that new use of antipsychotic drugs among older persons represents a situation that can be associated with a considerable increase in the risk of hyperglycemia, at least among patients with preexisting diabetes.
Our study found that the risk of hyperglycemia was increased with the use of all antipsychotic agents. Previous research among patients with schizophrenia has yielded inconsistent results, suggesting greater metabolic risks with the use of olanzapine13 or atypical agents14 compared with traditional antipsychotic drugs. There has been recent interest in the association between therapy with atypical agents and diabetes, in part owing to greater weight gain with olanzapine therapy.7 However, the first reports of an association between diabetes and antipsychotic therapy were noted more than 40 years ago, when typical agents were associated with the development of diabetes.20,21 Indeed, use of the term phenothiazine diabetes is an indication of the historical recognition of this association.31 In our study, first-time use of both atypical and typical antipsychotic drugs was associated with a large early increase in hyperglycemia events. In contrast, although the number of patients taking typical agents may have been too small to draw conclusions, only long-term atypical antipsychotic use was associated with an ongoing substantial increase in the risk of hyperglycemia. These findings support the possibility of differing short- and long-term effects of these drugs. There is some evidence of a role for dopamine in the central control of glycemia based on studies showing that the use of central dopamine agonists can improve glucose control.32,33 Therefore, our finding in first-time users of all antipsychotic agents may reflect an acute disruption of central dopamine-mediated control of glucose homeostasis or some other mechanism or prescribing situation common to all antipsychotic drugs. In contrast, it is possible that the small but significant association found primarily with the use of chronic atypical agents may stem in part from the associated weight gain uniquely observed with atypical antipsychotic agents,7 leading to more long-term deterioration in glycemic control. Further studies are needed to explore the mechanisms underlying these findings.
Despite adjustment for a number of potential confounders, the main limitation of our study is the possibility of confounding by factors associated with both antipsychotic use and hyperglycemia. For example, diabetic patients requiring new antipsychotic treatment might have other underlying risk factors for hyperglycemia, such as acute psychiatric stress, nonadherence to diabetes medication, or concurrent medical illness. We attempted to address these concerns in several ways. First, we confined our cohort to new antipsychotic users and compared events between current and previous users to ensure that all study patients had an indication for antipsychotic treatment at some point. Second, we limited our outcome to hospital visits for which the primary reason was hyperglycemia or diabetes rather than other medical illnesses. Third, to address a possible effect of nonadherence to diabetes therapy, we compared events between first-time users who did and did not have a current diabetes drug prescription at the time of their event, and we found similar results. Fourth, to control for the effect of diabetes severity, we stratified our analyses by baseline diabetes treatment, and we found a persistent effect across all diabetes treatment groups. Nonetheless, it is important to note that a causal relationship between antipsychotic agents and hyperglycemia cannot be concluded based on our observational study. Rather, our findings highlight an important increase in hospitalizations during this period that needs to be further addressed.
The prevalence of diabetes and the prescribing of antipsychotic drugs have increased dramatically over the last decade.1,11 Antipsychotic drugs remain popular despite several recent warnings regarding higher risk of stroke and death with these agents.2- 5 Older patients also have the largest burden of diabetes,11 and those with dementia may have their diabetes less aggressively managed, leaving them more susceptible to glycemic deterioration if challenged.34 Moreover, dementia may be more prevalent in patients with diabetes35; therefore, a greater proportion of patients with diabetes may receive antipsychotic agents for behavioral symptoms of dementia. Also, given the widespread perception of overuse and questionable cost-effectiveness of these drugs, our findings raise awareness of potential increasing harms associated with these agents4- 6 such that more careful consideration of the risk to benefit ratio of these drugs is warranted in more vulnerable patients.
In summary, we found that seniors with diabetes and newly prescribed antipsychotic drugs had a higher risk of hospitalizations for hyperglycemia, with a dramatic increase early in the course of therapy. Further studies are needed to determine whether this finding represents a causal effect of antipsychotic drug use. In the meantime, other options to manage behavioral symptoms of dementia should be considered among older persons with diabetes. Patients and their families should be alerted to observe for signs of glycemic decompensation when treatment with an antipsychotic agent is initiated, and enhanced glucose monitoring is recommended for all patients for whom an antipsychotic drug is prescribed, particularly after treatment initiation.
Correspondence: Lorraine L. Lipscombe, MD, MSc, Women's College Hospital, Women's College Research Institute, 790 Bay St, Room 741, Toronto, ON M5G 1N8, Canada (Lorraine.Lipscombe@wchospital.ca).
Accepted for Publication: May 2, 2009.
Author Contributions:Study concept and design: Lipscombe, Lévesque, Gruneir, Fischer, Gill, Herrmann, Hux, and Rochon. Analysis and interpretation of data: Lipscombe, Lévesque, Gruneir, Fischer, Juurlink, Gill, Herrmann, Hux, and Anderson. Drafting of the manuscript: Lipscombe and Lévesque. Critical revision of the manuscript for important intellectual content: Lipscombe, Lévesque, Gruneir, Fischer, Juurlink, Gill, Herrmann, Hux, Anderson, and Rochon. Statistical analysis: Lévesque, Gruneir, and Juurlink. Obtained funding: Anderson and Rochon. Administrative, technical, and material support: Lévesque and Fischer. Study supervision: Anderson and Rochon.
Financial Disclosure: Dr Fischer was employed by Bayer Inc from September 2003 to October 2004. Dr Herrmann has received research support and speaker's honoraria from Janssen Ortho, Eli Lilly, Pfizer, and Novartis.
Funding/Support: This work was supported by a Team Grant (OTG-88591) from the Canadian Institutes of Health Research (CIHR) and by a CIHR Interdisciplinary Capacity Enhancement Grant (HOA-80075). Dr Lipscombe was supported in part by the Team Grant and is supported by a Canadian Diabetes Association/CIHR Clinician Scientist Award. Dr Gruneir is supported by a CIHR Fellowship Award. Dr Juurlink is supported by a CIHR New Investigator Award. Dr Gill is supported by an Ontario Ministry of Health and Long-Term Care Career Scientist Award.
Role of the Sponsor: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Previous Presentation: This study was presented in part at the Canadian Diabetes Association Conference; October 16, 2008; Montreal, Quebec, Canada.
Disclaimer: The opinions, results and conclusions are those of the authors and no endorsement by the Ministry of Health and Long-Term Care or by the Institute for Clinical Evaluative Sciences is intended or should be inferred.
Additional Contributions: Cindy Huo, Tara Gomes, MSc, and Xuesong Wang (Institute for Clinical Evaluative Sciences) assisted with statistical analyses and data presentation, and Pierina Cheung, BSc (Women's College Hospital), assisted with manuscript preparation.