Antipsychotics and diabetes: observational data
A link between schizophrenia and diabetes mellitus was postulated in the 1920s, prior to the development of antipsychotic medications , however, a temporal association between first-generation (“typical”) antipsychotic drug treatment and hyperglycemia was first reported in the 1950s , and current wisdom is that exposure to some second-generation (“atypical”) antipsychotics can considerably increase the risk of development of diabetes mellitus. The American Diabetes Association/American Psychiatric Association consensus statement on antipsychotic drugs and obesity and diabetes notes that aripiprazole and ziprasidone are the two second-generation antipsychotic agents with the lowest risk for metabolic disturbances, although data are limited .
Research demonstrates that people with schizophrenia are at a higher risk of developing diabetes mellitus than the general population, and that prevalence is on the increase. For example, in a report on the New York State Office of Mental Health psychiatric hospital system, the prevalence of people with diabetes mellitus doubled from 6.9% of 10,091 patients in 1997 to 14.5% of 7,420 patients in 2004 irrespective of antipsychotic use . Several theories about the linkage between schizophrenia and diabetes have emerged, including a potential genetic link between the heritability of diabetes and the heritability of schizophrenia. Persons with schizophrenia have a higher rate of family history of diabetes than the general population  and a positive family history of diabetes may increase the risk of developing diabetes in individuals with schizophrenia and other serious and persistent mental illnesses up to threefold.
Insulin resistance can take more than a decade to manifest as diabetes. Someone destined to develop diabetes often has alterations in their ability to handle glucose that have existed for years—alterations that may remain asymptomatic in individuals who remain unaware of having this problem until they eventually have their blood glucose levels checked.
There are clear differences in body weight gain among the different antipsychotics, as evidenced in a recent meta-analysis . However, the development of diabetes mellitus is multi-factorial and increase in body weight may or may not lead to clinically relevant increases in insulin resistance in an individual person.
What can pharmacoepidemiological studies tell us?
Different study designs have been used to examine the possible connection between antipsychotic selection, diabetes, and schizophrenia . The published pharmacoepidemiological literature has relied on administrative databases to determine whether or not there is an association between antipsychotic exposure and the new onset of diabetes mellitus. Although a cross-sectional study can assess the burden of a disease in a population, this study design cannot answer questions regarding causality. Ideally, a prospective cohort study examining antipsychotic exposure and the development of diabetes mellitus would best address causality, but this study design is expensive to execute and takes a lengthy amount of time to complete; to date, no large-scale prospective cohort studies have been conducted on diabetes and antipsychotic use. Thus, retrospective analyses have been preferred for logistical reasons and two types have been conducted: retrospective cohort and retrospective case control. Both designs are limited by the quality of the data source. For example, a prescription drug database may not clearly indicate whether a patient is taking multiple medications (including medicines known to be diabetogenic), nor does it reflect whether a patient is adhering to a prescribed regimen. The database might indicate which people are being prescribed antidiabetic medications, but it probably will not show who actually has been diagnosed with diabetes or who is controlling diabetes through diet and exercise instead of medication. The database will not indicate people with undiagnosed diabetes. Administrative databases do not usually include information regarding family history of diabetes—perhaps one of the strongest risk factors for the disease. The databases are also generally silent regarding weight, exercise, or dietary habits. Perhaps the most obvious flaw in these studies is treatment assignment bias - the factors that influence a doctor to prescribe one antipsychotic instead of another might be based on a patient’s preexisting risk factors for diabetes (e.g., family history, obesity, lack of exercise).
What the Data Show
In a systematic review of 25 observational pharmacoepidemiologic studies, all retrospective, sufficient information was provided in 15 reports to estimate the attributable risk comparing second-generation with first-generation antipsychotics . However, the studies differed in how much information about past history was made available and used to eliminate people with preexisting diabetes mellitus. Definitions of what constituted an index case for antipsychotic exposures also differed from study to study. The definition of development of diabetes mellitus differed from study to study as well and included the new prescription of an antidiabetic agent, a recorded diagnosis of diabetes mellitus, or both. Rarely were blood glucose levels used to identify incident cases. None of the studies controlled for familial history of diabetes, levels of physical activity, or diet. The principal second-generation antipsychotics available for study at the time of the review were clozapine, risperidone, olanzapine, and quetiapine. The attributable risk for developing incident diabetes did not show a consistent advantage for typical compared with atypical agents, however, there was a signal regarding added risk with clozapine. All but one of the studies that compared clozapine with first-generation antipsychotics demonstrated a higher potential risk for clozapine. For the studies that demonstrated a higher risk for clozapine, number needed to harm (NNH) for the outcome of new-onset diabetes mellitus in studies that compared clozapine with first-generation antipsychotics ranged from 19 to 179. Results for olanzapine, quetiapine, and risperidone were less consistent, with multiple studies suggesting that the second-generation antipsychotics carry a higher potential risk but with other studies suggesting that first-generation antipsychotics carry a higher potential risk (nine studies found that patients treated with at least one of the second-generation antipsychotics had a lower risk of developing diabetes than did patients treated with a first-generation antipsychotic). Regardless of the direction of the difference in risk, effect sizes were small, as demonstrated by large values of NNH, ranging in absolute magnitude from 19 to 333 for clozapine, 22 to 3333 for risperidone, 21 to 747 for olanzapine, and 24 to 500 for quetiapine. Seldom were the differences statistically significant and the confidence intervals themselves were relatively wide. Results for the studies where the length of follow-up postexposure was one year or more were not any more consistent in their outcomes.
The association of clozapine with new-onset diabetes mellitus may be particularly difficult to interpret, as patients prescribed that drug are likely more severely ill and may have also been receiving clozapine and other antipsychotics for longer periods of time than patients who have never received clozapine. The mandatory surveillance for agranulocytosis for patients on clozapine may also increase the likelihood of additional tests for other abnormalities, such as hyperglycemia.
Newly published observational data
A retrospective cohort study examined the risk associated with use of aripiprazole compared to other individual second-generation antipsychotic agents for major cardiovascular outcomes including myocardial infarction, stroke, heart failure, and coronary artery bypass grafting or percutaneous transluminal coronary angioplasty, as well as new-onset diabetes mellitus . Aripiprazole initiators were matched 1:1 to 9,917 olanzapine, 14,935 quetiapine, 10,192 risperidone, and 5,696 ziprasidone initiators. Incidence of diabetes ranged from 18 to 21 events per 1,000 person-years in each cohort, and did not differ significantly between these second-generation antipsychotics.
The development of type 2 diabetes mellitus is typically an insidious process that extends over many years. Data from retrospective observational pharmacoepidemiologic studies are very difficult to interpret because they involve people with largely unknown risk factors (eg, weight, family history), largely unknown presence of other diabetogenic medications, unknown change in exercise or dietary habits, unknown reasons for selection of particular antipsychotic medicine, unknown adherence to antipsychotic medicine, and unknown frequency and mechanism for screening of diabetes.
^ Lorenz WF. Sugar tolerance in dementia praecox and other mental disorders. Arch Neurol Psychiatry 1922;8:184-96.
^ Charatan FB, Bartlett NG. The effect of chlorpromazine (largactil) on glucose tolerance. J Ment Sci 1955;101:351-3.
^ American Diabetes Association, American Psychiatric Association, American Association of Clinical Endocrinologists, et al. Consensus development conference on antipsychotic drugs and obesity and diabetes. Diabetes Care. 2004;27(2):596–601.
^ Citrome L, Jaffe A, Levine J, Martello D. Incidence, prevalence, and surveillance for diabetes in New York State psychiatric hospitals, 1997-2004. Psychiatr Serv 2006;57(8):1132-9.
^ Mukherjee S, Schnur DB, Reddy R. Family history of type 2 diabetes in schizophrenic patients [letter]. Lancet. 1989;1(8636):495.
^ Cheta D, Dumitrescu C, Georgescu M, et al. A study on the types of diabetes mellitus in first degree relatives of diabetic patients. Diabetes Metab. 1990;16:11-5.
^ Lamberti JS, Crilly JF, Maharaj K, et al. Prevalence of diabetes mellitus among outpatients with severe mental disorders receiving atypical antipsychotic drugs. J Clin Psychiatry 2004;65:702-6.
^ Leucht S, Cipriani A, Spineli L, et al. Comparative efficacy and tolerability of 15 antipsychotic drugs in schizophrenia: a multiple-treatments meta-analysis. Lancet. 2013;382(9896):951-62.
^ Citrome LL. The increase in risk of diabetes mellitus from exposure to second-generation antipsychotic agents. Drugs of Today. 2004;40:445-64.
^ Citrome LL, Holt RI, Zachry WM, Clewell JD, Orth PA, Karagianis JL, Hoffmann VP. Risk of treatment-emergent diabetes mellitus in patients receiving antipsychotics. Ann Pharmacother. 2007 Oct;41(10):1593-603.
^ Citrome L, Collins JM, Nordstrom BL, et al. Incidence of cardiovascular outcomes and diabetes mellitus among users of second-generation antipsychotics. J Clin Psychiatry. 2013;74(12):1199-206.