An Epidemiological Approach to “Steroid Psychosis”

Prescription glucocorticoids such as prednisone and methylprednisolone have been used for more than 60 years. Their immunosuppressant and anti-inflammatory properties make them useful for a variety of illnesses, including allergies, asthma, chronic obstructive pulmonary disease, arthritis, systemic lupus erythematosus (SLE), and inflammatory bowel disease, to name a few. They are also used as part of medication regimens to prevent transplant rejection. Despite their potentially lifesaving properties, glucocorticoids are associated with numerous side effects, including bone loss, glaucoma, cataracts, infection, bruising, and diabetes. With long-term use, patients taking glucocorticoids can also develop Cushingoid features, including buffalo hump, moon facies, and truncal obesity.

Soon after glucocorticoids were first used, reports of psychiatric symptoms began to appear in the literature. Case reports suggested that depression, mania, and even psychosis and delirium could develop during glucocorticoid therapy (1). Consistent with these anecdotal reports, the limited longitudinal data also suggested that mood changes were fairly common with glucocorticoid treatment, especially at high dosages (2, 3). Missing in the literature were data from large epidemiological samples, with the exception of a report on depression with glucocorticoid treatment (4) based on data from over 73,000 patients. In this issue of the Journal, Fardet et al. (5), using a remarkable data set from The Health Improvement Network, analyze data from over 300,000 patients treated with glucocorticoids in the United Kingdom and more than 1.2 million comparison patients to explore the prevalence of psychiatric disorders in patients treated with glucocorticoids. The article makes an important contribution to the literature.

Consistent with reports in the literature, the risk of depression, mania, psychosis/delirium, and (to a lesser extent) panic disorder increased with glucocorticoid exposure. Perhaps the most concerning finding was a sixfold increase in the risk of suicide or suicide attempt. An earlier review of the literature observed that 3% of reported cases of steroid-induced psychiatric syndromes were suicides (1), and one of the first case reports involved suicide (6). The Fardet et al. study confirms that suicide attempt is a concern with glucocorticoid treatment (5). Younger people appeared to be at the greatest risk.

The article also addresses the important topic of risk factors. The findings of the Boston Collaborative Drug Surveillance Program first demonstrated that psychiatric symptoms with glucocorticoid treatment were dose dependent (7), and the Fardet et al. study confirms this finding. Women have been thought to be at greater risk for psychiatric side effects with glucocorticoid treatment than men, and more reports of symptoms in women than in men have appeared in the literature. However, because many of the illnesses requiring glucocorticoid therapy (e.g., SLE) are more common in women, it was not at all clear whether gender was actually related to vulnerability to psychiatric symptoms with glucocorticoids. Fardet and colleagues' results suggest that, as with other depressive disorders, women are more likely than men to develop depression with glucocorticoid treatment. Men, however, are more likely than women to develop mania and delirium. These findings suggest interesting gender differences with glucocorticoids, although it is important to note that the gender differences for all of the disorders examined were fairly modest. The role of age as a risk factor was also interesting. The risk of suicide or suicide attempt and panic disorder decreased substantially with age, while the risk of mania and depression increased modestly with age. However, delirium showed a greater than 10-fold increase in patients over age 70 as compared with young persons. This finding, combined with the large hazard ratio for delirium/psychosis in the glucocorticoid-exposed group, suggests that caution and careful observation are indicated when using glucocorticoids in older patients.

Findings from this study regarding prior glucocorticoid exposure also advance the field. Subsequent exposure was associated with a lower overall risk of psychiatric symptoms, but those who had psychiatric symptoms during a prior course were more likely to have symptoms during a subsequent course. This suggests that physicians may avoid prescribing glucocorticoids to patients with a history of psychiatric symptoms during prior courses. Among such patients who do receive a subsequent course, the risk of developing symptoms is increased, albeit fairly modestly (32% greater). However, based on Fardet and colleagues' findings, the answer to the question of whether side effects will inevitably develop in a subsequent course in those with a prior history is no. These findings have very practical clinical implications.

Another previously unanswered question was whether medical illness had a role in the development of symptoms with glucocorticoid treatment. Speculation had included the possibility that symptoms were due to improvement or worsening of the medical illness as a result of glucocorticoid therapy. Perhaps mild euphoria was just a response to improvement in pain or disability. Direct CNS effects of illnesses such as SLE might also be a contributing factor. Fardet et al. shed some light on this issue by using two comparison groups—a general comparison group and one matched for medical illness. Matching for medical illness slightly decreased the hazard ratios, but the findings remained significant. Thus, the data demonstrate differences in patients treated with glucocorticoids even when compared with patients with similar illnesses who were not exposed to glucocorticoids.

The Fardet et al. study has many strengths; however, given the limitations of the data, it is not perfect. The study's strengths and limitations are those inherent in most studies using large cohorts. The diagnoses were not necessarily made on the basis of careful assessment by psychiatrists or structured diagnostic interviews and do not precisely fit DSM-IV terminology. Because of the small number of actual suicides, suicidal behavior is assessed largely by suicide attempts, so the study does not provide specific information about the risk of completed suicide. Nonetheless, the report gives us a fresh look at an old problem and provides clinically important information for psychiatrists, internists, and other medical professionals.