subjects who received the highest level of anticholinergic unit exposure (AUE) from antipsychotic medications were more likely to develop delirium (OR 15.42, significance level 0.001). Lower levels of AUE from antipsychotics were statistically significant at a P value £ 0.05 but did not strongly predict delirium. There were no consistent correlations between estimates of anticholinergic exposure for antidepressant or antiparkinsonian medications and the incidence of delirium. We concluded that a predictive algorithm based on anticholinergic exposures calculated from clinical data could not be used as a preventive intervention.

Project data sources included various published estimates of each medication’s average relative anticholinergic effect (ARAE) and hospital chart data describing medications taken and their dosages. Error sources included anticholinergicity estimates and documentation errors in clinical records. Predictive estimates derived from error-prone data sources are vulnerable to misclassification bias (3). Anticholinergic exposure categorization entails 3 complications: errors in recording of type, dosage, and frequency of medications; possible patient noncompliance with medication; and crude estimations of each medication’s ARAE from diverse literature sources. These estimations do not incorporate pharmacokinetic factors, such as rates of absorption, metabolism, and elimination of medications. This misclassification of exposure status probably occurred with the same frequency in both case and control

subjects (nondifferential misclassification bias) and, therefore, caused a dilution of the observed effect. Ultimately, this resulted in a bias toward the null and a dilution of  the formula’s predictive capacity.

Investigators considering use of electronic health care data should be aware that powerful mathematical factors may result in a dilution of predictive effects. In our example, this dilution apparently resulted in poor utility of a proposed pharmacy-based preventive strategy. In studies comparing outcome, the dilution due to nondifferential misclassification bias could result in diminished apparent differences among interventions.

References

1. Patten SB, Williams JVA, Haynes L, McCruden J, Arboleda-Flórez J. The incidence of delirium in psychiatric inpatient units. Can J Psychiatry 1997;42:858–63.
2. Patten SB, Williams JVA, Petcu R, Oldfield R. Delirium in psychiatric inpatients: a case-control study. Can J Psychiatry. 2001;46:162–6.
3. Patten SB. Integrating administrative and research data in pharmacoepidemiological research. Can J Clin Pharmacol 1998;5(2):92–7.

Kristin Bristow, BMR(OT)
Scott Patten, MD, FRCPC, PhD
Calgary, Alberta

Dear Editor:

Psychiatric reactions occur in 3% to 60% of steroid-treated patients (1). Treatment with steroid taper, neuroleptics, selective serotonin reuptake inhibitors (SSRIs), or electroconvulsive therapy (ECT) is

generally effective (2). Previous reports suggest that prophylactic treatment with lithium carbonate (3), valproate (4), or lamotrigine (5) might reduce steroid-induced psychiatric symptoms. We report the successful use of the anticonvulsant gabapentin in the prevention of steroid-induced mania.

Case Report

A 59-year-old woman with dissociative identity and panic disorders, but no history of mania or hypomania, was stable on a regimen of sertraline 200 mg daily, buspirone 30 mg twice daily, and alprazolam 4 to 5 mg daily. For nearly 3 months, she had not experienced any panic attacks. Her medical history included hyperthyroidism, hypertension, hypercholesterolemia, and mitral valve prolapse. Other medications included amlodipine 5 mg daily, atorvastatin 10 mg daily, levothyroxine 125 mcg every other day (alternating with 150 mcg every other day), atenolol 25 mg daily, and transdermal estradiol 0.1 mg daily.

In March 2000, the patient experienced a flare-up of asthma that was treated with prednisone 40 mg daily, tapered over a 12-day period. Within 48 hours, she developed a manic reaction. She felt “hyper,” as if her body were “going to explode, like a hand grenade,” likening it to “adrenaline multiplied by 100.” She was unable to focus, make decisions, sleep, or write. She experienced the sensation of bugs crawling on her skin. She put her sunglasses in the freezer and her eyeglasses in the laundry basket. Worried that she was “going crazy,” she felt completely dysfunctional. Her symptoms resolved several days after discontinuing prednisone.