The Frequency of Cognitive Impairment in Patients With Anxiety, Depression, and Bipolar Disorder: An Unaccounted Source of Variance in Clinical Trials

Background: Patients with anxiety, depression, and bipolar disorder are known to be impaired relative to healthy controls on neurocognitive tests, but the degree of impairment may be obscured if the data are analyzed in terms of group means.

Method: Patients and controls were administered a comprehensive neurocognitive assessment that measured performance in 5 domains: memory, psychomotor speed, reaction time, attention, and cognitive flexibility. Clinic patients diagnosed per DSM-IV-TR criteria with generalized anxiety disorder (N = 63), major depressive disorder (N = 285), and bipolar I or II disorder (N = 96) were compared with 907 controls. Subjects’ age range was 18 to 65 years. Patients had no comorbid psychiatric disorders and no medical, neurologic, or developmental conditions that might affect cognition (e.g., attention-deficit/hyperactivity disorder, brain injury, mild cognitive impairment, chronic pain). Data on patients and controls (collected from March 2003 through February 2007) were taken from a clinical database that also contained neurocognitive test scores.

Results: There were small differences between patients and controls, between different patient groups, and between treated and untreated patients when neurocognitive results in terms of group means were compared. Comparisons of results in terms of the frequency with which patients and controls fell below certain cutoff scores amplified the importance of these differences. Only 4% of controls fell below a standard score of 70 (2 standard deviations below the mean) on 2 or more cognitive domains, but 19% of anxiety patients, 21% of depressed patients, and 30% of bipolar patients fell below the standard score.

Conclusions: Substantial numbers of patients with anxiety, depression, and bipolar disorder are cognitively impaired. A score that is 2 standard deviations below the mean is usually clinically important, and 2 domain scores in that range is cause for serious concern. The importance of this finding is discussed, with respect to clinical trials, in terms of establishing a homogeneous trial population and minimizing the placebo response rate.