The relationship between anxiety disorder and depressive disorder has been a topic of great interest in psychiatry. Comorbidity of anxiety and depressive disorders has been shown to influence many clinical aspects of these disorders, including onset, duration, response to treatment, and severity, when compared with each condition alone.1—4 Thus, the comorbid condition appears to have important implications for both long-term outcome and response to treatment.
During the past few years, we have been examining the clinical and neuropathological correlates of anxiety disorder following acute stroke.5,6 These studies have identified the frequent comorbidity of depressive disorders and anxiety disorders in patients with acute stroke. Comorbid major depression and generalized anxiety disorder were associated with left frontal cortical lesions.5 Similarly, anxiety disorder, but not comorbid anxiety and depression, was associated with increased frequency of alcohol abuse and right hemisphere lesions.6 These data, as well as other similar findings, suggest that comorbid anxiety and depression may have a different etiological basis than depression alone or anxiety without depressive disorder.
Previous studies have shown that poststroke depressive disorders influence the cognitive7,8 and physical9 consequences of stroke. Major depressive disorder is associated with a significantly greater degree of cognitive impairment than can be explained by the presence of the lesion alone.10,11 Parikh et al.,9 as well as Sinyor et al.,12 demonstrated that physical and language recovery from stroke were adversely affected by the existence of depressive disorder. We have not, however, previously examined the influence of comorbid anxiety and depressive disorder on cognitive or physical recovery from stroke or the longitudinal course of depression. Therefore, the present study examined the hypothesis that the comorbid existence of anxiety and depressive disorder would be associated with a greater degree of impairment in cognitive function and a poorer recovery from depression and physical impairment than would depressive disorder without comorbid anxiety disorder.
A consecutive series of 142 patients who were admitted to University of Maryland Hospital, Baltimore, MD, with a diagnosis of either acute intracerebral hemorrhage or acute cerebral infarction and who were seen in the follow-up clinic were examined for psychiatric disorder after providing informed consent. Patients were excluded only if they were unable to undergo a verbal interview. Nearly one-third of patients admitted with acute stroke were unable to complete an interview because of decreased level of consciousness or aphasia with impaired comprehension. Of these 142 patients, 110 patients (77.5%) were examined at either 3-month or 6-month follow-up (short-term follow-up), and 102 patients (71.8%) were examined at either 12-month or 24-month follow-up (long-term follow-up). Patients were lost to follow-up due to mortality (15%) or failure to show up for appointments (10%).
Psychiatric and Neurological Evaluations
After obtaining informed consent, we examined the patients by using a structured mental status examination, the Present State Examination13 (PSE). We made diagnoses of major depression (MDD) or generalized anxiety disorder (GAD) by using the symptoms elicited on the PSE and the DSM-IV14 symptom criteria for major depressive disorder and generalized anxiety disorder. The diagnostic criteria for GAD were slightly modified from DSM-IV criteria as follows: patients had to acknowledge the presence of worry or anxious foreboding, and duration criteria were excluded. In addition, the severity of depression was measured with the Hamilton Rating Scale for Depression15 (Ham-D). Cognitive functioning was evaluated with the Mini-Mental State Examination16 (MMSE). Scores range from 0 to 30, with lower scores indicating greater impairment. Activities of daily living were quantified by use of the Johns Hopkins Functioning Inventory17 (JHFI). Scores range from 0 to 27, with higher scores indicating greater impairment. Social functioning was assessed by use of the Social Functioning Exam18 (SFE). Scores range from 0.00 to 1.00, with higher scores indicating greater perception of impaired social function.
Computed tomography scans were obtained from 73 (51.4%) of the patients included in the study and were evaluated by a neurologist who was blind to the psychiatric findings. All CT scans had consistent 10-mm slice thickness and consistent angle to the canthomeatal line. Methods for determining lesion location (lesion volume, interrater reliability, cortical-subcortical location) have been described in a previous report.5
A 2×2 design was used in this study. Analysis of variance (ANOVA) and repeated-measures ANOVA were performed to analyze the relation between dependent variables such as MMSE scores and each of the dependent variables such as the existence of depression or GAD. If the ANOVA was significant, each group was compared by using post hoc planned tests (Scheffé). Chi-square tests were performed to analyze frequency distributions. Yates' correction was used to adjust for expected cell sizes below 5, and two-tailed P-values were used.
Subjects were divided into four groups: control (n=100), generalized anxiety disorder only (n=15), major depressive disorder only (n=9), and MDD plus GAD (n=18).
The background characteristics of these groups are shown in T1. There were significantly more female patients in the MDD and MDD plus GAD groups than in the non-MDD groups (χ2=12.9, df=3, P=0.049);.
Of the 142 patients in this study, 48 had single lesions restricted to cortical or subcortical structures as assessed by CT scan. In all cases, brain lesions were found by a neurologist to be compatible with findings on neurologic examination. Cortical lesions were found in 13 of 27 patients (48.1%) in the control group, 0 of 4 patients (0%) in the MDD-only group, and 4 of 5 patients (80%) in the MDD plus GAD group. These were statistically significant intergroup differences in the frequency of cortical lesions (χ2=7.4, df=2, P=0.03).
Relationship to Diagnostic Outcome
We next examined diagnostic outcome among groups. There were significant intergroup differences in the frequency of major depression at short-term follow-up (control: 5 of 73; MDD only: 0 of 8; MDD+GAD: 9 of 14; χ2=24.7, df=2, P<0.0001). Most patients with MDD plus GAD remained depressed at short-term follow-up, whereas none of the MDD-alone patients remained depressed (MDD only vs. MDD+GAD, χ2=6.3, P=0.02). There were no significant intergroup differences, however, in the persistence of major depression at long-term follow-up (control: 12 of 78; MDD only: 1 of 4; MDD+GAD: 0 of 9; χ2=3.2, df=2, P>0.1).
We also examined changes in the severity of depression (Ham-D scores) during the follow-up evaluation. A one-way ANOVA of Ham-D scores for the patients with 3- to 6-month follow-up showed that patients who had MDD plus GAD in-hospital had significantly higher depression scale scores at both time evaluations than did patients with MDD only or control subjects (initial: F=87.8, df=2,86, P<0.0001; short-term follow-up: F=15.4, df=2,86, P<0.0001). Post hoc test (Scheffé) demonstrated that patients with MDD plus GAD had significantly higher depression scale scores at both time evaluations than did patients with MDD only (initial: P<0.001; short-term follow-up: P<0.05). In contrast, one-way ANOVA of Ham-D scores for the patients with 12- to 24-month follow-up showed significantly higher in-hospital depression scores for the MDD plus GAD patients than for patients with MDD only or control subjects, but there were no significant differences among the groups at long-term follow-up (initial: F=48.4, df=2,86, P<0.0001; long-term follow-up: F=1.6, df=2,86, P>0.1).
Effect of MDD and GAD on Impairment and Recovery
We found that non-MDD patients tended to be older than patients with MDD (non-MDD, 59.2±12.4, vs. MDD, 54.0±13.0: t=1.93, df=140, P=0.056), and non-GAD patients tended to have larger lesion volumes than patients with GAD (non-GAD, 7.4±7.3, vs. GAD, 3.4±2.8: t=1.97, df=65, P=0.054). Values are reported as means and standard deviations.
For the purpose of excluding factors other than MDD or GAD that were known to effect stroke outcome, the following analyses were done, using patients who were matched for age (±2 years), education (±2 years), and lesion volume (±2% of total brain volume).
Relationship to Cognitive Impairment and Recovery:
In short-term follow-up, repeated-measures ANOVA of MMSE scores (factor 1: presence or absence of MDD; factor 2: presence or absence of GAD) showed significant effects for MDD (F=21.7, df=1,56, P<0.0001) and time (F=7.23, df=1,56, P=0.009), but no significant effect for an MDD×time interaction and no significant effect for GAD (T2). A two-way ANOVA at each time showed a significant effect for MDD at both initial and short-term follow-up (initial: F=29.68, df=1,56, P<0.0001; short-term: F=8.04, df=1,56, P=0.006), but no effect for GAD and no MDD×GAD interaction. Similarly, in long-term follow-up, repeated-measures ANOVA of MMSE scores showed a significant effect of MDD×time interaction (F=6.57, df=1,58, P<0.01), but not of GAD or other interaction. A two-way ANOVA at each time showed a significant effect of MDD at initial evaluation (F=5.16, df=1,58, P=0.03), but not at long-term follow-up (F=0.10, df=1,58, P=0.8). In contrast, there was no significant effect of GAD, and there were no significant interaction effects at either time.
Relationship to Physical Impairment and Recovery:
Repeated-measures ANOVA of JHFI scores for the patients with 3- to 6-month follow-up showed a significant effect for MDD (F=8.78, df=1,57, P=0.004) and time (F=25.60, df=1,57, P<0.0001), but no significant effect for GAD or MDD×GAD interaction (T3). A two-way ANOVA at each time showed no significant effect at the initial evaluation. In contrast, however, patients with major depression had significantly lower JHFI scores at short-term follow-up than those who were not depressed (MDD: F=16.27, df=1,57, P=0.0002).
Repeated-measures ANOVA of JHFI scores for patients with 12- to 24-month follow-up showed significant effects for MDD (F=4.50, df=1,55, P=0.04) and MDD×GAD interaction (F=5.63, df=1,55, P=0.02), but no significant effect for GAD or other interaction. A two-way ANOVA at each time showed significant effects for MDD and an interaction with GAD at long-term follow-up (MDD: F=5.94, df=1,55, P=0.02; MDD×GAD: F=4.45, df=1,40, P=0.04). Thus, patients who had both MDD and GAD in-hospital were more impaired in their activities of daily living at long-term follow-up than were control patients or patients with GAD alone.
Relationship to Social Functioning:
Repeated-measures ANOVA of SFE scores for the patients with 3- to 6-month follow-up showed significant effects for MDD (F=28.43, df=1,54, P<0.0001), GAD (F=16.93, df=1,54, P<0.0001), GAD×time interaction (GAD×time: F=4.88, df=1,54, P=0.03) and MDD×GAD×time interaction (MDD×GAD×time: F=6.56, df=1,54, P<0.01), but no significant effect for MDD×GAD interaction (T4). A two-way ANOVA at each time showed a significant effect for MDD at both times (initial: F=13.48, df=1,54, P<0.001; short-term: F=21.52, df=1,54, P<0.0001). There were significant effects for GAD and interaction with MDD at short-term follow-up (GAD: F=25.90, df=1,54, P<0.0001; MDD×GAD: F=12.06, df=1,54, P<0.001).
Repeated-measures ANOVA of SFE scores for patients with 12- to 24-month follow-up showed significant effects for MDD (F=11.65, df=1,56, P<0.001), GAD×time (F=5.79, df=1,56, P=0.02), and MDD×GAD×time interaction (F=5.20, df=1,56, P=0.03). A two-way ANOVA at each time showed a significant effect for MDD at initial evaluation (F=12.20, df=1,56, P=0.0009). In contrast, there were significant effects for GAD (F=6.80, df=1,56, P<0.01) and MDD×GAD interaction (F=6.46, df=1,56, P<0.01) at long-term follow-up, but not at initial evaluation. Thus, patients with MDD plus GAD in the hospital were more socially impaired at long-term follow-up than control patients or patients with MDD alone.
The present study was undertaken to evaluate the effect of comorbid anxiety disorder and depressive disorder on impairment following acute stroke. Our findings demonstrated for the first time that anxiety disorder following stroke influenced the severity and course of recovery from stroke. Anxiety disorder interacted with depression to influence the degree of impairment in activities of daily living and the course of recovery in social functioning at long-term follow-up. On the other hand, depression, but not an interaction with GAD, explained our previously demonstrated finding that depression affects cognitive function following acute stroke.7,11
Before further discussing these findings, we should address several methodological limitations. First, cognitive function was evaluated by using only the brief, language-dominated MMSE. Because of the limited number of cognitive functions evaluated by MMSE, we were unable to determine whether there may have been a subtle effect of GAD on cognitive function. This will require a study using a more detailed neuropsychological assessment.11 Second, patients with severe comprehension deficits were excluded, and the population of this study was predominantly black and of lower socioeconomic status. Therefore, the results of this study may not apply to a more general population of poststroke patients. Another limitation was the potential effect of medications on cognitive function or recovery. Benzodiazepines or other sleeping medications are commonly prescribed in this population and may contribute to cognitive impairment or slower speed of recovery. When we examined the frequency of various psychotropic medications at the in-hospital evaluation, we did not find significant differences in the frequency of antidepressant medications among groups, and we did not find any patients who were treated with other psychotropics such as benzodiazepines (T1). However, we did not examine the frequency of medication use at follow-up. Thus, there may have been medication effects that were missed. Another limitation is that the number of follow-up patients who were initially diagnosed with GAD only or MDD only was small. Thus, the power of our analysis was limited, and some significant effect of GAD may have been masked by this. In addition, mortality or failure to comply with follow-up evaluation led to some attrition at each follow-up (24% at short-term and 29% at long-term follow-up). About 20% of this study population died within the 2-year follow-up period.19
The most surprising finding in this study was that neither GAD nor an interaction of GAD with MDD influenced cognitive impairment either in-hospital or during follow-up. The failure of GAD to worsen the impairment did not support our original hypothesis. This finding also suggests that the mechanism of cognitive impairment associated with depression is not altered by comorbid anxiety disorder, implying perhaps the independence of anxiety and depressive disorders. This finding also supports our group designation of MDD plus GAD. It might be argued, however, that these are not "comorbid" disorders but rather are a type of anxious depression. This issue will require further research.
Perhaps the most significant finding was that patients with MDD plus GAD had greater impairment in activities of daily living at long-term follow-up than did patients with MDD alone, and the course of recovery in social functioning was significantly worse in patients with GAD plus MDD compared with control subjects or patients with MDD or GAD alone. These findings suggest that anxiety disorder is an important variable affecting long-term prognosis of depression following stroke. We have previously reported that major depression adversely affects the course of physical recovery.9 Thus, one possible explanation is that since, as discussed below, MDD with GAD has a longer duration of depression than MDD alone, this prolonged depression may have led to these adverse physical and social functioning outcomes.
Another interesting finding in this study was that patients with MDD plus GAD had significantly longer and more severe depressions than patients with MDD only. The most obvious explanation is that cortical lesions (associated with anxious depressions) in some way produced more refractory depressions. We previously reported that patients who failed to recover from depression during the first 6 months following stroke had mainly cortical lesion location.20 Feeney and Baron21 reported that cortical lesions produced long-lasting hypometabolic areas of brain, distant from the site of the lesion. These distant effects may be more extensive and persistent after cortical lesion than after subcortical lesion and may lead to longer-lasting depressions. We previously found that depressed mood following stroke was associated with an increased risk of mortality over the first 10 years poststroke.19 These findings suggest that MDD plus GAD, and perhaps MDD only, may be associated with increased risk of mortality. However, it remains to be determined whether anxious depression has different survival probability than depression alone or no mood disorder and whether this risk may be improved by treatment with anxiolytic or antidepressant medications. Although depression appears to be the major factor in cognitive and physical impairment following stroke, the intriguing therapeutic issue raised by this study is whether anxiolytic treatment might improve physical and psychosocial as well as emotional recovery from stroke.