Depression and Antidepressant Use After Stroke and Transient Ischemic Attack
Background and Purpose—Patients with stroke and transient ischemic attack (TIA) often have comparable comorbidities, but it is unclear whether they have similar rates of depression or antidepressant use.
Methods—This study was a secondary analysis of a prospective cohort registry that enrolled subjects from 2006 to 2008 in the United States. Depression (defined by the Patient Health Questionnaire-8 score ≥10) and medication use were prospectively assessed 3 and 12 months after hospitalization in 1450 subjects with ischemic stroke and 397 subjects with TIA.
Results—The proportional frequency of depression after stroke and TIA was similar at 3 months (17.9% versus 14.3%, P=0.09) and at 12 months (16.4% versus 12.8%, P=0.08). The rates of newly identified depression between 3 and 12 months were also similar (8.7% versus 6.2%, P=0.12). Persistent depression (defined as Patient Health Questionnaire-8 score ≥10 at both 3 and 12 months) was present in 134 (9.2%) of those with stroke and in 30 (7.6%) of those with TIA. Younger age, greater stroke-related disability, and inability to work at 3 months were associated with persistent depression in subjects with stroke. Among subjects with persistent depression, 67.9% of those with stroke and 70.0% of those with TIA were not using antidepressants at either time point (P=0.920).
Conclusions—Stroke and TIA subjects had a similar frequency of depression at 3 and 12 months after hospitalization and similar rates of newly identified depression between 3 and 12 months. A high proportion of those with persistent depression was untreated.
- behavioral neurology
- cerebral infarct
- cerebrovascular disease
- stroke care
- transient ischemic attack
Depression is the most common psychiatric disorder affecting patients with stroke and may contribute to poststroke morbidity and mortality.1 The frequency of poststroke depression varies considerably across studies depending on cohort characteristics and diagnostic criteria but is considerably higher than control populations matched for age and sex.2–4 The pathophysiology of poststroke depression is likely multifactorial and influenced by the location and extent of brain injury, vascular comorbidities, and reaction to new functional disability.5–7 Patients with transient ischemic attack (TIA) share comorbid conditions with those who have had an ischemic stroke, and although approximately 30% to 40% may have radiographically demonstrated brain injury, by definition, a TIA is not associated with a long-lasting functional impairment. Nonetheless, there is a paucity of studies assessing the proportional frequency of depression and antidepressant use among patients with TIA.8,9 Furthermore, there are little patient-level longitudinal data reflecting the presence or absence of depression and antidepressant use within the year after hospitalization for stroke or TIA. Such analyses are needed to inform screening recommendations and to assess the adequacy of current treatment approaches.
Various reports on depression care in the general US population suggest racial and ethnic disparities in the use of psychiatric resources, including the use of antidepressants.10–12 Other factors such as financial stress, lack of health insurance coverage, and the presence of concurrent medical conditions may also affect antidepressant use.11,12 It is not clear whether the same factors affect antidepressant use in patients with stroke and depression.
The primary objectives of the current study were to compare the proportional frequency of depression, newly identified depression between 3 and 12 months and antidepressant use in the year after hospitalization for stroke or TIA. The secondary objectives were to identify factors associated with undertreatment of depression, depression at 12 months posthospitalization, and persistent depression (depression at both 3 and 12 months) in patients with stroke.
Study Population and Design
This is a secondary analysis of a multicenter prospective cohort registry (Adherence eValuation After Ischemic stroke Longitudinal [AVAIL] study), which was designed to assess adherence to stroke prevention medications from hospital discharge to 1 year in patients admitted with stroke or TIA. The AVAIL study methodology has been previously published.13 Briefly, 2889 patients with stroke or TIA were recruited from hospitals participating in the Get With The Guidelines (GWTG)–Stroke program from July 2006 through July 2008. The 12-month follow-up was completed in August 2009. Outcome Sciences, Inc served as the data collection and coordinating center for GWTG–Stroke. The Duke Clinical Research Institute served as the data analysis center for both GWTG and AVAIL. Each participating site obtained Institutional Review Board approval before enrolling patients into AVAIL.
The inclusion criteria for the AVAIL registry were: age ≥18 years; hospitalization with a primary diagnosis of acute ischemic stroke or TIA based on the GWTG–Stroke identified International Classification of Diseases, Ninth Edition Codes; direct admission based on physician evaluation or arrival through the emergency department; patient or legally authorized representative consent to participate; and inclusion in the GWTG–Stroke program. AVAIL subjects were excluded from the present analysis if baseline data were missing; the 3- or 12-month Patient Health Questionnaire-8 (PHQ-8) was not completed; if antidepressant doses were missing; or if the subject died before 12 months. The enrolling sites were both academic (73.3%) and nonacademic (26.3%) hospitals; 30.3% were in the Midwest, 23.6% in the Northeast, 25.3% in the South, and 18.2% in the West. Duke Clinical Research Institute research personnel conducted telephone interviews at 3 months and 12 months after hospital discharge using standardized scripts.13
Outcomes and Covariates
The primary outcomes for this analysis were depression (assessed with the PHQ-8) and antidepressant use (at doses at or above those recommended for treatment of depression). The PHQ-8, based on symptoms within the prior 2 weeks, includes 8 of the 9 criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition for diagnosis of major depressive disorder.14 The PHQ-8 yields a score from 0 to 24 with a score of ≥10 indicating a clinically significant depressive disorder.15 This cutoff has a sensitivity and specificity of 88% and positive predictive value of 57% for major depression.15,16 Persistent depression was defined as PHQ-8 ≥10 at both 3 and 12 months regardless of antidepressant use. Incident depression was defined as having a PHQ-8 ≥10 at 12 months in those with PHQ-8 <10 at 3 months.
Medication use was subject-reported and ascertained through a series of questions aimed at comparing current and previous use. Antidepressants used at an approved dosage were identified based on a previously validated algorithm (online-only Data Supplemental Table I).17 Undertreatment was identified when subjects were not using antidepressants at 3 or 12 months but had PHQ-8 ≥10 at both time points.
Trained hospital personnel abstracted baseline time-independent demographic characteristics from inpatient medical records as part of the GWTG–Stroke admission. Time-dependent subject characteristics were self-reported by subjects on follow-up at each time point as part of the AVAIL protocol.13 The modified Rankin Scale (mRS) was scored at each follow-up based on the standardized telephone interview.13,18 A cutoff of 3 was used to differentiate mild from moderate/severe stroke, because a mRS <3 is usually associated with the ability to maintain independent living.19
Frequency distribution was used for categorical variables. Median and interquartile ranges were calculated for continuous variables. Chi-squared tests were used to compare categorical variables and Wilcoxon rank-sum tests to compare continuous variables. Key demographic variables were compared between included and excluded subjects. The observed proportional frequency of depression was compared between stroke and TIA at 3 and 12 months before and after adjusting for gender, age, disability (mRS), previous history of stroke/TIA, and history of smoking using multivariable logistic regressions. The modified Wald method was used to calculate CIs. To account for the possible confounding effect of previous stroke/TIA on the frequency of depression, the proportional frequency of depression was also compared between stroke and TIA after excluding subjects with a history of cerebrovascular events. The frequency of newly identified depression was calculated by dividing the number of those with both PHQ-8 ≥10 at 12 months and PHQ-8 <10 at 3 months by the sample size within each group. Subject-level longitudinal data about depression and antidepressant medications were described.
Multivariate logistic modeling was used to assess the association of prespecified variables with persistent depression and depression at 12 months in subjects with stroke. The variables included: sex, race/ethnicity (white versus other), age, mRS at 3 months (range, 0–5), work status at 3 months (work, home by choice, versus home not by choice), living situation at 3 months (with someone versus other), use of antidepressants at 3 months, and history of coronary artery disease/myocardial infarction at baseline. Depression at 3 months was included only as a predictor variable for depression at 12 months. Backward selection of covariates with a probability value of <0.1 was performed in both models. The rate of missing covariate data was low with the values imputed. A similar analysis in subjects with TIA was not performed because the number in the TIA group was too small to support a valid multivariate analysis. The association of prespecified variables with undertreatment of persistent depression was assessed with univariate logistic analysis; a multivariable logistic regression was not planned because of sample size limitations for this outcome.
Although the primary end point was based on self-reported symptoms of depression, subjects using antidepressants may carry a diagnosis of “depression” even if symptoms are adequately controlled by treatment. To address this possibility, a sensitivity analysis was performed by redefining depression as PHQ-8 ≥10 and/or the use of antidepressants at an appropriate dose. All probability values are 2-sided with P<0.05 considered statistically significant. All analyses were performed using SAS software Version 9.2.
Comparison of Included and Excluded Subjects
Of AVAIL subjects (n=2889), 1042 (36%) were excluded from the present analysis because baseline data were missing, the 3- or 12-month PHQ-8 was not completed, antidepressant doses were missing, or the subject died before 12 months (Figure) leaving a total of 1847 AVAIL subjects from 99 hospitals. Subjects who were excluded were more likely to be older (median age, 69 years; interquartile range, 58–80 years, versus 65 years; interquartile range, 55–74.5 years; P<0.0001), to have been discharged to an institution (45.7% versus 26.8%; P<0.0001), to be living in an institution at 3 months (2.3% versus 0.4%; P<0.001), to have less than a college education (63.4% versus 52.4%; P<0.001), to report that household income did not meet needs (7.9% versus 5.2%; P<0.0001), to be unmarried (44.4% versus 38.9%; P<0.004), to report their work status as “home not by choice” (13.4% versus 10.3%; P<0.0001), to have a history of previous stroke/TIA (29.4% versus 23.1%; P=0.0002), and to have a history of coronary artery disease/myocardial infarction (29.1% versus 24.3%; P=0.005). In addition, because subjects rather than a proxy had to complete the PHQ-8, those with aphasia or significant cognitive impairments were more likely to be excluded. The excluded group did not differ with respect to sex (female 46.5% versus 45.5%; P=0.734), race (white versus nonwhite 83.0% versus 80.6%; P=0.108) or medical history of carotid stenosis (4.7% versus 5.6%; P=0.296), diabetes mellitus (28.7% versus 31.6%; P=0.088), peripheral vascular disease (3.7% versus 4.2%; P=0.4691), hypertension (78.7% versus 79.2%; P=0.444), smoking (25.0% versus 26.9%; P=0.227), or dyslipidemia (48.4% versus 45.6%; P=0.254).
Comparison of Stroke and TIA Subjects
Subjects with stroke (N=1450) were more likely than those with TIA (N=397) to be men (55% versus 45%; P=0.002), younger (median age, 64 versus 68 years; P=0.003), smokers (27.6% versus 15.1%; P<0.0001) and have a mRS ≥3 at 3 months (23.0% versus 14.2%; P=0.0001). Subjects with TIA were more likely to have a history of stroke/TIA (30.3% versus 21.2%; P=0.0003; Table 1).
Proportional Frequency and Frequency of Newly Identified Depression and Antidepressant Use in Subjects With Stroke and TIA
The proportional frequency of depression was similar in subjects with stroke and TIA at 3 months (17.9% versus 14.3%; P=0.09) and at 12 months (16.4% versus 12.8%; P=0.08), even after adjustment for disability, history of stroke/TIA, history of smoking, and baseline demographic variables (Table 2). After excluding subjects with previous stroke/TIA in both groups, the proportional frequency of depression was higher after stroke compared with TIA at 3 months (17.4% versus 12.4%, respectively; P=0.038) but was not different at 12 months (15.9% versus 12.0%, respectively; P=0.100). The frequency of newly identified depression between 3 and 12 months was similar after stroke and TIA (8.7% versus 6.2%, respectively; P=0.12). When depression was redefined as PHQ-8 ≥10 and/or use of antidepressants, the proportional frequency of depression and the frequency of newly identified depression between 3 and 12 months (11.3% versus 7.8%, respectively; P=0.08) remained comparable (Table 2).
The proportion of antidepressant use at 3 months (13.9% versus 15.4%, respectively; P=0.45) and 12 months (16.0% versus 14.6%; P=0.50) and new antidepressant use between 3 and 12 months (5.1% versus 3.6%; P=0.237) were similar between stroke and TIA (Table 2).
Longitudinal Follow-Up of Depression and Antidepressant Medications in Subjects With Stroke and TIA
The pattern of depression and antidepressant use was similar after stroke and TIA (online-only Data Supplemental Figures I–II). The proportional frequency of depression over the year after stroke and TIA was constant due to a similar proportion of subjects who had resolving depression and subjects who had new-onset depression between 3 and 12 months (resolving versus new onset: 7.1% versus 8.7% in stroke; 6.8% versus 6.2% in TIA). Of the 134 subjects with stroke and 30 subjects with TIA who were persistently depressed (PHQ-8 ≥10 at both 3 and 12 months), 67.9% of subjects with stroke and 70.0% of subjects with TIA were not using antidepressants at either time point (P=0.920). Only a minority of subjects with stroke (N=28 [1.9%]) and TIA (6 [1.5%]) had persistent depression despite antidepressant use. Of the 205 subjects with stroke and 43 with TIA who were depressed and not using antidepressants at 3 months, 112 (54.6%) and 22 subjects (51.2%), respectively, continued to have depressive symptoms and/or were treated with antidepressants at 12 months, whereas the rest had spontaneous resolution of depressive symptoms. Of those subjects who did not have spontaneous resolution of depressive symptoms, only 18.7% of subjects with stroke and 4.5% of TIA (P=0.04) were using antidepressants at 12 months, suggesting a low rate of depression recognition and treatment. Among those using antidepressants at both 3 and 12 months (N=168 with stroke versus 46 with TIA), 62.5% of subjects with stroke and 65.2% of subjects with TIA were not depressed at either time point (P=0.862).
Factors Associated With Lack of Antidepressant Medication Use in Subjects With Stroke and Persistent Depression
Among subjects with stroke who were persistently depressed and not using antidepressants (N=91), there was no association between the lack of antidepressant use and age (P=0.516), sex (P=0.092), race (P=0.274), living situation (P=0.612), mRS (P=0.215), medication insurance status (P=0.894), or interim follow-up with a primary care physician (P=0.363) or a neurologist (P=0.844) when compared with those who were depressed and untreated at 3 months but using antidepressants by 12 months (N=21).
Factors Associated With Depression at 12 Months and Persistent Depression in Subjects With Stroke
Univariate analyses showed that depression at 12 months (N=238) in subjects with stroke was associated with younger age (P<0.0001), female sex (P=0.031), nonwhite race (P=0.031), inability to work (P<0.0001), higher mRS at 3 months (P<0.0001), using an antidepressant at 3 months (P=0.004), and being depressed at 3 months (P<0.0001). There was no association with a history of coronary artery disease/myocardial infarction (P=0.139) or living situation at 3 months (P=0.292). Persistent depression (N=134) was associated with the same variables. Multivariate analysis confirmed that younger age, a higher mRS at 3 months, and inability to work at 3 months were associated with depression at 12 months and with persistent depression (Table 3). In addition, depression at 3 months was associated with depression at 12 months. Both models had good discriminate validity (C=0.82 and C=0.81, respectively).
In this large longitudinal study, the proportional frequency and the frequency of newly identified depression between 3 and 12 months posthospitalization were similar after stroke and TIA when assessed with the PHQ-8 scale or by the PHQ-8 and/or use of antidepressants. The frequency of poststroke depression varies between published studies depending on the setting, patient population, diagnostic criteria, time period, and inclusion and exclusion criteria. According to a meta-analysis published in 2005, the pooled frequency of depressive symptoms among stroke survivors at any time during follow-up was 33% (95% CI, 29%–36%). The pooled frequencies varied when grouped by method of mood assessment; it was lowest in studies that used a single simple question to determine depression status (14%; 95% CI, 14%–15%) and highest when using standardized questionnaires such as the Montgomery Asberg depression rating scale (41%; 95% CI, 23%–60%).2 None of the included studies used the PHQ-8 scale.
Consistent with previous reports, younger age, poor functional outcome, depression, and inability to work at 3 months were associated with poststroke depression at 12 months.6,20 We further found that the same variables were associated with persistent depression. The similar frequency of depression after stroke and TIA could be explained by the nature of the stroke and TIA cohorts included in the current study. The majority of subjects with stroke had only mild disability (mRS <3), and approximately 15% of subjects with TIA had severe disability (mRS ≥3), presumably related to previous or intercurrent stroke, or other comorbidities; 30.3% of subjects with TIA had also a history of stroke/TIA. However, the frequency of depression remained comparable between the 2 groups even after adjusting for disability, history of stroke/TIA, and baseline demographic factors. After excluding subjects with a history of stroke/TIA, the frequency of depression was higher after stroke compared with TIA at 3 months indicating that prior cerebrovascular events partly explained the high frequency of depression after TIA.8,9 At 12 months, the frequency of depression after excluding previous cerebrovascular events was similar after stroke and TIA, suggesting that other factors play a more important role in late (12 months) versus early depression (3 months).
Consistent with previous reports, these results also indicate that the risk of depression after even mild stroke or TIA was higher than the general population with a comparable age distribution.8 For example, the Behavioral Risk Factor Surveillance Survey data from 2006 and 2008 indicated that the overall frequency of “current” depression in the general population using the PHQ-8 diagnostic algorithm was 9.0% (95% CI, 8.7–9.3); the rates were 10.0% (95% CI, 9.5–10.5) in the 45- to 65-year-old population and 6.8% (95% CI, 6.4–7.2) in those ≥65 years old.21
The longitudinal data from AVAIL demonstrate, similar to previous studies, that the frequency of depression is constant in the year after stroke and TIA, which argues against the hypothesis that the increased risk of depression is limited to the first few months after a cerebrovascular event.2 In light of the higher frequency of depression at 3 and 12 months, screening for depression appears to be warranted in the first year after stroke or TIA at both time points. Although depression at 3 months was a strong predictor of depression at 12 months, it had resolved by 12 months in approximately half of the subjects without the use of antidepressants. This was possibly due to the self-limited nature of depressive symptoms, a lower level of depression severity at baseline, or antidepressant use in the interim that was stopped before the 12-month follow-up. However, the frequency of newly identified depression between the 3- and 12-month follow-up was also significant. Subjects with stroke who are young, unable to work, and have significant disability may need more vigilance because of their higher risk for long-term and persistent depression.
Various hypotheses regarding the etiology of poststroke depression include the relative contribution of brain injury related to stroke, reaction to new functional disability and overall health status, including vascular comorbidities. In the current study, we were not able to correlate depression with stroke location or volume, because radiographic data were not readily available. Previous studies found a comparable incidence of depression among patients with stroke and those with similar vascular comorbidities (eg, myocardial infarction) but a higher prevalence of depression among patients with stroke compared with those with orthopedic disease with the same level of physical disability and suggested a “vascular etiology” of depression.5,22,23 Our finding that the frequency of depression was similar between subjects with stroke and TIA is also consistent with this hypothesis. It is possible, however, that contributors to depression may differ between patients with stroke and TIA. For example, in our study, functional disability and inability to work were independently associated with depression among stroke subjects. Because TIA is not associated with residual functional impairment, these factors would not be anticipated to play a significant role in the TIA cohort except in those with previous or intercurrent stroke. Additional studies are needed to investigate determinants of depression among patients with TIA.
An important finding from the current study is that undertreatment of depression was common in subjects with either stroke or TIA, because more than two thirds of those with persistent depression were not appropriately treated with antidepressants. Undertreatment was not associated with any of the study's prespecified variables; however, this conclusion is limited by the small sample size of subjects for this outcome and warrants replication. Nonetheless, antidepressant use was similar for both subjects with stroke and those with TIA and higher than the reported rates in the general population (approximately 10%).24,25 The majority of subjects using antidepressants did not have symptoms of depression, which implies that depression in subjects with stroke and TIA is amenable to treatment or, alternatively, that antidepressants are used nonselectively in this population.20 In addition, we cannot exclude the possibility that some of the antidepressant drugs were being used for other purposes besides depression.
The strengths of this study were the large cohort with complete follow-up that prospectively and longitudinally evaluated both depressive symptoms and antidepressant use in both subjects with stroke and those with TIA. AVAIL subjects were recruited from geographically diverse hospitals participating in the American Heart Association GWTG–Stroke program. GWTG–Stroke hospitals do tend to be larger, urban and teaching centers. Despite these differences, a study of the representativeness of the GWTG–Stroke registry indicates that the data are generally representative of national fee-for-service Medicare ischemic stroke populations providing support for external validity.26 However, there are several limitations. Potential inception cohort bias is always a concern when consent is required. The demographics of the AVAIL population, however, did not differ substantially from the overall GWTG–Stroke cohort.13,27,28 Overall 36% of AVAIL subjects were excluded from the present analysis (Figure). The baseline data of subjects who were excluded from this analysis do differ in several ways. Because subjects rather than proxies were required to complete the PHQ-8, the results are also likely biased toward those with mild disability including those without aphasia or significant cognitive impairments. These differences need to be considered and limit the generalizability of the results. Despite these limitations, this study reveals that, even in patients with minor disability, the frequency of depression was high after stroke or TIA. Some of the subjects with “depression” in this study may not have needed treatment with antidepressants, because a PHQ-8 score of ≥10 reflects “current depression,” which includes major depression, other depressive disorders, depressed mood, or anhedonia.15 However, it is very likely that those who were characterized as persistently depressed may have had clinically significant depression that warranted treatment. Other nonpharmacological treatments for depression such as cognitive behavioral therapy were not captured in this study, thus possibly overestimating the frequency of “undertreatment.”
Antidepressant use was self-reported based on structured interviews and not confirmed by alternative methods; however, this method of medication assessment has good concordance with claims records and is likely to be valid for assessing current antidepressant use for depression therapy.29 Prestroke depression, shown to be strongly associated with poststroke depression, was not assessed in our study.30 Given this limitation, we cannot exclude the possibility that a proportion of subjects in the cohort had a history of depression. However, we were able to determine the frequency of newly identified depression between 3 and 12 months by using the 3-month assessment as the baseline. All participants were discharged from hospitals for a stroke or TIA, limiting the cerebrovascular event to a subset of events that were serious enough to warrant hospitalization. Lastly, a baseline assessment of stroke severity, as determined by National Institutes of Health Stroke Scale, was missing in a large proportion of GWTG–Stroke subjects and therefore was not included in our analyses. Instead we used the mRS as an indication of functional outcome, which correlates well with stroke severity but can be confounded by depression.
Both subjects with stroke and those with TIA had a similarly high proportional frequency and frequency of newly identified depression between 3 and 12 months after hospitalization, but a high proportion of persistent depression in these 2 groups was not appropriately treated. Systematic evaluation for depression in patients with stroke or TIA may improve detection and treatment of this condition.
Sources of Funding
Dr El Husseini was fully supported and Dr Goldstein and Dr Laskowitz were partially supported by an American Stroke Association-Bugher Foundation Stroke Prevention Research Center award. The AVAIL project was supported by unrestricted funds from Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership and conducted through collaboration with the American Heart Association and the “Get With The Guidelines–Stroke” program. AVAIL analyses were also supported in part by the Agency for Healthcare Research and Quality cooperative agreement U18HS016964.
Dr Laskowitz serves as an associate editor for the Critical Care research and Practice from 2009 to the present (no compensation). He has obtained funding from (National Institutes of Health) 1D43-TW008308-01 “Training for the prevention and treatment of stroke.” Dr Goldstein serves as a consultant for Allergan, Pfizer, Boheringer-Ingelheim, Merck, and BMS. He also served on the Neurology Executive Committee RESPECT Trial (Patent foramen ovale closure, AGA Medical Corporation), Site Oversight Committee, Asymptomatic Carotid Trial (Abbott), and Steering Committee for the Stroke Prevention by Aggressive Reduction in Cholesterol Levels (SPARCL) Trial (Pfizer). Dr Williams received funds for other research from the Agency for Healthcare Research and Quality, Veterans' Affair, and the Foundation for Informed Medical Decision Making, Department of Defense (FIMDM). The Veterans Affairs and are on mental health topics including depression but not directly related to the article. Dr Bushnell was a co-Principal Investigator of AVAIL and received research salary support from Bristol Myers Squibb-Sanofi Joint Partnership for her role, but the funding ended in 2009.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.643130/-/DC1.
- Received November 3, 2011.
- Revision received January 17, 2012.
- Accepted February 1, 2012.
- © 2012 American Heart Association, Inc.
- Hackett ML,
- Yapa C,
- Parag V,
- Anderson CS
- House A,
- Dennis M,
- Mogridge L,
- Warlow C,
- Hawton K,
- Jones L
- Wade DT,
- Legh-Smith J,
- Hewer RA
- Hackett ML,
- Anderson CS
- Luijendijk HJ,
- Stricker BH,
- Wieberdink RG,
- Koudstaal PJ,
- Hofman A,
- Breteler MM,
- et al
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, IV ed. Washington, DC: American Psychiatric Association; 1994.
- Van Swieten JC,
- Koudstaal PJ,
- Visser MC,
- Schouten HJ,
- Van Gijn J
- Uyttenboogaart M,
- Stewart RE,
- Vroomen PC,
- De Keyser J,
- Luijckx GJ
- Eriksson M,
- Asplund K,
- Glader EL,
- Norrving B,
- Stegmayr B,
- Terent A,
- et al
- Folstein MF,
- Maiberger R,
- McHugh PR
- Aben I,
- Verhey F,
- Strik J,
- Lousberg R,
- Lodder J,
- Honig A
- Stagnitti MN
- Reeves MJ,
- Fonarow GC,
- Smith EE,
- Pan W,
- Olson D,
- Hernandez AF,
- et al
- Fonarow GC,
- Reeves MJ,
- Smith EE,
- Saver JL,
- Zhao X,
- Olson DW,
- et al