Michellereinlieb.files.wordpress.com

Change in Cognitive Functioning Following
Acute Antidepressant Treatment in
Late-Life Depression
Michelle E. Culang, B.S., Joel R. Sneed, Ph.D., John G. Keilp, Ph.D.,
Bret R. Rutherford, M.D., Gregory H. Pelton, M.D.,
D. P. Devanand, M.D., Steven P. Roose, M.D.
Objective: Selective Serotonin Reuptake Inhibitors (SSRIs) are the most commonly
prescribed medications for geriatric depression. The association of late-life depression
and cognitive impairment has been well documented. However, there have been few placebo-controlled trials examining the impact of SSRIs on cognitive functioning. De-
sign: Prepost neuropsychological (NP) data collected as part of an 8-week, double-blind,
placebo-controlled trial of citalopram in depressed patients aged 75 years and older were
used to examine change in cognitive functioning. Setting: University-affiliated outpa-
tient psychiatry clinics. Participants: One hundred seventy-four community-dwelling
men and women aged 75 years or older with nonpsychotic unipolar depression. Mea-
surements: NP assessments included mental status (Mini-Mental State Examination),
psychomotor speed (Wechsler Adult Intelligence Scale-III Digit Symbol Subtest), reaction
time (Choice Reaction Time), visual-spatial skill (Judgment of Line Orientation), execu- tive functioning (Stroop Color/Word Test), and memory (Buschke Selective Reminding Test). Results: Differences in the pattern of change by treatment group depended on
responder status. Citalopram nonresponders were the only group to decline on verbal
learning and psychomotor speed. Citalopram responders showed significant improve- ment in visuospatial functioning, when compared with nonresponders in either condi- tion, but their improvement was not greater than responders on placebo. Citalopram responders showed greater improvement on psychomotor speed than citalopram nonre- sponders, but their improvement was not greater than placebo responders or nonre- sponders. Conclusions: Medication may have a deleterious effect on some aspects of
cognition among patients aged 75 years and older who have not responded. This suggests
that patients should not be maintained on a medication if they have not had an adequate response. (Am J Geriatr Psychiatry 2009; 17:881–888) Key Words: Cognitive functioning, cognitive impairment, geriatric depression, late-life
depression, citalopram
Received November 3, 2008; revised May 4, 2009; accepted June 1, 2009. From the Department of Psychology, Queens College, City Universityof New York (MC, JRS); and the Department of Geriatric Psychiatry, Columbia University and the New York State Psychiatric Institute (MC, JRS,JGK, BRR, GHP, DPD, SPR), NY. Send correspondence and reprint requests to Michelle Culang, B.S., Department of Psychology, Queens Collegeof the City University of New York, 65-30 Kissena Blvd, Flushing, NY 11367. e-mail: [email protected] 2009 American Association for Geriatric Psychiatry Am J Geriatr Psychiatry 17:10, October 2009 SelectiveSerotoninReuptakeInhibitors(SSRIs)are been few placebo-controlled trials examining this is- the first-line of treatment in the geriatric de- sue. In one study, nortriptyline and phenelzine pro- pressed due to the efficacy, safety, and tolerability of duced no change in cognition in depressed older its class. Cognitive impairment is common in late-life adults, when compared with placebo, and this effect depression (LLD), particularly in memory,1–3 vi- did not depend on responder status.8 However, the suospatial functioning,2,4 information processing small sample size and limited number of responders speed,5,6 and executive functioning.5,7 It is important made it difficult to determine the impact of responder to consider the impact that antidepressant treatment status on change in cognition. Furthermore, this study can have on cognition when treating depressed older was restricted to the use of a monoamine oxidase in- hibitor and tricyclic antidepressant. In another study, Research that has examined the impact of SSRIs on patients taking duloxetine showed significant improve- the cognitive functioning of depressed older adults ment in verbal learning and memory, when compared has been inconclusive because most studies have with the placebo group.15 Therefore, it is unclear what been limited by methodological constraints includ- impact medication, SSRIs in particular, has on cogni- ing small sample size or lack of an age-matched control group for comparison.8–11 For instance, treat- The purpose of this study was to examine the ment of LLD with sertraline led to an improvement impact of antidepressant treatment on change in cog- in short- and long-term memory storage and re- nitive functioning. To accomplish this aim, we used trieval and speed of processing.11 Although these neuropsychological (NP) data collected as part of the results suggest that some aspects of cognition (i.e., Old-Old Depression Study,16 a large (N ϭ 174), ran- memory and processing speed) improve with anti- domized, double-blind, placebo-controlled trial of depressant treatment, it is difficult to determine citalopram in depressed older adults (age Ͼ75 whether the improvement was a function of repeat years). These data provided us with the methodolog- testing or medication because the design lacked a ical strength to address two questions: 1) do patients treated with citalopram show differential change in Studies using age-matched controls have shown cognitive functioning over the 8 weeks, when com- that cognitive functioning of depressed older adults pared with patients treated with placebo? 2) Does does not improve beyond the expected practice effect change in cognitive performance depend on re- (practice effects refer to improvement due to repeat sponder status? To our knowledge, this is the first testing and are defined by the performance of a attempt to approach these issues using a placebo- comparison condition, either an age-matched control controlled trial of an SSRI in an old-old (Ͼ75 years) group12–14 or a placebo comparison group,8,15 that is not being treated with medication).12–14 For example,working and episodic memory, attention shifting,and processing speed did not improve after treat-ment with paroxetine to a greater degree than nor- mal controls did with practice, regardless of re-sponder status.14 Similarly, cognitive functioning The procedures used in the multisite, randomized, showed no improvement beyond a practice effect placebo-controlled trial (RCT) have been previously among responders to either nortriptyline or parox- described.16–18 Briefly, 174 community-dwelling men etine.13 These studies suggest that depressed older and women aged 75 years or older meeting Diagnos- adults show little improvement as a function of treat- tic and Statistical Manual of Mental Disorders, ment and cognitive impairment persists after an ad- Fourth Edition criteria (based on Structured Clinical equate trial of antidepressant medication.
Interview for Diagnostic and Statistical Manual of Although such designs allow us to determine Mental Disorders-III-R interview) for nonpsychotic whether cognition changes as a function of antide- unipolar depression (single or recurrent) with a base- pressant treatment, it does not allow us to conclude line 24-item Hamilton Rating Scale for Depression that the change (if any) is a result of treatment due to (HRSD) score Ն20 participated in this 8-week RCT.
a lack of a placebo condition. However, there have All patients began the trial with a 1-week single- Am J Geriatr Psychiatry 17:10, October 2009 blind placebo lead-in with the baseline visit con- rules26,28 to generate valid statistical inferences that ducted at the end of the lead-in. Patients were ran- reflect uncertainty due to missing values and im- domized to citalopram 20 mg/day or matched placebo only if they continued to meet inclusion andexclusion criteria at the end of the placebo lead-in. At Statistical Analyses
the end of Week 4, patients with a HRSD score Ͼ10had the dose increased to two pills per day, i.e., 40 Before testing for differences in change in NP test mg of citalopram or 2 placebo pills.
performance, we used the PROC REG and PROCLOGISTIC procedures in SAS to test for differences NP Test Battery
at baseline between the two treatment conditionsand the four treatment condition by responder status The test battery was designed to assess a number groups (see later). There were no differences on age, of cognitive functions pertinent to aging and major education, gender, baseline depression severity, re- depression including mental status, psychomotor sponder status, or on any of the NP tests with the speed, reaction time, visual-spatial skill, attention, exception of baseline scores on the MMSE and Digit and memory. Three of the tests (Choice Reaction Symbol subtest of the Wechsler Adult Intelligence Time [CRT], Judgment of Line Orientation [JOLO], Scale-III. Therefore, we adjusted for baseline Digit and Stroop) were presented on a Macintosh laptop Symbol and MMSE scores in the two treatment computer and were written in the PsyScope pro- group analyses. When comparing the four patient gramming language,19 whereas the other three tests groups (responder status by treatment condition), (Mini-Mental State Exam, Buschke Selective Remind- we found differences on education and baseline ing Test [SRT], and Digit Symbol) were administered MMSE, CRT, and Buschke SRT scores. Therefore, we by hand. The tests included the 30-item Folstein included these variables as covariates in the four Mini-Mental State Examination (MMSE)20 to esti- group analyses. We also adjusted for site of study in mate global cognitive functioning, the Wechsler all analyses, which we know from previous reports Adult Intelligence Scale-III Digit Symbol Subtest21 as a measure of psychomotor speed, the CRT test To test for differences in change in NP test perfor- adapted from Thorne et al.,22 the Stroop Color/Word mance, we used data from the multiply imputed Test23 to assess the response inhibition component of datasets and adopted a partial or regressed change executive functioning, the JOLO24 as a measure of approach to analyzing two time-point data29 using spatial judgment, and the Buschke SRT25 as a mea- the PROC REG procedure in SAS. According to this approach, the endpoint NP test score is treated as theoutcome variable, and the baseline test score istreated as a covariate. This effectively removes all Missing Data
correlation of the endpoint score from the baseline Missing data at baseline ranged from 0.6% on the score and represents an improvement over simple MMSE and Buschke SRT to 9.8% on the JOLO and change scores (subtracting baseline from endpoint), from 11.5% on the MMSE to 19.0% on the Stroop at which tend to overcorrect the endpoint score by the follow-up. To accommodate missing data, we used baseline score due to unreliability of measurement.29 multiple imputation using the PROC MI and We first tested for differences in endpoint scores MIANALYZE procedures in SAS. Multiple imputa- between treatment conditions using a dummy coded tion is a simulation technique that replaces each (citalopram ϭ 1 and placebo ϭ 0) variable. To test missing datum with a set of m Ͼ1 plausible values.26 whether change in NP test performance depends on This report is based on five imputed datasets (m ϭ 5), responder status (50% reduction from baseline which is sufficient to obtain excellent results unless HRSD score), we again used a dummy coded vari- rates of missing data are exceptionally high.27 The able to designate the four patient groups (citalopram imputed datasets are analyzed using standard statis- responders, citalopram nonresponders, placebo re- tical analyses, and results from the analyses from the sponders, and placebo nonresponders). Each covari- m complete datasets are combined using Rubin’s ate was centered at its respective mean, so the inter- Am J Geriatr Psychiatry 17:10, October 2009 cept corresponded to the mean of the reference cline. As can be seen in Fig. 1, the citalopram group group at endpoint and the unstandardized regres- improved on some tests and declined on others.
sion weights reflected the difference between thegroups included in the model and the reference Hypothesis Testing
group (excluded from the model). All significancetests were evaluated at the 5% level.
Table 2 shows the unadjusted means and standard deviations for all NP tests both pre and posttreat-ment for the citalopram and placebo groups and the four patient groups (treatment group by responderstatus). Adjusting for site and baseline MMSE andDigit Symbol, there was a statistically significant dif- Descriptive Statistics
ference between the placebo and citalopram condi- Table 1 presents baseline demographic and clinical tions at endpoint on the Buschke SRT. Specifically, characteristics of the total sample, placebo and cita- patients treated with citalopram scored lower at end- lopram groups, and the four groups of patients clas- point than patients treated with placebo (B ϭ Ϫ2.74, sified by treatment condition and responder status.
SE ϭ 1.41, t[1,087] ϭ Ϫ1.94, 95% CI: Ϫ5.52 to 0.03, The average study participant was 79.57 years and completed about 2 years of college. Approximately We next compared the four groups of patients 58% of the sample were women, average baseline classified according to treatment condition and re- depression severity was 24.32 on the 24-item HRSD, sponder status on endpoint NP test performance. As and 40% of the sample was classified as responders.
can be seen in Table 2, citalopram responders scored The average MMSE score of the sample at baseline significantly higher than both citalopram nonre- was 27.99, and 6.9% had a score of 24 or below.
sponders (B ϭ Ϫ2.54, SE ϭ 0.97, t[80.46] ϭ Ϫ2.54, To facilitate interpretation of the pattern of change 95% CI: Ϫ4.38 to Ϫ0.53, p ϭ 0.01) and placebo non- in NP test performance as a function of treatment responders (B ϭ Ϫ2.47, SE ϭ 0.89, t(217.31) ϭ Ϫ2.77, group, all NP scores were converted to z-scores 95% CI: Ϫ4.23 to Ϫ0.71, p ϭ 0.01) on the JOLO at based on mean values at baseline in the total sample.
endpoint. However, citalopram responders were not As can be seen in Fig. 1, NP test performance im- statistically significantly different than placebo re- proved on each test in the placebo group, which is sponders at endpoint (B ϭ Ϫ1.81, SE ϭ 1.02, t[111.56] ϭ consistent with a practice effect. For the purposes of Ϫ1.78, 95% CI: Ϫ3.83 to 0.20, p ϭ 0.08). Looking at this report, practice effects refer to improvement in endpoint performance on the Buschke SRT, citalo- performance due to repeat testing and are defined on pram nonresponders were the only group to decline the basis of the performance of the placebo group.
from pretest to posttest. Specifically, citalopram non- Differences from the placebo group (both positive and responders scored lower (3.64 points) than placebo negative) reflect deviations from a practice effect and nonresponders at study end (B ϭ Ϫ3.64, SE ϭ 1.83, represent either improvement beyond a practice effect t(472.15) ϭ Ϫ1.99, 95% CI: Ϫ7.23 to Ϫ0.05, p ϭ 0.05).
or the absence of a practice effect, possibly even de- Similarly, citalopram nonresponders were the only Baseline Clinical and Demographic Characteristics of the Total Sample, the Citalopram and Placebo Conditions, and
the Four Patients Groups Classified by Treatment Condition and Responder Status
Total Sample
Citalopram
Citalopram
Citalopram
Variable
(n ؍ 174)
Responders
Nonresponders
Responders
Nonresponders
Am J Geriatr Psychiatry 17:10, October 2009 FIGURE 1.
Change in Cognitive Performance From Pre to Posttreatment in the (a) Placebo Condition and (b) Citalopram
Condition Across Six Neuropsychological Tests
group to decline from pretest to posttest on the Digit improvement was not greater than placebo respond- Symbol. In particular, citalopram nonresponders ers or nonresponders. The findings indicate that the scored lower than citalopram responders at endpoint practice effect is impaired in some domains among (B ϭ Ϫ5.62, SE ϭ 2.65, t(233.31) ϭ Ϫ2.12, 95% CI: nonresponders on medication. Therefore, these find- ings suggest that patients should not be maintainedon a medication if they have not had an adequateresponse.
One possible explanation for the observed decline DISCUSSION
in verbal learning and psychomotor speed is that the This was the first study to examine the impact of overall level of cognitive functioning in the sample antidepressant treatment on change in cognitive was low and the citalopram group had a dispropor- functioning in depressed adults aged 75 years and tionately high number of cognitively impaired pa- older using data from an 8-week RCT. Although the tients. This might explain why there was inconsistent placebo group showed a distinct practice effect from improvement in the citalopram condition and why baseline to endpoint on all NP tests, the citalopram this study differs from previous placebo-controlled group improved on some tests but declined on oth- trials.8,15 However, the average MMSE score at base- ers. However, the pattern of change depended on line for the sample was 28, which is well within responder status. Specifically, citalopram nonre- normal limits. Moreover, although there was a sig- sponders were the only group to decline in perfor- nificant difference in MMSE scores between the treat- mance on verbal learning (Buschke SRT) and psy- ment groups at baseline, it was the citalopram group that scored higher at baseline than the placebo responders showed significant improvement in visuospatial functioning (JOLO), when compared Another possibility is that brain lesions, which are with nonresponders in either condition, but their associated with age30 and other risk factors such as improvement was not greater than responders on hypertension and diabetes,31 were disproportion- placebo. Similarly, citalopram responders showed ately represented in the citalopram condition. White greater improvement on psychomotor speed (Digit matter hyperintensities (WMH) that are characteris- Symbol) than citalopram nonresponders, but their tic of LLD may interrupt frontal-striatal pathways Am J Geriatr Psychiatry 17:10, October 2009 that mediate cognitive functions that are commonly impaired in LLD. Furthermore, cognitive impair- ment is associated with the presence of WMH in LLD and deficits worsen as the lesions become more se- vere.2,5 However, there were no statistically signifi- Nonresponders
cant differences in the percentage of patients in the citalopram group and the placebo group classified as having high lesion load, which was defined as a deep WMH rating of 2 or a subcortical gray matter rating of 3 on the Fazekas modified Coffey Rating Scale for Subsamples
The decline in verbal learning may be particularly Responders
attributed to the anticholinergic effects of SSRIs. SSRIs have unique nonserotonergic pharmacologic profiles that are associated with distinct effects on cognitive functioning.33 Paroxetine, for example, may causeimpairment in delayed verbal recall in healthy middle- aged adults and elderly subjects, whereas sertraline is Citalopram
associated with improvement in immediate and de- layed verbal recall and verbal fluency.34,35 Although Citalopram
Nonresponders
administration of citalopram is associated with im- provement in working memory in depressed adults36 and increased memory consolidation in healthy adults,37 Endpoint
it is still unclear what effect citalopram can have on cognitive functioning of the geriatric depressed, a pop- ulation that is especially vulnerable to the adverse ef- Baseline
Citalopram
Responders
The observed decline in verbal learning and psy- Responder
chomotor speed in the citalopram group is consistent with a recent report from an epidemiological study of elderly depressed patients examining the relation- ship among depressive symptoms, cognitive impair- ment, and antidepressant use.39 Findings revealed Condition
that baseline depression scores predicted future mild Performance
cognitive impairment (MCI) but only among those using antidepressant medications at baseline. Taken Treatment
together, these findings support the contention that nonresponders should not be maintained on medi-cation that may have a negative impact on some aspects of cognitive functioning, which may facilitate According
This study should be interpreted in the context of Neuropsychological
several limitations. First, there were statistically sig- Classified
Citalopram
nificant differences between the two treatment con- ditions and the four patient groups at baseline on Unadjusted
several NP tests. However, these differences were adjusted for in the statistical models by including those tests as covariates. Second, it may be possible that including a small number of MCI patients Am J Geriatr Psychiatry 17:10, October 2009 (MMSE Յ24) in this study (N ϭ 12) might have and psychomotor speed among patients who do not influenced our results. However, we ran the analyses respond on medication. Although responders on with and without this group of patients, and the medication may improve in some domains, their results were not different. Third, there was missing improvement does not exceed the expected practice data, as is typically the case in clinical trials, and we effect observed in patients randomized to placebo.
accommodated for missing data using multiple im- This raises the important clinical issue that, although putation, a far superior method compared with tra- two treatments may be equivalent with regard to ditional approaches using mean substitution or com- response, they may have differential effects on cog- plete case analysis. Fourth, a somewhat limited NP nitive functioning, especially in a cognitively vulner- battery was used. Only one aspect of executive func- able population. Our findings suggest that nonre- tioning (i.e., response inhibition) was evaluated, and sponders should not be maintained on medication no formal test of attention was included in the study.
that may have a negative effect on some aspects of However, these limitations are balanced by using data from the only randomized, placebo-controlledclinical trial of antidepressant treatment among de- This work was supported by a grant from Forest pressed patients aged 75 years or older. Moreover, Laboratories and National Institute of Mental Health unlike other studies, there were an approximately grants T32 MH20004 (to SPR) and K23 MH075006 (to equal number of responders in both treatment con- JRS) and from Eli Lilly and Novartis (to DPD). ditions, allowing for an adequate test of whether SPR has received consultant fees from Forest Labo- change in cognitive function across two treatments ratories, Wyeth Pharmaceuticals, Sanofi-Anventis, Pfizer, and Sierra Pharmaceuticals and DPD has received con- Our findings indicate that citalopram may inter- sultant fees from Glaxosmithkline, Acadia, and Sanofi- fere with the normal practice effect in verbal learning References
1. Salloway S, Malloy P, Kohn R, et al: MRI and neuropsychological 11. Doraiswamy PM, Krishnan KR, Oxman T, et al: Does antide- differences in early- and late-life-onset geriatric depression. Neu- pressant therapy improve cognition in elderly depressed pa- tients? J Gerontol A Biol Sci Med Sci 2003; 58:M1137–M1144 2. Kramer-Ginsberg E, Greenwald BS, Krishnan KRR, et al: Neuro- 12. Portella MJ, Marcos T, Rami L, et al: Residual cognitive impair- psychological functioning and mri signal hyperintensities in geri- ment in late-life depression after a 12-month period follow-up. Int atric depression. Am J Psychiatry 1999; 156:438 – 444 J Geriatric Psychiatry 2003; 18:571–576 3. Gallassi R, Di Sarro R, Morreale A, et al: Memory impairment in 13. Butters MA, Becker JT, Nebes RD, et al: Changes in cognitive patients with late-onset major depression: the effect of antide- functioning following treatment of late-life depression. Am J Psy- pressant therapy. J Affect Disord 2006; 91:243–250 4. Butters MA, Whyte EM, Nebes RD, et al: The nature and determi- 14. Nebes RD, Pollock BG, Houck PR, et al: Persistence of cognitive nants of neuropsychological functioning in late-life depression.
impairment in geriatric patients following antidepressant treat- ment: a randomized, double-blind clinical trial with nortriptyline 5. Lesser I, Boone K, Mehringer C, et al: Cognition and white matter and paroxetine. J Psychiatr Res 2003; 37:99 –108 hyperintensities in older depressed patients. Am J Psychiatry 15. Raskin J, Wiltse CG, Siegal A, et al: Efficacy of duloxetine on cognition, depression, and pain in elderly patients with major 6. Nebes RD, Butters MA, Mulsant BH, et al: Decreased working depressive disorder: an 8-week, double-blind, placebo-controlled memory and processing speed mediate cognitive impairment in trial. Am J Psychiatry 2007; 164:900 –909 geriatric depression Psychol Med 2000; 30:679 – 691 16. Roose SP, Sackeim HA, Krishnan KRR, et al: Antidepressant phar- 7. Lockwood KA, Alexopoulos GS, van Gorp WG: Executive dys- macotherapy in the treatment of depression in the very old: a function in geriatric depression.[see comment]. Am J Psychiatry randomized, placebo-controlled trial. Am J Psychiatry 2004; 161: 8. Georgotas A, McCue RE, Reisberg B, et al: The effects of mood 17. Sneed JR, Roose SP, Keilp JG, et al: Response inhibition predicts changes and antidepressants on the cognitive capacity of el- poor antidepressant treatment response in very old depressed derly depressed patients. Int Psychogeriatr 1989; 1:135–143 patients. Am J Geriatr Psychiatry 2007; 15:553–563 9. Nebes RD, Pollock BG, Mulsant BH, et al: Cognitive effects of 18. Sneed JR, Keilp JG, Brickman AM, et al: The specificity of neuro- paroxetine in older depressed patients. J Clin Psychiatry 1999; psychological impairment in predicting antidepressant non-re- sponse in the very old depressed. Int J Geriatr Psychiatry 2008; 10. Bondareff W, Alpert M, Friedhoff AJ, et al: Comparison of sertra- line and nortriptyline in the treatment of major depressive disor- 19. Cohen JD, MacWhinney B, Flatt M, et al: PsyScope: an interactive der in late life. Am J Psychiatry 2000; 157:729 –736 graphic system for designing and controlling experiments in the Am J Geriatr Psychiatry 17:10, October 2009 psychology laboratory using Macintosh computers. Behav Res 31. Murray AD, Staff RT, Shenkin SD, et al: Brain white matter hyper- Methods Instruments Comput 1993; 25:257–271 intensities: relative importance of vascular risk factors in nonde- 20. Folstein MF, Folstein SE, McHugh PR: Mini-mental state: a practi- mented elderly people. Radiology 2005; 237:251–257 cal method for grading the state of patients for the clinician.
32. Krishnan KR, Hays JC, Blazer DG: MRI-defined vascular depres- sion. Am J Psychiatry 1997; 154:497–501 21. Wechsler D: Wechsler Adult Intelligence Scale 3rd Revision 33. Chew ML, Mulsant BH, Pollock BG, et al: Anticholinergic activity (WAIS-III). San Antonio, Psychological Corporation, 1997 of 107 medications commonly used by older adults. J Am Geriatr 22. Thorne DR, Genser SG, Sing HC, et al: The Walter reed perfor- mance assessment battery. Neurobehav Toxicol Teratol 1998; 34. Schmitt JAJ, Kruizinga MJ, Riedel WJ: Non-serotonergic pharma- cological profiles and associated cognitive effects of serotonin 23. MacLeod C: A half-century of research on the Stroop effect: an reuptake inhibitors. J Psychopharmacol 2001; 15:173–179 integrative review. Psychol Bull 1991; 109:163–203 35. Furlan PM, Kallan MJ, Ten Have T, et al: Cognitive and psychomo- 24. Benton AL, Sivan AB, Hamsher K, et al: Contributions to Neuro- tor effects of paroxetine and sertraline on healthy elderly volun- psychological Assessment. New York, Oxford, 1983 teers. Geriatr Psychiatry 2001; 9:429 – 438 25. Buschke H, Fuld P: Evaluating storage, retention, and retrieval in 36. Zobel AW, Schulze-Rauschenbach S, von Widdern OC, et al: disordered memory and learning. Neurology 1974; 24:1019 –1025 Improvement of working but not declarative memory is corre- 26. Schafer JL, Olsen MK: Multiple imputation for multivariate miss- lated with HPA normalization during antidepressant treatment.
ing-data problems: a data analyst’s perspective. Multivariate Be- 37. Harmer CJ, Bhagwagar Z, Cowen PJ, et al: Acute administration of 27. Schafer JL: Multiple imputation: a primer. Stat Methods Med Res citalopram facilitates memory consolidation in healthy volun- teers. Psychopharmacology 2002; 163:106 –110 28. Schafer JL, Graham JW: Missing data: our view of the state of the 38. Baldwin D, Johnson FN: Tolerability and safety of citalopram. Rev 29. Cohen J, Cohen P, West SG, et al: Applied Multiple Regression/ 39. Ravaglia G, Forti P, Lucicesare A, et al: Prevalent depressive Correlation for the Behavioral Sciences. 3rd ed. Mahwah, NJ, symptoms as a risk factor for conversion to mild cognitive im- Lawrence Erlbaum Associates Publishers, 2003 pairment in an elderly Italian cohort. Am J Geriatr Psychiatry 30. Taylor WD, MacFall JR, Steffens DC, et al: Localization of age-associated white matter hyperintensities in late-life depres- 40. Devanand DP, Pelton GH, Marston K, et al: Sertraline treatment of sion. Prog Neuropsychopharmacol Biol Psychiatry 2003; 27: elderly patients with depression and cognitive impairment. Int J Am J Geriatr Psychiatry 17:10, October 2009

Source: http://michellereinlieb.files.wordpress.com/2011/10/culang-et-al-2009-ajgp.pdf