Suicidal Thinking and Behavior DuringTreatment With Sertraline in Late-Life

Depression

J. Craig Nelson, M.D., Kevin Delucchi, Ph.D., Lon Schneider, M.D.

Objective: To determine effects of sertraline on suicidal thinking and behavior inpatients with late-life major depression. Methods: This was a secondary analysisof an eight-week, placebo-controlled, double-blind randomized trial of a multi-center trial at 66 clinical sites. Outpatients �60 years of age with major depres-sion and a Hamilton Depression Rating (HAMD) score �18 were included.Intervention was sertraline 50–100 mg/day or placebo for eight weeks. Measure-ments were 17-item HAMD administered at baseline and two-week intervals.HAMD Item 3 used to assess suicidal ideation (SI) and behavior. Reports ofserious adverse events (SAEs), adverse events (AEs) leading to discontinuation,and spontaneously reported adverse events reviewed. Results: A total of 747patients received at least one dose of medication, and 728 had at least oneposttreatment assessment. Mean age (�SD) was 69.8 � 6.7 years (range: 59–97years) and 56% were female. There were no completed suicides or suicide at-tempts during the double-blind trial. One SAE, hospitalization, was associatedwith SI in a patient on sertraline. No other AEs were associated with SI orbehavior. HAMD Item 3 ratings progressively declined during the trial withsignificantly lower values for sertraline than placebo (Z�2.41, p �0.02). In 248patients with HAMD Item 3 of zero at baseline, the percentage of patients whoseItem 3 ratings increased during treatment did not differ in the two groups (22.4%versus 25.8% for sertraline and placebo, respectively.) Conclusion: Sertraline wasassociated with significantly lower HAMD Item 3 scores than placebo duringtreatment. There was no evidence of greater emergent suicidal thinking orbehavior with sertraline than placebo. (Am J Geriatr Psychiatry 2007; 15:573–580)

The issue of suicide in adolescents and childrenhas recently received considerable attention.

Concerns that antidepressant medications increasesuicidal thinking during treatment, led the U.S. Food

and Drug Administration (FDA) to include a black-box warning about this potential risk.1,2 Althoughthis risk in children and adolescents has been themajor focus of attention, the rate of completed sui-

Received October 31, 2006; revised February 9, 2007; accepted February 19, 2007. From the Department of Psychiatry, University of CaliforniaSan Francisco, San Franciso, CA (JCN, KD); and the Department of Psychiatry and Behavioral Sciences, Keck School of Medicine and the LeonardDavis School of Gerontology, University of Southern California, Los Angeles, CA (LS). Send correspondence and reprint requests to J. CraigNelson, M.D., 401 Parnassus Ave, Box 0984-F, San Francisco, CA 94143. e-mail: craign@lppi.ucsf.edu.

© 2007 American Association for Geriatric Psychiatry

Am J Geriatr Psychiatry 15:7, July 2007 573

cide remains highest in older adults. For example, inthe United States in 2003, the suicide rate in adults 65years and older was 14.61 per 100,000 versus 10.29per 100,000 among individuals less than 65 years ofage.3 In the oldest group reported, those 85 andolder, the rate rose to 16.93. The higher rates in olderadults were accounted for by elevated rates in men.In women the rate of suicide per 100,000 was 4.33 inthose under 65 years of age and 3.79 in women 65years and older. In men, however, rates per 100,000were 16.19 in those under 65 years, 29.79 in those 65and older, and 47.75 in men 85 and older. The rate inmen 85 years and older was 3 times the rate inyounger men and more than 10 times the rate inwomen.

The effects of antidepressants on suicidal think-ing and behavior in older adults have not receivedmuch attention. The Prevention of Suicide in Pri-mary Care Elderly: Collaborative Trial (PROS-PECT) study in older adults with depression haspreviously demonstrated that both “usual care”and an intervention that included treatment guide-lines and a care manager, were both associatedwith a decline in suicidal thinking over a 12-monthtreatment period in a primary care setting.4 Szantoand colleagues also reported declining suicidalthinking during acute treatment of depressionwith nortriptyline or paroxetine.5 They noted sim-ilar rates of emergence of suicidal thinking orthoughts of death, 7.8% or 8.0%, with nortriptylineand paroxetine respectively. Neither of these stud-ies, however, was placebo-controlled. Thus inter-pretations of the role of medications in improvingor aggravating suicidal thinking were limited.

To our knowledge, a comparison of the acuteeffects of antidepressant treatment with placebo onsuicidal thinking in older adults with depressionhas not been reported. In this study, we addressthis question by examining the effects of sertralineon suicidal thinking in the largest placebo-con-trolled clinical trial reported in older adults withdepression.6 In that trial, there was a significantbut modest advantage for sertraline on responserates defined as 50% improvement on the Hamil-ton Depression Rating Scale7 (HAMD) (35% versus26%) that was similar to the difference we ob-served for other second-generation antidepres-sants in a meta-analysis of placebo-controlledstudies in late-life depression.8

METHODS

This study is a secondary analysis of an eight-week,placebo-controlled, double-blind study of sertralinein patients 60 years of age and older with Diagnosticand Statistical Manual of Mental Disorders, Fourth Edi-tion major depression, nonpsychotic, single episodeand recurrent, with a duration of at least four weeksand a HAMD score �18. Details of the study havebeen previously published.6 Briefly, patients wererandomized to eight weeks of sertraline 50–100 mg/day or placebo. Outcome was assessed at baselineand every two weeks with the 17-item HAMD. Pa-tients judged to be at significant suicide risk wereexcluded. The manufacturer that conducted the orig-inal trial (Pfizer, U.S. Pharmaceuticals Group) pro-vided all requested data relevant to this report to theinvestigators. Analysis of the data was conducted bythe authors.

Narrative reports of all serious adverse events(SAEs; FDA definition) and adverse events (AEs)leading to discontinuation were reviewed by ratersblind to treatment assignment to determine if theywere associated with suicidal ideation (SI), suicideattempts, or self-injurious behavior. All spontane-ously reported adverse events were inspected byblinded raters to determine if an increase in suicidalthinking or self-injurious behavior was reported. Inaddition spontaneously reported adverse eventswere searched using the text string (e.g., attempt, cut,gas, hang, etc.) employed by the FDA to identifypossible suicide related AEs.2

Item 3 from the HAMD was used to assess suicidalthinking. This rating was assessed in all subjects whoreceived at least one dose of medication and had oneposttreatment assessment. Item 3 is rated on a 0 to 4ordinal scale with anchor points for each rating (0�no suicidal thinking, 1� feels life not worth living,2�wishes he or she were dead, 3�active thoughtsof suicide, 4�suicide attempt). Because a rating of 1does not specifically indicate suicidal thinking, inthis report we to refer to “Item 3 scores” rather thansuggesting that any positive rating was an indicationof suicidal thinking or that this item rates suicidality.

Other measures included sex and ethnicity, whichare associated with suicide rates.3 Because 93% of theparticipants were coded as white, the ethnicity vari-able was recoded into white versus other. Baseline

Suicide and Sertraline in Late-Life Depression

Am J Geriatr Psychiatry 15:7, July 2007574

measures included the baseline Hamilton Depres-sion Rating total score, computed with Item 3 omit-ted, and a baseline anxiety score derived using theanxiety/somatization factor from the Hamiltonscale.7

Statistical Methods

Preliminary analysis consisted of describing thebaseline measures and the responses on Item 3 overthe eight weeks of the study. The percentage of miss-ing data were also examined and compared betweentreatment conditions and in relationship to the base-line Item 3 rating.

In addition to the examination of adverse eventdata, two primary analyses were planned—an as-sessment of Item 3 scores by treatment group overtime and comparison of Item 3 by treatment group atendpoint. The first analysis was a comparison of theresponses to Item 3 in the two treatment groups(placebo or sertraline) by estimating and testing amixed-effects model for ordered, categorical longitu-dinal data with the intercept as a random variableusing the MIXOR program.9 This methodology al-lowed us to use all of the observed data and notrestrict the analysis to completers. It also allowedbetter control over the total number of statistical testsas opposed to conducting multiple tests at each as-sessment. Additional covariates in the model in-cluded ethnicity, sex, assessment week, whether theparticipant completed the study or not, the baselineHamilton total depression score (minus Item 3), andthe baseline anxiety factor score.

For the second primary analysis, we examinedendpoint Item 3 scores by treatment group usingordinal logistic regression with sex, ethnicity, thebaseline Item 3 scores, the baseline HAMD totalscore (minus Item 3), and the baseline anxiety factorscore entered as covariates.

As secondary analyses we examined emergent sui-cidal thinking on Item 3 in patients who scored zeroon Item 3 at baseline by comparing Item 3 ratings inthe two groups during treatment. Because some in-dividual patients reported an Item 3 score greaterthan zero at more than one visit, we determined thenumber of patients who reported an increase in Item3 at any visit. For patients who discontinued thestudy before completion, we examined all ratingsavailable. For the number at risk, we used the num-

ber of patients who received at least one dose ofmedication or placebo and who had at least oneposttreatment rating.

For patients whose Item 3 rating was greater thanzero at baseline, we examined change in Item 3scores over time in the two treatment groups usingthe mixed-effects model described above.

RESULTS

A total of 752 patients were enrolled in the study, 747received at least one dose of medication and 728patients had at least one posttreatment assessment.The mean age of the sample was 69.8 years (SD: 6.7,range: 59–97); 56% were female, and 93% werewhite. More details on the participants can be foundin the original publication.6

Serious Adverse Events

Examination of all SAEs indicated that there wereno completed suicides or suicide attempts during thedouble-blind trial. During the trial, one SAE (hospi-talization) was associated with suicidal thinking.This was a 60-year-old man who was hospitalized 13days after starting sertraline because of worseningdepression that included suicidal ideation. Althoughthe narrative report described an increase in suicidalthinking, his Item 3 rating on the day of admissionwas unchanged from baseline (Item 3�2).

Discontinuations Due to Adverse Events

Sixty-eight patients discontinued treatment be-cause of adverse events. Review of the narrativereports revealed no patients other than the 60-year-old man mentioned above who discontinued treat-ment because of suicidal ideation or behavior.

Spontaneously Reported Adverse Events

A total of 2,374 adverse events (AEs) were re-ported during the eight-week trial. Two AEs wereexamined in detail as possibly related to suicidalideation or self-injurious behavior. One of these wasthe hospitalized patient cited above. A second pa-tient, a 65-year-old woman on placebo, was noted tohave worsening depression, but her Item 3 score never

Nelson et al.

Am J Geriatr Psychiatry 15:7, July 2007 575

exceeded 1 and she completed the trial. Spontaneousadverse event reports were also reviewed using thetext string employed by the FDA.2 No other AEreports were identified that indicated an increase insuicidal thinking or behavior.

Analysis of the Hamilton Item 3 Ratings

Missing Data. The percentages of the 752 subjectswith Item 3 data available were 92%, 89%, 83%, and80% at weeks 2, 4, 6, and 8, respectively. The propor-tions dropping out by week 8 (24% and 16% forsertraline and placebo) differed significantly by treat-ment condition (Pearson’s �2�6.91, df�1, p�0.009)but there was no observed relationship between pro-viding data and the Item 3 rating at each assessment.

Change on Item 3 Over 8 Weeks by Treatment Group.Ratings on Item 3 over time are shown in Figure 1.The percentage of patients scoring zero on Item 3increased from 35.7% in the placebo condition and34% in the sertraline condition at baseline to 56.5%and 57.2%, respectively, by week 8. (Percentagesbased on observed 8-week data/number per treatmentgroup with at least one posttreatment rating.) No rat-ings of 4 were observed and 3s were uncommon.

Table 1 summarizes the results of fitting themixed-effects model to the data. The effect for treat-ment group, controlling for the other covariates inthe model, was statistically significant indicating thatover time, participants on sertraline reported lesssuicidal thinking than those on placebo. As with allmultivariate models this effect is interpreted as con-trolling for the other covariates in the model. Thestrongest predictive covariate of suicidal thinkingduring treatment was, as expected, the baseline Item3 score. Significant effects on Item 3 scores were alsofound for time (declining scores), and ethnicity(lower scores in whites than nonwhites). Althoughnot reaching the 0.05 criterion, study completion(lower scores in completers) and sex (higher scores inmen) were close to significance with p values of�0.06 and �0.07, respectively. Baseline scores on theHAMD total (minus Item 3) and the anxiety factorscore were not significant predictors of Item 3 scoresduring treatment.

Analysis of Item 3 Scores at Observed Endpoint. Re-gardless of when the last observation was recorded,endpoint ratings on Item 3 were lower in the sertra-line group than in the placebo group, (Wald �2�

4.33, df�1, p�0.04, adjusted OR�0.70 using ordinallogistic regression).

Secondary Analyses

Emergent Suicidal Thinking. At baseline, 262(34.8%) of the patients scored zero on Item 3. Ofthese, 116 patients on sertraline and 132 patients onplacebo had at least one posttreatment rating. Be-cause some individual patients reported an Item 3score greater than zero at more than one visit, wecounted the total number of patients reporting anincrease in Item 3 at any time point. Among allpatients rating zero on Item 3 at baseline, only one, apatient on placebo, developed active suicidal think-ing (Item 3�3) during treatment. Ten patients (4patients on sertraline and 6 patients on placebo) re-ported the wish to die, (Item 3�2) and 50 patients(22 patients on sertraline and 27 patients on placebo)reported that life is not worth living (Item 3�1).Among those patients with Item 3�0 at baseline andat least one posttreatment rating, a total of 26 of 116(22.4%) patients on sertraline and 34 of 132 (25.8%)patients on placebo reported at least one posttreat-ment rating on Item 3 �0, a difference that was notsignificant (Pearson’s �2�0.37, df�1, p�0.54).

We also examined emergent active suicidal think-ing in patients who scored 1 or 2 on Item 3 at base-line. In this group, six patients (two on sertraline andfour on placebo) developed active suicidal thinking(Item 3�3) during treatment.

Change in Item 3 Over Time in Patients With Item 3Scores �1 at Baseline. A similar mixed-effects modelanalysis was performed using only the data from the476 patients who reported Item 3 scores �0 at base-line. The results of the analysis were similar to thosefor the whole sample except that the difference be-tween sertraline and placebo was slightly greater(estimate of treatment effect�0.65, Z�2.48, p�0.02)than in the whole sample (estimate of effect�0.55,Z�2.41, p�0.02). Also, as would be expected giventhe baseline restriction, the random intercept termwas no longer significant.

DISCUSSION

These data indicate that the use of sertraline in olderdepressed patients did not increase suicidal thinking

Suicide and Sertraline in Late-Life Depression

Am J Geriatr Psychiatry 15:7, July 2007576

or behavior. During the double-blind trial, no patientcompleted suicide or attempted suicide. Overall,treatment was associated with significantly lowerHAMD Item 3 scores. In the patients who reportedthese thoughts before treatment, the difference be-tween treatment with sertraline and placebo waseven greater.

Typical of clinical trials, patients judged by thestudy site investigators to be at “significant suiciderisk” were excluded. As a result, we could not exam-ine the effects of treatment on high-risk patients.Moreover, exclusion of high-risk patients results in atruncated distribution of scores on Item 3 as shownin Figure 1, which may make it more difficult todemonstrate a treatment effect. (Patients who begintreatment with Item 3 score of zero cannot showfurther improvement). It is unlikely, however, thatexclusion of high-risk patients would effect our as-sessment of emergent suicidal thinking in patientswithout evidence of suicidal thinking at baseline.

One 78-year-old woman made a suicide attempt 29days after completing the double-blind study. Shehad begun open-label treatment with sertraline afterdouble-blind sertraline treatment and appeared to beimproving but discontinued the medication after 24

FIGURE 1. Distribution of HAMD Item 3 ratings (suicidal thinking and behavior) at baseline and during treatment in eachtreatment group. A mixed-effects repeated measures analysis demonstrated lower Item 3 ratings with sertralinethan placebo (estimate of effect�0.55, Z�2.41, p�0.02).

TABLE 1. Parameter Estimates, Standard Errors, and Z-testsof Terms in Final Model of Hamilton Item 3Responses Over Time

Variable Estimate Standard Error Z Score p Value

Intercept �2.85 1.163 �2.45 0.014Group 0.55 0.227 2.41 0.016Time �0.43 0.049 �8.80 �0.0001White 1.02 0.481 2.12 0.034Female �0.42 0.230 �1.81 0.070Baseline item 3 2.52 0.150 16.83 �0.0001Baseline HAMDa 0.02 0.048 0.47 0.639Baseline anxiety 0.01 0.079 0.16 0.874Completer �0.65 0.342 �1.90 0.058

a Total Hamilton Depression Scale score computed without item 3.

Nelson et al.

Am J Geriatr Psychiatry 15:7, July 2007 577

days. Five days later she ingested an overdose ofalprazolam, was hospitalized for 24 hours, dis-charged, and restarted on sertraline. We did not in-clude this patient in the results because the eventoccurred after the double-blind, placebo-controlledperiod of the study. The FDA generally considersSAEs occurring within 30 days of the trial or ofdiscontinuing double-blind treatment as being asso-ciated with that treatment. We think that is an un-likely explanation in this case. In any case, if we hadincluded this patient among all randomized patients,the rate of suicide attempts in the sertraline group, 1in 374 (0.3%), would not have differed significantlyfrom the rate in the placebo group, 0 in 378 (0%).

In evaluating the possibility that sertraline mightaggravate suicidal thinking, it is essential to controlfor fluctuations in the depressive disorder and forthat reason the placebo comparison group is critical.In this trial, patients that scored zero on Item 3 atbaseline and were receiving sertraline were not morelikely than those on placebo to experience an increasein thoughts that life is not worth living, the wish todie, or active suicidal thoughts. Among the 248 pa-tients with Item 3 rated zero at baseline only onepatient (a patient on placebo) reported new activesuicidal thinking and only 10 reported the wish todie (four on sertraline and six on placebo). Amongpatients without active suicidal thinking at baseline(Item 3 �3), two patients on sertraline and five pa-tients on placebo developed active suicidal thinking(Item 3 � 3) during treatment. Given these low ratesof emergent suicidal thinking over the eight weeks ofthe trial, it is not possible to conclude that there is noeffect of sertraline on emergent suicidal thinking.The data, however, do not suggest that sertralineincreases emergent suicidal thinking. Coupled withthe findings of significantly lower HAMD Item 3endpoint ratings and lower HAMD Item 3 scoresover time in the sertraline group, these data supporta conclusion that sertraline had a beneficial effect onsuicidal thinking in this sample.

It is important in interpreting the meaning of thesedata to consider how Item 3 is scored. A rating of oneis given if the patient feels life is not worth living.This rating does not necessarily indicate suicidalthinking. A rating of two indicates the patient has awish to die, but does not indicate active planning. Inorder to be clear about the meaning of Item 3 scores,we have avoided the suggestion that any positive

rating indicates suicidal thinking. We have alsoavoided the use of the term “suicidality” which hasbeen criticized for grouping together behaviors thatimply an association with completed suicide even ifthat association has not been established.10

The value of assessing feelings that life is notworth living, the wish to die, or active thoughts ofsuicide is related to the possible predictive value ofthese items for completed suicide. The literature inolder adults documents through retrospective stud-ies or “psychological autopsies” that prior suicideattempts are associated with completed suicide.11,12

However, as Beautrais has reported,11 the predictivevalue of suicide attempts for completed suicide islow. In their sample, 7% of 300 subjects who made asuicide attempt committed suicide during a fol-low-up period, a 14 to 1 rate. In adolescents the ratioof attempts to completed suicide appears to be quitehigh. Langley and Bayatti estimated the ratio to be ashigh as 200 to 1.13 In older adults this ratio declinesand may be as low as 4 to 1.14 Thus in older adultssuicide attempts may be a better predictor or proxyfor completed suicide. Suicidal ideation is more com-mon than suicide attempts, but its prevalence inthose who commit suicide is difficult to determineretrospectively and it was not evaluated in the onlyprospective case-control study in older adults report-ed.15 As a result, the predictive value of suicidalideation for completed suicide is not well estab-lished.

The presence of missing data does not appear to bea limiting factor in these results. We found no evi-dence of a relationship of study completion withbaseline Item 3 scores. The statistical model allowedretention of all of the collected data and an indicatorof study completer status was a covariate in themodel. This covariate was nearly statistically signif-icant (Z��1.90, p�0.058) with a negative sign, in-dicating a trend for those with higher Item 3 scoresduring treatment to be less likely to complete thetrial. As noted in the results, however, we found onlyone patient in the narrative reports who discontin-ued treatment prematurely because of suicide relatedthinking or behavior.

There are limitations to the current study. Al-though this was a prospective clinical trial, the as-sessment of suicidal thinking was not planned pro-spectively. The principle measure used to assesssuicidal thinking was Item 3 on the Hamilton scale;

Suicide and Sertraline in Late-Life Depression

Am J Geriatr Psychiatry 15:7, July 2007578

however, the Hamilton scale was designed for use asa 17-item scale and the validity of individual items isnot well established. Yet, the anchor points for theitem have reasonable face validity and using similaranchor points, we previously reported high interra-ter reliability for this item with an intraclass correla-tion of 0.90.16

It might also be argued that even though this wasthe largest placebo-controlled trial conducted inolder depressed adults, it still might have insufficientpower to detect differences in rarely occurringevents. It should be noted, however, that the Na-tional Institute of Mental Health sponsored study oftreatment for adolescents with depression study(TADS) found 33 harm-related adverse events in 439adolescent subjects treated for 12 weeks (5,268 pa-tient-weeks).17 These events included 24 suicide-re-lated events and 9 harm-related events without ap-parent suicidal intent, such as nonlife threateningcutting. Our current sample of over 747 subjectstreated for eight weeks (5,976 patient-weeks) shouldbe sufficient to detect a similar signal if present inolder patients. It was noteworthy that a careful re-view of all serious adverse events, adverse eventsleading to discontinuation, and all spontaneouslyreported adverse events failed to demonstrate anyevidence of nonsuicidal harm-related events in theolder patients.

A possible difference between the child-adolescentstudies and this older sample is possible inclusion ofundeclared or undiagnosed bipolar patients in theyounger samples. Patients with bipolar depressionmay be at greater risk for “activation” during anti-depressant treatment that might present as emergentsuicidal thinking. Older patients, who have passedthrough most of the age of risk for mania, are less likelyto have undeclared or unrecognized bipolar disorder.

Recently the FDA reported a review of 372 antide-pressant clinical trials in nearly 100,000 individualsof all ages.18 They found that when antidepressantswere compared with placebo, rates of suicidal think-ing and behavior were higher during treatment inpersons under 25 years of age, did not differ fromplacebo in adults 25 to 64 years, and were lower withtreatment relative to placebo in adults 65 years andolder. Our results are consistent with their findingsfor older patients.

Methods in the FDA study differed from ours. Thevery large FDA database increased power for detect-

ing differences in rarely occurring events such ascompleted suicide. And because they included trialsof individuals of all ages, they could assess the effectof age on their findings. Their analysis of suicidalthinking, however, was based on spontaneous re-ports. We examined spontaneous reports but, in ad-dition, we examined HAMD Item 3 scores, whichwere obtained in all patients. Both the FDA data andours support the beneficial effects of treatment inolder adults.

Our data and the FDA data differ from a recentpopulation-based case-controlled study from On-tario that compared risk of completed suicide inpatients 65 and older receiving SSRIs with the risk onother antidepressants.19 Consecutive suicides (N�1,138) were identified from coroner reports over anine-year period and compared with 4,552 subjectsidentified using prescription records, billing claims,and hospitalization data. Comparison subjects werematched using propensity scores. The authors founda fivefold higher risk of suicide with selective sero-tonin reuptake inhibitors (SSRIs) than other agents.The other agents included tricyclics (TCAs), amox-apine, maprotiline, trazodone, nefazodone, bupro-pion, and venlafaxine. One important difference ofthis dataset is that patients at significant suicide riskwere not excluded unlike the clinical trial describedhere or in the FDA report. Limitations of this obser-vational study, however, include lack of random as-signment and that despite propensity matching,there may have been important differences betweenthe SSRI and comparison group. For example, it isnot clear to what extent clinicians selected SSRIsinstead of TCAs for patients judged to be at highersuicide risk because of overdose concerns. In addi-tion, only one-third of the cohort had depression,only one third received antidepressants, and it isunclear if the SSRIs, TCAs, and trazodone were ad-ministered for the same indications, which couldhave included anxiety, insomnia, and pain as well asdepression. Finally the study by design could notassess the effects of medication on suicidal thinking.

In summary, sertraline treatment had a beneficialeffect on Item 3 of the HAMD, which rates suicidalthinking and behavior, the wish to die, and feelingthat life is not worth living. There was no evidencethat sertraline aggravated suicidal thinking or behav-ior when compared with placebo, nor was there ev-idence of an increase in self-injurious behavior. Al-

Nelson et al.

Am J Geriatr Psychiatry 15:7, July 2007 579

though the predictive value of suicidal thinking forcompleted suicide is not established, the high rates ofcompleted suicide in older individuals coupled withthe beneficial effects of sertraline observed here sug-gest that the use of effective antidepressant treat-ments should be encouraged in older adults.

These data were obtained in a clinical trial conducted byPfizer during 1999–2001. That RCT was previously pub-lished. The data for this report was requested by theauthors and send to them by Pfizer. The authors analyzed

the data, prepared this report, and received no financialsupport for any aspect of the preparation of this paper.

Dr. Nelson has received honoraria from GlaxoSmith-Kline and Pfizer and has participated in consulting forAbbott, Biovail, Bristol Myers Squibb, Corcept, Eli Lilly,GlaxoSmithKline, Orexigen, Organon, Pfizer, Sepracor,and Shire. Dr. Schneider has received honoraria fromAstra Zeneca, Eli Lilly, and Forest Laboratories and hasparticipated in consulting for Astra Zeneca, Forest Labo-ratories, Johnson & Johnson, and Pfizer.

References

1. Food and Drug Administration. Black Box warning for antidepres-sants in children and adolescents. Available at http://www.fda.gov/cder/drug/antidepressants/PI_template.pdf. Accessed Febru-ary 23, 2006

2. Hammad TA, Laughren T, Racoosin J: Suicidality in pediatricpatients treated with antidepressant drugs. Arch Gen Psychiatry2006;63:332–339

3. National Center for Injury Prevention and Control, United StatesSuicideDeaths and Rates per 100,000. Available at: http://webapp.cdc.gov/sasweb/ncipc/mortrate.html. Accessed February 23,2006

4. Bruce ML, Ten Have TR, Reynolds CF, et al: Reducing suicidalideation and depressive symptoms in depressed older primarycare patients. JAMA 2004; 291:1081–1091

5. Szanto K, Mulsant BH, Houck P, et al: Occurrence and course ofsuicidality during short-term treatment of late-life depression.Arch Gen Psychiatry 2003; 60:610–617

6. Schneider LS, Nelson JC, Clary CM, et al: An eight-week, multi-center, parallel-group, double-blind, placebo-controlled study ofsertraline in elderly outpatients with major depression. Am J Psy-chiatry 2003; 160:1277–1285

7. Hamilton M: A rating scale for depression. J Neurol NeurosurgPsychiatry 1960; 23:56–61

8. Nelson JC, Delucchi K, Schneider LS: Efficacy of second-genera-tion antidepressants in late life depression: a meta-analysis. Pre-sented at the 45th Annual Meeting of the American College ofNeuropsychopharmacology, Hollywood, FL, December 3-7, 2006

9. Hedeker D, Gibbons RD: Mixor: a computer program for mixed-

effects ordinal regression analysis. Comp Meth Prog Biomed1996; 49:157–176

10. Klein DF: The flawed basis for FDA post-marketing safety deci-sions: the example of antidepressants and children. Neuropsy-chopharmacology 2006; 31:689–699

11. Beautrais AL: A case control study of suicide and attemptedsuicide in older adults. Suic Life-Threaten Behav 2002; 32:1–9

12. Conwell Y, Lyness JM, Duberstein P, et al: Completed suicideamong older patients in primary care practices: a controlledstudy. JAGS 2000; 48:23–29

13. Langley GE, Bayatti NN: Suicides in Exe Vale Hospital, 1972-1981.Br J Psychiatry 1984; 145:463–467

14. Parkin D, Stengel E: Incidence of suicide attempts in an urbancommunity. Br Med J 1965; 2:133–138

15. Ross RK, Bernstein L, Trent L, et al: A prospective study of riskfactors for traumatic deaths in a retirement community. Prev Med1990; 19:323–334

16. Mazure C, Nelson JC, Price LH: Reliability and validity of thesymptoms of major depressive illness. Arch Gen Psychiatry 1986;43:451–456

17. March J, Silva S, Petrycki S, et al: Fluoxetine, cognitive-behaviortherapy, and their combination for adolescents with depression.JAMA 2004; 292:807–820

18. Stone MB, Jones ML: FDA Clinical Review: relationship betweenantidepressant drugs and suicidality in adults. Washington, DC:U.S. Food and Drug Administration,2006:1–64

19. Juurlink DN, Mamdani MM, KoppA, Redelmeier DA: The risk ofsuicide with selective serotonin reuptake inhibitors in the el-derly. Am J Psychiatry 2006; 163:813–821

Suicide and Sertraline in Late-Life Depression

Am J Geriatr Psychiatry 15:7, July 2007580