Evidence-Based Psychotherapies forSuicide Prevention

Future DirectionsGregory K. Brown, PhD, Shari Jager-Hyman, PhD

Psychotherapeutic interventions targeting suicidal thoughts and behaviors are essential for reducingsuicide attempts and deaths by suicide. To determine whether specific psychotherapies areefficacious in preventing suicide and suicide-related behaviors, it is necessary to rigorously evaluatetherapies using RCTs. To date, a number of RCTs have demonstrated efficacy for severalinterventions focused on preventing suicide attempts and reducing suicidal ideation. Althoughthese studies have contributed greatly to the understanding of treatment for suicidal thoughts andbehaviors, the extant literature is hampered by a number of gaps and methodologic limitations.Thus, further research employing increased methodologic rigor is needed to improve psychother-apeutic suicide prevention efforts. The aims of this paper are to briefly review the state of the sciencefor psychotherapeutic interventions for suicide prevention, discuss gaps and methodologiclimitations of the extant literature, and suggest next steps for improving future studies.(Am J Prev Med 2014;47(3S2):S186–S194) & 2014 American Journal of Preventive Medicine

Introduction

The development and implementation of effectiveinterventions are imperative for reducing rates ofsuicide and related behaviors. In response to the

ongoing need for effective treatments aimed at prevent-ing suicide and self-directed violence, the NationalAction Alliance for Suicide Prevention’s (Action Alli-ance) Research Prioritization Task Force (RPTF)1 hasproposed the following Aspirational Goal focused onpsychotherapeutic interventions: “…develop widelyavailable, more effective and efficient psychosocial inter-ventions targeted at individuals, families, and communitylevels.”The current paper has three main aims in discussing

this Aspirational Goal. First, with a focus on RCTs, thestate of the science for evidence-based psychotherapyinterventions for suicidal ideation and behavior isreviewed. Second, limitations of the current researchand suggestions for future research are discussed. Finally,a step-by-step pathway for evaluating psychotherapyinterventions for suicide prevention is proposed.

State of the Science of Evidence-BasedTreatments for Suicide PreventionSeveral RCTs2–5 have demonstrated promising results inreducing suicide attempts and self-directed violence. Acomprehensive review of the literature is beyond thescope of this paper; however, reviews2–5 were used toidentify studies to include in this brief review. A selectionof studies yielding positive effects will be highlighted andpresented in Table 1. Briefly, cognitive therapy for suicideprevention (CT-SP)6; cognitive–behavioral therapy(CBT)7; dialectical behavior therapy (DBT)8; problem-solving therapy (PST)9; mentalization-based treatment(MBT)10; and psychodynamic interpersonal therapy(PIT)11 have all evidenced positive effects for preventingsuicide attempts or self-directed violence in adults.More specifically, recent suicide attempters who

received CT-SP were 50% less likely to reattempt thanparticipants who received enhanced usual care (EUC)with tracking and referrals.6 CBT plus treatment as usual(TAU) also reduced self-harming behaviors relative toTAU alone.7 For individuals with borderline personalitydisorder (BPD), DBT demonstrated a greater reduction insuicide attempts relative to community treatment byexperts.8 However, DBT was not statistically more effec-tive than a manualized general psychiatric managementcondition, consisting of case management, dynamicallyinformed psychotherapy, and medication management.12

Also focused on BPD, MBT, a psychoanalyticallyoriented partial hospitalization program, was more

From the Perelman School of Medicine of the University of Pennsylvania,Philadelphia, Pennsylvania

Address correspondence to: Gregory K. Brown, PhD, Department ofPsychiatry, University of Pennsylvania, 3535 Market Street, Room 2032,Philadelphia PA 19104-3309. E-mail: gregbrow@mail.med.upenn.edu.

0749-3797/$36.00http://dx.doi.org/10.1016/j.amepre.2014.06.008

S186 Am J Prev Med 2014;47(3S2):S186–S194 & 2014 American Journal of Preventive Medicine � Published by Elsevier Inc.

Table1.Su

mmaryof

select

RCTs

Autho

rsSa

mple

Stud

yinterven

tion

Con

trol

cond

ition

Outcome

varia

bles

Follow-up

intervals

Mainfind

ings

Batem

anan

dFo

nagy

(1999)10

Adults

with

BPD

referred

topsychiatric

unit

Partial

hospita

lization

(n¼1

9)

Stan

dard

psychiatric

care

(n¼1

9)

Suicide

attempts

3,6

,9,1

2,1

5,

18mon

ths

Patie

ntswho

received

thestud

yinterven

tion

expe

rienced

asign

ificant

redu

ctionin

attempts

from

admission

to18mon

ths(Ken

dall’sW¼0

.59,

χ2(3)¼

33.5,p

o0.001)

Blum

etal.

(2008)13

Adults

with

BPD

STEP

PSplus

TAU(n¼6

5)

TAU(n¼5

9)

Suicide

attempts

1,3

,6,9

,12

mon

ths

Nodifferen

cesin

timeto

firstsuicideattempt

betwee

nST

EPPS

þTA

Uan

dTA

Ugrou

ps;

χ2(1)o

0.1,p

¼0.994

Brownet

al.

(2005)6

Adults

recruited

from

EDfollowing

asuicideattempt

CT(n¼6

0)

EUC(n¼6

0)

Suicidal

idea

tion,

suicide

attempts

1,3

,6,1

2,1

8mon

ths

At6mon

ths,usingtheKap

lan–

Meier

metho

d,estim

ated

reattempt-free

prob

ability:C

Tgrou

p¼0.86(95%

CI¼

074,0

.93);usua

lcare

¼0.68(95%

CI¼

0.54,0

.79)

At18mon

ths,estim

ated

reattempt-free

prob

ability:C

T¼0.76(95%

CI¼

0.62,0

.85);usua

lcare¼0

.58(95%

CI¼

0.44,0

.70)

Patie

ntsin

theCTcond

ition

hadasign

ificantly

lower

reattempt

rate

(Waldχ2¼3

.9,p¼0

.049)a

ndwere50%

less

likelyto

reattempt

than

theusua

lcare

grou

p(hazardratio¼0

.51,95

%CI¼0

.26,0.99

7)Th

erewereno

sign

ificant

grou

pdifferen

cesin

suicidal

idea

tion

Bruce

etal.

(2004)21

Dep

ressed

olde

rad

ults

recruited

from

prim

arycare

Structured

,team

-based

interven

tion

includ

ing

citalopram

þpsycho

therap

y(n¼3

20)

TAU(n¼2

78)

Suicidal

idea

tion

4,8

,12

mon

ths

Rates

ofsuicidal

idea

tionde

clined

faster

forthe

interven

tiongrou

p(12.9%

declinefrom

baseline)

than

theTA

Ugrou

p(3.0%

declinefrom

baseline;

p¼0.01foralld

epressed

patie

nts,p¼

0.006for

patie

ntswith

MDD)

Com

toiset

al.

(2011)17

Adults

evalua

ted

forsuicide

attempt

orim

minen

tris

kbu

tjudg

edsafe

for

discha

rge

CAM

S(n¼1

6)

E-CAU

(n¼1

6)

Suicide

attempts,

suicidal

idea

tion

2,4

,6,1

2mon

ths

Participan

tswho

received

CAM

Smad

efewer

suicideattemptsthan

thosewho

received

E-CAU

at2-,4-,an

d6-m

onth

follow-ups

a

Suicidal

idea

tionim

proved

sign

ificantlyforCAM

Spa

tients,

reaching

89%

redu

ctionat

12mon

ths,

RR¼0

.11,9

5%

CI¼

0.04,0

.30;a

t12mon

ths,

E-CAU

patie

ntsrepo

rted

sign

ificantlyworse

suicidal

idea

tionthan

CAM

Spa

tients(RR¼4

.81,

95%

CI¼

1.61,1

4.33)

(con

tinue

don

next

page

)

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194 S187

September 2014

Table1.S

ummaryof

select

RCTs

(con

tinue

d)

Autho

rsSa

mple

Stud

yinterven

tion

Con

trol

cond

ition

Outcome

varia

bles

Follow-up

intervals

Mainfind

ings

Davidsonet

al.

(2006)14

Adults

with

BPD

andan

episod

eof

DSH

with

inthe

past

12mon

ths

CBTþ

TAU

(n¼5

3)

TAU(n¼4

9)

Suicidal

acts

6,1

2,1

8,2

4mon

ths

After24mon

ths,therewas

agrea

terredu

ctionin

numbe

rof

suicidal

acts

intheinterven

tiongrou

pcompa

redto

theTA

Ugrou

p(m

eandifferen

ce¼

–0.91,p

¼0.020)

Diamon

det

al.

(2010)19

Adolescents

iden

tified

assuicidal

byscreen

ingdu

ring

prim

arycare

orED

visits

ABFT

(n¼3

5)

EUC(n¼3

1)

Suicidal

idea

tion

6,1

2,2

4wee

ksAt

the12-wee

kassessmen

t,pa

tientsreceiving

ABFT

demon

stratedasign

ificantlygrea

terrate

ofim

provem

entin

suicidal

idea

tionthan

patie

nts

receivingEU

C,F

(1,6

4)¼

12.60,p

¼0.001

ABFT

hadasign

ificant

effect

onclinical

recovery

(SIQ-JRr13)ofsuicida

lide

ationat

alltim

epo

ints;

at6wee

ks,6

9.7%of

ABFT

patie

ntsan

d40.7%of

EUCpa

tientsrepo

rted

suicidal

idea

tionin

the

norm

ativerang

e,OR¼3

.35,9

5%

CI¼

1.15,9

.73,

χ²(1)¼

5.07,p

¼0.02;a

t12wee

ks,8

7.1%of

ABFT

patie

ntsan

d51.7%

ofEU

Cpa

tientsrepo

rted

idea

tionin

theno

rmativerang

e,OR¼6

.30,9

5%

CI¼

1.76,2

2.61,χ²(1

)¼8.93,p

¼0.003;a

t24

wee

ks,7

0%

ofAB

FTpa

tientsan

d34.6%

ofEU

Cpa

tientsrepo

rted

idea

tionin

theno

rmativerang

e,OR¼4

.41,9

5%

CI¼

1.43,1

3.56,χ²(1

)¼7.01,

p¼0.008

Guthrie

etal.

(2001)11

Adults

presen

ting

toED

afterself-

poison

ing

Psycho

dy-

namic

interpersona

ltherap

yde

livered

inho

me(n¼5

8)

TAU(n¼6

1)

Suicidal

idea

tion

1,6

mon

ths

Atthe6-m

onth

follow-upassessmen

t,pa

tients

receivingthestud

yinterven

tionrepo

rted

lower

levelsof

suicidal

idea

tioncompa

redto

those

receivingTA

U(differen

cesbe

twee

nmea

ns¼–4.9,

95%

CI¼

–8.2,–

1.6,p

¼0.005)

Hatcher

etal.

(2011)9

Adults

presen

ting

toaho

spita

lafter

self-ha

rm

PST(n¼5

22)

Usual

care

(n¼5

72)

Self-ha

rm3,1

2mon

ths

Fewer

patie

ntsreceivingPS

Trepo

rted

repe

atep

isod

esof

self-ha

rmat

the12-m

onth

assessmen

tthan

thosereceivingusua

lcare

(RR¼0

.39,9

5%

CI¼

0.07,0

.60,p

¼0.03)

Hue

yet

al.

(2004)16

Youthfollowing

EDvisitfor

suicide

attempt,ide

ation,

orplan

ning

MST

bStan

dard

trea

tmen

tbSu

icidal

idea

tion,

suicide

attempts

4,1

6mon

ths

MST

was

sign

ificantlymoreeffectivethan

stan

dard

trea

tmen

tat

redu

cing

suicideattempts

over

16mon

ths,t(linea

r)¼2

.61,p

o0.01,t

(qua

dratic)¼

3.60,p

o0.001

Therewereno

sign

ificant

grou

pdifferen

cesfor

suicidal

idea

tion

(con

tinue

don

next

page

)

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194S188

www.ajpmonline.org

Table1.S

ummaryof

select

RCTs

(con

tinue

d)

Autho

rsSa

mple

Stud

yinterven

tion

Con

trol

cond

ition

Outcome

varia

bles

Follow-up

intervals

Mainfind

ings

Line

hanet

al.

(2006)8

Wom

enwith

BPD

with

Z2ep

isod

esof

self-ha

rmin

the

past

5years,

includ

ingZ1

with

inthepa

st8

wee

ks

DBT(n¼5

2)

Com

mun

itytrea

tmen

tby

expe

rts

(n¼4

9)

Suicidal

idea

tion,

suicide

attempts

4,8

,12,1

6,

20,2

4mon

ths

Fewer

patie

ntsreceivingDBTha

dsuicideattempts

than

thosereceivingtrea

tmen

tby

expe

rts(23.1%

vs46%

,hazardratio

¼2.66,p

¼0.005,

NNT¼

4.24,

95%

CI¼

2.40,1

8.07);themea

nprop

ortio

nsof

suicideattempterspe

rtreatmen

tgroup

perp

eriod

were6.2%

(95%

CI¼

3.1%,1

1.7%)a

nd12.2%

(95%

CI¼

7.1%,2

0.3%)for

theDBTan

dcontrol

grou

ps,respe

ctively

Fewer

patie

ntsreceivingDBTthan

commun

itytrea

tmen

tby

expe

rtsha

dno

n-am

bivalent

suicide

attempts(5.8%

vs13.3%,p

¼0.18,F

ishe

r’sexact

test

andNNT¼

13.3,9

5%

CI¼

5.28,2

5.41)

Therewereno

sign

ificant

grou

pdifferen

cesfor

suicidal

idea

tion

McM

ainet

al.

(2009)12

Adults

with

BPD

with

Z2suicidal

orno

n-suicidal

self-injurio

usep

isod

esin

the

past

5years,Z1

episod

ein

the

past

3mon

ths

DBT(n¼9

0)

Gen

eral

psychiatric

man

agem

ent

(n¼9

0)

Freq

uencyan

dseverityof

suicidal

episod

es

4,8

,12

mon

ths

Therewereno

sign

ificant

grou

pdifferen

cesfor

suicidal

episod

es

Slee

etal.

(2008)7

Adults

who

recentlyen

gage

din

deliberateself-

poison

ingor

self-

injury

CBTþ

TAU

(n¼4

0)

TAU(n¼4

2)

Self-ha

rm,

suicidal

cogn

ition

3,6

,9mon

ths

At9mon

ths,pa

tientswho

received

CBTþ

TAU

hadsign

ificantlygrea

terredu

ctions

inself-ha

rmthan

thosewho

received

TAUalon

e(po0.05)

CBTþT

AUpa

tientsha

dsign

ificantlyde

crea

sed

suicidalcogn

ition

sas

compa

redto

TAUpa

tientsat

the3-(po

0.05),6-(po

0.05),an

d9-m

onth

(po0.01)a

ssessm

ents

Stew

artet

al.

(2009)18

Adults

intrea

tmen

tfollowingasuicide

attempt

CBT(n¼1

1),

PST(n¼1

2)

TAU(n¼9

)Su

icidal

idea

tion,

suicide

attempts

4wee

ks(PST

),7wee

ks(CBT),

2mon

ths

(TAU

)

CBTwas

themoste

ffectivetrea

tmen

tfor

redu

cing

suicideattempts;pa

tientsreceivingCBTmad

eno

attemptsdu

ringthestud

y,whe

reas

patie

nts

receivingPS

Tan

dTA

Umad

ean

averag

eof

0.33

attemptsan

d0.22attempts,respectively

Suicidal

idea

tionde

crea

sedwith

both

CBT(z¼

�2.32,p

o0.05,r¼0

.49)a

ndPS

T(z¼�

2.39,

po0.05,r¼0

.49);de

crea

sesin

suicidal

idea

tion

weregrea

terfor

thePS

Tthan

TAUgrou

p(U¼2

6.5,

pr0.05,r¼0

.49)

(con

tinue

don

next

page

)

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194 S189

September 2014

effective than general psychiatric servicesin reducing suicidal and self-mutilatoryacts.10 Similarly, relative to TAU alone,PST plus usual care resulted in a decreasein repeat hospitalizations for self-harm inindividuals with a history of previous self-harm.9 Finally, four home-based sessionsof interpersonal therapy were more effec-tive than TAU in reducing suicidal idea-tion and repeated self-harm in individualswho self-poisoned.11

Although a number of suicide-prevention interventions have evidencedefficacy, other interventions, includingsystems training for emotional predict-ability and problem solving13 and CBT forCluster B personality disorders,14 havenot been supported empirically. For acomprehensive review of negative find-ings, please see previous reviews.2–5

Fewer studies15,16 have demonstratedefficacy for psychotherapy interventionsin reducing self-directed violence in ado-lescents. Wood and colleagues15 found thatadolescents who received developmentalgroup therapy (consisting of componentsof CBT, DBT, and psychodynamic grouptherapy) plus TAU were less likely toengage in repeated deliberate self-harmon two or more occasions than thosewho received TAU alone. Finally, multi-systemic therapy, an intensive family-based treatment, reduced the frequencyof suicide attempts compared to treatmentreceived during inpatient hospitalization.16

In addition to psychosocial interven-tions designed to prevent suicideattempts, several psychotherapy treat-ments directly target suicidal ideation.Specifically, collaborative assessment andmanagement of suicidality (CAMS),17

CBT,18 PST,18 and PIT11 have resultedin the reduction of suicidal ideation inadults. CAMS, a therapeutic frameworkfocused on identifying causes of suicidalideation and treatment goals for reducingsuicidal ideation, was associated withsignificantly greater and sustained reduc-tion of suicidal ideation at 12 monthspost-treatment compared to TAU.17 Sim-ilarly, both PST and CBT resultedin greater reduction of suicidal ideationthan TAU.16 Attachment-based familyTa

ble1.S

ummaryof

select

RCTs

(con

tinue

d)

Autho

rsSa

mple

Stud

yinterven

tion

Con

trol

cond

ition

Outcome

varia

bles

Follow-up

intervals

Mainfind

ings

Unü

tzer

etal.

(2006)20

Older

adults

with

MDDor

dysthymia

IMPA

CT

interven

tion

(n¼9

06)

Usual

care

(n¼8

95)

Suicidal

idea

tion

6,1

2,1

8,2

4mon

ths

Fewer

patie

ntsreceivingtheIM

PACTinterven

tion

than

usua

lcarerepo

rted

thou

ghts

ofsuicideat

6(OR¼0

.54,9

5%

CI¼

0.37,0

.78,p

¼0.001),12

(OR¼0

.54,9

5%

CI¼

0.40,0

.73,p

o0.001),18

(OR¼0

.52,9

5%

CI¼

0.36,0

.75,p

o0.001),an

d24

(OR¼0

.65,95

%CI¼0

.46,0.91

,po0.01

)mon

ths

Fewer

patie

ntsreceivingtheIM

PACTinterven

tion

repo

rted

thou

ghts

ofde

athor

dyingat

6(OR¼0

.62,9

5%

CI¼

0.49,0

.78,p

o0.001),12

(OR¼0

.44,9

5%

CI¼

0.35,0

.56,p

o0.001),18

(OR¼0

.62,9

5%

CI¼

0.49,0

.79,p

o0.001),an

d24

(OR¼0

.72,95

%CI¼0

.57,0.92

,p¼0

.01)

mon

ths

Woo

det

al.

(2001)15

Adolescents

referred

tomen

tal

health

services

afterde

liberate

self-ha

rm

Group

therap

y(n¼3

2)

Rou

tinecare

(n¼3

1)

Rep

etition

ofself-ha

rm,

suicidal

thinking

6wee

ks,7

mon

ths

Participan

tswho

received

grou

ptherap

ywereless

likelyto

repe

atself-ha

rmthan

thosewho

received

routinecare

(OR¼6

.3,6

5%

CI¼

1.4,2

8.7)

Therewereno

sign

ificant

grou

pdifferen

ces

rega

rdingsuicidal

thinking

Note:

Onlyou

tcom

esrelatedto

suicideidea

tionan

dattemptsarerepo

rted

.aNostatistical

analyses

werepe

rformed

fortheseresults.

bStud

ydidno

tindicate

thenu

mbe

rof

participan

tspe

rcond

ition

.AB

FT,attachm

ent-b

ased

family

therap

y;BPD

,borde

rline

person

ality

disorder;C

AMS,

collabo

rativeassessmen

tand

man

agem

ento

fsuicida

lity;CBT,cogn

itive–be

havioraltherap

y;CT,cogn

itive

therap

y;DBT,

dialectical

beha

vior

therap

y;DSH

,deliberateself-ha

rm;E

-CAU

,enh

ancedcare

asusua

l;ED

,emerge

ncyde

partmen

t;EU

C,e

nhan

cedusua

lcare;

IMPA

CT,

ImprovingMoo

d:Prom

otingAccess

toCollabo

rativeTrea

tmen

t;MDD,m

ajor

depressive

disorder;M

ST,m

ultisystemictherap

y;NNT,

numbe

rne

eded

totrea

t;PS

T,prob

lem-solving

therap

y;RR,riskratio

;SIQ-JR,S

uicide

Idea

tion

Que

stionn

aire–Junior;S

TEPP

S,system

straining

forem

otiona

lpredictab

ility

andprob

lem

solving;

TAU,treatmen

tas

usua

l

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194S190

www.ajpmonline.org

therapy, which focuses on strengthening the parent–adolescent attachment bond, has also demonstratedpromise in reducing suicidal ideation in suicidal adoles-cents relative to EUC.19

Finally, to our knowledge, two studies have demon-strated efficacy in reducing suicidal ideation in depressedolder adults in primary care settings.20,21 The ImprovingMood: Promoting Access to Collaborative Treatmentstudy determined that a collaborative, team-basedapproach to treating depression resulted in a greaterreduction of suicidal ideation than usual care. ThePrevention of Suicide in Primary Care Elderly: Collabo-rative Trial intervention, consisting of a clinical algo-rithm for treating geriatric depression in primary caresettings and care management, was more effective inreducing suicidal ideation than EUC.

Limitations of the Current State of theScienceAlthough the aforementioned RCTs represent importantfirst steps in gaining a deeper understanding of effectivesuicide prevention strategies, several gaps and methodo-logic concerns limit conclusions that can be drawn fromthese studies. Several significant gaps in the literatureshould be noted. First, given the paucity of RCTs poweredto detect deaths by suicide, it is unknown whether deathby suicide (rather than suicide attempts) can be preventedby psychotherapy. Moreover, it is unclear as to whetherthe reduction of suicide attempts or ideation via psycho-therapy actually reduces deaths by suicide.Second, many studies focused on suicide prevention

exclude patients at imminent risk for suicide, making itimpossible to determine whether interventions that areefficacious for lower-risk patients are also efficacious forthose at highest risk.22 Third, there are limited psycho-therapy RCTs focused on preventing suicide attempts formany at-risk populations, including older adults; Veteransor military service members; lesbian, gay, bisexual, trans-gender, queer, and two-spirit (LGBTQ2) populations;Native Americans and other minority groups; and survi-vors of suicide or suicide attempts. It is unclear whether theresults of existing RCTs generalize to these populations.Additionally, the majority of psychotherapy interven-

tions for suicidal thoughts and behaviors have beenconducted in outpatient settings, and very few RCTs havebeen conducted in acute care settings, such as emergencydepartments, inpatient units, and crisis hotlines. Thedevelopment of interventions for these settings is partic-ularly important given that many high-risk patients onlypresent to acute care services and never receive additionalpsychosocial treatment. The dearth of knowledge abouteffective treatments for inpatient settings is especially

alarming given that the current standard of care is to admithigh-risk patients to inpatient units. This suggests thatpatients who are at high risk for suicide may not receiveappropriate evidence-based treatments to prevent suicide.A final gap in the extant research examining the

efficacy of psychotherapy interventions for suicide pre-vention is the failure to replicate studies in whichtreatments have been found to be efficacious. It isespecially critical that replication trials be conducted byindependent researchers, as in some cases replicationstudies conducted outside of the original research groupshave failed to demonstrate the same beneficial effects.12

A variety of methodologic limitations of the existingresearch hamper the ability to draw firm conclusionsregarding the effectiveness and generalizability of varioussuicide prevention efforts (limitations have been pub-lished elsewhere1–4). First, a lack of consensus regardingterms and operationalized definitions used to describesuicide, attempts, ideation, and other related behaviorslimits the ability to generalize across studies and replicatefindings. Researchers also often neglect to use reliableand validated measures of suicidal ideation and behav-iors, making it difficult to understand the specificbehaviors measured and targeted by the interventionsin question.In addition, many previously published RCTs do not

provide detailed psychotherapy manuals. The absence oftreatment manuals creates significant challenges for dis-semination and implementation efforts in the communityand precludes appropriate replication studies. Furthermore,researchers often neglect to include measures assessing theintegrity of the study intervention. It is important to assessthe extent to which study therapists adhere to the theoryand practice of the intervention of interest.An additional common methodologic problem is that

studies are underpowered to adequately detect treatmenteffects, causing potentially efficacious treatments to yieldnegative results owing to lack of power rather than lack ofefficacy. Moreover, very few studies include descriptions ofpower analyses, making it difficult to determine the reasonsfor failing to find positive effects. Other studies conductpower analyses based on unlikely or biased estimates ofeffects, leading to inadequate estimates of sample sizes.Conservative estimates are necessary to ensure that sam-ples are powered sufficiently to detect effects.Given that RCTs are generally longitudinal, attrition is

common and results in an additional methodologic issueof handling missing data. This is particularly problematicwhen dropout rates differ across treatment conditions,which may result in biased results.7 As recommended inthe CONSORT guidelines for reporting RCT results,intention-to-treat analysis is a helpful statistical approachto handling missing data to minimize bias.23

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194 S191

September 2014

Other methodologic limitations encountered in theextant literature include potential threats to externalvalidity by choosing highly selective samples11; failureto use blind investigators, assessors, or patients or specifywhether blinding was implemented; potential measure-ment bias (e.g., using differential measurement intervalsand methods for assessing primary outcomes in inter-vention and control groups10); failure to identify, meas-ure, and control for potential non-study co-interventions(e.g., pharmacotherapy); and analyses capitalizing ondifferences in baseline characteristics.16

It is also advised that researchers focus on a priorianalyses and refrain from making firm conclusions onthe basis of unplanned, underpowered subgroup analy-ses. Finally, stratified randomization is an important toolin preventing Type I errors and imbalance betweentreatment groups, particularly for smaller trials in whichknown factors influence treatment responsiveness.

Next Steps and Breakthroughs NeededAlthough the existing RCTs have created an importantjumping-off point for evaluating future psychotherapeuticinterventions for suicide attempts and ideation, muchwork remains. The adoption of the following recommen-dations may lead to increased methodologic rigor withwhich suicide research is conducted, and in turn, thedevelopment and dissemination of treatments that reducesuicidal ideation, suicide attempts, and ultimately, suicide.Given that the current lack of consensus of terms and

definitions leads to difficulty in interpreting results andaggregating findings across studies, an important short-term goal is to adopt an agreed-upon nomenclature forall studies addressing suicide-relevant thoughts andbehaviors, such as the self-directed violence nomencla-ture proposed by the CDC’s National Center for InjuryPrevention and Control.24 It is then essential to employvalid and reliable measures to assess these constructs.The Columbia Suicide Severity Rating Scale (C-SSRS25)

is one such measure endorsed by the U.S. Food and DrugAdministration for use in pharmaceutical trials. It wouldalso be beneficial to use an agreed-upon measure forpsychotherapy trials. Furthermore, to achieve continuityacross studies, it would be helpful for all studies to use thesame endpoints in reporting outcomes, thereby increasingthe ease with which results can be aggregated acrossstudies via meta-analyses.There is also a need for methods to address ambiguous

suicide behavior that may not neatly fit into a specificcategory of suicidal thoughts or behaviors. One potentialsolution to this problem is to form suicide adjudicationboards to review ambiguous behaviors and reach aconsensus regarding appropriate classification.26

An additional short-term goal is to develop interven-tions designed for high-risk populations, including olderadults, Veterans or military service members, LGBTQ2individuals, minority groups, and survivors of suicide orsuicide attempts as indicated by empirical research. Thereis also a need for methods to screen and treat high-riskindividuals in acute care settings, including emergencydepartments, crisis hotlines, and inpatient units.As previously mentioned, many studies assessing the

efficacy of treatments for suicide prevention are under-powered. Although preliminary studies to determineacceptability and feasibility of specific interventions arenecessary, large-scale RCTs that are adequately poweredto detect treatment effects are also imperative. This is truefor studies assessing treatments focused on reducingsuicidal thoughts, suicide attempts, and other self-directed violence, as well as those designed to evaluatetreatments for the prevention of deaths by suicide.Because suicide is a low base rate behavior, very large

samples are required to conduct adequately poweredtrials. Multi-site collaborations allow the collection ofdata from large samples while reducing financial andorganizational burden on any one site. In addition, theuse of standardized outcome measures and data sharingmay facilitate meta-analytic approaches and circum-vent problems associated with inadequately poweredstudies.Further development and dissemination of treatments

specifically targeting suicidal ideation are also necessary,particularly for populations such as older men who havethe highest rates of suicide of any age group.27 Despitetheir increased rate of deaths by suicide, older adults areless likely to make suicide attempts than individuals inany other age group.28 Suicidal ideation may thus serve asthe only warning sign of future suicides in older adults,making it especially important to specifically targetsuicidal ideation in this population. As frequent attemptsare less common in this population, treatments focusedon preventing attempts may be less appropriate.Because suicidal ideation is a dimensional construct

that waxes and wanes over time, RCTs should includeappropriate measures for tracking fluctuations in suicidalideation. The use of ecological momentary assessment,for example, would provide much-needed insight intothe fluctuation of suicidal ideation and inform thedevelopment of timely interventions that specificallytarget changes in suicidal ideation.Very little is known about whether positive effects of

psychotherapies for suicide prevention extend beyondlaboratory settings. In addition to efficacy trials, effec-tiveness trials are also needed to assess whether specifictreatments work in real-world settings. Moreover, inorder to increase external validity of psychotherapy trials,

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194S192

www.ajpmonline.org

it is important that inclusion and exclusion criteria resultin samples that reflect patients as they present in the realworld (e.g., the exclusion of potential participants who donot misuse substances may result in a biased sample ofsuicide attempters10).There is a need to better develop mechanisms to ensure

that the individuals at risk of suicide have access totreatments that work. In designing interventions,researchers should consider ways to increase the feasi-bility and ease with which treatments can be disseminatedand adapted to various settings. For example, futurepsychotherapies that can be implemented in rural settingsusing telehealth technologies are needed.In addition, researchers are encouraged to clearly

communicate the specific treatment components neces-sary to successfully implement interventions in non-laboratory settings. Another potential approach toincreasing the availability of evidence-based treatmentsis to develop innovative electronic health interven-tions (e.g., smartphone applications, texting, web-basedinterventions, or chat rooms) as either widely availablestand-alone interventions or adjunctive treatments toface-to-face interventions. Finally, further research isneeded to determine the cost-effectiveness and costutility of psychotherapy studies for suicide prevention.As researchers continue to find support for treatments

that reduce suicidal thoughts and behaviors, it is necessaryto identify potential mechanisms of actions that account fortherapeutic change. Thus, in addition to asking whether atreatment works, it is essential to ask why a treatmentworks. This can be achieved by including measures assess-ing constructs underlying treatment effects, such asimprovements in hopelessness or emotion regulation.Identifying mechanisms of action will allow for the

development of more efficient, targeted treatments andmay provide insight into which treatments work bestfor whom.In addition to identifying treatments that are effective in

reducing suicide ideation and behaviors, it is also impor-tant to understand which treatments have not garneredsupport in psychotherapy trials. Systematic trial registra-tion is one method for reducing the “file-drawer effect” inwhich negative findings are not presented to the public.Given the gaps and methodologic flaws in the literature

focused on psychotherapy interventions for suicide pre-vention, additional research is needed to determine theefficacy of existing and future treatments. Thus, we proposea general step-by-step research pathway for conductingfuture RCTs with high-risk patients for examining theefficacy of new psychotherapy treatments (Figure 1).The first step of this paradigm is to identify high-risk

subjects by using agreed-upon nomenclature (e.g., CDCnomenclature) as well as validated and reliable assess-ment measures. These high-risk patients can be recruitedfrom a variety of settings including emergency depart-ments, inpatient units, mental health outpatient clinics,and primary care. Following recruitment and initialassessment to determine eligibility, it is recommendedthat patients be randomly assigned to either (1) theco-active intervention condition, which may includemedication, treatment as usual, a comparative therapy,or follow-up services, or (2) the same co-active interven-tion plus a suicide-specific study intervention condition.Alternatively, depending on the question of interest, it

may be more appropriate to omit the co-active inter-vention for participants who are randomized to thesuicide-specific study intervention condition. In orderto gain an understanding of the pathways by which

Figure 1. Proposed step-by-step research pathway for conducting RCTsED, emergency department; MH, mental health; PC, primary care; SDV, self-directed violence; Ss, subjects

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194 S193

September 2014

treatment affects the outcome of interest (i.e., suicidalideation, suicide attempts, or suicides), it is imperative toexamine moderators of the study treatment and potentialmechanisms of actions. Elucidating the moderators andmechanisms at play will inform the development of moreefficient and targeted future interventions. This paradigmwill also allow for increased understanding of the relationbetween reductions in suicidal ideation and reductions insuicide attempts or deaths by suicide.

ConclusionsDespite important advances in the development andevaluation of psychotherapeutic treatments for suicideprevention, additional research is needed to improve thecurrent state of the science. A focus on filling the gaps inthe literature and increasing methodologic rigor withwhich RCTs of suicide-prevention psychotherapies areconducted will lead to increasingly effective treatmentsfor reducing suicidal ideation, attempts, and deaths.

Publication of this article was supported by the Centers forDisease Control and Prevention, the National Institutes ofHealth Office of Behavioral and Social Sciences, and theNational Institutes of Health Office of Disease Prevention.This support was provided as part of the National Institute ofMental Health-staffed Research Prioritization Task Force ofthe National Action Alliance for Suicide Prevention.

No financial disclosures were reported by the authors ofthis paper.

References1. USDHHS Office of the Surgeon General and National Action Alliance

for Suicide Prevention. 2012 National Strategy for Suicide Prevention:Goals and Objectives for Action. Washington, DC: HHS, September2012. http://www.surgeongeneral.gov/library/reports/national-strategy-suicide-prevention/full-report.pdf.

2. O’Neil M, Peterson K, Low A, et al. Suicide prevention interventionsand referral/follow-up services: a systematic review. Washington, DC:Department of Veterans Affairs, 2012.

3. Mann J, Apter A, Bertolote J, et al. Suicide prevention strategies: asystematic review. JAMA 2005;294(16):2064–74.

4. National Institute for Health andClinical Excellence. Self-harm: longer-termmanagement. (Clinical Guideline CG133). guidance.nice.org.uk/CG133.

5. O’Connor E, Gaynes BN, Burda BU, et al. Screening for and treatmentof suicide risk relevant to primary care: a systematic review for the U.S.Preventive Services Task Force. Ann Intern Med 2013;158(10):741–54.

6. Brown GK, Ten Have T, Henriques GR, et al. Cognitive therapy for theprevention of suicide attempts: a randomized controlled trial. JAMA2005;294(5):563–70.

7. Slee N, Garnefski N, van der Leeden R, et al. Cognitive-behaviouralintervention for self-harm: randomized controlled trial. Br J Psychiatry2008;192(3):202–11.

8. Linehan MM, Comtois KA, Murray AM, et al. Two-year randomizedcontrolled trial and follow-up of dialectical behavior therapy vs therapyby experts for suicidal behaviors and borderline personality disorder.

Arch Gen Psychiatry 2006;63(7):757–66.9. Hatcher S, Sharon C, Parag V, Collins N. Problem-solving therapy for

people who present to hospital with self-harm: Zelen randomisedcontrolled trial. Br J Psychiatry 2011;199(4):310–6.

10. Bateman A, Fonagy P. Effectiveness of partial hospitalization in thetreatment of borderline personality disorder: a randomized controlledtrial. Am J Psychiatry 1999;156(10):1563–9.

11. Guthrie E, Kapur N, Mackway-Jones K, et al. Randomised controlledtrial of brief psychological intervention after deliberate self-poisoning.BMJ 2001;323(7305):135–8.

12. McMain SF, Links PS, Gnam WH, et al. A randomized trialof dialectical behavior therapy versus general psychiatric managementfor borderline personality disorder. Am J Psychiatry 2009;166(12):136–74.

13. Blum N, St John D, Pfohl B, et al. Systems training for emotionalpredictability and problem solving (STEPPS) for outpatients withborderline personality disorder: a randomized controlled trial and1-year follow-up. Am J Psychiatry 2008;165(4):468–78.

14. Davidson K, Norrie J, Tyrer P, et al. The effectiveness of cognitivebehavior therapy for borderline personality disorder: results from theBorderline Personality Disorder Study of Cognitive Therapy(BOSCOT) trial. J Pers Disord 2006;20(5):450–65.

15. Wood A, Trainor G, Rothwell J, et al. Randomized trial of a grouptherapy for repeated deliberate self-harm in adolescents. J Am AcadChild Adolesc Psychiatry 2001;40(11):1246–53.

16. Huey S, Henggeler S, Rowland M, et al. Multisystemic therapy effectson attempted suicide by youths presenting psychiatric emergencies. JAm Acad Child Adolesc Psychiatry 2004;43(2):183–90.

17. Comtois K, Jones D, O’Connor S, et al. Collaborative assessment andmanagement of suicidality (CAMS): feasibility trial for next-dayappointment services. Depress Anxiety 2011;28(11):963–72.

18. Stewart C, Quinn A, Plever S, Emmerson B. Comparing cognitivebehavior therapy, problem solving therapy, and treatment as usual in ahigh-risk population. Suicide Life Threat Behav 2009;39(5):538–47.

19. Diamond G,WintersteenM, Brown GK, et al. Attachment-based familytherapy for adolescents with suicidal ideation: a randomized controlledtrial. J Am Acad Child Adolesc Psychiatry 2010;49(2):122–31.

20. Unützer J, Tang L, Oishi S, et al. Reducing suicidal ideation in depressedolder primary care patients. J Am Geriatr Soc 2006;54(10):1550–6.

21. Bruce M, Ten Have T, Reynolds Cr, et al. Reducing suicidal ideationand depressive symptoms in depressed older primary care patients: arandomized controlled trial. JAMA 2004;291(9):1081–91.

22. Pearson J, Stanley B, King C, Fisher C. Intervention research withpersons at high risk for suicidality: safety and ethical considerations. JClin Psychiatry 2001;62(25S):S17–S26.

23. Moher D, Schulz KF, Altman DG, CONSORT Group. Consoli-dated standards of reporting trials. Ann Intern Med 2001;134(8):657–62.

24. Crosby AE, Ortega L, Melanson C. Self-directed violence surveillance:uniform definitions and recommended data elements, version 1.0.Atlanta GA: CDC, 2011. www.cdc.gov/violenceprevention/pdf/Self-Directed-Violence-a.pdf.

25. Posner K, Brown GK, Stanley B, et al. The Columbia-Suicide SeverityRating Scale: initial validity and internal consistency findings fromthree multisite studies with adolescents and adults. Am J Psychiatry2011;168(12):1266–77.

26. Meltzer HY, Alphs L, Green AI, et al. Clozapine treatment forsuicidality in schizophrenia: International Suicide Prevention Trial(InterSePT). Arch Gen Psychiatry 2003;60(1):82–91.

27. Hoyert S, Smith B, Murphy S, Kochanek M. Deaths: final data for 1999.Natl Vital Stat Rep 2001;49(8):1–113.

28. CDC, National Center for Injury and Prevention Control. Web-BasedInjury Statistics Query and Reporting System (WISQARS). webappa.cdc.gov/sasweb/ncipc/mortrate10_sy.html.

Brown and Jager-Hyman / Am J Prev Med 2014;47(3S2):S186–S194S194

www.ajpmonline.org