990 www.thelancet.com/psychiatry Vol 11 December 2024
Articles
Psychological interventions to prevent the onset of major
depression in adults: a systematic review and individual
participant data meta-analysis
Claudia Buntrock, Mathias Harrer, Antonia A Sprenger, Susan Illing, Masatsugu Sakata, Toshi A Furukawa*, David D Ebert†, Pim Cuijpers*†,
on behalf of the IPD-PrevDep Consortium‡
Summary
Background Psychological interventions are increasingly discussed as a method to prevent major depressive disorder
(MDD) in adults who already experience subthreshold depressive symptoms. In this individual participant data meta-
analysis, we quantify the effect of preventive interventions against control on MDD onset in this population, and
explore effect modifiers.
Methods In this systematic review and individual participant data meta-analysis, we screened full-texts of eligible
studies within the Metapsy research domain for articles on psychological interventions for depression, from
database inception to May 1, 2023, published in English, German, Spanish, and Dutch. We included individual
participant data of randomised trials comparing psychological interventions with a control group regarding their
effects on MDD onset in adults with subthreshold depressive symptoms but no MDD at baseline, confirmed by
standardised diagnostic interviews. Risk of bias was assessed using the RoB 2 tool. Effect on the onset of MDD (the
primary outcome) and moderators were analysed using one-stage individual participant data meta-analysis. Survival
analyses were conducted to examine effects on time to MDD onset within 12 months. We involved people with
related lived experience in the study design and implementation. This study is registered with PROSPERO,
CRD42017058585.
Findings 
30 of 42 eligible randomised controlled trials with 7201 participants (2227 [30·9%] male, 4957 [68·9%]
female, and 17 [0·2%] preferred not to report their sex) were included in our analysis (3697 participants had
intervention and 3504 participants had control). The mean age of participants was 49·9 years (SD 19·2). Of the
3152 participants with reported ethnicity, 1608 (51·0%) were White. Five studies received a high risk of bias rating.
Psychological interventions were associated with significantly reduced MDD incidence at post-treatment (incidence
rate ratio [IRR] 0·57 [95% CI 0·35–0·93]; τ²=0·29; 18 studies), within 6 months (0·58 [0·39–0·88]; τ²=0·11; 18 studies),
and within 12 months (0·67 [0·51–0·88]; τ²=0·05; 19 studies). No significant effect was observed at 24 months
(IRR 1·16 [95% CI 0·66–2·03]; τ²=0·10; six studies). Preventive effects were stronger for individuals who had not
previously had psychotherapy (IRR 0·39 [95% CI 0·25–0·62]) compared with those who had received psychotherapy
before (0·92 [0·61–1·36]; p=0·029; seven studies). Although no overall linear association was identified, higher
baseline depressive (Patient Health Questionnaire-9) and anxiety symptom (Generalized Anxiety Disorder-7) scores
were associated with greater reductions in MDD onset risk. On the study level, delivery type appeared to moderate
outcomes, with conference telephone calls being more effective than delivery via face-to-face, internet-based, and
other formats (p=0·002), albeit based on only two studies of conference telephone calls with four comparisons. Other
factors (eg, age) showed no significant differential effects.
Interpretation Our findings show the effectiveness of preventive psychological interventions for subthreshold
depressive symptoms. Tailoring interventions to consider participant-level and study-level factors could help to
increase the impact of such interventions on a population level.
Funding None.
Copyright © 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND
4.0 license.
Introduction
Major depressive disorder (MDD) ranks among the most
prevalent mental health conditions globally,
1 
and its
worldwide prevalence has increased in the past 20 years.
2
MDD accounts for 7·5% of all years lived with disability,
3
and imposes a substantial economic burden on society,
estimated at US$326·2 billion annually in the USA
alone.
4
Despite the availability of effective first-line treatments
for MDD, including pharmacotherapy, psychotherapy,
and digital interventions (eg, programme delivered via
the web or mobile applications), their impact in adults is
Lancet Psychiatry 2024;
11: 990–1001
See Comment page 947
*Toshi Furukawa and
Pim Cuijpers are retired
†Contributed equally as last
authors
‡Members of the IPD-PrevDep
Consortium are listed at the end
of this Article
Institute of Social Medicine
and Health Systems Research,
Medical Faculty, Otto-von-
Guericke-University
Magdeburg, Magdeburg,
Germany (C Buntrock PhD,
A A Sprenger 
MSc); Psychology
and Digital Mental Health Care
Research Unit, School of
Medicine and Health, Technical
University of Munich, Munich,
Germany (M Harrer 
MSc,
S Illing BA, D D Ebert PhD);
Department of
Neurodevelopmental
Disorders, Nagoya City
University, Nagoya, Japan
(M Sakata 
PhD); Department of
Health Promotion and Human
Behavior, Kyoto University
Graduate School of Medicine
and School of Public Health,
Kyoto, Japan (M Sakata); Kyoto
University Office of
Institutional Advancement and
Communications, Kyoto, Japan
(Prof T A Furukawa PhD);
Department of Clinical Neuro-
and Developmental
Psychology, VU University
Amsterdam, Amsterdam,
Netherlands
(Prof P Cuijpers PhD)
Correspondence to:
Dr Claudia Buntrock, Institute of
Social Medicine and Health
Systems Research, Medical
Faculty, Otto-von-Guericke-
University Magdeburg,
39120 Magdeburg, Germany
claudia.buntrock@med.ovgu.
de

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 991
modest.
5,6 
A modelling study suggests that even with
optimal coverage of evidence-based treatments, the
overall disease burden of MDD can only be reduced by
one-third.
7 
Similarly, real-world evidence indicates that
only about 16·5% of patients with depression, even in
high-income countries, receive minimally adequate
treatment.
8
Effective prevention approaches, with the potential to
mitigate initial episodes and decrease recurrent episodes
of MDD, are needed to reduce the overall prevalence of
MDD.
9 
Indicated prevention, targeting individuals
already exhibiting subthreshold symptoms of depression,
has garnered particular attention. This focus is driven by
the estimated 11% prevalence rate of subthreshold
symptoms of depression in the general population,
10
coupled with its comparable adverse effects on quality of
life as MDD.
11 
Notably, individuals with subthreshold
symptoms of depression have about a three-times greater
risk of developing MDD compared with individuals
without symptoms of depression.
10
Psychological interventions reduce the relative
incidence of MDD in adults by 19% within a year, but a
notable proportion of participants still develop MDD.
12
Advancements in prevention efforts can be made by
identifying participant characteristics to refine and
personalise interventions, improving outcomes for
individuals with subthreshold depression symptoms.
However, existing evidence is conflicting; a systematic
review found moderate evidence that effectiveness of
psychological interventions for depression prevention
increased as participants’ age decreased.
13 
By contrast, an
individual participant data meta-analysis
14 
examined
internet-based preventive interventions for subthreshold
depressive symptoms and found that the effects on
symptom severity became more pronounced with
increasing age. In addition, the effectiveness of these
interventions increased with the severity of the initial
depressive symptoms.
14 
However, this analysis did not
assess effect modifiers for MDD onset and included only
seven studies, limiting its generalisability beyond internet
interventions. Our study aimed to investigate the effects
of various psychological interventions versus control
conditions on both MDD onset and symptom severity in
adults with subthreshold depressive symptoms assessed
in randomised controlled trials, while assessing effect
modifiers for MDD onset at both participant and study
levels.
Methods
Search strategy and selection criteria
This systematic review and individual participant
data meta-analysis is registered with PROSPERO,
CRD42017058585). Analyses are reported according to the
Preferred Reporting Items for Systematic Review and
Research in context
Evidence before this study
We searched the National Library of Medicine via PubMed, from
database inception to May 1, 2024, for peer-reviewed
publications in English using the search string
(“depression”[MeSH Terms] OR “depressive disorder”[MeSH
Terms]) AND (“prevention”[Title] OR “preventing”[Title] OR
“prevent”[Title] OR “subthreshold”[Title] OR “subclinical”[Title])
AND (“psychological”[Title/Abstract] OR “intervention”[Title/
Abstract]) AND “meta-analysis”[Title]. We identified
68 records, of which five were systematic reviews with meta-
analyses that, based on title and abstract, were eligible to report
on the effects of psychological interventions for depression
prevention outcomes (eg, odds ratio, relative risk, or hazard
ratio). Although these studies evaluated the effects of various
psychological interventions on the onset of major depressive
disorder (MDD), only one study used an individual participant
data meta-analysis to explore effect modifiers related to
depressive symptom severity, rather than MDD onset. This
analysis was restricted to internet-based interventions. As a
result, there was a scarcity of evidence from previous research
on effect modifiers for different types of psychological
interventions on MDD onset.
Added value of this study
To our knowledge, this is the first systematic review and
individual participant data meta-analysis to identify effect
modifiers of different types of psychological interventions on
MDD onset at both the individual level and study level. We
included data from 30 trials with 7201 participants. Our individual
participant data meta-analysis showed that psychological
interventions significantly reduced the incidence of MDD at post-
treatment and within 6 months and 12 months after the
intervention. However, the effect waned by the 24-month follow-
up. Furthermore, we found that no previous psychotherapy
experience resulted in greater effects in preventing MDD.
Individual characteristics, such as higher baseline depressive
(Patient Health Questionnaire-9 ≥10) or anxiety symptoms
(Generalised Anxiety Disorder-7 ≥10), appeared to be associated
with stronger preventive effects. Conference telephone calls,
although assessed in only two studies, appeared more effective
than face-to-face, internet-based, or other formats.
Implications of all the available evidence
Psychological interventions have the potential to prevent MDD
onset in individuals not currently experiencing a depressive
episode. Further investigations are needed to determine the
threshold at which depressive symptoms, at the lower end of
the severity spectrum, become persistent enough to warrant
preventive interventions. Future research and policy efforts
should prioritise strategies to integrate these interventions into
routine care settings, addressing the current implementation
gap.

Articles
992 www.thelancet.com/psychiatry Vol 11 December 2024
Meta-Analyses of individual participant data statement
(appendix pp 1–4).
15 
The rationale and methods of this
study are described in greater detail in a published
protocol.
16 
Deviations from the planned approach are
reported in the appendix (p 5). Ethical approval and
participant consent were obtained within the context of the
primary studies. We involved people with related lived
experience in the study design and implementation.
In this individual participant data meta-analysis, we
included randomised trials in which a psychological
intervention was compared with a comparison group
(waitlist, care as usual, placebo, or antidepressant
medication) with regard to effects on MDD onset (ie, the
incidence of new episodes of MDD during the study
period as confirmed by a standardised diagnostic
interview), in adults aged 18 years and older with
subthreshold depressive symptoms but no MDD at
baseline, as confirmed by a standardised diagnostic
interview. Studies that met the eligibility criteria outlined
in the study protocol but used self-report measures
instead of diagnostic interviews during follow-up were
excluded, as they did not provide data on the primary
outcome (ie, onset of MDD). Having subthreshold
symptoms of depression was defined as scoring higher
than a cutoff score on a self-rating depression
questionnaire; scoring higher than a cutoff score on a
clinician-rated instrument; or meeting criteria for minor
depression according to the DSM-IV, or the ICD.
Psychological intervention was defined as the “application
of psychological mechanisms and interpersonal stances
derived from psychological principles for the purpose of
assisting people to modify their behaviours, cognitions,
emotions and/or other personal characteristics in
directions that the participants deem desirable”.
16
In studies without an elevated depressive symptom
threshold for inclusion, only participants with at least
mild depressive symptoms at baseline (Patient Health
Questionnaire [PHQ]-9 score ≥5) were included. Scores
from other symptom questionnaires were converted to
PHQ-9 values using a common metric.
17
To identify eligible studies, two independent researchers
screened the full texts of the Metapsy database on
psychological interventions for depression. The Metapsy
database includes articles in English, German, Spanish,
and Dutch. Disagreements were resolved by a senior
researcher. The database is updated three times a year
through systematic searches of PubMed, Embase,
PsycINFO, and Cochrane Central. Full search strings are
provided in the appendix (pp 6–9). In each update,
two independent researchers screened titles, abstracts, and
full texts of eligible studies, resolving disagreements by
consensus. Additionally, previous systematic reviews and
meta-analyses on MDD prevention were reviewed, and
experts were consulted for other relevant studies. Studies
published up to May 1, 2023, were included.
Corresponding authors of all eligible articles were
contacted to request permission to use their data, with
reminders sent after 2 weeks and 1 month. If there was
no response, the trial was excluded. Responding authors
provided data on demographic, clinical, outcome, and
intervention-related characteristics, with variables
selected based on a pre-defined list of predictors of
long-term outcomes in depression (appendix p 10).
16 
Data
were collated centrally by independent researchers (MH,
AS, and SI), and cross-checked with the trial publications.
Depressive symptom severity measures were trans-
formed into a common metric using a partial credit
model to facilitate joint analyses.
17 
The harmonised
individual participant data were merged into a single
dataset after a standardised protocol. Post-intervention
assessments were treated as one assessment point, and
follow-ups were categorised by length (up to 6 months,
12 months, or 24 months). For studies that did not
provide individual participant data, suitable outcome
data (ie, number of participants with and without MDD
at any given timepoint after intervention), were extracted
for a conventional aggregate data meta-analysis if
available in the published report.
Risk of bias was planned to be assessed using version 1
of the Cochrane risk of bias (RoB) tool. An updated
version of the tool (RoB 2
18
) has been released and was
used to assess the randomisation process, deviation from
the intervention, missing outcome data, measurement of
the outcome, and selection of the reported results. All
studies were rated as being at low risk of bias for missing
outcome data, since multiple imputation with auxiliary
variables could be used to handle missing data
consistently in this individual participant data meta-
analysis, and missingness at random was assumed to be
plausible.
Outcomes
The primary outcome was the onset of a MDD (first or
recurrent episode). MDD cases after treatment and
during follow-up (up to 6 months, 12 months, and
24 months) had to be confirmed using clinical interviews.
Additionally, we examined effects on time to MDD onset.
Secondary outcomes were depressive symptom severity
(transformed into common metrics), 50% symptom
reduction compared with baseline, near symptom-free
status (defined as scores equivalent to PHQ-9 <5
19
), as
well as reliable improvement and reliable deterioration
in depressive symptoms, which were determined using
the reliable change index (RCI). We did not include
quality of life, anxiety, or suicidal thoughts and behaviour
as secondary outcomes due to insufficient available data.
Data analysis
All analyses followed the intention-to-treat principle.
Missing data were handled using multiple imputation
(fully conditional specification; MICE algorithm) under
the missing at random assumption. Multilevel two-stage
imputation models with heteroscedastic errors were used
to account for the nested data structure
20 
(appendix p 11).
See Online for appendix
For more on psychological
interventions for depression
from Metapsy see https://docs.
metapsy.org/databases/
depression-psyctr/

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 993
Highly collinear variables and variables with systematically
missing information (structural zeros) were excluded as
predictors. A total of m=50 imputation sets were generated.
For moderator analyses, substantive model compatible
fully conditional specification was used separately for each
study and putative moderator.
21 
These models included a
treatment-covariate interaction with the examined
moderator variable as well as auxiliary variables.
Using one-stage individual participant data meta-
analysis, we calculated pooled effects on MDD onset,
depressive symptom severity, 50% symptom reduction,
near symptom-free status, reliable improvement and
reliable deterioration (according to RCI) at post-treatment
and follow-up (ie, up to 6 months, 12 months, and
24 months). Generalised linear mixed models were used
for all analyses. Poisson models were used for MDD onset,
with the study-specific observation period serving as an
offset. A normal linear model with stratified trial intercepts
and trial-specific error terms was used for depressive
symptom severity, and a binomial logit-link model was
used for the other outcomes. All models were adjusted for
baseline symptom severity, centred around trial means,
and fitted in the multiply imputed data, with final estimates
aggregated using Rubin’s rules. Marginal incidence rate
ratios (IRRs) and relative risk (RR) estimates were obtained
using G-computation,
22 
and standardised mean differences
were calculated for depressive symptom severity using the
pooled endpoint SD. Number-needed-to-treat values were
derived from standardised mean differences
23 
with the
control group event rate obtained from the individual
participant data. We used I² to quantify the between-study
heterogeneity (appendix p 11). As a sensitivity analysis, we
also used two-stage individual participant data meta-
analysis models with a log-normal model and a continuity
correction of 0·5, using the restricted maximum likelihood
estimator for the between-study heterogeneity variance τ².
Conventional meta-analyses, including studies without
individual participant data, were also conducted for MDD
onset and depressive symptom severity, with subgroup
analyses comparing studies that provided data with those
that did not. Survival analysis was conducted for time to
MDD onset (expressed in weeks). The Kaplan–Meier
estimator was used to calculate survival functions by
treatment status within a 12-month study period.
MDD-free participants were censored at the time of study
termination or loss to follow-up, whichever occurred first.
Non-informative censoring was assumed. Mixed-effects
Cox regression models with stratified trial intercepts were
fitted to estimate the pooled treatment effect on time to
MDD onset. Hazard ratios (HRs) and 95% CIs were
derived based on the model’s estimated intervention effect.
We evaluated the proportional hazards assumption by
examining the scaled Schoenfeld residuals. We used a
delta-adjustment approach to control for the possibility
that data are missing not at random. Data were imputed
under the assumption that, in each trial, the incidence of
MDD among participants lost to follow-up who had
received the intervention was 5% to 30% higher than
predicted by the main imputation model.
We examined moderators of the effect on MDD onset by
including treatment-covariate interaction terms (including
main effects) into the Poisson model. Moderator analyses
were only conducted for MDD onset at the first available
assessment. As putative effect modifiers on participant
level, sex, age, ethnicity, education, employment status,
relationship status, baseline depressive symptom severity
(PHQ-9) and anxiety symptom severity (Generalized
Anxiety Disorder [GAD]-7), presence of chronic medical
conditions, history of MDD, intake of anti-depressive
medication, and previous psychotherapy were explored.
After identifying eligible studies, we screened the articles
for prespecified effect modifiers and included only those
for which we expected to obtain sufficient data. We used
continuous variables in their original form to preserve the
full range of information, as dichotomising or categorising
them can lead to a loss of statistical power and precision.
However, in some instances, variables were dichotomised
to simplify the analysis, necessitated by inconsistencies in
data collection across studies. As study-level variables, we
examined if the country of origin, publication year, type of
delivery, intervention type, type of control condition, or
risk of bias predicted differential intervention effects. For
significant continuous moderators, we also estimated the
conditional treatment effects on MDD incidence at
representative values of the covariate (ie, psychometrically
validated cut-off scores). Analyses were conducted using R
(version 4.2.0).
Role of the funding source
There was no funding source for this study.
Results
Of 1013 full-text articles screened, 42 were determined to
be eligible for this study figure 1. Individual participant
data could be obtained from 30 (71·43%) of all eligible
trials. No major issues were identified during the
assessment of individual participant data. Of the studies
from which no individual participant data could be
obtained, aggregate data were available for ten studies.
This finding corresponds to 1190 additional participants
(599 with intervention and 591 with control) who were
eligible for combined analysis of individual participant
data and aggregate data. References of the included
studies are provided in the appendix (pp 12–15).
Study characteristics are summarised in the
appendix (pp 16–30). 26 studies (86·7%) were conducted
in high-income countries, nine (30·0%) in general adult
populations, nine (30·0%) in older adults (with
two studies conducted in nursing home residents),
three (10·0%) in pregnant women, four (13·3%) in
university students, three (10·0%) in non-professional
caregivers, one (3·3%) in employees, and one (3·3%) in
patients with diabetes. Most interventions were based on
cognitive behavioural therapy (16 [53·3%] studies) and

Articles
994 www.thelancet.com/psychiatry Vol 11 December 2024
Figure 1: Study selection
MDD=major depressive disorder.
33 825 studies identified through database searching
23 896 studies after duplicates removed
23 896 studies screened for eligibility
1013 studies included in database
42 studies for which individual participant data were
sought
30 studies for which individual participant data were
provided
7767 participants for whom data were provided
0 participants for whom
no data were provided
Individual participant data
30 studies included in analysis
7201 participants included in analysis
566 participants excluded
566 depressive symptoms too low, MDD at
baseline, or age <18 years
22 883 reports excluded
951 companion papers
587 depression not an inclusion criterion
27 dissertations
120 effect sizes cannot be estimated
143 maintenance trial
78 no control condition
320 no psychotherapy for depression
92 no random assignment
101 not available
254 other
23 other language
410 protocol paper
19 
777 title and abstract not suitable
975 reports excluded
3 companion papers
10 full-text unavailable
645 MDD at baseline
9 no clinically relevant depressive symptoms
105 no comparison of psychotherapy versus
control
77 no diagnostic interview at baseline
31 no diagnostic interview at follow-up
10 no randomised controlled trial
7 other
33 sample size without MDD too small
45 studies with children or adolescents
12 studies for which individual participant data were
not provided
1 communication interrupted or ongoing
4 data rejected
7 no reply
1422 participants
10 studies for which aggregate data were available
1190 participants
Aggregate data
10 studies included in analysis
1190 participants included in analysis
0 participants excluded
65 records identified from hand search and expert
consultation
61 studies after duplicates removed
61 studies screened for eligibility
57 reports excluded
1 data not available
32 MDD at baseline
6 no adequate measurement instrument
1 no clinically relevant depressive symptoms
6 no diagnostic interview at baseline
1 sample size without MDD too small
10 studies with children or adolescents
142 additional studies identified through other sources

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 995
were provided in a face-to-face format (12 [40·0%]
studies). We included 7201 participants (3697 with
intervention and 3504 with control) with a mean age
of 49·9 years (SD 19·2). Of the participants,
2227 (30·9%) were male, 4957 (68·9%) female, and
17 (0·2%) preferred not to record their sex. Of the
3152 participants with reported ethnicity, 1608 (51·0%)
were White. Of the 6605 participants who reported their
educational level, 3305 (50·0%) had a higher education.
Participant characteristics at baseline are provided in the
appendix (p 31).
Risk of bias assessments for each included study are
displayed in the appendix (pp 16–22). The overall risk of
bias was low. With one exception, all studies used an
adequate sequence generation mechanism (domain 1)
and showed low risk of bias due to deviations from the
intended interventions (domain 2). An intention-to-treat
analysis could be performed in all trials (domain 3).
Effect size
(95% CI)*
Number
of
studies
τ² 
I² t 
ν 
p† 
FMI 
NNT (95% CI) 
Event rate
Control* Intervention*
MDD incidence (incidence rate ratio)
Post-treatment 0·57
(0·35 to 0·93)
18 0·292 40·8% –2·588 872·8 0·010 0·239 19·93
(8·72 to 69·87)
15·0%
(3·7 to 26·3)
10·0%
(0·8 to 19·3)
Up to 6 months
follow-up
0·58
(0·39 to 0·88)
18 0·110 21·6% –2·340 258·5 0·020 0·440 37·81
(18·43 to 728·83)
8·7%
(3·3 to 14·2)
6·1%
(0·6 to 11·6)
Up to 12 months
follow-up
0·67
(0·51 to 0·88)
19 0·052 16·3% –2·954 872·0 0·003 0·239 20·54
(11·74 to 82·09)
14·0%
(8·1 to 19·9)
9·1%
(5·0 to 13·2)
Up to 24 months
follow-up
1·16
(0·66 to 2·03)
6 0·098 19·4% 0·398 6337·7 0·691 0·088 388·52
(23·39 to 20·88)
7·7%
(2·4 to 13·0)
7·9%
(3·8 to 12·1)
Symptom severity, g
Post-treatment –0·49
(–0·66 to –0·32)
28 0·176 90·5% –5·572 
20 988·4 <0·001 0·048 7·00
(4·88 to 11·57)
– –
Up to 6 months
follow-up
–0·26
(–0·41 to –0·11)
23 0·082 82·3% –3·412 1455·1 0·001 0·185 20·68
(12·01 to 53·49)
– –
Up to 12 months
follow-up
–0·27
(–0·40 to –0·14)
23 0·077 79·4% –4·027 
38 377·9 <0·001 0·036 14·39
(9·12 to 29·90)
– –
Up to 24 months
follow-up
–0·14
(–0·32 to 0·04)
11 0·055 66·9% –1·523 2985·4 0·128 0·129 44·30
(17·15 to 165·32)
– –
50% symptom reduction (RR)
Post-treatment 1·86
(1·43 to 2·41)
28 0·772 89·4% 5·225 
27 127·9 <0·001 0·043 5·19
(8·48 to 3·74)
24·5%
(18·9 to 30·1)
43·8%
(37·3 to 50·3)
Up to 6 months
follow-up
1·47
(1·23 to 1·75)
23 0·266 79·1% 4·109 1816·4 <0·001 0·165 7·45
(13·32 to 5·17)
31·1%
(25·5 to 36·8)
44·6%
(38·0 to 51·1)
Up to 12 months
follow-up
1·37
(1·17 to 1·60)
23 0·254 78·6% 3·757 
13 590·3 <0·001 0·060 8·31
(15·98 to 5·62)
36·0%
(30·2 to 41·9)
48·1%
(42·7 to 53·4)
Up to 24 months
follow-up
1·27
(1·04 to 1·54)
11 0·193 74·1% 1·059 4776·3 0·290 0·102 9·71
(51·98 to 5·36)
45·1%
(35·3 to 54·9)
55·4%
(47·1 to 63·7)
Near symptom-free status (RR)
Post-treatment 1·71
(1·33 to 2·20)
28 0·930 94·5% 4·506 
35 615·8 <0·001 0·037 5·96
(10·73 to 4·13)
34·3%
(27·2 to 41·3)
51·0%
(45·1 to 57·0)
Up to 6 months
follow-up
1·34
(1·13 to 1·59)
23 0·404 90·4% 3·020 4361·1 0·003 0·106 9·02
(20·73 to 5·77)
41·7%
(34·9 to 48·6)
52·8%
(45·8 to 59·8)
Up to 12 months
follow-up
1·32
(1·12 to 1·55)
23 0·431 88·2% 3·141 
42 319·5 0·002 0·034 9·09
(21·31 to 5·78)
45·9%
(38·7 to 53·2)
56·9%
(50·7 to 63·2)
Up to 24 months
follow-up
1·23
(1·03 to 1·48)
11 0·241 81·3% 1·226 2730·9 0·220 0·135 9·12
(37·21 to 5·19)
56·9%
(43·2 to 70·7)
67·9%
(57·6 to 78·2)
Reliable improvement (RR)
Post-treatment 1·72
(1·38 to 2·14)
28 0·850 91·2% 5·282 
49 702·5 <0·001 0·031 5·37
(8·89 to 3·85)
30·7%
(25·9 to 35·4)
49·3%
(41·3 to 57·4)
Up to 6 months
follow-up
1·46
(1·26 to 1·70)
23 0·192 73·7% 5·023 1891·7 <0·001 0·162 8·01
(12·83 to 5·83)
35·3%
(29·7 to 40·9)
47·8%
(41·4 to 54·1)
Up to 12 months
follow-up
1·42
(1·26 to 1·61)
23 0·120 65·2% 5·721 4351·7 <0·001 0·107 8·10
(12·18 to 6·06)
38·2%
(32·9 to 43·5)
50·5%
(44·5 to 56·6)
Up to 24 months
follow-up
1·33
(1·07 to 1·67)
11 0·234 77·4% 1·645 4178·1 0·100 0·109 10·41
(40·87 to 5·96)
42·5%
(34·3 to 50·7)
52·1%
(43·7 to 60·4)
(Table 1 continues on next page)

Articles
996 www.thelancet.com/psychiatry Vol 11 December 2024
Four (13·3%) studies showed high risk of bias in
measurement of the outcome (domain 4). With
one exception, all studies had a low risk for selective
reporting (domain 5).
Proportions of missing outcome data are provided in
the appendix (pp 32–34). An overview of results is
presented in table 1. We found that psychological
interventions reduced the incidence of major depression
after treatment (IRR 0·57 [95% CI 0·35–0·93]; 18 studies),
within 6 months (0·58 [0·39–0·88]; 18 studies), and
within 12 months (0·67 [0·51–0·88]; 19 studies). This
finding means that interventions reduced the incidence of
depression by 43%, 42%, and 33%, respectively, compared
with control. The few studies which recorded the incidence
of depression up to 24 months did not yield a significant
pooled effect (IRR 1·16 [95% CI 0·66–2·03]; six studies). A
forest plot detailing the results on depression incidence is
provided in figure 2.
Similar findings emerged for all other outcomes. At post-
treatment (n=28 studies), within 6 months (23 studies),
and within 12 months (23 studies), we found that
interventions led to a reduction in depressive symptom
severity (standardised mean difference –0·49 [95% CI–0·66
to –0·32] to –0·26 [–0·41 to –0·11]), an increase in the
number of participants with 50% symptom reduction
(RR 1·37 [95% CI 1·17–1·60] to 1·86 [1·43–2·41]),
symptom-free status (1·32 [1·12–1·55] to 1·71 [1·33–2·20]),
and reliable improvement (1·42 [1·26–1·61] to 1·72
[1·38–2·14]), as well as a decrease in reliable symptom
deterioration (0·48 [0·36–0·63] to 0·65 [0·50–0·86]),
compared with controls. No significant effects were
observed at 24-month follow-up (all p≥0·05; 11 studies).
Between-study heterogeneity was moderate to high in
most analyses (I² range 19·4% to 94·5%; table 1).
Analyses using two-stage models closely mirrored the
main findings (appendix p 35). Results from combined
analyses of individual participant data and aggregate data
studies showed largely similar effect estimates, for both
for MDD onset and depressive symptom severity
(appendix p 36). The effect on depression incidence up to
24 months was not significant (IRR 0·83 [95% CI
0·61 to 1·13]; eight studies). Effects of studies that
provided individual participant data did not differ
significantly from those of studies that did not for all
analysed assessment points and outcomes (all p>0·05).
Effects on 12-month incidence remained robust under all
examined missing-not-at-random scenarios. At post-test
and up to 6 months, effects remained significant under
minor deviations from missingness at random (appendix
pp 37–38).
Eight (26·7%) of the 30 studies provided time to MDD
onset data suitable for a pooled survival analysis
(2670 participants; 1337 with intervention and 1333 with
control). In the intervention groups, 227 (17·0%)
participants experienced MDD onset during the study
period, compared with 301 (22·6%) in the control groups.
Kaplan–Meier survival curves for the intervention and
control groups across all trials are presented in figure 3.
The 6-month incidence rate of MDD onset was
13·4% (95% CI 11·5–15·6) in the intervention groups
versus 17·9% (15·8–20·3) in the control groups; whereas
the 12-month incidence rate in the intervention groups
was 21·4% (19·1–24·1) compared with 27·4% (24·8–30·2)
in the control groups. Using mixed-effects Cox regression
controlling for baseline depressive symptom severity, the
pooled preventive effect of the interventions was
estimated at an HR of 0·76 (95% CI 0·64–0·90). This
effect was significant (p<0·001) and means that, given
individuals with the same initial depressive symptom
severity, provision of the intervention reduces the hazard
of MDD onset within 12 months by 24%.
On a participant level, individuals who had not
previously undergone psychotherapy showed stronger
preventive effects against MDD onset than individuals
Effect size
(95% CI)*
Number
of
studies
τ² 
I² t 
ν 
p† 
FMI 
NNT (95% CI) 
Event rate
Control* Intervention*
(Continued from previous page)
Reliable deterioration (RR)
Post-treatment 0·48
(0·36 to 0·63)
28 0·152 32·5% –5·222 1987·1 <0·001 0·158 18·46
(13·17 to 30·87)
11·2%
(8·9 to 13·5)
5·8%
(3·9 to 7·7)
Up to 6 months
follow-up
0·65
(0·50 to 0·86)
23 0·163 39·8% –2·632 1368·2 0·009 0·190 38·79
(21·28 to 219·09)
10·5%
(8·1 to 12·8)
7·9%
(4·7 to 11·0)
Up to 12 months
follow-up
0·59
(0·46 to 0·76)
23 0·029 8·8% –4·135 686·2 <0·001 0·269 27·30
(18·70 to 50·59)
9·8%
(7·5 to 12·1)
6·1%
(4·3 to 8·0)
Up to 24 months
follow-up
0·59
(0·34 to 1·03)
11 0·432 53·1% –1·366 2216·1 0·172 0·149 23·20
(11·00 to 212·02)
10·3%
(4·8 to 15·8)
6·0%
(3·1 to 8·8)
FMI=fraction of missing information due to non-response. MDD=major depressive disorder. NNT=number needed to treat. RR=risk ratio. τ
2
=between-study heterogeneity variance. I
2
=between-study
heterogeneity. ν=degrees of freedom (multiple imputation large-sample approximation by Rubin, 1987, equation 3.1.6). *Calculated using regression standardisation (G-computation). CIs around these
marginal effect estimates are generated using the delta method and can diverge in their interpretation of significance from the t-test of the intervention effect as measured by the treatment indicator coefficient.
†Test of the treatment indicator coefficient (one-stage individual participant data meta-analysis model)
Table 1: Pooled effects on MDD incidence, symptom severity, response, and reliable deterioration

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 997
who had received psychotherapy before (seven studies,
e
β 
2·292 [SE 0·38], p=0·029). A marginal IRR of
0·92 (95% CI 0·61–1·36) was computed for patients with
psychotherapy experience, compared with an IRR of
0·39 (95% CI 0·25–0·62) for those without. No
moderator effects were found for sex, age, ethnicity,
education, employment, relationship status, chronic
medical conditions, MDD history, or antidepressant use.
No significant linear relationship was found between
baseline depressive symptom severity and effects on
MDD onset (30 studies; e
β 
1·016 [SE 0·128]; p=0·900;
table 2). However, estimated marginal intervention
effects appeared stronger for higher baseline PHQ-9
scores: IRR 0·72 (95% CI 0·53–0·99) at a score of 5 (mild
depressive symptoms), 0·56 (0·41–0·76) at score 10
(moderate depressive symptoms), and 0·59 (0·45–0·78)
Incidence rate ratio
(95% CI)
Post-test
Vázquez et al, 2017b
Le et al, 2011
Yang et al, 2015
Vázquez et al, 2022b
Albert et al, 2019
Basanovic et al, 2019
Vázquez et al, 2022a
Otero et al, 2014
Irwin et al, 2022
Lara et al, 2010
Vázquez et al, 2017a
Rovner et al, 2007
Dozeman et al, 2012
van't Veer−Tazelaar et al, 2009
Klein et al, 2016
Muñoz et al, 2007
Cook et al, 2019
Konnert et al, 2009
One−stage model
Two−stage model
Up to 6 months follow-up
Cook et al, 2019
Imamura et al, 2014
Otero et al, 2014
Irwin et al, 2022
Buntrock et al, 2015
Yang et al, 2015
Pols et al, 2017
Dozeman et al, 2012
Batterham et al, 2017
Rovner et al, 2007
Lara et al, 2010
Apil et al, 2014
Almeida et al, 2020
Zhang et al, 2014
Vázquez et al, 2012
Basanovic et al, 2019
Konnert et al, 2009
Albert et al, 2019
One-stage model
Two-stage model
29
217
77
109
102
202
110
173
181
315
32
206
179
170
723
41
159
58
159
421
173
181
406
77
223
179
1149
206
315
91
307
240
133
202
58
102
Number of
participants
0·11 (0·00–2·44)
0·13 (0·01–2·63)
0·20 (0·07–0·58)
0·20 (0·03–1·32)
0·28 (0·02–5·05)
0·30 (0·04–2·61)
0·32 (0·05–2·19)
0·37 (0·12–1·14)
0·40 (0·06–2·75)
0·53 (0·17–1·61)
0·53 (0·01–25·91)
0·54 (0·24–1·21)
0·84 (0·16–4·32)
0·86 (0·28–2·61)
0·91 (0·54–1·53)
1·15 (0·46–2·88)
1·42 (0·38–5·33)
1·82 (0·07–44·51)
0·57 (0·35–0·93)
0·58 (0·41–0·82)
0·14 (0·01–2·38)
0·27 (0·08–0·91)
0·27 (0·09–0·82)
0·40 (0·06–2·75)
0·46 (0·18–1·19)
0·46 (0·17–1·23)
0·60 (0·18–2·02)
0·64 (0·07–5·90)
0·70 (0·30–1·63)
0·76 (0·40–1·44)
0·77 (0·15–4·00)
0·80 (0·31–2·08)
0·91 (0·21–3·90)
0·98 (0·10–9·59)
1·07 (0·33–3·48)
1·17 (0·21–6·58)
1·31 (0·08–22·51)
3·66 (0·23–59·41)
0·58 (0·39–0·88)
0·62 (0·47–0·83)
Up to 12 months follow-up
Yang et al, 2015
Otero et al, 2014
Irwin et al, 2022
Muñoz et al, 2007
Sander et al, 2020
Wong et al, 2018
Almeida et al, 2020
Zhang et al, 2014
van't Veer−Tazelaar et al, 2009
Buntrock et al, 2015
Willemse et al, 2004
Klein et al, 2016
Basanovic et al, 2019
Albert et al, 2019
Konnert et al, 2009
Karyotaki et al, 2022
Allart et al, 2007
Pols et al, 2017
Apil et al, 2014
One-stage model
Two-stage model
Up to 24 months follow-up
Apil et al, 2014
Cook et al, 2019
Irwin et al, 2022
Reynolds et al, 2014
Albert et al, 2019
Zhang et al, 2014
One-stage model
Two-stage model
77
173
181
41
295
231
307
240
170
406
216
723
202
102
58
48
111
223
91
91
159
181
247
102
240
0·41 (0·12–1·43)
0·42 (0·18–0·95)
0·42 (0·09–1·93)
0·43 (0·09–2·17)
0·48 (0·28–0·85)
0·49 (0·20–1·21)
0·52 (0·11–2·44)
0·55 (0·11–2·62)
0·57 (0·25–1·29)
0·63 (0·29–1·39)
0·67 (0·32–1·40)
0·84 (0·44–1·59)
0·98 (0·25–3·82)
1·00 (0·11–8·93)
1·00 (0·04–27·98)
1·03 (0·12–9·14)
1·08 (0·45–2·59)
1·22 (0·32–4·59)
1·31 (0·42–4·12)
0·67 (0·51–0·88)
0·64 (0·51–0·80)
0·70 (0·24–2·07)
0·71 (0·18–2·76)
0·93 (0·30–2·85)
1·07 (0·45–2·56)
2·25 (0·14–36·69)
2·56 (0·58–11·24)
1·16 (0·66–2·03)
1·02 (0·62–1·67)
0·01 
0·1 
1 
10 
100
0·01 
0·1 
1 
10 
100
Incidence rate ratio
(95% CI)
Number of
participants
Favours control group
Favours intervention group
Favours control group
Favours intervention group
Figure 2: Forest plot for effects on MDD onset at post-treatment and up to 6 months, 12 months, and 24 months
Vázquez et al, 2017a and Vázquez et al, 2020a compared behavioural activation and control. Vázquez et al, 2017b and Vázquez et al, 2020b compared cognitive behavioural therapy and control. Full
details of all included studies are in the appendix (pp 16–22).

Articles
998 www.thelancet.com/psychiatry Vol 11 December 2024
at score 15 (moderately severe depressive symptoms).
Similarly, higher baseline GAD-7 scores were associated
with stronger effects: IRR 0·71 (95% CI 0·42–1·21) at
score 5, 0·59 (0·36–0·98) at score 10, and 0·52 (0·30–0·91)
at score 15. On a study level, delivery type significantly
affected preventive effects (interaction p=0·002), with
conference telephone calls (IRR 0·24 [95% CI 0·11–0·53])
being potentially more effective than other methods
(IRR 0·63–0·71), although based on a small sample
(two studies with four comparisons). Geographical
region, intervention type, publication year, control
condition, and risk of bias showed no significant
differences in preventing MDD onset (table 3).
Discussion
This systematic review and individual participant data
meta-analysis synthesised data from 30 randomised
controlled trials with 7201 participants, mostly from
high-income countries. The overall study quality was
satisfactory, although five studies had a high risk of bias.
Psychological interventions showed strong evidence of
positive outcomes up to 12 months, including reduced
MDD incidence, depressive symptom severity, increased
proportions of participants with 50% symptom reduction,
near symptom-free status, and reliable improvement,
along with decreased reliable symptom deterioration.
Psychological interventions were more effective in
preventing MDD onset in individuals without previous
psychotherapy experience and appeared more effective in
those with initial moderate to moderately severe
depressive or anxiety symptoms (PHQ-9 or GAD-7 ≥10),
indicated by smaller IRRs. However, the overlapping CIs
and non-significant interaction term warrant caution in
interpreting this result.
Our findings support previous individual participant
data meta-analysis findings that highlight the potential of
psychological interventions to prevent MDD onset in
individuals not currently experiencing a depressive
episode. Our one-stage individual participant data meta-
analysis showed a significantly reduced incidence of
depression by 42% up to 6 months, which was maintained
with a 33% reduction up to 12 months. This finding
exceeds the results of previous conventional meta-
analyses by Cuijpers and colleagues
12 
(19% reduction)
and Huang and colleagues
24 
(22% reduction). Individual
participant data meta-analyses examine individual-level
data, accounting for variability and potential biases,
providing more precise estimates of intervention effects.
Our results align with an individual participant data
meta-analysis of internet-based interventions
14 
and meta-
analyses of online psychological and psychoeducational
interventions,
25 
and self-guided internet interventions.
26
However, the two meta-analyses
25,26 
did not require
diagnostic interviews at baseline for assessing depression
diagnosis. The secondary outcome findings from our
individual participant data meta-analysis largely align
with those of previous meta-analyses.
14,27,28
Our moderation analyses suggesting that preventive
effects are more pronounced in the subgroups with
initial moderate and moderately severe depressive
symptoms partially validate previous research on
preventive interventions for depression.
13 
Of eight studies
in a systematic review on symptom severity as a
moderator, half supported this effect, whereas one found
Figure 3: Time to depression onset
The mean duration of the interventions was 13·3 weeks (range 2–52 weeks).
++
+
+
+
++
+
+++
+ 
+
+ 
+
+
+ 
+ 
+
+ 
+
+
+
++ 
+ +
+ 
+ 
+
++++
+ 
+ 
+
+ 
+ +
++ 
+
+
++
+++
+++
+++
+
++
+
++
++
++
++
++++++ +
+++++
++
++
+++
++
++
+
++
+
+
+
+
+++
++
++
++ 
+
+++
+
+
++ +
+
++
+++
++
+
++ 
+
++++
 
+++
++
+
+
++ +
+
++++
+
++ 
+ +
 
++
+
+++
++++
+++
++
+++
+++
+
+
+++++
++++++++++
+++++
+++++
Time (weeks)
Survival probability
Number at risk (number
censored)
Control
Intervention
1333 (162)
1337 (207)
0 
10 20 
30 
40 
50
0·6
0·7
0·8
0·9
1·0
1053 (177)
1014 (228)
937 (232)
907 (299)
840 (275)
831 (340)
788 (280)
779 (356)
698 (339)
714 (402)
Intervention
Control
0
Number
of
studies
β √V ̂(β) τ² (95% CI) 
t 
p value
Depressive symptom
severity
30 0·016 0·128 0·253 (0·000–1·017) 0·126 0·900
Anxiety symptom
severity
10 –0·195 0·260 0·403 (0·000–1·425) –0·837 0·402
Sex, male 25 0·073 0·242 0·000 (0·000–0·000) 0·293 0·770
Age, years 30 0·129 0·116 0·195 (0·000–0·731) 1·046 0·296
Ethnicity, non-White 8 0·726 0·532 0·000 (0·000–2·519) 1·027 0·304
Education, higher 21 0·196 0·270 0·000 (0·000–0·100) 0·663 0·507
Employment, yes 16 –0·037 0·293 0·000 (0·000–1·014) –0·130 0·897
Relationship, yes 27 –0·105 0·233 0·000 (0·000–0·908) –0·475 0·635
Chronic medical
condition, yes
10 0·013 0·585 0·000 (0·000–0·853) 0·022 0·982
History of MDD, yes 11 0·020 0·442 0·000 (0·000–0·450) 0·045 0·964
Antidepressive
medication, yes
8 –0·103 0·366 0·000 (0·000–1·250) –0·296 0·767
Previous
psychotherapy, yes
7 0·829 0·381 0·000 (0·000–1·499) 2·179 0·029
MDD=major depressive disorder.
Table 2: Results of participant-level moderator analyses

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 999
lower symptom burden increased efficacy.
13 
Our finding
that higher initial anxiety symptoms (GAD ≥10) improve
intervention outcomes aligns with previous research.
13
Unlike previous studies,
13,14 
we did not find age to be a
moderating factor. However, previous research focused
on specific populations (eg, female caregivers) and
delivery formats (eg, internet interventions).
Our findings support guidelines recommending
psychological interventions as a treatment option in the
management of subthreshold persistent depressive
symptoms.
29 
Moreover, our results underscore the
importance of public health messaging that emphasises
the effectiveness of preventive psychological interventions
for depression, contributing to de-stigmatisation and
encouraging widespread engagement in mental health
practices. Future research should explore strategies for
integrating preventive interventions into routine care
settings, acknowledging the current gap in implementation
of preventive interventions.
30
Our findings suggest that the provision of subgroup-
specific, tailored interventions based on factors such as
sex, age, ethnicity, education, employment status,
relationship status, presence of chronic medical
conditions, history of MDD, and intake of anti-depressive
medication might not be warranted. These insights
might support a universal applicability of psychological
interventions across diverse demographic and clinical
groups. However, ethnicity is often used as a proxy for
culture, although they are distinct. Culturally tailored
interventions improve engagement, retention, and
overall effectiveness.
31 
Targeting individuals’ specific
challenges and tailoring content to their experiences is
crucial for successful preventive interventions. Additional
research is essential to evaluate the potential of preventive
psychological interventions in subgroups with mild
initial symptom severity. Specifically, investigations are
needed to identify the point at which depressive and
anxiety symptoms at the lower end of the severity
spectrum become persistent enough to warrant
preventive interventions. Although a history of MDD did
not act as an effect modifier, it was assessed in only
11 (37%) of 30 studies, and we could not assess if a
threshold of previous episodes affects intervention
effectiveness. Further research is warranted to explore
methods for enhancing preventive effects in individuals
with previous exposure to psychotherapy. Our findings
suggest that a preventive effect might not extend up to
24 months, although available data are scarce. This
finding is consistent with the evidence observed in
children and adolescents.
32 
Future research is needed to
explore how interventions might be revised or refined to
stabilise their impact. Although conference calls showed
stronger preventive effects, the low number of studies
(two) prevents drawing definitive conclusions about their
superiority. The possibility that various delivery formats
and intervention types might have similar effects in
preventing MDD might motivate a shared
decision-making process and a personalised, patient-
centred approach to intervention selection based on
individual needs, preferences, and considerations of
accessibility and convenience. However, more research is
needed to determine whether these different formats and
intervention types truly have similar effects.
Our systematic review and individual participant data
meta-analysis has several limitations. First, we were not
able to retrieve data from all eligible randomised
controlled trials, with data obtained from 30 of 42 studies.
Second, between-study heterogeneity was moderate to
large in many analyses, indicating that true effects might
differ across contexts. Third, not all studies provided data
at all assessment points, which means that systematic
differences among trials could cause different effects
across time. Trials reporting outcomes at 24 months
Number
of
studies
Incidence rate
ratio (95% CI)
τ² p value
Region
North America 7 0·76 (0·48–1·21) 0·000 0·252
Europe 17 0·68 (0·54–0·87) 0·000 ··
Australia 3 0·68 (0·34–1·36) 0·000 ··
Asia 4 0·34 (0·19–0·62) 0·000 ··
Other 1 0·53 (0·17–1·61) - ··
Intervention type
Problem-solving
therapy
4 0·62 (0·36–1·06) 0·158 0·374
Cognitive behaviour
therapy
16 0·69 (0·54–0·89) 0·083 ··
Behavioural
activation
4 0·54 (0·27–1·08) 0·000 ··
Stepped care 5 0·78 (0·45–1·37) 0·000 ··
Other 3 0·32 (0·15–0·69) 0·207 ··
Delivery type
Face to face 12 0·71 (0·52–0·97) 0·000 0·002
Other 8 0·63 (0·41–0·96) 0·000 ··
Conference call* 2 0·24 (0·11–0·53) 0·000 ··
Internet 8 0·63 (0·43–0·92) 0·186 ··
Publication year
Until 2010 7 0·77 (0·54–1·11) 0·000 0·453
Until 2015 10 0·54 (0·35–0·84) 0·337 ··
After 2015 15 0·63 (0·47–0·84) 0·090 ··
Control group
Care as usual 24 0·68 (0·54–0·84) 0·000 0·337
Other 8 0·52 (0·32–0·85) 0·372 ··
Risk of bias
Low 24 0·62 (0·50–0·77) 0·000 0·366
Some concerns 3 0·94 (0·52–1·69) 0·000 ··
High 5 0·50 (0·21–1·19) 0·578 ··
*Each of the two studies encompasses two distinct comparisons: Vázquez et al,
2017a and Vázquez et al, 2020a compared behavioural activation and control.
Vázquez et al, 2017b and Vázquez et al, 2020b compared cognitive behavioural
therapy and control.
Table 3: Results of study-level moderator analyses

Articles
1000 www.thelancet.com/psychiatry Vol 11 December 2024
often did not provide efficacy data for earlier timepoints.
Fourth, we could not include certain putative moderators
in our analyses due to insufficient data in the included
studies, such as childhood adversity, quality of life,
mastery (ie, locus of control), or racial, ethnic, and
cultural diversity. Improved consistency in assessing
relevant effect modifiers in depression prevention
studies would enable more sophisticated treatment-
covariate interaction analyses, for example by using core
outcome sets.Establishing core outcome sets for
psychological depression prevention could be a
promising direction for future research. Last, the
sample’s demographic composition, with half of
participants having higher education and 70% being
female, might limit the generalisability of the findings.
However, neither education nor sex were identified as
effect modifiers.
We conclude that psychological interventions for
depression prevention might serve as an effective strategy
to alleviate the disease burden of depression at an
individual and societal level. Our findings contribute to
estimating the benefits of preventive interventions for
clinicians and policy makers. Given the significant burden
of depression, clinicians and policy makers should
consider preventive psychological interventions as a viable
option for individuals with subthreshold depression.
IPD-PrevDep Consortium
Marcel C Adriaanse (Department of Health Sciences and Amsterdam
Public Health Research Institute, Vrije Universiteit Amsterdam,
Amsterdam, Netherlands), Steven M Albert (Department of Behavioral
and Community Health Sciences, Graduate School of Public Health,
University of Pittsburgh, PA, USA), Esther Allart (Allart Supervisie,
Nijmegen, Netherlands), Osvaldo P Almeida (Medical School, University
of Western Australia, WA, Australia; Institute for Health Research,
University of Notre Dame Australia, Fremantle, WA, Australia),
Julian Basanovic (Department of Psychology, University of Exeter, Exeter,
UK; School of Psychological Science, University of Western Australia, WA,
Australia), Philip J Batterham (Centre for Mental Health Research, College
of Health and Medicine, Australian National University, Canberra, ACT,
Australia), Harald Baumeister (Department of Clinical Psychology and
Psychotherapy, Ulm University, Ulm, Germany), Thomas Berger (Institute
of Psychology, University of Bern, Bern, Switzerland), Vanessa Blanco
(Department of Evolutionary and Educational Psychology, University of
Santiago de Compostela, Santiago de Compostela, Spain), Robin Casten
(Sidney Kimmel Medical College of Thomas Jefferson University,
Philadelphia, PA, USA), Dicken Chan (School of Public Health and
Primary Care, Chinese University of Hong Kong, Hong Kong, China),
Helen Christensen (Black Dog Institute, Sydney, Australia; School of
Psychiatry, Faculty of Medicine, University of New South Wales, Sydney,
NSW, Australia), Marketa Ciharova (Department of Clinical Neuro- and
Developmental Psychology, VU University Amsterdam, Amsterdam,
Netherlands), Lorna Cook (SMART Lab, Mood Disorders Centre,
Department of Psychology, University of Exeter, Exeter, UK),
Keith S Dobson (Department of Psychology, University of Calgary, Calgary,
AB, Canada), Elsien Dozeman (GGZ inGeest, Specialized Mental Health
Care, Amsterdam, Netherlands), Kotaro Imamura (Department of Digital
Mental Health, Graduate School of Medicine, University of Tokyo, Tokyo,
Japan), Michael R Irwin (Cousins Center for Psychoneuroimmunology,
Jane and Terry Semel Institute for Neuroscience and Human Behavior,
David Geffen School of Medicine, University of California, Los Angeles,
CA, USA; Department of Psychiatry and Biobehavioral Sciences,
David Geffen School of Medicine, University of California, Los Angeles,
CA, USA), Norito Kawakami (Department of Digital Mental Health,
Graduate School of Medicine, University of Tokyo, Tokyo, Japan),
Eirini Karyotaki (Department of Clinical Neuro- and Developmental
Psychology, VU University Amsterdam, Amsterdam, Netherlands),
Jan Philipp Klein (Department of Psychiatry, Psychosomatics, and
Psychotherapy, University of Lübeck, Lübeck, Germany; Center for Brain,
Behavior, and Metabolism, University of Lübeck, Lübeck, Germany),
Candace Konnert (Department of Psychology, University of Calgary,
Calgary, AB, Canada), María Asunción Lara (Dirección de Investigaciones
Epidemiológicas y Psicosociales, Instituto Nacional de Psiquiatría Ramón
de la Fuente Muñiz, Ciudad de México, México), Huynh-Nhu Le
(Department of Psychological and Brain Sciences, George Washington
University, Washington, DC, USA), Dirk Lehr (Department of Health
Psychology and Applied Biological Psychology, Institute of Psychology,
Leuphana University Lueneburg, Lueneburg, Germany), Steffen Moritz
(Department of Psychiatry and Psychotherapy, University Medical Center
Hamburg-Eppendorf, Hamburg, Germany), Ricardo F Muñoz
(Department of Psychology, Palo Alto University, Palo Alto, CA, USA;
Department of Psychiatry and Behavioral Sciences, University of
California, San Francisco, CA, USA), Richard Olmstead (Cousins Center
for Psychoneuroimmunology, Jane and Terry Semel Institute for
Neuroscience and Human Behavior, David Geffen School of Medicine,
University of California, Los Angeles, CA, USA; Department of Psychiatry
and Biobehavioral Sciences, David Geffen School of Medicine, University
of California, Los Angeles, CA, USA), Patricia Otero (Department of
Psychology, University of A Coruña, A Coruña, Spain),
Charles F Reynolds III (Department of Psychiatry, University of Pittsburgh
School of Medicine, Pittsburgh, PA, USA), Barry W Rovner
(Sidney Kimmel Medical College of Thomas Jefferson University,
Philadelphia, PA, USA), Lasse B Sander (Medical Psychology and Medical
Sociology, Faculty of Medicine, University of Freiburg, Freiburg,
Germany), Filip Smit (Department of Mental Health and Prevention,
Trimbos Institute, Utrecht, Netherlands; Department of Epidemiology and
Biostatistics, Amsterdam University Medical Centers, Amsterdam,
Netherlands), Philip Spinhoven (Department of Clinical Psychology,
Institute of Psychology, Leiden University, Netherlands), Liza Stelmach
(Carewest, Calgary, AB, Canada), Yannik Terhorst (Department of
Psychology, Ludwig Maximilian University of Munich, Munich, Germany
and German Center for Mental Health (DZPG), Partner-Site Munich-
Augsburg, Munich, Germany), Fernando L Vázquez (Department of
Clinical Psychology and Psychobiology, University of Santiago de
Compostela, Santiago de Compostela, Spain), Ed Watkins (SMART Lab,
Mood Disorders Centre, Department of Psychology, University of Exeter,
Exeter, UK), Godelief R W M Willemse (Vilans, Utrecht, Netherlands),
Wenhui Yang (Department of Psychology, Hunan Normal University,
Changsha, Hunan, China), Samuel Y S Wong (School of Public Health
and Primary Care, Chinese University of Hong Kong, Hong Kong, China).
Contributors
CB, DDE, and PC conceived and designed the study. CB, MH, AAS, and
SI selected the studies and extracted data. CB, DDE, PC, and members of
the IPD-PrevDep Consortium contributed the individual participant data.
MH, AAS, and SI verified the data. MH analysed the data. CB, MH, MS,
TAF, and PC interpreted the results. CB and MH wrote the first draft of
the manuscript. All authors had access to all the data and provided critical
input and revisions to the draft manuscripts and approved the final
manuscript. CB and PC had final responsibility for the decision to submit
for publication.
Data sharing
All extracted data are available in the manuscript and appendix, as well
as on the Metapsy website. Individual-level data cannot be made
available due to confidentiality agreements in the original studies.
Declaration of interests
DDE is a stakeholder of the Institute for health training online (GET.ON),
which aims to implement scientific findings related to digital health
interventions into routine care. MS is employed through a grant provided
by the Nagoya City. TAF reports personal fees from Boehringer-
Ingelheim, Daiichi Sankyo, DT Axis, Kyoto University Original, Shionogi,
SONY, and UpToDate, and a grant from DT Axis and Shionogi, outside of
the submitted work; TAF has a patent (7448125), a pending patent
(2022–082495), and intellectual properties for Kokoro-app licensed to
Mitsubishi-Tanabe. All other authors declare no competing interests.
For the Metapsy website see
https://docs.metapsy.org/
databases/depression-psyctr/

Articles
www.thelancet.com/psychiatry Vol 11 December 2024 1001
Acknowledgments
The authors thank Stella Wernicke and Svenja Kratzke for help in the
literature search, assistance in full text acquisition, screening, data
extraction and data cleaning, and risk of bias assessment.
References
1 World Health Organization. Depression and other common mental
disorders: global health estimates. Geneva: World Health
Organization, 2017.
2 Richter D, Wall A, Bruen A, Whittington R. Is the global prevalence
rate of adult mental illness increasing? Systematic review and meta-
analysis. Acta Psychiatr Scand 2019; 140: 393–407.
3 Vos T, Abajobir AA, Abate KH, et al. Global, regional, and national
incidence, prevalence, and years lived with disability for
328 diseases and injuries for 195 countries, 1990–2016: a systematic
analysis for the Global Burden of Disease Study 2016. Lancet 2017;
390: 1211–59.
4 Greenberg PE, Fournier AA, Sisitsky T, et al. The economic burden
of adults with major depressive disorder in the United States
(2010 and 2018). PharmacoEconomics 2021; 39: 653–65.
5 Cipriani A, Furukawa TA, Salanti G, et al. Comparative efficacy and
acceptability of 21 antidepressant drugs for the acute treatment of
adults with major depressive disorder: a systematic review and
network meta-analysis. Lancet 2018; 391: 1357–66.
6 Cuijpers P, Quero S, Noma H, et al. Psychotherapies for
depression: a network meta-analysis covering efficacy, acceptability
and long-term outcomes of all main treatment types.
World Psychiatry 2021; 20: 283–93.
7 Andrews G, Issakidis C, Sanderson K, Corry J, Lapsley H. Utilising
survey data to inform public policy: comparison of the cost-
effectiveness of treatment of ten mental disorders. Br J Psychiatry
2004; 184: 526–33.
8 Thornicroft G, Chatterji S, Evans-Lacko S, et al. Undertreatment of
people with major depressive disorder in 21 countries.
Br J Psychiatry 2017; 210: 119–24.
9 Herrman H, Patel V, Kieling C, et al. Time for united action on
depression: a Lancet–World Psychiatric Association Commission.
Lancet 2022; 399: 957–1022.
10 Zhang R, Peng X, Song X, et al. The prevalence and risk of
developing major depression among individuals with subthreshold
depression in the general population. Psychol Med 2023;
53: 3611–20.
11 Volz HP, Stirnweiß J, Kasper S, Möller HJ, Seifritz E. Subthreshold
depression—concept, operationalisation and epidemiological data.
A scoping review. Int J Psychiatry Clin Pract 2023; 27: 92–106.
12 Cuijpers P, Pineda BS, Quero S, et al. Psychological interventions to
prevent the onset of depressive disorders: a meta-analysis of
randomized controlled trials. Clin Psychol Rev 2021; 83: 101955.
13 Conejo-Cerón S, Bellón JÁ, Motrico E, et al. Moderators of
psychological and psychoeducational interventions for the
prevention of depression: a systematic review. Clin Psychol Rev 2020;
79: 101859.
14 Reins JA, Buntrock C, Zimmermann J, et al. Efficacy and
moderators of internet-based interventions in adults with
subthreshold depression: an individual participant data meta-
analysis of randomized controlled trials. Psychother Psychosom 2021;
90: 94–106.
15 Stewart LA, Clarke M, Rovers M, et al. Preferred Reporting Items
for Systematic Review and Meta-Analyses of individual participant
data: the PRISMA-IPD statement. JAMA 2015; 313: 1657–65.
16 Ebert DD, Buntrock C, Reins JA, Zimmermann J, Cuijpers P.
Efficacy and moderators of psychological interventions in treating
subclinical symptoms of depression and preventing major
depressive disorder onsets: protocol for an individual patient data
meta-analysis of randomised controlled trials. BMJ Open
2018; 8: e018582.
17 Wahl I, Löwe B, Bjorner JB, et al. Standardization of depression
measurement: a common metric was developed for 11 self-report
depression measures. J Clin Epidemiol 2014; 67: 73–86.
18 Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for
assessing risk of bias in randomised trials. BMJ 2019; 366: l4898.
19 Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief
depression severity measure. J Gen Intern Med 2001; 16: 606–13.
20 Resche-Rigon M, White IR. Multiple imputation by chained
equations for systematically and sporadically missing multilevel
data. Stat Methods Med Res 2018; 27: 1634–49.
21 Bartlett JW, Seaman SR, White IR, Carpenter JR. Multiple
imputation of covariates by fully conditional specification:
accommodating the substantive model. Stat Methods Med Res 2015;
24: 462–87.
22 Morris TP, Walker AS, Williamson EJ, White IR. Planning a
method for covariate adjustment in individually randomised trials:
a practical guide. Trials 2022; 23: 328.
23 Furukawa TA, Leucht S. How to obtain NNT from Cohen’s d:
comparison of two methods. PLoS One 2011; 6: e19070.
24 Huang K, You S, Yue X, et al. Effects of non-pharmacological
interventions on depressive symptoms and risk of major depressive
disorder in adults with subthreshold depression: a systematic
review and meta-analysis. Psychiatry Res 2023; 326: 115333.
25 Rigabert A, Motrico E, Moreno-Peral P, et al. Effectiveness of online
psychological and psychoeducational interventions to prevent
depression: systematic review and meta-analysis of randomized
controlled trials. Clin Psychol Rev 2020; 82: 101931.
26 Edge D, Watkins ER, Limond J, Mugadza J. The efficacy of self-
guided internet and mobile-based interventions for preventing
anxiety and depression—a systematic review and meta-analysis.
Behav Res Ther 2023; 164: 104292.
27 Leng LL, Yin XC, Ng SM. Mindfulness-based intervention for
clinical and subthreshold perinatal depression and anxiety:
a systematic review and meta-analysis of randomized controlled
trial. Compr Psychiatry 2023; 122: 152375.
28 van der Feltz-Cornelis C, Allen SF, Holt RIG, Roberts R, Nouwen A,
Sartorius N. Treatment for comorbid depressive disorder or
subthreshold depression in diabetes mellitus: systematic review and
meta-analysis. Brain Behav 2021; 11: e01981.
29 Kendrick T, Pilling S, Mavranezouli I, et al. Management of
depression in adults: summary of updated NICE guidance. BMJ
2022; 378: o1557.
30 Campion J, Javed A, Lund C, et al. Public mental health: required
actions to address implementation failure in the context of
COVID-19. Lancet Psychiatry 2022; 9: 169–82.
31 Harper Shehadeh M, Heim E, Chowdhary N, Maercker A,
Albanese E. Cultural adaptation of minimally guided interventions
for common mental disorders: a systematic review and meta-
analysis. JMIR Ment Health 2016; 3: e44.
32 Ssegonja R, Nystrand C, Feldman I, Sarkadi A, Langenskiöld S,
Jonsson U. Indicated preventive interventions for depression in
children and adolescents: a meta-analysis and meta-regression.
Prev Med 2019; 118: 7–15.