The clinical e ectiveness of sertraline in primary care and the role of depression severity and duration (PANDA):
a pragmatic, double-blind, placebo-controlled randomised trial
Gemma Lewis, Larisa Du y, Anthony Ades, Rebekah Amos, Ricardo Araya, Sally Brabyn, Katherine S Button, Rachel Churchill, Catherine Derrick, Christopher Dowrick, Simon Gilbody, Christopher Fawsitt, William Hollingworth, Vivien Jones, Tony Kendrick, David Kessler, Daphne Kounali, Naila Khan, Paul Lanham, Jodi Pervin, Tim J Peters, Derek Riozzie, George Salaminios, Laura Thomas, Nicky J Welton, Nicola Wiles,
Rebecca Woodhouse, Glyn Lewis
Background Depression is usually managed in primary care, but most antidepressant trials are of patients from secondary care mental health services, with eligibility criteria based on diagnosis and severity of depressive symptoms. Antidepressants are now used in a much wider group of people than in previous regulatory trials. We investigated the clinical e ectiveness of sertraline in patients in primary care with depressive symptoms ranging from mild to severe and tested the role of severity and duration in treatment response.
Methods The PANDA study was a pragmatic, multicentre, double-blind, placebo-controlled randomised trial of patients from 179 primary care surgeries in four UK cities (Bristol, Liverpool, London, and York). We included patients aged 18 to 74 years who had depressive symptoms of any severity or duration in the past 2 years, where there was clinical uncertainty about the bene t of an antidepressant. This strategy was designed to improve the generalisability of our sample to current use of antidepressants within primary care. Patients were randomly assigned (1:1) with a remote computer-generated code to sertraline or placebo, and were strati ed by severity, duration, and site with random block length. Patients received one capsule (sertraline 50 mg or placebo orally) daily for one week then two capsules daily for up to 11 weeks, consistent with evidence on optimal dosages for e cacy and acceptability. The primary outcome was depressive symptoms 6 weeks after randomisation, measured by Patient Health Questionnaire, 9-item version (PHQ-9) scores. Secondary outcomes at 2, 6 and 12 weeks were depressive symptoms and remission (PHQ-9 and Beck Depression Inventory-II), generalised anxiety symptoms (Generalised Anxiety Disorder Assessment 7-item version), mental and physical health-related quality of life (12-item Short-Form Health Survey), and self-reported improvement. All analyses compared groups as randomised (intention-to-treat). The study is registered with EudraCT, 2013-003440-22 (protocol number 13/0413; version 6.1) and ISRCTN, ISRCTN84544741, and is closed to new participants.
Findings Between Jan 1, 2015, and Aug 31, 2017, we recruited and randomly assigned 655 patients—326 (50%) to sertraline and 329 (50%) to placebo. Two patients in the sertraline group did not complete a substantial proportion of the baseline assessment and were excluded, leaving 653 patients in total. Due to attrition, primary outcome analyses were of 550 patients (266 in the sertraline group and 284 in the placebo group; 85% follow-up that did not di er by treatment allocation). We found no evidence that sertraline led to a clinically meaningful reduction in depressive symptoms at 6 weeks. The mean 6-week PHQ-9 score was 7·98 (SD 5·63) in the sertraline group and 8·76 (5·86) in the placebo group (adjusted proportional di erence 0·95, 95% CI 0·85–1·07; p=0·41). However, for secondary outcomes, we found evidence that sertraline led to reduced anxiety symptoms, better mental (but not physical) health- related quality of life, and self-reported improvements in mental health. We observed weak evidence that depressive symptoms were reduced by sertraline at 12 weeks. We recorded seven adverse events—four for sertraline and three for placebo, and adverse events did not di er by treatment allocation. Three adverse events were classi ed as serious— two in the sertraline group and one in the placebo group. One serious adverse event in the sertraline group was classi ed as possibly related to study medication.
Interpretation Sertraline is unlikely to reduce depressive symptoms within 6 weeks in primary care but we observed improvements in anxiety, quality of life, and self-rated mental health, which are likely to be clinically important. Our ndings support the prescription of SSRI antidepressants in a wider group of participants than previously thought, including those with mild to moderate symptoms who do not meet diagnostic criteria for depression or generalised anxiety disorder.
Lancet Psychiatry 2019
Published Online September 19, 2019 https://doi.org/10.1016/ S2215-0366(19)30366-9
See Online/Comment https://doi.org/10.1016/ S2215-0366(19)30367-0
Division of Psychiatry, University College London, London, UK (Ge Lewis PhD,
L Duffy BSc, G Salaminios MSc, Prof Gl Lewis PhD); Bristol Medical School, University of Bristol, Bristol, UK (A Ades PhD, C Derrick BSc, C Fawsitt PhD, Prof W Hollingworth PhD,
V Jones BA, Prof D Kessler PhD, DKounaliPhD,
Prof T J Peters PhD,
L Thomas MSc,
Prof N J Welton PhD,
N Wiles PhD); Institute of Psychology Health and Society, University of Liverpool, Liverpool, UK (R Amos BSc,
Prof C Dowrick MD, N Khan PhD); Institute of Psychiatry, Psychology and Neuroscience, King’s College London,
London, UK (Prof R Araya PhD); Department of Health Sciences (S Brabyn MSc, ProfRChurchillPhD,JPervinBSc, R Woodhouse MSc) and Centre for Reviews and Dissemination (Prof S Gilbody PhD, D Riozzie), University of York, York, UK; Department of Psychology, University of Bath, Bath, UK
(K S Button PhD); Primary Care and Population Sciences, Faculty of Medicine, University of Southampton, Aldermoor Health Centre,
(Prof T Kendrick MD); School of Nursing and Health Studies, University of Dundee,
Dundee, UK; and Department of Liberal Arts, Durham University, Durham, UK (P Lanham BA)
￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼
Correspondence to: Dr Gemma Lewis, Division of Psychiatry, University College London, London W1T 7NF, UK email@example.com
Depression is a major contributor to the global burden of disease and by 2030 is predicted to be the leading cause of disability in high-income countries.1 People with depression are usually managed in primary care and SSRI antidepressants are often the rst-line treatment.2,3 Prescriptions for antidepressants have risen dramatically in high-income countries over the past decade (70·9 million items were prescribed in England in 2018), leading to concerns that they are overprescribed.4
Existing evidence on the clinical e ectiveness of anti- depressants has been comprehensively summarised in a recent network meta-analysis, which reported small bene ts of all antidepressants compared with placebo.5 However, the constituent trials had many methodological limitations, were usually done decades ago for regulatory purposes, and generally were of poor quality, with 82% at moderate or high risk of bias.5 Larger, more recent placebo-controlled trials reported smaller antidepressant e ect sizes, perhaps re ecting more rigorous methods.5
A more important criticism of the existing evidence is that most trials have used eligibility criteria based on
diagnoses and severity. Primary care physicians see patients whose depressive symptoms range from mild to severe.2,6 Milder depressive symptoms can still cause distress and impairment,7 and it is common for primary care physicians to prescribe antidepressants when diagnostic criteria are not met.2,8 The use of anti- depressants has increased and broadened since the original regulatory trials were done. Most antidepressants are now prescribed in primary care without any standardised assessment of depressive symptoms. It is di cult to generalise existing results to patients currently receiving treatment for depression.
Some studies have argued that antidepressants are more e ective for patients with more severe symptoms9,10 but most,11–14 although not all,15 meta-analyses of individual patient data do not support this. Systematic reviews on the e ectiveness of antidepressants for patients with less severe depression have led to contradictory ndings.16,17 The THREAD study was done in UK primary care and found small bene cial e ects of SSRIs in patients with mild to moderate depressive symptoms, but comparison was with usual care rather than placebo.18
Copyright © 2019 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.
Research in context
Evidence before this study
Existing evidence on the e ectiveness of antidepressants has been comprehensively summarised in a recent network meta-analysis by Cipriani and colleagues, which reported small bene ts of all antidepressants compared with placebo.
This meta-analysis included double-blind, randomised controlled trials published up to Jan 8, 2016, of adults meeting diagnostic criteria for depression. We updated this search for articles published up to Jan 31, 2019. We also searched for trials including patients not meeting diagnostic criteria for depression. We searched PubMed and the Cochrane Central Register of Controlled Trials, for studies published in English, with the search terms “depress*” OR “dysthymi*” OR “adjustment disorder”
OR “mood disorder” OR “a ective disorder” OR “depress* symptoms” AND “antidepress*”. We did not nd any individual trials of antidepressants compared with placebo in people with depression published since the review by Cipriani and colleagues. We found two systematic reviews of patients with subthreshold depressive symptoms. Most trials were done in secondary care, with patients who met diagnostic criteria and speci ed severity thresholds. These trials had many methodological limitations, were usually done several decades ago for regulatory purposes, and were generally of poor quality, with 82% rated by Cipriani and colleagues as having a moderate or high risk of bias. Some primary care trials have been done, but with similar limitations. Most studies of patients with subthreshold depression were low quality and reported inconsistent ndings.
Added value of this study
To our knowledge, the PANDA trial is the largest placebo-controlled trial of an antidepressant not funded by the pharmaceutical industry. We did not restrict eligibility by specifying lower or higher thresholds of depression severity but relied instead upon clinical uncertainty. This improved generalisability to the relevant population in primary care.
In our primary analyses, we did not nd convincing evidence that sertraline led to clinically important reductions in depressive symptoms within six weeks, although there was weak evidence of a small bene t at 12 weeks. In secondary analyses, we found evidence that sertraline led to reduced anxiety symptoms, better mental health-related quality of life, and self-reported improvements in mental health.
Implications of all the available evidence
Sertraline leads to reduced anxiety symptoms and self-reported improvements in mental health within 6 weeks, but any e ect on depressive symptoms takes longer to emerge and is more modest. Our ndings support the prescription of SSRI antidepressants in a wider group of participants than previously thought, including those with mild to moderate symptoms who do not meet diagnostic criteria for depression or generalised anxiety disorder. Clinicians and patients should be aware of the symptoms that are likely to improve so that they can consider alternative management of depressive symptoms that might not respond.
Duration of depression, as well as severity, might help determine whether antidepressants could be bene cial.19 Antidepressants can be e ective for dysthymia— depressive symptoms not meeting diagnostic criteria but present for 2 years or more.20 Dysthymia has been criticised for con ating severity and duration.19 To our knowledge, no previous trial has investigated the association between depression duration and anti- depressant response.
In this study, we report on the clinical e ectiveness of sertraline in a large sample of patients who presented to primary care with depressive symptoms. We also explored the in uence of depression severity and duration. We did not restrict eligibility by specifying lower or higher thresholds of depression severity but relied instead upon clinical uncertainty,21,22 aiming to improve generalisability to the population receiving antidepressants in primary care.
Study design and participants
The PANDA study was a pragmatic, multicentre, double- blind, placebo-controlled randomised trial of patients from 179 primary care surgeries in four UK cities (Bristol, Liverpool, London, and York).23 Patients were eligible for inclusion if they were aged 18 to 74 years and there was uncertainty (from general practitioner [GP] and patient) about the possible bene t of an antidepressant. Use of clinical uncertainty as an entry criterion avoids diagnostic or severity criteria and leads to a sample that is more generalisable to the population currently receiving antidepressant medication. Clinical uncertainty relies upon variation between doctors in how they make decisions and, with a su cient number of doctors, this variation in practice leads to a wide and heterogeneous group of participants in the trial, re ecting clinical equipoise within the eld.22 In our case, because primary care physicians do not use standardised assessments of depression, we expected trial participants to have a range of depressive symptom severity. Clinical uncertainty is also the ethical basis for randomised controlled trials.22,24,25 Exclusion criteria were antidepressant treatment in past 8 weeks; comorbid psychosis, schizophrenia, mania, hypomania, bipolar disorder, dementia, eating disorder, or major alcohol or substance abuse; and medical contraindications for sertraline.23
There were two methods of recruitment to the study. First, patients were referred during GP consultation. Second, GP practices searched computerised records for patients who presented with depression or depressive symptoms in the past two years. Patients were told that they had been referred to the study because their GP had identi ed them as having depression or low mood. A researcher then contacted patients by telephone to con rm eligibility. Since it was possible that some patients might not currently have depressive symptoms, researchers asked “ rstly, I need to ask you do you want
treatment for your depression at the moment?”. If the answer was no, the patient was excluded. Before randomisation, the patient’s GP was asked to approve inclusion in the trial and could exclude the patient if there was not clinical uncertainty.
All participants provided written informed consent. Ethics approval was from the National Research Ethics Service committee, East of England—Cambridge South (ref 13/EE/0418). The results presented here focus on the treatment e ect of sertraline in primary care and are part of a larger research programme (NIHR RP-PG-0610-10048).
Randomisation and masking
Participants were randomly assigned (1:1) to sertraline or placebo by PRIMENT Clinical Trials Unit (CTU) with a remote computer-generated code, and were strati ed by severity, duration, and site with random block length. Prespeci ed thresholds for strati cation were total score (0–11, 12–19, and ≥20) and depression duration (<2 years or ≥2 years) at baseline on the Clinical Interview Schedule—Revised (CIS-R).26 Sertraline 50 mg tablets and placebo were encapsulated and were identical. Researchers and statisticians were masked to treatment allocation until completion of analyses. Primary care physicians were told the treatment allocation when participants nished their nal assessment or withdrew from study medication, so they could discuss further treatment.
Patients received one capsule (sertraline 50 mg or placebo orally) daily for one week then two capsules daily for up to 11 weeks, consistent with evidence on optimal dosages for e cacy and acceptability.27 After six weeks, medication could be increased to three capsules in consultation with the local principal investigator. Trial medication was sent by a trial pharmacy to the patient’s primary care physician (or home) after randomisation. Baseline assessments before randomisation and follow- up assessments at 2 weeks, 6 weeks, and 12 weeks after randomisation took place in patients’ homes, primary care surgeries, or universities.23
We assessed patients’ depressive symptoms with the Patient Health Questionnaire, 9-item version (PHQ-9; available range 0–27, higher scores indicating more severe symptoms), a self-report measure of depressive symptoms in the past 2 weeks that is widely used in primary care.28 Self-report questionnaires avoid the possibility of observer bias associated with clinically rated outcomes29,30 and the PHQ-9 might be more responsive to change than other self-report measures.31,32 We also assessed depressive symptoms with the Beck Depression Inventory (BDI-II). The BDI-II is a self report 21-item scale that assesses the severity of depressive symptoms in the past 2 weeks. Scores range from 0–63, with higher scores indicating more severe symptoms. We assessed generalised anxiety symptoms with the
For more on the software used for randomisation see https://sealedenvelope.com
For the ISCRTN listing see https://doi.org/10.1186/ ISRCTN84544741
For the uploaded analysis plan see http://discovery.ucl.ac. uk/10041458/
Generalised Anxiety Disorder Assessment 7-item version (GAD-7). The GAD-7 is a self-report measure of the severity of anxiety symptoms in the past 2 weeks, with possible scores ranging from 0–21 and higher scores indicating more severe symptoms. We derived separate physical and mental health-related quality of life scores with the 12-item Short-Form Health Survey (SF-12). Possible scores range from 0 to 100, with higher scores indicating better quality of life.
We used a ve-item self-report adherence scale developed for the CoBalt study and a binary variable indicating at least 80% adherence.33 Patients were also asked about serious adverse events at each follow-up visit by use of open-ended questions. Physical symptoms that could be SSRI side-e ects were recorded with a modi ed Toronto scale.34
The CIS-R is a computerised, self-administered, fully structured interview measuring 14 common mental disorder symptom groups.26 The CIS-R generates diagnoses meeting ICD-10 criteria for depressive or anxiety episodes, a total common mental disorders score, and a depression severity score (available range 0–21) created by the sum of the following ve symptoms: depression, depressive ideas, fatigue, concentration, and sleep problems. The total score in three categories was used for strati cation. The severity score was used for the hypothesis that the treatment e ect varied according to the severity of depressive symptoms at baseline. The CIS-R asks about duration of depressive symptoms with the following categories: less than 2 weeks, between 2 weeks and 6 months, between 6 months and 1 year, between 1 year and 2 years; between 2 years and 5 years; between 5 years and 10years,andmorethan10years.Thesecategorieswere dichotomised into less than 2 years and 2 years or more.
The primary outcome was total score at 6 weeks on the PHQ-9.
Secondary outcomes were analysed as repeated measures across 2-week, 6-week, and 12-week follow-up assessments and included PHQ-9 and BDI-II total scores and remission (de ned as a score of <10 on each measure), GAD-7 total scores, mental and physical health-related quality-of-life scores with the SF-12, and self-reported improvement in mental health. Self- reported improvement was assessed by asking patients, “compared to when we last saw you, how have your moods and feelings changed?”. Responses were “I feel a lot better”, “I feel slightly better”, “I feel about the same”, “I feel slightly worse”, or “I feel a lot worse”.35,36 We created a binary variable of feeling the same or worse (0) and feeling better (1). Patients also completed the ve-level EQ-5D instrument (for economic analyses) and emotional processing tasks, the results of which will be reported separately.
When originally devised, the primary aim of the PANDA trial was to investigate the severity and duration of depressive symptoms associated with a clinically important response to sertraline, as stated in the protocol paper23 and on ISCRTN. However, towards the later stages of designing the trial and when formulating the detailed analysis plan (uploaded before any analyses were done to UCL Discovery and approved by the trial steering committee), it became apparent that we would have insu cient statistical power to estimate plausible inter- action e ects that would allow us to investigate those aims. Therefore, our power calculation23 and primary analysis (as stated in the analysis plan) are based on a primary aim to examine the clinical e ectiveness of sertraline versus placebo. Interactions between severity and duration at baseline and treatment allocation were planned as exploratory. Our protocol paper has described the power calculation in detail.23 Assuming a two-sided α of 5%, power of 90%, and 10% attrition, 547 participants were required to detect an 11% (relative) di erence in scores between sertraline and placebo.
Analyses and reporting were in line with CONSORT guidelines. Primary and secondary outcomes were described in the approved protocol before the study started (appendix pp 59). The measure of self-reported improvement was added as a secondary outcome after the study started but before any analyses were done (appendix pp 25). When the study was registered on ISCRTN, the primary outcome was listed. Secondary outcomes were added later and were described and approved in the study protocol. All analyses compared groups as randomised (intention-to-treat). We compared the proportion of patients in each group who were categorised as adherent with Pearson’s χ2 test. Physical symptoms that could be SSRI side-e ects were compared by group with independent t-tests. Analyses of adherence and physical symptoms were done at each follow-up visit. We report serious adverse events during the trial according to treatment allocation.
The primary analysis was a linear regression of log- transformed PHQ-9 scores at 6 weeks, adjusted for baseline PHQ-9 score and strati cation variables (severity assessed by CIS-R in three categories, duration in twocategories,andsite).Regressioncoe cientsfromthis model were the di erence in geometric means of log- transformedPHQ-9scoresbetweenrandomisedgroups. We exponentiated this regression coe cient to report the treatment e ect as a ratio of the (geometric) mean PHQ-9 score at 6 weeks in those allocated to sertraline to the (geometric) mean PHQ-9 score at 6 weeks in those allocated to placebo. The null value was 1 and proportions below 1 indicated fewer symptoms in the sertraline group. For example, a coe cient of 0·89 re ects PHQ-9 scores that were 11% lower in the sertraline than placebo group.
The proportional scale was chosen on the basis of model comparison between di erent scales. This choice
See Online for appendix
35 719 patients identified as potentially eligible via record search
31 645 sent an invitation letter
4074 removed from the initial list by GPs
30 616 excluded
3399 “no” replies
27 217 no response
1029 “yes” replies and checked for GP suitability
110 not suitable
20 GP suitability confirmations not received 6 did not proceed to GP screen
427 patients referred to study at consultation
1320 checked for eligibility
113 not eligible
376 declined to take part
160 could not be reached for telephone screening
671 eligible and proceeded to consenting and randomisation
667 consented and completed baseline assessment
4 refused to consent
12 refused randomisation
655 randomly assigned
￼ ￼ ￼ ￼ ￼ ￼ ￼
329 assigned to placebo
328 completed PHQ-9 at baseline
1 did not complete PHQ-9 at baseline
326 assigned to sertraline
325 completed PHQ-9 at baseline
1 did not complete PHQ-9 at baseline
2 did not complete a substantial proportion of the assessments
￼ ￼ ￼ ￼ ￼
22 not interested
4 no time
1 did not want to
2 lost to follow-up
22 not interested
7 no time
2 did not want to
2 lost to follow-up
8 missed 2-week follow-up
￼ ￼ ￼ ￼ ￼ ￼
292 attended 2-week follow-up 292 completed PHQ-9
279 attended 2-week follow-up 277 completed PHQ-9
2 did not complete PHQ-9
12 missed 2-week follow-up
7 not interested 3 no time
3 did not want to
1 lost to follow-up
11 not interested
6 no time
1 lost to follow-up
￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼
1 missed 6-week follow-up
6 missed 6-week follow-up
￼ ￼ ￼ ￼ ￼ ￼ ￼
285 attended 6-week follow-up and included in primary analysis
285 completed PHQ-9
267 attended 6-week follow-up and included in primary analysis
267 completed PHQ-9
￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼ ￼
265 attended 12-week follow-up 263 completed PHQ-9
15 not interested
2 no time
1 did not want to
4 not interested 3 no time
￼ ￼ ￼ ￼
264 attended 12-week follow-up 262 completed PHQ-9
2 did not complete PHQ-9
3 missed 12-week follow-up
2 did not complete PHQ-9
2 missed 12-week follow-up
Figure: Trial pro le GP=general practitioner. PHQ-9=Patient Health Questionnaire, 9-item version.
was supported by previous work showing bene ts for choosing the scale. These bene ts include variance stabilisation and overcoming the issue of baseline dependence.35,37 A proportional model is also plausible, as it seems unlikely that a treatment would lead to the same di erence in means in those starting at a high score compared with those with a low score.
Given our interest in the relationship between severity and treatment e ect, we included baseline depression CIS-R scores in the primary analysis. The primary analysis was also done without the CIS-R depression score, to examine the possible in uence of collinearity. We investigated how robust ndings were to the model speci cation by comparison with a Poisson model.
We calculated linear multilevel models for continuous secondary outcomes (PHQ-9, BDI-II, GAD-7, and SF-12 physical and mental health quality of life) with log- transformed PHQ-9, BDI-II, and GAD-7 scores and raw SF-12 scores, as stated in the analysis plan. Logistic multilevel models were calculated for binary secondary outcomes (remission on PHQ-9 and BDI-II, and self- reported improvement). Models included continuous baseline measurement of the outcome, CIS-R depression severity score, and strati cation variables. For primary and secondary outcomes, we report interactions between treatment allocation and baseline CIS-R depression severity score and duration.
For sensitivity analyses, we reported primary and secondary outcomes after adjustment for variables that were possibly imbalanced at baseline, which were identi ed using descriptive statistics. We investigated possible clustering within surgeries by calculating the intraclass correlation coe cient using multilevel models, with a random intercept for surgeries. We reported the treatment e ect from this multilevel model to assess whether clustering had an in uence.
We did two sensitivity analyses to investigate the possible e ect of missing data. First, we adjusted for baseline variables associated with missing outcome data. Second, we used multiple imputation by chained equations to replace missing data. We assumed missingness was dependent on observed data and imputed 50 datasets. Imputation models included variables from our models and variables associated with missing data, with pooled results obtained using Rubin’s rules.38 Our post-hoc analyses are detailed in the appendix (p 4).
The study was overseen by a data monitoring committee. All analyses were done in STATA version 15. The study is registered with EudraCT, 2013-003440-22 (protocol number 13/0413; version 6.1), and ISRCTN, ISRCTN84544741.
Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to
131 (40%) 58 (18%) 69 (21%) 66 (20%)
218 (67%) 106 (33%)
203 (63%) 121 (37%)
144 (45%) 179 (55%)
116 (36%) 152 (47%) 55 (17%)
180 (56%) 101 (31%) 42 (13%)
259 (80%) 64 (20%)
134 (41%) 58 (18%) 73 (22%) 64 (19%)
221 (67%) 108 (33%)
181 (55%) 148 (45%)
155 (47%) 174 (53%)
139 (42%) 144 (44%) 46 (14%)
184 (56%) 103 (31%) 42 (13%)
263 (80%) 66 (20%)
(Table 1 continues on next page)
Liverpool London York
CIS-R depression duration (years)* <2
ICD-10 CIS-R anxiety diagnosis† Yes
Married or living as married Single
Separated, divorced, or widowed
Financial di culty†
Living comfortably or doing alright Just about getting by
Finding it di cult or very di cult
Depression in the past† Yes
Patient Health Questionnaire, 9-item version total score (possible range 0–27)
265 (41%) 116 (18%) 142 (22%) 130 (20%)
439 (67%) 214 (33%)
384 (59%) 269 (41%)
299 (46%) 353 (54%)
255 (39%) 296 (45%) 101 (15%)
364 (56%) 204 (31%) 84 (13%)
522 (80%) 130 (20%)
CIS-R total score*
ICD-10 CIS-R depression diagnosis†
In paid employment
Highest educational quali cation†
A Level or higher
GCSE, standard grade, or other
No formal quali cation
Antidepressant in the past†
CIS-R total score (possible range 0–64)
all the data in the study and had nal responsibility for the decision to submit for publication.
Between Jan 1, 2015, and Aug 31, 2017, we recruited and randomly assigned 655 patients—326 (50%) to sertraline and 329 (50%) to placebo ( gure). Two patients in the sertraline group did not complete a substantial proportion of the baseline assessment and were excluded, leaving 653 patients in total included in analyses. Patients were recruited from 179 GP surgeries (appendix p 7). Patients who were invited to participate and those who participated were similar in age and sex (appendix p 8).
The proportions of patients attending each follow-up visit were similar in both trial groups and we found no statistical evidence for a di erence in attrition by group ( gure). Missing follow-up data were more likely in patients who were recruited in London, younger, ethnic minority, single, having nancial di culties, and those with no formal quali cations, and higher baseline PHQ-9, BDI-II, GAD-7, and life-event scores (appendix pp 9–10).
Baseline characteristics of the sample were similar between treatment groups (table 1). Overall, the mean patient age was 39·7 years (SD 14·96), 384 (59%) participants were female, and 433 (66%) were employed. On the CIS-R, 355 (54%) patients met ICD-10 criteria for depression, 299 (46%) patients met generalised anxiety criteria, 197 (30%) met both depression and generalised anxiety disorder criteria, and 96 (15%) met criteria for mixed anxiety and depressive disorder (scoring 12 or more but not meeting diagnostic criteria for depression or anxiety).39 99 (15%) patients did not meet any diagnostic criteria for depression or anxiety. The mean PHQ-9 score was 12·0 (SD 5·8) and 405 (62%) patients exceeded a score of 10 (the recommended threshold for clinical use). 391 (60%) patients had received anti- depressants previously and 522 (80%) patients reported previous depression. In the sertraline group, there was a higher proportion of women, and lower proportions of those who were married or living as married and meeting ICD-10 depression criteria.
563 (86%) of 653 randomised patients provided the date they started medication (median 8 days after randomisation, IQR 6–11, range 1–70). There were delays with follow-up assessments so that these were carried out a median of 1 week, 5 weeks, and 11 weeks after patients started medication.
Adherence was greater than 80% at all timepoints and we observed no evidence of a di erence in adherence between treatment groups (appendix p 11). More patients in the sertraline group than the placebo group thought they were taking sertraline at 2 weeks, 6 weeks, and 12 weeks (appendix p 11).
Due to attrition and missing data, 550 patients were included in the analysis of the primary outcome. Mean PHQ-9 scores at 6 weeks were 7·98 (SD 5·63) in patients
allocated to sertraline and 8·76 (5·86) in patients allocated to placebo (table 2). After adjustment for baseline scores and strati cation variables, the adjusted proportional di erence between sertraline and placebo
(Continued from previous page)
CIS-R depression severity score (possible range 0–21)‡
Generalised Anxiety Disorder Assessment, 7-item version score (possible range 0–21)
SF-12 mental health subscale (possible range 0–100)
Number of life events in past 6 months
Frequency of physical symptoms (possible range 55–112)§
Beck Depression Inventory, second edition total score (possible range 0–63)
Social support score (possible range 1–24)
SF-12 physical health subscale (possible range 0–100)
Number of physical symptoms in past 2 weeks
Data are n (%) or mean (SD). One person in the placebo group was missing the PHQ-9 score at baseline and one person in the sertraline group was missing the CIS-R depression score at baseline. CIS-R=Clinical Interview Schedule—Revised. SF-12= Short-Form Health Survey. *The total CIS-R score assesses the severity of symptoms of common mental disorders. The total CIS-R score in three categories was a strati cation variable at randomisation: 0–11 (minimal symptoms),
12–19 (moderate to severe symptoms), and ≥20 (severe symptoms). †A CIS-R diagnosis used the criteria and threshold required to meet an ICD-10 clinical diagnosis of depression or anxiety. CIS-R data were missing for one person. ‡The CIS-R depression severity score (possible range 0–21) assesses the severity of depressive symptoms. §How often during the past 2 weeks the patient had each symptom: 1 (not at all), 2 (several days), 3 (more than half the days), and 4 (nearly every day).
Table 1: Baseline characteristics
￼ ￼ ￼
Model 1† Model 3§ Model 5||
n Mean (SD)
266 7·98 (5·63) 266 7·98 (5·63) 324 8·25 (5·81)
n Mean (SD)
284 8·76 (5·86) 284 8·76 (5·86) 329 8·88 (5·87)
Adjusted proportional di erence* (95% CI)
0·95 (0·85–1·07) 0·95 (0·85–1·05) 0·96 (0·86–1·08)
0·41 0·31 0·50
Means (SDs) are for non-log-transformed PHQ-9 scores at 6 weeks. One person in the placebo group was missing the PHQ-9 score at baseline and one person in the sertraline group was missing the CIS-R depression score at baseline. These individuals were excluded from the regression model of the primary outcome at 6 weeks. PHQ-9=Patient Health Questionnaire, 9-item version. CIS-R=Clinical Interview Schedule—Revised. *All models use a log-transformed
PHQ-9 score as the outcome. Adjusted proportional di erence can be interpreted as the di erence in scores between randomised groups expressed as a proportion (or percentage). †Primary analysis model: intention-to-treat analysis adjusted for baseline PHQ-9 and CIS-R depression scores and strati cation variables (total CIS-R score in three categories, duration of depressive episode in two categories, and site; n=550). ‡Model 1 adjusted for variables showing imbalance at baseline (sex, ICD-10 depression diagnosis, marital status; n=550). §Model 1 adjusted for possible clustering e ects using a random e ect for general practices (n=550). ¶Model 1 adjusted for baseline variables associated with missing outcome data (age, ethnicity, marital status, nancial di culty, ICD-10 anxiety diagnosis, used antidepressants in the past, Beck Depression Inventory, second edition score, number of life events, Generalised Anxiety Disorder Assessment, 7-item version score; n=550). ||Model 1 with missing data replaced by multiple imputation (n=653). **The imputed model (model 5) adjusted for variables showing imbalance at baseline (sex, ICD-10 depression diagnosis, marital status; n=653).
Table 2: Analyses of primary outcome (log-transformed PHQ-9 scores at 6 weeks)
n Mean (SD)
Placebo Adjusted proportional di erence* (95% CI)
n Mean (SD)
￼ ￼ ￼
Follow-up assessment (weeks)
0·96 (0·87 to 1·07)
0·96 (0·86 to 1·07)
0·87 (0·79 to 0·97)
0·93 (0·86 to 1·01)
Group by time interaction
Beck Depression Inventory, second edition
SF-12 mental health
SF-12 physical health
Follow-up assessment (weeks)
0·99 (0·97 to 1·10)
0·95 (0·85 to 1·07)
0·84 (0·74 to 0·95)
0·93 (0·84 to 1·21)
Group by time interaction
Follow-up assessment (weeks)
0·91 (0·82 to 1·03)
0·79 (0·70 to 0·89)
0·77 (0·68 to 0·87)
0·83 (0·75 to 0·91)
Group by time interaction
Follow-up assessment (weeks)
1·58 (–0·10 to 3·26)
2·90 (1·17 to 4·63)
2·85 (1·12 to 5·47)
2·41 (1·14 to 3·69)
Group by time interaction
Follow-up assessment (weeks)
–0·71 (–1·75 to 0·34)
–0·36 (–1·44 to 0·71)
–0·89 (–1·96 to 0·19)
–0·66 (–1·48 to 0·17)
Group by time interaction
PHQ-9=Patient Health Questionnaire, 9-item version. SF-12= Short-Form Health Survey. CIS-R=Clinical Interview Schedule—Revised. *These models used a log-transformed PHQ-9 score as the outcome. Adjusted proportional di erences can be interpreted as the di erence in scores between randomised groups expressed as a proportion
(or percentage). All multilevel models were adjusted for baseline measure of each outcome (continuous), baseline CIS-R depression severity score, and strati cation variables (baseline total CIS-R score in three categories, duration of depressive episode in two categories, and site). †Calculated using data from all three follow-ups combined.
Table 3: Repeated measures analyses of continuous secondary outcomes at 2 weeks, 6 weeks, and 12 weeks
was 0·95 (95% CI 0·85–1·07; p=0·41; model 1; table 2). This result changed very little after adjustment for variables showing imbalance (model 2; table 2). We found little evidence of clustering by surgery (intracluster correlation coe cient 0·04, 95% CI 0·00–0·01) and no
evidence that clustering altered the treatment e ect (model 3; table 2). The treatment e ect was similar after considering missing data (models 4, 5, 6; table 2).
We observed an association between baseline CIS-R depression severity and PHQ-9 scores at 6 weeks, with a 3% increase in PHQ-9 scores for each unit increase in CIS-R depression score (adjusted proportional di erence 1·03, 95% CI 1·01–1·05; p=0·0046). We found no evidence of an association between baseline depression duration and PHQ-9 score at 6 weeks (1·08, 95% CI 0·96–1·22; p=0·19). We observed no evidence that treatment response varied with depression severity (p=0·9907) or duration (p=0·8203).
The result of the primary analysis did not change after we excluded baseline CIS-R depression scores to explore possible multicollinearity (adjusted proportional dif- ference 0·96, 95% CI 0·85–1·07; p=0·41). The result of the primary analysis was consistent with a negative binomial model (chosen as an extension of the Poisson model because PHQ-9 variance at 6 weeks was larger than the mean).
The adjusted proportional di erence in PHQ-9 scores across all timepoints was 0·93 (95% CI 0·86–1·01, p=0·11; table 3). At 12 weeks, PHQ-9 scores were 13% lower (0·87, 95% CI 0·79–0·97) in the sertraline group. We observed similar results for the BDI-II (table 3). We observed evidence that di erences in GAD-7 scores became larger over time (p=0·0075). At 6 weeks, GAD-7 scores were 21% lower (adjusted proportional di erence 0·79, 95% CI 0·70–0·89) in those allocated to sertraline than in those allocated to placebo. Mental health-related quality of life scores were higher (2·41, 95% CI 1·14–3·96, p=0·00021) in the sertraline group than in the placebo group, with no evidence that di erences varied over time (p=0·22). We observed no evidence of a di erence in physical health-related quality of life (table 3).
We found evidence that sertraline led to increased odds of remission at 12 weeks (but not 6 weeks) and this evidence was stronger for BDI-II than PHQ-9 (table 4; appendix p 24). Patients in the sertraline group had increased odds of reporting feeling better than the same or worse (adjusted OR 1·96, 95% CI 1·45–2·63; p<0·0001), with no evidence that this di erence varied over time (p=0·16).
Analysis of secondary outcomes was una ected by adjustment for variables that were imbalanced at baseline (appendix pp 13–14). We observed no evidence that treat- ment response for secondary outcomes varied according to severity or duration of symptoms. Results of our post-hoc analyses are presented in the appendix (pp 15–24).
We recorded seven adverse events—four in the sertraline group and three in the placebo group (appendix p 12). Three adverse events were classi ed as serious—two in the sertraline group and one in the placebo group. One serious adverse event in the sertraline group was classi ed as possibly related to study medication and two
in the placebo group were classi ed as not related or unlikely to be related (appendix p 12). We found no evidence of a di erence in the number of physical symptoms between groups (appendix p 12). 59 patients chose to increase their dose to 150 mg—37 (11%) in the placebo group and 22 (7%) in the sertraline group (p=0·047). We are aware of three incidents in which the capsule was broken. One of these incidents was intentional, although the participant drew the wrong conclusion about allocation. The other two incidents were inadvertent. The three a ected participants were withdrawn from study medication once we were aware of these incidents.
In our primary analysis, we did not nd convincing evidence that sertraline led to a greater reduction in depressive symptoms compared with placebo within 6 weeks. However, in our secondary analyses we found weak evidence that sertraline reduced depressive symptoms at 12 weeks compared with placebo. This e ect was small, but does not exclude the possibility of a clinically important e ect on depressive symptoms by 12 weeks.
In our secondary analyses, we found evidence that sertraline reduced generalised anxiety symptoms at 6 weeks and 12 weeks compared with placebo. We also found evidence that sertraline led to better mental (but not physical) health-related quality of life and self-rated improvement in mental health.
Although results from secondary analyses should be interpreted with caution, statistical evidence for all our secondary outcomes (except depression) was strong. In particular, self-reported improvements in mental health can be used to calculate minimal clinically important di erences and can be regarded as patient-centred indicators of clinically meaningful change.40 Depressive and anxiety symptoms often occur together and comorbid anxiety symptoms have been associated with poorer quality of life in people with depressive symptoms.41,42 We have found important evidence that sertraline can lead to clinical bene ts for patients in primary care with depressive symptoms. However, these bene ts occur mainly through improvements in anxiety symptoms and quality of life rather than reductions in depressive symptoms.
We found no evidence that severity or duration of depressive (or anxiety) symptoms a ected treatment response, although analyses using interaction terms lacked statistical power. We found that duration of symptoms was not associated with outcome once severity of symptoms had been accounted for, suggesting that an interaction with treatment response is unlikely.
Sertraline has a similar pharmacological pro le to other SSRIs and acts via a similar mechanism.5 Therefore, we would expect our results to apply to other SSRIs when used in this population. Generalising our ndings to antidepressants of other classes is more di cult, although there are similarities in the mechanism of action of all commonly used antidepressants.43
To our knowledge, the PANDA trial is the largest individual placebo-controlled trial of an antidepressant not funded by the pharmaceutical industry. Our trial had more attrition than predicted but recruited more participants than our target and had power to detect an 11% reduction in symptoms—enough to detect a clinically important di erence. Clinical trials are often criticised for using narrow inclusion criteria, which can reduce external validity.25 We used clinical uncertainty as an entry criterion, which avoids reliance upon diagnostic or severity criteria that have never been validated as an indication for antidepressants. Participants in our study ranged from those with very few depressive symptoms to those with severe depressive symptoms, therefore our results are more readily generalisable to the population currently receiving antidepressants in primary care than are the results of previous trials. Many people in our sample had very severe depression, suggesting that there is uncertainty about prescription of antidepressants in primary care at all levels of severity.
Use of clinical uncertainty as an entry criterion could have potential disadvantages. If patients with few depressive symptoms were less likely to respond to
Sertraline Placebo Adjusted* OR (95% CI)
n n (%) n n (%)
Patient Health Questionnaire, 9-item version remission
Beck Depression Inventory, second edition remission
Feeling better (self-rated improvement)
￼ ￼ ￼
Follow-up assessment (weeks)
Group by time interaction
Follow-up assessment (weeks)
Group by time interaction
Follow-up assessment (weeks)
Group by time interaction
CIS-R=Clinical Interview Schedule—Revised. *All multilevel models were adjusted for baseline measure of each outcome (continuous), baseline CIS-R depression severity score, and strati cation variables (baseline total CIS-R score in three categories, duration of depressive episode in two categories, and site). †Calculated using data from all three follow-ups combined.
Table 4: Repeated measures analyses of binary secondary outcomes at 2 weeks, 6 weeks, and 12 weeks
antidepressants, this might have reduced the treatment e ect, but we did not nd any suggestion of this. We also observed strong evidence of a treatment e ect for anxiety symptoms despite our broad inclusion criteria.
Reliance upon clinical uncertainty will exclude those for whom a clinician is certain that antidepressants are indicated. We would expect people with the most severe depression to be somewhat underrepresented in the trial. However, this e ect should be modest given the very wide range of depression severity in our sample. Use of clinical uncertainty as an entry criterion requires many clinicians to participate in a study to encompass the whole range of clinical practice. We included 179 practices from four UK cities, therefore it is unlikely that we are re ecting idiosyncratic practice. Instead, our sample most likely re ects the diverse decisions made by many doctors over a broad range of uncertainty.
There are marked similarities between the behaviour of doctors and patients in primary care within high- income countries, despite divergent health systems. Antidepressant prescribing has increased in all high- income countries over the past two decades.44 A similar trial in other high-income countries would probably identify participants over a wide range of depression severities and include many people with anxiety symptoms. Our main conclusions about the bene ts of sertraline on anxiety rather than depressive symptoms should be generalisable to other high-income countries.
In the PANDA trial, 115 (46%) participants on sertraline thought they were taking the active drug at 6 weeks compared with 52 (19%) participants on placebo. In other SSRI trials that have assessed patient beliefs about the medication that they are taking,45–48 only one study45 found evidence that patients can distinguish active treatment from placebo. We consid- ered two possible explanations for this nding in our trial. First, patients detected an improvement in their symptoms, therefore thought they were taking sertraline. Second, patients could detect being on sertraline because of physical or psychological e ects that we did not measure. We found no evidence that side-e ects di ered by study group, therefore it is unlikely that reported adverse e ects systematically led to a di erence in patients guessing whether they were on sertraline or placebo.
The possibility that adverse e ects of antidepressants could lead to unblinding and an inadvertent placebo e ect has been raised previously.49 Such a placebo e ect should lead to a di erence in all outcomes. In our trial, anxiety, quality of life, and self-reported improvement measures showed greater changes than depressive symptom measures.Thisobservationreducesthelikelihoodthat di erences between sertraline and placebo were explained purely by the possibility of unblinding.
Medication was encapsulated so participants could, in principle, break open the capsule and discover if there was a tablet included. We are aware of three incidents when
this occurred. All three participants stopped the study medication once this had occurred and we do not think the encapsulation of the medication had an overall e ect on the blinding of assessments in the trial.
Patients from record searches had less severe depressive and anxiety symptoms but were more likely to have taken antidepressants previously. These di erences had no in uence on treatment e ects (appendix pp 19–21). Current practice in UK primary care is to engage with patients’ views and to share decision making.50 Therefore, patients who present to clinicians with mental health problems and ask for antidepressants will usually be prescribed these. Patients we identi ed through record searches who wanted treatment for their depressive symptoms are also those for whom there might be clinical uncertainty. GPs could exclude anyone who was recruited through record searches.
Our nding that sertraline was not e ective for depressive symptoms at 6 weeks is inconsistent with previous studies.5 Antidepressants could have an e ect on depressive symptoms that is not large enough to be detected in a trial of this size. A meta-analysis by Cipriani and colleagues5 reported a standardised di erence in means of –0·27 (95% CI –0·34 to –0·21) for sertraline versusplaceboatamedianfollow-upof8weeks.Thise ect size is considerably larger than the –0·09 (–0·23 to 0·05) standardised di erence in means from our trial at 6 weeks so the explanation of our trial being too small to detect the e ect reported by Cipriani and colleagues5 seems unlikely. Our ndings at 12 weeks of a –0·18 standardised mean di erence (95% CI –0·33 to –0·03) are closer to the result reported by Cipriani and colleagues.5
Previous studies have used observer-rated scales to assess depression, such as the Hamilton Depression Scale.30 Unlike our self-reported scales, these measures allow interviewer discretion and increase the possibility of observer bias. Bene cial e ects of sertraline on anxiety and self-reported improvement might therefore lead to a so-called halo e ect, such that observers rate depressive symptoms as improved. There is evidence that observer rating can bias results of depression and anxiety assessments.30 The Hamilton Depression Scale also contains questions on the physical and psychological symptoms of generalised anxiety, which we found were a ected by sertraline.
Our results suggest that the main bene ts in the rst 6 weeks of treatment with sertraline are on reduction of anxiety symptoms, such as worry and restlessness, rather than an improvement in depressive symptoms. Any e ect on depressive symptoms takes longer to emerge and is more modest. However, an improvement in anxiety symptoms in someone presenting with depression could lead to a clinical bene t. In cases where there is uncertainty about prescribing an antidepressant, the presence of anxiety symptoms, such as worry and restlessness, could indicate an increased likelihood of bene t. Clinicians and patients should be aware of the
symptoms that are likely to improve so that they can consider alternative management of depressive symp- toms that might not respond.
Our ndings support the prescription of SSRI antidepressants in a wider group of participants than previously thought, including those with mild to moderate symptoms who do not meet diagnostic criteria for depression or generalised anxiety disorder.
We did not nd evidence that treatment response for either anxiety or depressive symptoms varied according to the severity of depressive or anxiety symptoms, even within this heterogenous group of participants. There could be subgroups within this population that did not bene t from the anxiety reducing e ect; however, in the absence of any evidence for such subgroups we conclude that our results should be generalisable to the broad group of people included in our study.
In conclusion, our nding that sertraline a ects anxiety more quickly than depressive symptoms has potential implications for understanding the mechanisms of antidepressant treatment. Depressive symptoms could take longer to reduce than anxiety symptoms and some of the reduction might be explained by the earlier e ects on anxiety. Our results also challenge our reliance on placebo-controlled studies that have been done primarily for regulatory purposes. Investigation of pharmacological treatments after regulatory approval is needed. SSRIs are among the most commonly prescribed medications in the world and yet we still have an imperfect knowledge of their clinical e ectiveness and indications for their use.
GlL, AA, RA, RC, Tim Croudace, CDo, SG, WH, TK, DK, PL,
Alice Malpass, TJP, Debbie Sharp, NW, and NJW were responsible for the original proposal and securing funding for the trial. GlL as chief investigator and LD as trial manager had overall responsibility for the writing of the detailed protocol, and management of the study and the London site. Co-investigators DK and NW had responsibility for the Bristol site, CDo for the Liverpool site, and SG for the York site. TJP, DK, GeL, NW, and GlL wrote the statistical analysis plan. CDe, LT, GS, NK, RA, SB, JP, and RW were responsible for data collection. GeL did the main analyses with input from TJP and GlL. GeL wrote the initial draft of the report. All authors contributed to re nement of the trial protocol and contributed to, and approved, the nal report.
Declaration of interests
CDo reports grants from P zer Independent Grants for Learning and Change outside the submitted work. GlL reports grants from the National Institute of Health Research (NIHR), during the conduct of the study and personal fees from Fortitude Law outside the submitted work. NJW reports grants from the NIHR during the conduct of the study and was principal investigator on a grant part-funded by P zer (the project was purely methodological, using historical data on pain relief).
AA reports grants from the NIHR during the conduct of the study. NW reports grants from the NIHR during the conduct of the study. All other authors report no con icts of interest.
De-identi ed individual participant data for the study, the study protocol, statistical analysis plan, and analytical code will be available to investigators for individual participant data meta-analyses that have been approved by independent review committees. Data will be available from the publication date of this Article, with no end date. Proposals for use of data and requests for access should be directed to firstname.lastname@example.org. To gain access, researchers will need to sign a data access agreement with the study sponsor (University College London, London, UK).
We thank the sta in participating general practitioner surgeries for their help with recruitment and all patients who participated in the study.
We also thank colleagues who have contributed to the study through recruitment, administrative help, and other advice—in particular, Bryony Thompson, Yvonne Donkor, and Faye Bacon. We have been supported
by the following clinical research networks (CRNs): North Thames CRN, CRN North West London, CRN South London, North West Coast CRN, Greater Manchester CRN, West Midlands CRN, West of England CRN, and Yorkshire and Humber CRN. We especially thank the following CRN sta : Khatiba Raja, Tara Harvey, Zara Prem, Yvonne Forshaw,
Lynsey Wilson, Ruth Leach, Maya Leach, Helen Permain, and Sally Gordon. We also acknowledge the support of the University College London Hospitals Biomedical Research Centre Mental Health theme.
We also thank Carolyn Chew-Graham, Ian Anderson, Evan Kontopantelis, Christopher Williams, Richard Bying, and Obi Ukoumunne for generously agreeing to sit on either the trial steering committee or data monitoring committee. The PANDA trial is independent research commissioned by the National Institute for Health Research (NIHR) Programme Grant for Applied Research (RP-PG-0610-10048, awarded to GlL, CDo, WH, NW, SG, TJP, RC, and RA). The views expressed in this publication are those of the author(s) and not necessarily those of the sponsor, UK National Health Service, NIHR, or UK Department of Health and Social Care.
1 Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006; 3: e442.
2 Kendrick T, Dowrick C, McBride A, et al. Management of depression in UK general practice in relation to scores on depression severity questionnaires: analysis of medical record data. BMJ 2009; 338: b750.
3 Olfson M, Kroenke K, Wang S, Blanco C. Trends in o ce-based mental health care provided by psychiatrists and primary care physicians. J Clin Psychiatry 2014; 75: 247–53.
4 NHS Digital. Prescription cost analysis: England 2018. Mar 28, 2019. https://digital.nhs.uk/data-and-information/publications/statistical/ prescription-cost-analysis/2018 (accessed Aug 15, 2019).
5 Cipriani A, Furukawa TA, Salanti G, et al. Comparative e cacy 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 Cameron IM, Cardy A, Crawford JR, et al. Measuring depression severity in general practice: discriminatory performance of the PHQ-9, HADS-D, and BDI-II. Br J Gen Pract 2011; 61: e419–26.
7 Cuijpers P, de Graaf R, van Dorsselaer S. Minor depression: risk pro les, functional disability, health care use and risk of developing major depression. J A ect Disord 2004; 79: 71–79.
8 Simon GE, Rossom RC, Beck A, et al. Antidepressants are not overprescribed for mild depression. J Clin Psychiatry 2015;
9 Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ, Johnson BT. Initial severity and antidepressant bene ts:
a meta-analysis of data submitted to the Food and Drug Administration. PLoS Med 2008; 5: e45.
10 Khan A, Leventhal RM, Khan SR, Brown WA. Severity of depression and response to antidepressants and placebo: an analysis of the Food and Drug Administration database. J Clin Psychopharmacol 2002;
11 Furukawa TA, Maruo K, Noma H, et al. Initial severity of major depression and e cacy of new generation antidepressants: individual participant data meta-analysis. Acta Psychiatr Scand 2018; 137: 450–58.
12 Gibbons RD, Hur K, Brown CH, Davis JM, Mann JJ. Bene ts from antidepressants: synthesis of 6-week patient-level outcomes from double-blind placebo controlled randomized trials of uoxetine and venlafaxine. Arch Gen Psychiatry 2012; 69: 572–79.
13 Rabinowitz J, Werbelo N, Mandel FS, Menard F, Marangell L, Kapur S. Initial depression severity and response to antidepressants v placebo: patient-level data analysis from 34 randomised controlled trials. Br J Psychiatry 2016; 209: 427–28.
14 Hieronymus F, Lisinski A, Nilsson S, Eriksson E. In uence of baseline severity on the e ects of SSRIs in depression: an item-based, patient-level post-hoc analysis. Lancet Psychiatry 2019; published online July 11. https://doi.org/10.1016/S2215-0366(19)30216-0.
15 16 17 18
Fournier JC, DeRubeis RJ, Hollon SD, et al. Antidepressant drug e ects and depression severity: a patient-level meta-analysis. JAMA 2010; 6: 47–53.
Barbui C, Cipriani A, Patel V, Ayuso-Mateos JL, van Ommeren M. E cacy of antidepressants and benzodiazepines in minor depression: systematic review and meta-analysis. Br J Psychiatry 2011; 198: 11–16. Cameron IM, Reid IC, MacGillivray SA. E cacy and tolerability of antidepressants for sub-threshold depression and for mild major depressive disorder. J A ect Disord 2014; 166: 48–58.
Kendrick T, Chatwin J, Dowrick C, et al. Randomised controlled trial to determine the clinical e ectiveness and cost-e ectiveness of selective serotonin reuptake inhibitors plus supportive care, versus supportive care alone, for mild to moderate depression with somatic symptoms in primary care: the THREAD (THREshold for AntiDepressant response) study. Health Technol Assess 2009; 13: 1–159. Cleare A, Pariante CM, Young AH, et al. Evidence-based guidelines for treating depressive disorders with antidepressants: a revision of the 2008 British Association for Psychopharmacology guidelines.
J Psychopharmacol 2015; 29: 459–525.
de Lima MS, Hotoph M, Wessely S. The e cacy of drug treatments
for dysthymia: a systematic review and meta-analysis. Psychol Med 1999; 29: 1273–89.
Peto R, Baigent C. Trials: the next 50 years. Large scale randomised evidence of moderate bene ts. BMJ 1998; 317: 1170–71.
No authors listed. MRC European Carotid Surgery Trial: interim results for symptomatic patients with severe (70–99%) or with mild (0·29%) carotid stenosis. European Carotid Surgery Trialists’ Collaborative Group. Lancet 1991; 337: 1235–43.
Salaminios G, Du y L, Ades A, et al. A randomised controlled trial assessing the severity and duration of depressive symptoms associated with a clinically signi cant response to sertraline versus placebo, in people presenting to primary care with depression (PANDA trial): study protocol for a randomised controlled trial. Trials 2017; 18: 496.
Peto R, Collins R, Gray R. Large-scale randomized evidence:
large, simple trials and overviews of trials. J Clin Epidemiol 1995;
Hotopf M, Churchill R, Lewis G. Pragmatic randomised controlled trials in psychiatry. Br J Psychiatry 1999; 175: 217–23.
Lewis G, Pelosi AJ, Araya R, Dunn G. Measuring psychiatric disorder in the community: a standardized assessment for use by lay interviewers. Psychol Med 1992; 22: 465–86.
Furukawa TA, Cipriani A, Cowen PJ, Leucht S, Egger M, Salanti G. Optimal dose of selective serotonin reuptake inhibitors, venlafaxine, and mirtazapine in major depression: a systematic review and dose- response meta-analysis. Lancet Psychiatry 2019; 6: 601–09.
Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001; 16: 606–13. Hotopf M, Lewis G, Normand C. Putting trials on trial—the costs and consequences of small trials in depression: a systematic review of methodology. J Epidemiol Community Health 1997; 51: 354–58. Lewis G. Observer bias in the assessment of anxiety and depression. Soc Psychiatry Psychiatr Epidemiol 1991; 26: 265–72.
Kounali DZ, Button KS, Lewis G, Ades AE. The relative responsiveness of test instruments can be estimated using a meta-analytic approach: an illustration with treatments for depression. J Clin Epidemiol 2016; 77: 68–77.
Titov N, Dear BF, McMillan D, Anderson T, Zou J, Sunderland M. Psychometric comparison of the PHQ-9 and BDI-II for measuring response during treatment of depression. Cogn Behav Ther 2011;
33 Wiles N, Thomas L, Abel A, et al. Cognitive behavioural therapy as an adjunct to pharmacotherapy for primary care based patients with treatment resistant depression: results of the CoBalT randomised controlled trial. Lancet 2013; 381: 375–84.
34 Crawford AA, Lewis S, Nutt D, et al. Adverse e ects from antidepressant treatment: randomised controlled trial of 601 depressed individuals. Psychopharmacology (Berl) 2014; 231: 2921–31.
35 Button KS, Kounali D, Thomas L, et al. Minimal clinically important di erence on the Beck Depression Inventory-II according to the patient’s perspective. Psychol Med 2015; 45: 3269–79.
36 Robinson J, Khan N, Fusco L, Malpass A, Lewis G, Dowrick C. Why are there discrepancies between depressed patients’ Global Rating of Change and scores on the Patient Health Questionnaire depression module? A qualitative study of primary care in England. BMJ Open 2017; 7: e014519.
37 Angst F, Aeschlimann A, Angst J. The minimal clinically important di erence raised the signi cance of outcome e ects above the statistical level, with methodological implications for future studies. J Clin Epidemiol 2017; 82: 128–36.
38 Sterne JAC, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ 2009; 338: b2393.
39 McManus S, Bebbington P, Jenkins R BT (eds). Mental health and wellbeing in England: Adult Psychiatric Morbidity Survey 2014. Leeds: NHS Digital, 2016.
40 Fischer D, Stewart AL, Bloch DA, Lorig K, Laurent D, Holman H. Capturing the patient’s view of change as a clinical outcome measure. JAMA 1999; 282: 1157–62.
41 Johansson R, Carlbring P, Heedman Å, Paxling B, Andersson G. Depression, anxiety and their comorbidity in the Swedish general population: point prevalence and the e ect on health-related quality of life. PeerJ 2013; 1: e98.
42 Zhou Y, Cao Z, Yang M, et al. Comorbid generalized anxiety disorder and its association with quality of life in patients with major depressive disorder. Sci Rep 2017; 7: 40511.
43 Harmer CJ, Duman RS, Cowen PJ. How do antidepressants work? New perspectives for re ning future treatment approaches.
Lancet Psychiatry 2017; 4: 409–18.
44 Gusmão R, Quintão S, McDaid D, et al. Antidepressant utilization and suicide in Europe: an ecological multi-national study. PLoS One 2013; 8: e66455.
45 Kranzler HR, Burleson JA, Korner P, et al. Placebo-controlled trial of uoxetine as an adjunct to relapse prevention in alcoholics.
Am J Psychiatry 1995; 152: 391–97.
46 Killen JD, Fortmann SP, Schatzberg AF, et al. Nicotine patch and paroxetine for smoking cessation. J Consult Clin Psychol 2000;
47 Narushima K, Kosier JT, Robinson RG. Preventing poststroke depression: a 12-week double-blind randomized treatment trial and 21-month follow-up. J Nerv Ment Dis 2002; 190: 296–303.
48 Devanand DP, Nobler MS, Cheng J, et al. Randomized, double-blind, placebo-controlled trial of uoxetine treatment for elderly patients with dysthymic disorder. Am J Geriatr Psychiatry 2005; 13: 59–68.
49 Moncrie J, Wessely S, Hardy R. Meta-analysis of trials comparing antidepressants with active placebos. Br J Psychiatry 1998;
50 Elwyn G, Frosch D, Thomson R, et al. Shared decision making: a model for clinical practice. J Gen Intern Med 2012; 27: 1361–67.