PROSPERO International prospective register of systematic reviews

The effects of selective serotonin reuptake inhibitors versus no intervention, placebo, or ‘active’ placebo in patients with major depressive disorder. A systematic review of randomised clinical trials with meta-analyses and trial sequential analyses.

Janus Christian Jakobsen, Jane Lindschou Hansen, Signe Hellmuth, Anne Schou, Jesper Krogh, Christian Gluud

Citation

Janus Christian Jakobsen, Jane Lindschou Hansen, Signe Hellmuth, Anne Schou, Jesper Krogh, Christian Gluud. The effects of selective serotonin reuptake inhibitors versus no intervention, placebo, or ‘active’ placebo in patients with major depressive disorder. A systematic review of randomised clinical trials with meta-analyses and trial sequential analyses..
PROSPERO
2013:CRD42013004420
Available from http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42013004420

Review question(s)

In a systematic review of all randomised clinical trials we want to assess the beneficial and harmful effects of the SSRIs: citalopram, escitalopram, sertraline, fluoxetine, paroxetine, and fluvoxamine versus no intervention, placebo, or ‘active’ placebo in the treatment of major depressive disorder using meta-analyses and trial sequential analyses.

Searches

We will search The Cochrane Library’s CENTRAL, MEDLINE, EMBASE, PsycInfo, and Science Citation Index Expanded. We also searched other relevant publications for references to relevant trials (see Search strategy). To be able also to assess results from unpublished trials we also included trials submitted to FDA.

According to the WHO, major depressive disorder, i.e., unipolar depression, is the second largest healthcare problem worldwide in terms of years lived with disability (YLD). It afflicts an estimated 17% of individuals during their lifetime at tremendous cost to the individual and to society. Roughly a third of all depressive disorders take a chronic course. Compared to other medical disorders, depressive illness causes the most significant deterioration in individual quality of life. Approximately 15% of depressive patients will commit suicide over a 10 to 20 year period.

Participants/ population

Participants must be 18 years or more, and the primary diagnosis must be major depressive disorder. The diagnosis of major depression must be made based on one of the standardised criteria, such as ICD 10, DSM III, DSM III-R, DSM IV, or Feighner criteria. Comorbidity with schizophrenia will be an exclusion criterion, while comorbidity with other psychiatric diagnoses will not. Trials exclusively including participants with a somatic disease and comorbid depression and trials on depression during or after pregnancy will be excluded.

No intervention: any control intervention with no treatment elements, e.g., ‘waiting list’.

Placebo: any ‘placebo’ substance containing no active substance.

‘Active’ placebo: any active substance employed to mimic the adverse effects of taking a SSRI.

Co-interventions

Trials comparing SSRI versus no intervention, placebo, or ‘active’ placebo as add-on therapy to any other kind of intervention (e.g., treatment as usual) will be included, but only if this co-intervention is described and delivered similarly in the different intervention groups.

Outcome(s)

Primary outcomes

• The difference between the mean values from the two intervention groups using the 17-item Hamilton Depression Rating Scale (HDRS).

• The standardised mean difference between the two intervention groups using the HDRS, any other form of the Hamilton Depression Rating Scale, Montgomery-Asberg Depression Rating Scale (MADRS), and Bech’s Depression Inventory (BDI).

• The proportion of participants achieving remission. We have, pragmatically, defined remission as a Hamilton score of less than 8, BDI less than 10, or MADRS less than 10.

• Adverse events during the intervention period and the follow-up period. We will classify adverse events as serious and non-serious. Serious adverse events are defined as medical events that are life threatening, result in death, disability or significant loss of function; that cause hospital admission or prolonged hospitalisation or a hereditary anomaly or foetal injury. All other adverse events (that is, events that have not necessarily had a causal relationship with the treatment, but that resulted in a change in- or cessation of the treatment) will be considered non-serious events.

We will estimate all outcomes at two time points:

• Outcome at end of treatment. Often after 6-18 weeks of treatment. The trial’s choice of end of treatment will be used. This is the most important outcome measure time point in this review.

• Outcome at maximum follow-up.

Secondary outcomes

• Number of suicides.

• Number of suicide attempts.

• Suicide ideation (any kind of assessment used by the trialists).

• Quality of life (any assessment scale used by the trialists).

We will estimate all outcomes at two time points:

• Outcome at end of treatment. Often after 6-18 weeks of treatment. The trial’s choice of end of treatment will be used. This is the most important outcome measure time point in this review.

• Outcome at maximum follow-up.

Data extraction, (selection and coding)

Selection of trials

Two of the review authors will independently select relevant trials, based on criteria described in the above. If a trial only has been identified by one of the two, it will be discussed whether the trial should be included. If the two review authors disagree, a third review author will decide if the trial should be included. All excluded trials are entered on a list, stating the reason for exclusion.

Data extraction

The following data will be extracted from the included trials:

1. Whether the trial is published or not.

2. Choice of antidepressant.

3. Whether a placebo washout period was used in the trial before inclusion of the participants.

4. Whether the participants are elderly (any definition used by the trialists).

5. Whether the participants have drug or alcohol dependence.

6. Whether ECT was used as co-intervention.

7. Whether the participants are chronically depressed or treatment resistant depressed (any definition used by the trialists).

8. Choice of control (no intervention, placebo, or ‘active’ placebo).

9. Whether the participants have a high baseline depression score (HDRS = 23).

10. Whether the participants have comorbid psychiatric diagnoses.

11. Whether the participants have borderline personality disorder.

12. Whether the experimental intervention is an add-on therapy on other antidepressants.

13. Whether the trial results are published or unpublished.

14. Length of intervention period and follow-up period.

15. Number of participants.

16. Distribution of age and sex.

17. Choice of outcomes (e.g., HDRS, BDI, suicides).

18. Outcomes (e.g., HDRS score, number of suicides).

19. The choice of method and an evaluation of the bias risk and choice of method (see below).

Risk of bias (quality) assessment

Risk of systematic error (bias)

We will use the instructions in The Cochrane Handbook for Systematic Reviews of Interventions in our evaluation of the methodology and hence bias risk of the included trials. Again, two review authors will assess the included trials independent of each other. We will evaluate the methodology in respect of generation of allocation sequence, allocation concealment, blinding of participants and treatment providers, blinding of outcome assessors, incomplete outcome data, selective outcome reporting, industry funding, and other bias sources. This is done because these components enable classification of randomised trials with low risk of bias and high risk of bias. The latter trials overestimate positive intervention effects and underestimate negative effects. We will classify the trials according to the components below:

Generating allocation sequence

‘Low risk of bias’: If randomising is performed by computer or a ‘random number table’. If the randomising is a random process, e.g., ‘heads or tails’ or a throw of a dice; and the person performing the procedure in no other way is involved in the trial.

Uncertain: If the procedure in respect of randomising is not sufficiently described.

‘High risk of bias’: If the trial uses, e.g., alternation for allocating the participants.

Allocation concealment

‘Low risk of bias’: If the allocation sequence is concealed from the investigators, treatment providers and participants, for example by central randomisation, and this procedure is described and documented.

Uncertain: If the procedure to conceal allocation is not sufficiently described.

‘High risk of bias’: If the investigators, treatment providers, and the participants are able to predict the allocation sequence. Such trials will be included only in the assessment of harms.

Blinding of participants and treatment providers

‘Low risk of bias’: If the participants and the treatment providers are blinded to treatment allocation and this is described. The placebo tablets should be identical to the antidepresssive tablets regarding appearance, colour, smell, taste, and solubility.

Uncertain: If the procedure of blinding is insufficiently described.

‘High risk of bias’: If blinding is not performed.

Blinding of outcome assessment

‘Low risk of bias’: If the trial investigators performing the outcome assessments, analyses and calculations are blinded to the treatment allocation and this is described. If some kind of ‘active’ placebo is used as control intervention this will be classified as ‘lowest risk of bias’.

Uncertain: If the procedure of blinding is insufficiently described.

‘High risk of bias’: If blinding is not performed.

Incomplete outcome data

‘Low risk of bias’: If dropouts following randomising can be described as being similar in the two intervention groups, and if the trial allows intention-to-treat analysis.

Uncertain: If dropouts are not stated, or if the reasons why the participants dropped out are unclear.

‘High risk of bias’: If the pattern of dropouts can be described as being different in the two intervention groups.

Selective outcome reporting

‘Low risk of bias’: If all outcome measures are stated in the results. And the hierarchy of the outcome measures are documented in a protocol before launch of randomisation.

Uncertain: If the method of choosing outcome measures is inadequately described.

‘High risk of bias’: If there is incongruence between the original protocol and the outcome measures used in the results, or if not all of the outcome measures are stated.

For profit bias

‘Low risk of bias’: If the trial is not financed by a company that might have an interest in a given result.

Uncertain: If there is no description of how the trial is financed.

‘High risk of bias’: If the trial is financed by a company that might have an interest in a given result.

Other sources of bias

If other sources of bias are evident these sources of bias will be presented and the implications will be discussed.

Overall assessment of risk of bias

A trial will be classified as ‘low risk of bias’ only if all of the bias components described in the above paragraphs are classified as ‘low risk of bias’ and. If one or more of the bias components are classified as ‘uncertain’ or ‘high risk of bias’ the trial will be classified as ‘high risk of bias’.

In case that we find no trials with low risk of bias or only find very few trials with low risk of bias, we plan to identify a group of trials with lower risk of bias according to the bias risk domains described in the above. This group of trials with lower risk of bias should at least include about 6% of the trial population in order to give the comparison of trials with lower risk of bias a certain power.

Assessment of reporting biases

Different types of reporting biases (e.g., publication bias, time lag bias, outcome reporting bias, etc.) will be handled following the recommendations of the Cochrane Handbook for Systematic Reviews and Interventions. On all outcomes, we will test for funnel plot asymmetry when there are at least ten trials included in the meta-analysis. For continuous outcomes with intervention effects measured as mean difference, the test proposed by Egger et al. will be used to test for funnel plot asymmetry. We will take into account that asymmetric funnel plots are not necessarily caused by publication bias, and publication bias does not necessarily cause asymmetry in a funnel plot.

Strategy for data synthesis

Statistical methods

We will undertake this meta-analysis according to the recommendations stated in The Cochrane Handbook for Systematic Reviews of Interventions. In analysing the primary outcomes we will use the mean difference (MD) on the 17-item HDRS with a 95% confidence interval. We will also use the standardised mean difference with a 95% confidence interval to analyse the results from the 17-item HDRS and other forms of the HDRS, BDI, and MADRS. We will use the odds ratio (OR) with a 95% confidence interval to estimate intervention effects on dichotomous outcomes. All meta-analyses will be performed both with a fixed-effect and a random-effects model.

The National Institute for Clinical Excellence (NICE) of the National Health Service in England has formerly defined a threshold for clinical significance as an effect size of 0.50 standardised mean difference (SMD) or a drug-placebo difference of three points on the 17-item HDRS. Others have suggested and used the following ‘rule of thumb’: 0.2 SMD represents a small effect, 0.5 SMD a moderate effect, and 0.8 SMD a large effect. We have chosen, as NICE has recommended and other reviewers have chosen, an effect size of 0.50 SMD or a drug-placebo difference of three points on the 17-item HDRS as the threshold for clinical significance.

We will also perform trial sequential analyses on the outcomes, in order to calculate the required information size and the cumulative Z-curve’s breach of relevant trial sequential monitoring boundaries. A more detailed description of trial sequential analysis can be found at http://www.ctu.dk/tsa/. For binary outcomes we will estimate the required information size based on the proportion of patients with an outcome in the control group, a risk ratio of 30%, an alpha of 5%, a beta of 20%, and the empirical diversity. For continuous outcomes we will estimate the required information size based on a mean difference of 3 HDRS points, the empirical SD, an alpha of 5%, a beta of 20%, and the empirical diversity.

Missing outcomes

Dichotomous outcomes

If missing dichotomous outcomes are not reported, we will in the primary analyses impute missing values assuming that the participants missing at follow-up have obtained ‘no remission’. Otherwise the number of analyzed participants will be used.

Continuous outcomes

We will primarily use follow-up scores. If only change values are reported the results will be analysed together with follow-up scores. If standard deviations (SD) are not reported the SDs will be calculated if this is possible using other data from the trial. If calculation is impossible the SDs will imputed from trials with similar characteristics.

We will perform two sensitivity analyses:

1. ’Best-worst-case’ scenario: It will be assumed that all participants lost to follow-up in the experimental group have a HDRS score = mean - 1 SD and mean + 2 SDs, have survived, had no suicide attempt, had no adverse event, and have obtained remission; and all those with missing outcomes in the control group have a HDRS score = mean + 1 SD and mean 2 + SDs, have not survived, had a suicide attempt, had a adverse event, and have obtained ‘no remission’.

2. ’Worst-best-case’ scenario: It will be assumed that all participants lost to follow-up in the experimental group have a HDRS score = mean + 1 SD and mean 2 + SDs, have not survived, had a suicide attempt, had a adverse event, and have obtained ‘no remission’; and that all those lost to follow-up in the control group have a HDRS score = mean - 1 SD and mean + 2 SDs, have survived, had no suicide attempt, had no adverse event, and have obtained remission.

Results from both scenarios will be presented in our publication.

Assessment of the statistical and clinical significance

According to the methodology described in the paragraphs above, we will use the following nine-step procedure to assess whether the thresholds for significance are crossed or not:

1. We will calculate and report the confidence intervals and P-values from the fixed-effect and random-effects meta-analyses. The most conservative result will be the primary.

2. We will explore the reasons behind substantial statistical heterogeneity by performing sensitivity analyses and meta-regression.

3. We will adjust the thresholds for significance (P-values and the confidence intervals from the meta-analyses and the risks of type I error in the trial sequential analysis) according the number of primary outcomes (a threshold for significance exactly midway between 0.05 and the Bonferroni adjusted P-value threshold).

4. We will calculate and report realistic diversity adjusted required information sizes and analyse all outcomes in the protocol with trial sequential analysis, and we will report if the trial sequential monitoring boundaries for benefit, harm, or futility are crossed. The trial sequential analysis will adjust the thresholds for significance by relating the accrued data to the required information size.

5. We will calculate and report Bayes factor for the primary outcome (or outcomes) based on a pre-specified anticipated intervention effect (same anticipated intervention effect as the one used in to estimate the required information size). A Bayes factor less than 0.1 (a ten fold higher likelihood of compatibility with the alternative hypothesis than the likelihood of compatibility with the null hypothesis) will be chosen as threshold for significance.

6. We will use sensitivity analyses and subgroup analysis to assess the potential impact of systematic errors (bias) (overall risk of bias, missing data).

7. We will use sensitivity analyses to assess the range of uncertainty of the estimated anticipated intervention effects.

8. We will assess the risk of publication bias (funnel plot).

9. We will assess and report clinical significance of the review results if the prior eight steps have shown indications of a statistically significant result.

Analysis of subgroups or subsets

Subgroup analyses

We have planned the following subgroup analyses on the primary outcomes:

1. Whether the intervention effects from trials with overall low risk of bias (or lower risk of bias) differ from trials with overall high risk of bias.

2. Whether the intervention effects from the trials using no intervention, placebo, or ‘active placebo’ differ.

3. Whether the results from trials using a placebo washout period before inclusion differ from the remaining trials.

4. Whether the intervention effects from trials assessing the effects of SSRIs in elderly depressive participants (defined by the trialists but often adults >= 65 years) differ from the remaining trials.

5. Whether the intervention effects from trials using ECT as co-intervention differ from the remaining trials.

6. Whether the intervention effects from trials assessing the effects of SSRIs in participants without comorbidities differ from trials including patients with drug and alcohol problems and from trials that include other psychiatric comorbidities.

7. Whether the intervention effects from trials assessing the effects from the six different SSRI antidepressants differ.

8. Former meta-analyses have shown that effects of antidepressive medication compared with placebo increased with increasing baseline depression severity (HDRS score). We will investigate whether the intervention effects from trials with participants with a baseline HDRS score of 23 or above differ from the remaining trials.

9. Whether the intervention effects from trials assessing the effects of SSRIs in chronically depressive patients or treatment resistant depression differ from the remaining trials.

10. Whether the intervention effects from trials assessing the effects of SSRIs in participants with and without borderline personality disorder differ.

11. Whether the intervention effects from trials assessing the effects of SSRIs as add-on therapy on any other antidepressant differ from the remaining trials.

12. Whether the intervention effect from the published trials differ from unpublished trials.

Contact details for further information

Janus Christian Jakobsen

Copenhagen Trial Unit

Blegdamsvej 9

2100 Copenhagen, Denmark

janusjakobsen@mac.com

Organisational affiliation of the review

Copenhagen Trial Unit

www.ctu.dk

Review team

Dr Janus Christian Jakobsen, Copenhagen Trial Unit Jane Lindschou Hansen, Copenhagen Trial Unit Signe Hellmuth, Copenhagen Trial Unit Anne Schou, Copenhagen Trial Unit Jesper Krogh, Department of Endocrinology, Herlev Hospital, Denmark Dr Christian Gluud, Copenhagen Trial Unit

Anticipated or actual start date

30 April 2013

Anticipated completion date

26 September 2014

Funding sources/sponsors

The Copenhagen Trial Unit has initiated the project. We have received no specific funding.

Formal screening of search results against eligibility criteria

Data extraction

Risk of bias (quality) assessment

Data analysis

PROSPERO This information has been provided by the named contact for this review. CRD has accepted this information in good faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, any associated files or external websites.