Antidepressants for depression in physically ill people
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
We aim to test the hypothesis that antidepressants are effective in depression in the context of a physical illness.
Background
Description of the condition
Depression is one of the main causes of disability worldwide (WHO 2007). It is characterised by depressed mood, a loss of interest or pleasure in everyday experiences, and a range of emotional, cognitive, physical and behavioural symptoms (NICE 2004).
Depression is common in the general population and affects around 121 million people in the world (WHO 2007). It is particularly common in patients with a concomitant physical illness. Studies of depression in medical inpatient populations have suggested that more than one third of patients report moderate depressive symptoms and 11‐26% suffer from a depressive syndrome (Rodin 1986). Similar rates have been found in patients with one or more chronic physical disease (Moussavi 2007). Depression in the physically unwell is also associated with poorer physical health outcomes. Patients with physical illness and depression have poorer functioning and higher levels of mortality and morbidity compared with patients with similar illnesses but without depression (Wulsin 1999).
There is no current consensus about whether depression in physically ill populations is a clinically different entity to that seen in those without physical illness (Cleare 2006; Creed 2007). Symptoms which may be interpreted as biological/somatic symptoms of depression can be seen in physically ill persons, and may be solely due to their physical disease. This provides a diagnostic challenge to clinicians aiming to manage depression in this patient group (MacHale 2002).
The growing epidemiological evidence that depression is a risk factor for poor prognosis of physical illnesses may be related to poor compliance or adherence to treatment. There is certainly evidence that depression has a major and consistent impact on compliance across a range of physical illnesses (DiMatteo 2000).
Given the importance of depression for prognosis in patients with physical illness, recent guidelines by the British National Institute of Clinical Excellence (NICE) suggest that the physically ill should be screened for depression (NICE 2004). A major public health question is whether improving the treatment of patients with depression would improve physical health as well as mental health outcomes.
Description of the intervention
Treatment for depression ranges from exercise programmes for milder depression to electroconvulsive therapy for severe cases that have not responded to other treatments (NICE 2004). Antidepressant drugs and psychological therapies are the most common treatments for depression, yet there is limited evidence for their efficacy in physically ill people. Psychological therapies such as cognitive behavioural therapy and interpersonal therapy may be effective, but these treatments will not be included in this review. A second systematic review of non‐pharmacological treatments is planned by this group to address this question.
Antidepressants are widely used in clinical practice and provide a feasible option for treating depression in the physically ill (MacHale 2002). There is however, a lack of consensus among clinicians regarding their use. It has been noted that clinicians often withhold potentially effective antidepressant drugs because they assume that the physical illness caused the depression, or they consider the physical illness to be a contraindication to antidepressant drug treatment (Roose 1992).
Why it is important to do this review
A previous Cochrane review (Gill 2000) found that use of antidepressants should at least be considered in those with both physical illness and depression. This review provided evidence that antidepressants cause improvement in depression in patients with a wide range of physical illnesses significantly more frequently than either placebo or no treatment (Gill 1999). Since the publication of this review, there have been a number of larger randomised controlled trials looking at the antidepressant treatment of depression in physical illness, including major depressive disorder, adjustment disorder and dysthymia
We aim to reassess the literature in the light of these studies in order to help clinicians make the best decisions about treatment with an antidepressant
Objectives
We aim to test the hypothesis that antidepressants are effective in depression in the context of a physical illness.
Methods
Criteria for considering studies for this review
Types of studies
We will identify all randomised controlled trials.
Types of participants
The participants will be adult males and females above 18 years of age with depression in the context of a physical illness. Depression includes diagnoses of Major Depressive Disorder, Adjustment Disorder and Dysthymic Disorder based on standardised criteria, such as the DSM‐IV (APA 1994) or the ICD‐10 (WHO 1992), and/or according to participant scores on validated tools such as the Hamilton Rating Scale for Depression (HRSD) (Hamilton 1960), the Montgomery‐Åsberg Depression Rating Scale (MADRS) (Montgomery 1979) or the Hospital Anxiety and Depression Scale (HAD) (Zigmond 1983), that are indicative of a diagnosis.
Trials will include participants with one or more physical diseases. We will exclude trials where the main comorbidity is for symptom‐based diagnoses such as chronic fatigue syndrome, fibromyalgia, chronic pain, irritable bowel syndrome and other "medically unexplained syndromes", since the nature of these disorders is contested and depression may play a specific role in their aetiology. We will exclude trials where antidepressants were prescribed primarily to treat other symptoms and not to treat depression. The term ‘comorbidity’ refers exclusively to comorbid physical illness, and does not include psychiatric comorbidity.
Types of interventions
Intervention
All types of antidepressants will be included, including:
Selective serotonin reuptake inhibitors, tricyclic antidepressants, monoamine oxidase inhibitors, serotonin noradrenaline reuptake inhibitors, noradrenergic specific serotonergic antidepressant, serotonin 2 antagonists, noradrenaline reuptake inhibitor, norepinephrine and dopamine reuptake blockers, tetracyclic antidepressants, heterocyclic antidepressants and psychostimulants
Control condition
The control condition will be placebo.
Main comparison
All types of antidepressants will be compared with placebo
Types of outcome measures
Primary outcome
The primary outcomes are:
1) Continuous measures of depression expressed as mean values at 6 to 8 weeks from randomisation, defined by validated measures such as the HDRS, the MADRS or the HADS
2) Binary outcome of individuals who attain a 50% improvement of depressive symptomatology at 6 to 8 weeks from randomisation, defined by validated measures such as the HDRS, the MADRS or the HADS.
Depression scores and symptomatology, defined by validated measures such as the HDRS, MADRS or HADS, will also be assessed at the following time points:
1) at less than 6 weeks
2) at 9 to 18 weeks
3) after 18 weeks
Secondary outcomes
1) Number of adverse events
2) Number of drop outs
3) Compliance with study medication
Search methods for identification of studies
Electronic searches
We will request the Cochrane Depression Anxiety and Neurosis Group to complete a search from the register of trials. We will use established methods to identify all placebo controlled trials of antidepressant treatment for depression. The search strategy for use in the CCDAN registers (CCDANCTR‐Studies and CCDANCTR‐References) is detailed in Appendix 1.
Searching other resources
Hansearching
Key journals will be identified with the help of the CCDAN Trials Search Coordinator and handsearched where appropriate.
Reference Lists
The bibliographies of all identified trials will be scanned for additional studies.
Correspondence
Key researchers in the field will be approached for published and unpublished trials not already identified. Information about unpublished trials will also be sought from pharmaceutical companies by the CCDAN Trials Search Coordinator.
Data collection and analysis
Selection of studies
We will assess all potentially eligible studies generated from our literature search using the inclusion criteria stated above. Selection of studies will involve an initial screening of titles and abstracts to determine whether each study might meet the eligibility criteria. If it is not clear from the title or abstract that the study should be rejected, the full text of the article will be obtained and reviewed. This process will be conducted independently by two authors (AE and KV) to reduce the possibility of relevant reports being rejected. Any disagreements about selection criteria will be resolved by discussion between reviewers, with final decisions being reached by MH, in consultation with IH.
Data extraction and management
Data will be extracted independently onto a specially designed, standardised data extraction form (attached). The data from identified papers will be extracted independently by two authors (AP and LR). We will extract data on study setting, diagnostic criteria used, number and characteristics of participants allocated, intervention and outcome measures, risk of bias, adverse events, drop outs and depression scores. On completion of data extraction, we will identify areas where there is disagreement, determine disagreement and reach a consensus under the supervision of MH, in consultation with IH.
Assessment of risk of bias in included studies
The methodological quality of the trials will be assessed independently by two authors (AP and LR) in parallel with data extraction. Risk of bias will be assessed using the Risk of bias table approach recommended in the Cochrane Handbook. The tool addresses six domains ‐ sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other issues. A description of what was reported to have happened in the study will be given for each domain, and a judgement relating to the risk of bias will be assigned (low risk, unclear risk and high risk). Any disagreements about methodological quality will be resolved by MH, in consultation with IH.
Measures of treatment effect
We anticipate that most trials will use validated depression rating scales as their main outcomes ‐the Hamilton Rating Scale for Depression (HRSD), the Montgomery‐Åsberg Depression Rating Scale (MADRS) or the Hospital Anxiety and Depression Scale (HADS), which was designed for patients with physical disease and may be used more in the trials we identify than in other contexts. Each of these scales generates a total score (continuous variable) and a binary outcome (recovered/not recovered). It is likely that not all trials will report using both methods, so we anticipate reporting both mean scores with standard deviations at different time points, and numbers (and percentages) of individuals recovered at each stage. We also anticipate (because of the conventions widely adhered to in antidepressant trials) that most trials will follow patients for 6 weeks, and depression outcome at 6 weeks will be the principal outcome for this review. We will record drop out from the trial, and where possible categorise this into drop out due to side effects or lack of efficacy.
Dichotomous data
For binary data, we will calculate odds ratios (OR) with 95% confidence intervals (CI).
Continuous data
For continuous data, will calculate standardised mean differences (SMD), using available mean values and their standard deviations, together with 95% CI
Dealing with missing data
Where statistics essential for analysis are missing because neither continuous nor binary outcome variables are reported, we will contact the authors to ask for the data. For participants who withdrew from the study before endpoint, it will be assumed that they did not respond to treatment. Where binary outcomes are not provided, but baseline and endpoint mean and standard deviation depression scores are given, we will impute the number of responding participants by assuming the normal distribution for the depression scores and calculating the number of participants below half the baseline score. This is a validated imputation method with strong empirical support (Furukawa 2005). Missing standard deviations will be calculated using the mean standard deviation from the other studies.
Assessment of heterogeneity
The I‐squared statistic will be used to quantify inconsistency across studies and assess the impact of heterogeneity on the meta‐analysis, with a score of 50% or higher indicating significant heterogeneity. We anticipate a priori that the following variables may contribute to heterogeneity: comorbid disease type (e.g. cancer, heart disease etc); type of measure used to score depression; class of antidepressant (e.g. SSRIs, tricyclics, other); time when outcome was measured; severity of depression; source of funding of the trial; definition of depression. We will gather information on each of these variables in the data extraction form.
Assessment of reporting biases
We will assess potential publication bias using funnel plots.
Data synthesis
For binary and continuous outcomes, data will be pooled in meta‐analyses using a random effects model, in Review Manager 5.0.
Subgroup analysis and investigation of heterogeneity
We plan to perform two sets of subgroup analyses, provided there are sufficient data.
1) We will present the analyses by each antidepressant and class of antidepressants (SSRIs, tricyclics, other).
2) We will describe results stratified by disease type: we expect these to be in the following categories: cancers, heart disease, neurological disease, respiratory disease, 'other ‐ infectious' and 'other ‐ non‐infectious'. However, these categories may change if there are large numbers of studies on specific disorders in the other categories.
We do not plan to stratify analysis by age.
Sensitivity analysis
An exploration of the robustness of the main results of the data synthesis will be performed. This will determine, for example, whether the findings are unduly influenced by a small number of studies with strong effect sizes, or studies in which there are significant concerns regarding methodological quality. We will also perform sensitivity analyses to determine whether any effect of antidepressants is restricted to studies with a strict definition of major depressive disorder.
Risk of bias will be summarised according to the Cochrane ‘summary assessment tables’ using low, unclear and high risk of bias. Studies judged to have a high risk of bias will not be entered into a meta analysis, and sensitivity analysis will be performed on studies with low risk of bias.
Drop‐outs will be assumed to have not responded to treatment, and sensitivity analyses will examine the effect of this assumption. Trials for which the response rate was imputed and those for which the mean standard deviation of the other studies was used will also be subject to sensitivity analysis.
