We’ve talked in the woodlands before about the relative benefits of individual patient data meta analysis (IPDMA) compared to standard meta-analyses (Knowles, 2013). IPDMA, rather than pooling the reported average intervention effects like a typical meta-analysis, actually puts together all the original data from the included studies; hence the phrase “individual patient data”, because the analysis has the data for every individual participant in the original study rather than the summary statistics.
This makes IPD a very powerful tool for exploring differences in the effect between different patient groups, because we can actually analyse the data according to different patient characteristics (for example, age, gender, or severity of illness). In particular this lets us look at whether those differences act as moderators of outcome; do they make an effect bigger or smaller?
In this study (Weitz et al, 2015), the team wanted to find out whether baseline depression severity moderated the effect of pharmacotherapy (ADM – antidepressant medication) compared to cognitive behavioural therapy (CBT). Simply put, is one treatment better than the other depending on how severe depression is to begin with?
Guidelines both in the US and the UK recommend that for severe depression pharmacotherapy is used, but there hasn’t previously been an analysis with sufficient statistical power to look at whether patients with severe depression really do better with pharmacotherapy compared to CBT. The IPD analysis allowed them to do just this.
- Trials were included that compared CBT to pharmacotherapy in patients with a diagnosis of depression
- They excluded studies that looked at participants under 18, who were receiving inpatient care, which looked at relapse prevention (rather than treatment during depression), or that included patients with comorbid medical disorders
- They assessed study quality using the Cochrane Collaboration tool, and quality was assessed by two authors independently
- They ran the analyses looking at outcomes on two standardised symptom questionnaires, the HAM-D and the BDI, and also looking at response (classified as a 50% reduction in the HAM-D score after treatment) and remission (a score of less than or equal to 7 on the HAM-D after treatment).
- The analysis included 1,700 patients, 906 receiving medication and 794 receiving CBT. This was data from 16 of the 24 trials identified as eligible.
- They checked if there was bias in terms of which studies provided data and which did not. The trials that provided data were not significantly different to those that didn’t in terms of effect size or outcomes.
- First they ran a ‘main effect’ analysis, just looking at the difference on outcomes between pharmacotherapy and CBT, without considering differences in baseline severity:
- They found that pharmacotherapy had better outcomes compared to CBT on the HAM-D symptom scale,
- But there wasn’t any difference when it came to rates of recovery or remission,
- And only a ‘non significant trend’ on the BDI.
- They then ran the analysis looking for a moderation effect, and didn’t find one:
- This means that whether the patient had more severe depression at baseline or not, didn’t make a difference in terms of the two treatments being more or less effective
- This finding contradicts guidelines that suggest more severely depressed patients should be given pharmacotherapy rather than CBT.
- They ran sensitivity analyses that showed the results held up when lower quality studies were removed. There was also no evidence of publication bias.
The authors say:
While this [analysis] shows that pharmacotherapy provides minor improvement in the treatment of depression relative to CBT in terms of the continuous measures, there is no indication that differences between the modalities were moderated by the degree of baseline depression severity. Therefore, the data are insufficient to recommend antidepressant medication over CBT in outpatients based on baseline severity alone.
The authors end their paper with the suggestion that further research should explore whether other clinical or demographic factors moderate the effectiveness of the two treatments. I wonder if alternatively individual preference for one treatment over another should be prioritised, if we fail to find hard and fast rules for determining which treatments are most effective for whom. Whilst this might lead to guidelines sounding rather vague, the recommendation “pick whichever you prefer” might actually help achieve those elusive goals of “personalised care” and, based on studies such as this, it’s supported by the evidence as well.
- This is more a general question about the usefulness of these kind of studies, rather than being a criticism of this particular study, but it’s standard in trials to parse treatments into either/or, and I wonder if in routine care it’s more common for people with severe depression to receive both medication and therapy? Should research look more at additive effects of treatments rather than trying to compare effects in isolation?
- The authors note that the findings might not generalise to other forms of therapy or to medications not used in the included trials.
Weitz ES, Hollon SD, Twisk J, et al. (2015) Baseline Depression Severity as Moderator of Depression Outcomes Between Cognitive Behavioral Therapy vs Pharmacotherapy: An Individual Patient Data Meta-analysis. JAMA Psychiatry. Published online September 23, 2015. doi:10.1001/jamapsychiatry.2015.1516. [Abstract]
Knowles S. (2013) Moderators of outcome in late-life depression: should we be prescribing antidepressants to older people? The Mental Elf, 26 Aug 2013.
About IPD meta-analyses. Cochrane Methods IPD meta analysis website, last accessed 14 Dec 2015.