I was going to write another post relating to perceptions of identity in mental disorder this week but then this Lancet paper was published today (Tyrer P et al. The Lancet 2013) It presents the results of a study assessing the impact of cognitive behavioural therapy, focussed on health anxiety (CBT-HA) on health anxiety scores in those attending outpatient clinics in a general, not mental health, hospital (cardiology, endocrine, gastroenterology, and respiratory medicine clinics). This got me thinking about a topic, outcome measures in psychiatric clinical trials, that I think can be very interesting and needs more focus from research.
I’ll start with a quick overview of the Lancet report and then move onto a discussion of outcome measures in clinical trials.
CBT for Health anxiety in physical healthcare patients
The primary outcome measure for the study was reduction in health anxiety severity at one year. Secondary outcomes measures included – health anxiety severity at shorter time intervals, depressive symptom severity, social functioning and quality of life. Control was treatment as usual. The intervention was delivered by non-expert therapists.
As you would expect the health anxiety rating and generalised anxiety rating decreased throughout the period of assessment for both intervention and control groups, there was a statistically significant difference in change in rating scale between the intervention and control groups at 12 months (2.98 (1.64-4.33)), the ratio of those in the intervention group compared to control group meeting “normal health anxiety levels” was 2.15 (95% CI 1.09-4.23). So, on average, you were twice as likely to reach what the authors defined as “normal” levels on the health anxiety rating scale at one year with the intervention then without. Whether this difference was maintained is not quoted, so likely not.
Ok, sorry that last paragraph was dull – bottom line is that on their primary outcome measure they demonstrated a difference with their intervention. This seems good news. However none of the secondary outcome measures showed any difference with intervention compared to control – notably no impact on quality of life, meaning that they could then not demonstrate the impact of their intervention to a level required by NICE. The discussion section mentions this difficulty in one paragraph, but I felt it required far more attention for reasons I’ll explain below.
I first heard Peter Tyrer presenting these findings in Edinburgh during the summer and I agree with him that this is an important field needing more research and treatment options. Health anxiety is debilitating and can have detrimental impact on the quality of response to physical health care. I believe the study has a number of limitations which are well summarised in an accompanying editorial in the same Lancet issue (Williams C, House A. Cognitive behaviour therapy for health anxiety. The Lancet 2013).
Outcome measures in psychiatric research
So the paper has its limitations, it’s a valuable addition to the field but would not have me personally rushing to commission such a service if I had that power until there was more evidence available.
What I think is more interesting however, and ties in more closely with my research interests, is the use of outcome rating scales in psychiatric treatment trials.
In physical healthcare trials, for example looking at drugs to target type II diabetes, the choice of outcome measure has received a lot of recent attention. In type II diabetes the outcomes that matter are risk to life (mortality) and limb (morbidity). These however take a long time to measure as thankfully events are rare and require large populations in clinical trial followed up over long periods of time. What is much simpler is to target a proxy measure of clinical outcome. In the case of diabetes an example would be measuring blood sugar levels, which can be represented by the amount of glucose binding to haemoglobin in the blood, this is the HbA1c measure. The argument is that HbA1c can be used as a measure of disease progression, and it is routinely used in treatment assessment. However, if a trial focusses only on HbA1c and does not consider other outcomes then the true impact of an intervention can be missed. Thus a drug may lower sugar levels, but increase death rates owing to heart disease (cardiovascular mortality), as was the case with Rosiglitazone (Cohen D. Rosiglitazone: what went wrong? BMJ 2010; 341: c4848). There is a fantastic article in the BMJ last year discussing this in a lot more detail (Yudkin JS, Lipska KJ, Montori VM. The idolatry of the surrogate. BMJ 2011; 343: d7995).
So what does this mean for psychiatric clinical trials, including psychological treatments? The difficulty with surrogate measures is that they provide no information to the person requesting help from the clinician, and may mask underlying effects (increased heart failure rates with Rosiglitazone). Now – psychiatric clinical trials, as in the example of Tyrer et al above, tend to focus on symptom severity scales in terms of outcome measure. These scales are valuable as symptoms of mental disorder can be highly distressing. However, we have a lot of literature from the service user movement suggesting that symptom resolution, while obviously beneficial, is not entirely necessary, or sufficient, for the individual to progress. This leads us to the concept of “personal recovery”. Personal recovery represents an individuals perspective on their experience of mental disorder, the care and support received as well as their hopes for the future. Pat Deegan who I referenced in my last blog has written some fantastic articles on this.
Symptom rating scales, as used in many clinical trials, have been shown to overestimate the treatment effect in comparison to a impression of clinical change noted by the clinician (Lepping P, Sambhi RS, Whittington R, Lane S, Poole R. Clinical relevance of findings in trials of antipsychotics: systematic review. The British Journal of Psychiatry 2011; 198: 341–5). There is also little overlap between the rating scales used and client defined concepts of personal recovery (Andresen R, Caputi P, Oades LG. Do clinical outcome measures assess consumer-defined recovery? Psychiatry Research 2010; 177: 309–17).
So do the outcome scales used in the majority of psychiatric clinical trials represent useful information to clinicians and service users making decisions in relation to treatment options? I would argue not: I believe that people seek help in relation to their experience of mental disorder because of the impact of their personal experiences on their quality of life, relationships and work. So does assessment of quality of life make a reasonable target for gauging the efficacy of services and treatments? A recent paper in Social Science and Medicine argues that serial monitoring of subjective well being measures can be used to model individual experience of care (Lee H, Vlaev I, King D, Mayer E, Darzi A. Subjective well-being and the measurement of quality in healthcare. Social Science & Medicine 2013). The measurement of quality of life in those with experience of mental disorder can be complex, particularly in those suffering with psychotic experiences (Brazier J. Is the EQ-5D fit for purpose in mental health? The British Journal of Psychiatry 2010; 197: 348–9), but I do not think this should deter us from trying to develop better measures.
Personal recovery from mental disorder is an individual process and can likely never be adequately reflected through the use of outcome scales. Is it possible to develop individually tailored outcome measures that are agreed with individual participants at the beginning of clinical trials? At present I believe that measures of quality of life likely represent a better, more representative, measure than severity rating scales, although, I acknowledge, they are far from perfect. We need far more research to understand what people seeking help from the mental health services want with regard to support. In the mean time I think that we need to be careful in our reading of a lot of the psychiatric research literature in relation to treatment options. I don’t want to dismiss the literature, there is valuable information in there but ultimately treatment decisions should be taken on an individual basis with continuous dialogue between the clinician and service user to ensure that the best outcome is reached.
Rant over, dismounts from soapbox
That was a lot longer and more of a rant than I intended – sorry for this, but I do think that we need to think seriously about this topic as we risk misleading clients in making treatment decisions.