Comments from Keith Laws and Peter Kinderman re antipsychotics and CBT.
Kinderman: Throughout the report we were careful to say that different things help different people, and that (and I quote) “professionals need to acknowledge that the only way someone can find out for sure what helps them personally, is to try things out”. The comment that people have taken objection to is this one: ‘… on average, people gain around as much benefit from CBT as they do from taking psychiatric medication’. I think this is a defensible statement. Of course it’s true that there has not been a series of direct RCT comparisons between antipsychotic medication and CBT; those comparisons have not been conducted (although perhaps they should be). That doesn’t mean it’s appalling misrepresentation to say that the effectiveness appears about the same – it means that we have to synthesise other forms of data.
So the effect sizes of CBT for helping people with psychotic experiences in well-conducted meta-analyses of RCTs are modest, but clearly significant. My judgement is that the effect sizes are broadly comparable with the effect sizes reported for anti-psychotic medication.
This is an authoritative, consensual report, read and double-checked by many people, including many of the people speaking on Thursday. Some people disagree with the conclusions we’ve drawn from published findings, and life would be dull if we all agreed. But honourable disagreements are the stuff of science. The available evidence-base include a wide range of meta-analyses (often using different statistical estimates of effect-sizes) of the effectiveness of a wide range of anti-psychotic medication. That converges on an effect-size for the acute and medium-term reduction of psychotic experiences of somewhere between 0.2 and 0.9 (using Hedges’ g)(see http://bjp.rcpsych.org/content/198/4/247.full for one example, I can’t offer a systematic review here for obvious reasons). Recent meta-analyses offer an effect-size for CBTp of maybe 0.33, with some people (in, for instance the recent Maudsley debate on this issue http://www.kcl.ac.uk/ioppn/news/maudsleydebates/debate-archive-31-50.aspx) converging on an average effect size for CBTp between 0.5 and 0.7. S
There has, it’s true, been no direct comparison, so we have to make reasonable judgements of the available literature and I, for one, stand by that statement. It’s also worth noting that, in that judgment, issues of longer-term adverse effects should, I believe, be part of the consideration. And it’s worth noting that some people (at least) can benefit from CBT in the absence of anti-psychotic medication.
My judgement is that this particular statement does indeed reflect the available evidence, even if that evidence is limited.
Laws: A quick reply to the key claim that Peter Kinderman makes.
He states “There has, it’s true, been no direct comparison, so we have to make reasonable judgements of the available literature and I, for one, stand by that statement.”
So let’s go with his suggestion of comparing effect sizes in separate meta-analyses for antipsychoitics and CBT
He argues that antipsychotics for the acute and medium-term produce a “reduction of psychotic experiences of somewhere between 0.2 and 0.9 (using Hedges’ g)(see http://bjp.rcpsych.org/content/198/4/247.full for one example
And that “Recent meta-analyses offer an effect-size for CBTp of maybe 0.33, with some people (in, for instance the recent Maudsley debate on this issue http://www.kcl.ac.uk/ioppn/news/maudsleydebates/debate-archive-31-50.aspx) converging on an average effect size for CBTp between 0.5 and 0.7.”
Re the claim about antipsychotic effect sizes, I’m not sure why he references the Farooq and Taylor editorial – its about clozapine and comparisons of clozapine vs antipsychotics – I also cannot locate the effect size range he quotes in that editorial.
Nevertheless, we can turn to the excellent recent meta analysis by Leucht et al looking at 15 antipsychotics (first and second generation) in blind placebo RCTs measuring total symptom scores http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0057475/
Data from 43, 049 participants showed that all drugs were significantly more effective than placebo. The standardised mean differences with 95% credible intervals were: clozapine 0•88, 0•73-1•03; amisulpride 0•66, 0•53-0•78; olanzapine 0•59, 0•53-0•65; risperidone 0•56, 0•50-0•63; paliperidone 0•50, 0•39-0•60; zotepine 0•49, 0•31-0•66; haloperidol 0•45, 0•39-0•51; quetiapine 0•44, 0•35-0•52; aripiprazole 0•43, 0•34-0•52; sertindole 0•39, 0•26-0•52; ziprasidone 0•39, 0•30-0•49; chlorpromazine 0•38, 0•23-0•54; asenapine 0•38, 0•25-0•51; lurasidone 0•33, 0•21-0•45; and iloperidone 0•33, 0•22-0•43.
So a range at lowest .33 to highest of .88 – all significantly better than placebo in blind trials
Lets turn now to CBT meta analyses – Of course, I could present just our own recent meta-analysis (Jauhar et al 2014), but here are 9 recent meta outcomes for total symptoms (which tend to be better than positive or negative separately in the case of CBT) where blind outcomes were assessed:
Zimmerman et al 05 0.37
Wykes et al 08 0.22
Lincoln et al 08 0.25
NICE 09 0.27
Lynch et al 10 0.08
Sarin et al 11 0.07
Newton-Howes 11 0.04
Jauhar et al 14 0.15
Turner et al 14 0.12
They average out at 0.17 – almost spot on with our meta analysis (Jauhar et al 2014)
And .17 clearly doesn’t get anywhere near the effect sizes for even the poorest outcome antipsychotics
Just to finish the irony is not lost on me that the 0.33 for CBT that Peter Kinderman quotes is actually from my own meta analysis (Jauhar etal 2014) for all RCTs regardless of blinding– which we know to be crucial, though he seems keen to ignore blinding for some reason (except when it comes to measuring drug effects). Notably the UPS document also ironically references the same paper of ours as evidence that CBT effects are comparable to drug effects
And last, his claim that a wide range of meta analyses converge “on an average effect size for CBTp between 0.5 and 0.7.” The reference he ascribes is to his colleague David Kingdon from the Maudsley Debate we had earlier in the year– In it you will see Kingdon says “Meta-analyses “have consistently shown an effect size of around 0.3” which although is itself incorrect (as you can see above) is nowhere near the .5 to .7 ascribed
The bottom line is that when we go with Peter Kinderman’s suggested comparison, CBT effect size is nowhere near those for any antipsychotic and doesn’t even approach the poorest performing antipsychotic.