This Misuse Of IAPT Data Will Justify Cramming Clients Into Group Therapy

a just published study  of IAPT clients by Fanous and Daniels (2020) doi:10.1017/S1754470X20000045 apparently reveals that group CBT for generalised anxiety disorder is as good as individual cbt. But its’ assessment process was like no other in the randomised controlled trials of CBT for GAD. Making comparison between this study  and meta analysis of GAD outcome studies by Zhu et al  (2014) http://dx.doi.org/10.11919/j.issn.1002-0829.214173 impossible.  Specifically:

  1. Entry into the study was not determined by any of the standardised diagnostic interviews used in other controlled trials of CBT for GAD. Thus there can be no certainty that the population studied is the same as that in other rcts, i.e that they were suffering from GAD as the primary disorder. 
  2. The interview that was used to determine entry was not reused post treatment. Thus preventing a determination of the proportion of clients no longer suffering from GAD at the end.
  3. It is claimed that those entering the study were assessed according to the ICD 10 criteria for generalised anxiety disorder. But  in the ICD 10, GAD is a diagnosis of exclusion, it requires that a wide range of disorders to have first been excluded including depressive disorder, panic disorder and a specific phobia. Such differential diagnosis requires a very extensive clinical interview, see Clinical Interviewing Sommers- Flanagan 2016. There is no evidence that such an interview took place and it is unlikely to have done so as IAPT’s mantra ( see IAPT Manual) is that it doesn’t make diagnoses. What has happened is that the gatekeepers to the study have de facto used the GAD7 (which is based on the DSM criteria) for entry into the study and as the major metric of recovery.  But DSM-5, p19 [American Psychiatric (2013)]  explicitly counsels against the use of symptom checklists thus: ‘ it is not sufficient to simply check off the symptoms in the diagnostic criteria to make a mental disorder diagnosis’.
  4. The primary outcome measure used was the GAD7, though popular in IAPT this measure was not used in any of the 12 rcts of CBT for GAD  considered by Zhu et al (2014)  http://dx.doi.org/10.11919/j.issn.1002-0829.214173. Indeed Fanous and Daniels (2020) make no reference to this meta analysis.
  5. There was no independent blind assessment of outcome in the Fanous and Daniels (2020) study. Half the studies in the Zhu et al (2014) meta analysis had blind assessors and their findings were not as flattering as the self report measures.
  6. The authors failed to consider whether the found changes could have occurred in any case with the passage of time, as clients typically present initially at their worst. The effect sizes in the Fanous and Daniels (2020) study were 0.90 in group CBT and 0.94 in individual CBT, i.e the average client treated improved by less than 1 standard deviation. This should have been set in the context of  an 0.38 within subjects effect size found in GAD patients on a waiting list for CBT, see Robinson et al (2010)  and effect sizes of 1.73 and 1.55 in the comparison different modes of internet delivered CBT treatment.
  7. The authors use a score of below 10 at the end of treatment as indicating recovery, but this is not the metric used in any other study.  Yet using this, in true IAPT style, they claim a 53% recovery in individual CBT and a 41% recovery in group cbt. It is extremely doubtful that one half of the IAPT clients would have regarded themselves as back to their usual self after treatment.
  8. The GAD7 refers only to functioning in the last 2 weeks, whereas the DSM criteria refer to functioning over the last 6 months, recognising that anxiety symptoms do wax and wane. Thus to be regarded as in remission from GAD requires a significant period [at least 8 weeks as operationalised by Bruce et al (2005)] without the disorder. There is nothing in this IAPT study to suggest anyone was remitted in any meaningful way. Nonetheless the authors of the study promote the virtues of allegedly group cbt  over individual cbt.
  9. There may be a case for preferring group cbt over individual cbt, see Simply Effective Group Cognitive Behaviour Therapy [Scott (2011) London: Routledge], but this study does nothing to advance the case.  It is likely to result in the cramming in of a wide range of IAPT clients into ‘group cbt’ for ‘GAD’. A meta analysis of CBT for anxiety disorders and PTSD by Carpenter et al (2018) http://DOI: 10.1002/da.22728 found that sufficient data was only available for social anxiety disorder and PTSD  and for both disorders individual cbt was superior to group cbt.
  10. The Fanous and Daniels (2020) study had a dropout rate of about a third in each arm, making its’ intention to treat statistical procedure questionable, usually such a procedure is only used when the dropout rate is less than 20%.
  11. The study is retrospective, so it is not detailed how many people declined group CBT, an all too familiar occurrence in the author’s experience see Scott (2011). This gives a misleading impression of the acceptability of group cbt and minimises the complexity of selling it.
  12. In the Fanous and Daniels (2020) study there were no fidelity checks on the High Intensity clinicians to gauge whether they were tackling GAD treatment targets with matching treatment strategies.   This reflects the more general problem in IAPT, the espousal of NICE recommended treatments together with a paucity of evidence as to actual delivery at the coalface.

 

Dr Mike Scott

Chair BABCP Group CBT Special Interest Group