A question from an ignorant med student... The impaired group was defined as having a Cantab score of -1.0 compared to healthy individuals; (while I have no idea what scale they are using) from my dim recollections, if this executive functioning score is anything like IQ, the impaired group is still within the range of normal functioning, if it's sitting at one standard deviation below the average...
My understanding from observation was that the issue with cognitive impairments in schizophrenia was not that people were within the normal range, but that they were noticeably impaired, more like below 3 standard deviations below... I'm not sure how relevant this research/study population is to the patient with schizophrenia who needs help?
For what are probably understandable reasons, you're a bit mistaken about how cognitively impaired the average person with schizophrenia is. There have been quite a few studies that show that mean IQ for schizophrenics is somewhere around 85-90 (i.e. ~1 standard deviation below the healthy mean, see: https://pmc.ncbi.nlm.nih.gov/articles/PMC9118026/).
You're right that this is not what we would consider to be an intellectual disability, but 1 s.d. below the mean puts these individuals in the bottom ~16% of cognitive functioning for the population. That would be a significant disadvantage for most healthy people, nevermind ones with SMI, no?
I think you do get at an interesting question, though. What about that fraction of schizophrenics who are >3 s.d. from the mean? If you look at the last graph I included in the "Outcomes" section, I think this gets at your question about how relevant this population is to the severely cognitively impaired. The effect size gets *larger* the more impaired the average z-score of the group is, not less.
About the backronym - they're being self-referential; "Cantab" is the abbreviation that comes after degrees from Cambridge uni eg MA Hons (Cantab). Which is, I dunno.
More importantly, thanks for writing this - hopefully at some point NICE will approve this and I can see if it's helpful for our more impaired "chronic" schizophrenia patients in particular
I had no idea. Is that something unique to Cambridge, or do all universities in the UK do that? In the US we prefer to try and collect as abbreviatory letters after our names as possible.
It is an important abbreviation to put after some Cambridge degrees (esp. MAs) because if you complete a bachelor's degree at Cambridge, after a certain period of time you can fill out an application and pay a fee and automatically receive a Master of Arts degree (no field specified) without having to do literally any other academic work. It gives you various privileges in the university but the designation lets anyone who is familiar with the Oxbridge system (Oxford does it too) know that "this is not a real degree".
This is an incorrect understanding of effect size. You're right that for a fixed effect size, the sicker patients will have larger gains in absolute terms. It's not a given that effect size will increase in sicker groups, since effect size is a relative measure.
"The phenomena being measured might in some sense be more akin to mean regression than clinical efficacy."
If that was the case, why was the p-value between the controls and the treatment group in the impaired subgroup so large?
"It could be an imposition of order on random data, since it was done post facto by motivated researchers."
Yes, this is always the problem with post-hoc analysis, but I don't think it means you can never ever trust post-hoc analyses. Would you be willing to agree that there exists some threshold of statistical robustness that a post-hoc analysis could demonstrate that would make it highly credible?
I had never heard of the ICEMAN Guidelines before you mentioned them here - seems really useful to help evaluate subgroup analysis. Thanks for mentioning them
A question from an ignorant med student... The impaired group was defined as having a Cantab score of -1.0 compared to healthy individuals; (while I have no idea what scale they are using) from my dim recollections, if this executive functioning score is anything like IQ, the impaired group is still within the range of normal functioning, if it's sitting at one standard deviation below the average...
My understanding from observation was that the issue with cognitive impairments in schizophrenia was not that people were within the normal range, but that they were noticeably impaired, more like below 3 standard deviations below... I'm not sure how relevant this research/study population is to the patient with schizophrenia who needs help?
Ignore if I'm totally confused!
For what are probably understandable reasons, you're a bit mistaken about how cognitively impaired the average person with schizophrenia is. There have been quite a few studies that show that mean IQ for schizophrenics is somewhere around 85-90 (i.e. ~1 standard deviation below the healthy mean, see: https://pmc.ncbi.nlm.nih.gov/articles/PMC9118026/).
You're right that this is not what we would consider to be an intellectual disability, but 1 s.d. below the mean puts these individuals in the bottom ~16% of cognitive functioning for the population. That would be a significant disadvantage for most healthy people, nevermind ones with SMI, no?
I think you do get at an interesting question, though. What about that fraction of schizophrenics who are >3 s.d. from the mean? If you look at the last graph I included in the "Outcomes" section, I think this gets at your question about how relevant this population is to the severely cognitively impaired. The effect size gets *larger* the more impaired the average z-score of the group is, not less.
About the backronym - they're being self-referential; "Cantab" is the abbreviation that comes after degrees from Cambridge uni eg MA Hons (Cantab). Which is, I dunno.
More importantly, thanks for writing this - hopefully at some point NICE will approve this and I can see if it's helpful for our more impaired "chronic" schizophrenia patients in particular
I had no idea. Is that something unique to Cambridge, or do all universities in the UK do that? In the US we prefer to try and collect as abbreviatory letters after our names as possible.
All UK unis do it, though most make more sense because they aren't Latin!
It is an important abbreviation to put after some Cambridge degrees (esp. MAs) because if you complete a bachelor's degree at Cambridge, after a certain period of time you can fill out an application and pay a fee and automatically receive a Master of Arts degree (no field specified) without having to do literally any other academic work. It gives you various privileges in the university but the designation lets anyone who is familiar with the Oxbridge system (Oxford does it too) know that "this is not a real degree".
As a former Cantabrigian myself, it is an abbreviation of Cantabrigia (hence the adjective). It's a Latinization of the Old English Cantebrigge.
Oxford's "Oxon" is similarly an abbreviation of the Latin name for the place (Oxonia), because of course it is.
(you may well already know this, in which case apologies for being all WELL ACTUALLY...)
This is an incorrect understanding of effect size. You're right that for a fixed effect size, the sicker patients will have larger gains in absolute terms. It's not a given that effect size will increase in sicker groups, since effect size is a relative measure.
Here's an example in an analysis of MDD that showed just that: https://pubmed.ncbi.nlm.nih.gov/29611870/
Here's another for SAD, OCD, and PTSD: https://pubmed.ncbi.nlm.nih.gov/29659102/
Even without studies it's easy to imagine the counterfactual: patients with treatment resistant schizophrenia.
No worries! I think most of us (myself included) have this intuition about effect size until someone tells us otherwise
Well sure, but all the examples you gave (BP and Diabetes meds) actually work…?
Aren’t you saying these do too?
"The phenomena being measured might in some sense be more akin to mean regression than clinical efficacy."
If that was the case, why was the p-value between the controls and the treatment group in the impaired subgroup so large?
"It could be an imposition of order on random data, since it was done post facto by motivated researchers."
Yes, this is always the problem with post-hoc analysis, but I don't think it means you can never ever trust post-hoc analyses. Would you be willing to agree that there exists some threshold of statistical robustness that a post-hoc analysis could demonstrate that would make it highly credible?
Mean regression sounds great but it’s placebo controlled, so mean regression applies to the placebo as well.
Imposition of order on random order is a concern I give you that, but it seems a reasonable enough subgroup to assess.
I had never heard of the ICEMAN Guidelines before you mentioned them here - seems really useful to help evaluate subgroup analysis. Thanks for mentioning them
I say all this but also your intuition of being skeptical about the data is probably spot on!!