As described in the last issue of Mind the Brain, peaceful post-publication peer reviewers (PPPRs) were ambushed by an author and an editor. They used the usual home team advantages that journals have – they had the last word in an exchange that was not peer-reviewed.
As also promised, I will team up in this issue with Magneto to bust them.
Attacks on PPPRs threaten a desperately needed effort to clean up the integrity of the published literature.
The attacks are getting more common and sometimes vicious. Vague threats of legal action caused an open access journal to remove an article delivering fair and balanced criticism.
In a later issue of Mind the Brain, I will describe an horrendous incident in which authors modified a data set available on the web without notice after PPPRs used the data for re-analyses. The authors then used the modified data to redo analyses in an attempt to humiliate the PPPRs with claims that they did not know what they were doing. Fortunately, the PPPRs retained time stamped copies of both data sets. You may like to think that such precautions are unnecessary, but just imagine what critics of PPPR would be saying if they had not saved this evidence.
Until journals get more supportive of post publication peer review, we need repeated vigilante actions, striking from Twitter, Facebook pages, and blogs. Unless readers acquire basic critical appraisal skills and take the time to apply them, they will have to keep turning to the social media for credible filters of all the crap that is flooding the scientific literature.
I’ve enlisted Magneto because he is a mutant. He does not have any extraordinary powers of critical appraisal. To the contrary, he unflinchingly applies what we should all acquire. As a mutant, he can apply his critical appraisal skills without the mental anguish and physiological damage that could beset humans appreciating just how bad the literature really is. He doesn’t need to maintain his faith in the scientific literature or the dubious assumption that what he is seeing is just a matter of repeat offender authors, editors, and journals making innocent mistakes.
Humans with critical appraisal risk demoralization and too often shirk from the task of telling it like it is. Some who used their skills too often were devastated by what they found and fled academia. More than a few are now working in California in espresso bars and escort services.
Thank you, Magneto. And yes, I again apologize for having tipped off Jim Coan about our analyses of his spinning and statistical manipulations of his work to get newsworthy finding. Sure, it was an accomplishment to get a published apology and correction from him and Susan Johnson. I am so proud of Coan’s subsequent condemnation of me on Facebook as the Deepak Chopra of Skepticism that I will display it as an endorsement on my webpage. But it was unfortunate that PPPRs had to endure his nonsensical Negative Psychology rant, especially without readers knowing what precipitated it.
The following commentary on the exchange in Journal of Nervous and Mental Disease makes direct use of your critique. I have interspersed gratuitous insults generated by Literary Genius’ Shakespearean insult generator and Reocities’ Random Insult Generator.
How could I maintain the pretense of scholarly discourse when I am dealing with an author who repeatedly violates basic conventions like ensuring tables and figures correspond to what is claimed in the abstract? Or an arrogant editor who responds so nastily when his slipups are gently brought to his attention and won’t fix the mess he is presenting to his readership?
As a mere human, I needed all the help I could get in keeping my bearings amidst such overwhelming evidence of authorial and editorial ineptness. A little Shakespeare and Monty Python helped.
The statistical editor for this journal is a saucy full-gorged apple-john.
Cognitive Behavioral Techniques for Psychosis: A Biostatistician’s Perspective
Domenic V. Cicchetti, You may be, as your website claims
A psychological methodologist and research collaborator who has made numerous biostatistical contributions to the development of major clinical instruments in behavioral science and medicine, as well as the application of state-of-the-art techniques for assessing their psychometric properties.
But you must have been out of “the quintessential role of the research biostatistician” when you drafted your editorial. Please reread it. Anyone armed with an undergraduate education in psychology and Google Scholar can readily cut through your ridiculous pomposity, you undisciplined sliver of wild belly-button fluff.
You make it sound like the Internet PPPRs misunderstood Jacob Cohen’s designation of effect sizes as small, medium, and large. But if you read a much-accessed article that one of them wrote, you will find a clear exposition of the problems with these arbitrary distinctions. I know, it is in an open access journal, but what you say is sheer bollocks about it paying reviewers. Do you get paid by Journal of Nervous and Mental Disease? Why otherwise would you be a statistical editor for a journal with such low standards? Surely, someone who has made “numerous biostatistical contributions” has better things to do, thou dissembling swag-bellied pignut.
More importantly, you ignore that Jacob Cohen himself said
The terms ‘small’, ‘medium’, and ‘large’ are relative . . . to each other . . . the definitions are arbitrary . . . these proposed conventions were set forth throughout with much diffidence, qualifications, and invitations not to employ them if possible.
Cohen J. Statistical power analysis for the behavioural sciences. Second edition, 1988. Hillsdale, NJ: Lawrence Earlbaum Associates. p. 532.
Could it be any clearer, Dommie?
You suggest that the internet PPPRs were disrespectful of Queen Mother Kraemer in not citing her work. Have you recently read it? Ask her yourself, but she seems quite upset about the practice of using effects generated from feasibility studies to estimate what would be obtained in an adequately powered randomized trial.
Pilot studies cannot estimate the effect size with sufficient accuracy to serve as a basis of decision making as to whether a subsequent study should or should not be funded or as a basis of power computation for that study.
Okay you missed that, but how about:
A pilot study can be used to evaluate the feasibility of recruitment, randomization, retention, assessment procedures, new methods, and implementation of the novel intervention. A pilot study is not a hypothesis testing study. Safety, efficacy and effectiveness are not evaluated in a pilot. Contrary to tradition, a pilot study does not provide a meaningful effect size estimate for planning subsequent studies due to the imprecision inherent in data from small samples. Feasibility results do not necessarily generalize beyond the inclusion and exclusion criteria of the pilot design.
A pilot study is a requisite initial step in exploring a novel intervention or an innovative application of an intervention. Pilot results can inform feasibility and identify modifications needed in the design of a larger, ensuing hypothesis testing study. Investigators should be forthright in stating these objectives of a pilot study.
Dommie, although you never mention it, surely you must appreciate the difference between a within-group effect size and a between-group effect size.
- Interventions do not have meaningful effect sizes, between-group comparisons do.
- As I have previously pointed out
When you calculate a conventional between-group effect size, it takes advantage of randomization and controls for background factors, like placebo or nonspecific effects. So, you focus on what change went on in a particular therapy, relative to what occurred in patients who didn’t receive it.
Turkington recruited a small, convenience sample of older patients from community care who averaged over 20 years of treatment. It is likely that they were not getting much support and attention anymore, whether or not they ever were. The intervention that Turkington’s study provided that attention. Maybe some or all of any effects were due to simply compensating for what was missing from from inadequate routines care. So, aside from all the other problems, anything going on in Turkington’s study could have been nonspecific.
Recall that in promoting his ideas that antidepressants are no better than acupuncture for depression, Irving Kirsh tried to pass off within-group as equivalent to between-group effect sizes, despite repeated criticisms. Similarly, long term psychodynamic psychotherapists tried to use effect sizes from wretched case series for comparison with those obtained in well conducted studies of other psychotherapies. Perhaps you should send such folks a call for papers so that they can find an outlet in Journal of Nervous and Mental Disease with you as a Special Editor in your quintessential role as biostatistician.
Douglas Turkington’s call for a debate
Doug, as you requested, I sent you a link to my Google Scholar list of publications. But you still did not respond to my offer to come to Newcastle and debate you. Maybe you were not impressed. Nor did you respond to Keith Law’s repeated request to debate. Yet you insulted internet PPPR Tim Smits with the taunt,
You congealed accumulation of fresh cooking fat.
I recommend that you review the recording of the Maudsley debate. Note how the moderator Sir Robin Murray boldly announced at the beginning that the vote on the debate was rigged by your cronies.
Do you really think Laws and McKenna got their asses whipped? Then why didn’t you accept Laws’ offer to debate you at a British Psychological Society event, after he offered to pay your travel expenses?
High-Yield Cognitive Behavioral Techniques for Psychosis Delivered by Case Managers…
Dougie, we were alerted that bollacks would follow with the “high yield” of the title. Just what distinguishes this CBT approach from any other intervention to justify “high yield” except your marketing effort? Certainly, not the results you have obtained from an earlier trial, which we will get to.
Where do I begin? Can you dispute what I said to Dommie about the folly of estimating effect sizes for an adequately powered randomized trial from a pathetically small feasibility study?
I know you were looking for a convenience sample, but how did you get from Newcastle, England to rural Ohio and recruit such an unrepresentative sample of 40 year olds with 20 years of experience with mental health services? You don’t tell us much about them, not even a breakdown of their diagnoses. But would you really expect that the routine care they were currently receiving was even adequate? Sure, why wouldn’t you expect to improve upon that with your nurses? But would you be demonstrating?
The PPPR boys from the internet made noise about Table 2 and passing reference to the totally nude Figure 5 and how claims in the abstract had no apparent relationship to what was presented in the results section. And how nowhere did you provide means or standard deviations. But they did not get to Figure 2 Notice anything strange?
Despite what you claim in the abstract, none of the outcomes appear significant. Did you really mean standard error of measurement (SEMs), not standard deviations (SDs)? People did not think so to whom I showed the figure.
And I found this advice on the internet:
If you want to create persuasive propaganda:
If your goal is to emphasize small and unimportant differences in your data, show your error bars as SEM, and hope that your readers think they are SD.
If our goal is to cover-up large differences, show the error bars as the standard deviations for the groups, and hope that your readers think they are a standard errors.
Why did you expect to be able to talk about effect sizes of the kind you claim you were seeking? The best meta analysis suggests an effect size of only .17 with blind assessment of outcome. Did you expect that unblinding assessors would lead to that much more improvement? Oh yeh, you cited your own previous work in support:
That intervention improved overall symptoms, insight, and depression and had a significant benefit on negative symptoms at follow-up (Turkington et al., 2006).
Let’s look at Table 1 from Turkington et al., 2006.
A consistent spinning of results
Don’t you just love those three digit significance levels that allow us to see that p =.099 for overall symptoms meets the apparent criteria of p < .10 in this large sample? Clever, but it doesn’t work for depression with p = .128. But you have a track record of being sloppy with tables. Maybe we should give you the benefit of a doubt and ignore the table.
But Dougie, this is not some social priming experiment with college students getting course credit. This is a study that took up the time of patients with serious mental disorder. You left some of them in the squalor of inadequate routine care after gaining their consent with the prospect that they might get more attention from nurses. And then with great carelessness, you put the data into tables that had no relationship to the claims you were making in the abstract. Or in your attempts to get more funding for future such ineptitude. If you drove your car like you write up clinical trials, you’d lose your license, if not go to jail.
The 2014 Lancet study of cognitive therapy for patients with psychosis
Forgive me that I missed until Magneto reminded me that you were an author on the, ah, controversial paper
Morrison, A. P., Turkington, D., Pyle, M., Spencer, H., Brabban, A., Dunn, G., … & Hutton, P. (2014). Cognitive therapy for people with schizophrenia spectrum disorders not taking antipsychotic drugs: a single-blind randomised controlled trial. The Lancet, 383(9926), 1395-1403.
But with more authors than patients remaining in the intervention group at follow up, it is easy to lose track.
You and your co-authors made some wildly inaccurate claims about having shown that cognitive therapy was as effective as antipsychotics. Why, by the end of the trial, most of the patients remaining in follow up were on antipsychotic medication. Is that how you obtained your effectiveness?
In our exchange of letters in The Lancet, you finally had to admit
We claimed the trial showed that cognitive therapy was safe and acceptable, not safe and effective.
Maybe you should similarly be retreating from your claims in the Journal of Nervous and Mental Disease article? Or just take refuge in the figures and tables being uninterpretable.
No wonder you don’t want to debate Keith Laws or me.
A retraction for High-Yield Cognitive Behavioral Techniques for Psychosis…?
The Turkington article meets the Committee on Publication Ethics (COPE) guidelines for an immediate retraction (http://publicationethics.org/files/retraction%20guidelines.pdf).
But neither a retraction nor even a formal expression of concern has appeared.
Maybe matters can be left as they now are. In the social media, we can point to the many problems of the article like a clogged toilet warning that Journal of Nervous and Mental Disease is not a fit place to publish – unless you are seeking exceeding inept or nonexistent editing and peer review.
Vigilantes can periodically tweet Tripadvisor style warnings, like
Now, Dommie and Dougie, before you again set upon some PPPRs just trying to do their jobs for little respect or incentive, consider what happened this time.
Special thanks are due for Magneto, but Jim Coyne has sole responsibility for the final content. It does not necessarily represent the views of PLOS blogs or other individuals or entities, human or mutant.