Clinician-trialist rounds: 6. Testing for blindness at the end of your trial is a mug's game
Bibliographic record
Abstract
Today’s definition: Mug’s game: a useless or illadvised venture carried out by a gullible person. Today’s case: We’d just completed the first-ever RCT showing that aspirin (but not sulfinpyrazone) reduced the risk of stroke and death among patients with transient ischemic attacks.* We were elated with our primary results and already dreaming of a lead article in the New England Journal of Medicine. There were just a few odds and ends to attend to, and one of them was analyzing an end-of-study questionnaire we’d given to our collaborating neurologists to confirm that our efforts to keep them blind had been successful. Our trial had employed a ‘double-dummy’ factorial design in which patients were randomized to both active drugs, to active aspirin and placebo sulfinpyrazone, to placebo aspirin and active sulfinpyrazone, or to both placebos. Consequently, when we asked our neurologists which regimen they thought each of their patients had received, they would have guessed correctly for 25% of them on the basis of chance alone. Any big increase in this rate of correct responses would be worrisome, and a statistically significant difference would suggest that our attempts to blind them had failed. ‘I felt the bullet enter my heart’ [1] when our co-PI statistician tracked me down on the ward to tell me that our clinicians’ correct guesses were, indeed, statistically significantly different from 25%. Had our triumphant lead article just been reduced to an apologetic Letter to the Editor? And why did my co-PI have a big grin on his face? I have lots of textbooks on how to do ‘doubleblind’y RCTs [2], and (except for one recent revision) all of them recommend end-of-study tests for blindness on both patients and providers. They go on to warn that greater-than-chance correct guesses raise real concerns about whether blinding was successful, and that when this occurs, trial reports should admit these failures and temper their conclusions accordingly. Moreover, reviews of published RCTs have found that these textbook recommendations are rarely reported. Isabelle Boutron led a review of 90 trials obtained from several bibliographic databases and concluded: ‘Methods of assessing the success of blinding, analysis and reporting the results were inconsistent and questionable.’[3]. Testing for blinding was reported in only 8% of the random sample of 199 general medicine and psychiatry RCTs published in 1998–2001 assessed by Dean Fergusson and his colleagues [4] and in only 2% of a random sample of 1599 RCTs published in 2001 and drawn from the Cochrane Central Register of Controlled Trials by Asbjorn Hrobjartsson’s team [5]. Gloomier still (at least in these authors’ eyes), in the rare instances in which tests were carried out, blinding was judged to have been successful only one-third to one-half of the time. As with cointervention in our previous Round, understanding the measurement of blinding requires the synthesis of methodological and clinical competence. The key question in this Round is: What are you really measuring when you measure ‘blindness’ at the end of your trial? And the quick
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.727 | 0.871 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.012 | 0.011 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.003 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.016 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; both teacher heads agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".