What are the best meta-study papers/threads that critique pre-existing studies for quality? ["peer review failures"]

AlexKChen · July 9, 2025, 6:49am

Positive response bias (esp curcumin) is a massive problem, even more in the area of nutraceuticals and supplements (and neurofeedback/tACS)

Eg

https://x.com/minjunesh/status/1940589653410959784?s=19

Or physionic is good at reviewing papers critically
Matt kaeberlein also is

alternative phrasing: “If you were to train a classifier on noise/BS/sensationalist papers, what would you include in the training data?”

A_User · July 12, 2025, 9:33am

Avi Bitterman (generalize really well). He debunked the MAHA report here: MAHA Report (May 22, 2025) - #25 by A_User

AlexKChen · July 17, 2025, 9:06am

https://x.com/cremieuxrecueil/status/1945617415473082625?s=19

The King of Curcumin: a case study in the consequences of large-scale research fraud

Bharat B. Aggarwal is an Indian-American biochemist who worked at MD Anderson Cancer Center from 1989 to 2015. His research focused on potential anti-cancer effects and therapeutic applications of herbs and spices. Aggarwal was particularly drawn to curcumin, a non-toxic compound found in turmeric that has long been staple in Ayurvedic systems of medicine. … MD Anderson Cancer Center initially appeared to be fully on board with Aggarwal’s work. At one point, their website’s FAQ page recommended visitors buy curcumin wholesale from a company for which Aggarwal was a paid speaker (see “Spice Healer”, Scientific American). However, in 2012 (following observations of image manipulation raised by pseudonymous sleuth Juuichi Jigen), MD Anderson Cancer Center launched a research fraud probe against Aggarwal which eventually led to 30 of Aggarwal’s articles being retracted. Only some of these studies were about curcumin specifically, but most concerned similar natural products.

(1) Science of Science on X: “How much misconduct/fraud is there in the academic literature? About 0.2% of papers get retracted, but that’s obviously a severe underestimate. Probably the best estimate comes from a manual (!!!) inspection of 20K (!!!) Western blot images. Estimate is 3.8% (1/2) https://t.co/oLHKHEeteY” / X