As I noted in a recent tweetorial, Raymond Pearl reported a lower frequency of cancer in those with evidence of tuberculosis.
This finding led Pearl and others to treat patients with tuberculin.
Unfortunately, Pearl's original study methods suffered from bias.
This finding led Pearl and others to treat patients with tuberculin.
Unfortunately, Pearl's original study methods suffered from bias.
https://twitter.com/tony_breu/status/1426587830373978113
More specifically, Pearl's study sample contained an overrepresentation of exposed controls (i.e., control subjects who had died from tuberculosis).
This led to an incorrect conclusion that tuberculosis is associated with decreased rates of cancer.
This led to an incorrect conclusion that tuberculosis is associated with decreased rates of cancer.
Pearl published a "retraction" in Science.
While arguing that "any serious student of the matter" would agree that TB and cancer are rarely found together in the same person, he admits that concluding a mechanistic connection "may have been erroneous".
pubmed.ncbi.nlm.nih.gov/17777405/
While arguing that "any serious student of the matter" would agree that TB and cancer are rarely found together in the same person, he admits that concluding a mechanistic connection "may have been erroneous".
pubmed.ncbi.nlm.nih.gov/17777405/
This story was highlighted well in a great review by Kim Sutton-Tyrrell entitled "Assessing Bias in Case-Control Studies".
@LFGarabedian and @rbganatra - you'll like this one
pubmed.ncbi.nlm.nih.gov/1853415/
@LFGarabedian and @rbganatra - you'll like this one
pubmed.ncbi.nlm.nih.gov/1853415/
Unfortunately, most papers that reference Pearl's original study do not mention its flaws.
Once something with profound implications is "out there" what follows - even a retraction - is often lost in its wake.
Once something with profound implications is "out there" what follows - even a retraction - is often lost in its wake.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
