I haven't read the appendix in full detail, but how is it okay to run a linear regression here (instead of logistic regression) when the outcome is binary?
I do get a sense of p-value massaging after paging through the appendices.
The authors claim in the appendix that "linear models were estimated due to failure of the maximum likelihood algorithm to converge for logistic models."
Cancer must have been cured, as there's obviously nothing left to study at Harvard Medical School.
I feel your comment is politically tinged. This is far from a superfluous topic of study.
It's similar to the study where patient outcomes are more positive during medical conferences.