They should have consulted with a statistician and this paper should have been strongly flagged during peer review.
See Table 1.
[ATTACH=full]20813[/ATTACH]
[ATTACH=full]20814[/ATTACH]
Once you see the result in Table 2 you stop there. Going any further with linear regression (see Tables 3, 4, and 5) and a mix of one important categorical and other continuous variables is a big no no unless you look for potential interactions and covariance between age and TT.
What do the distributions for age, TT and Hct look like between the two groups? Normal, bimodal?
See Table 4. The authors conclude TRT modality is significant. Where's the potential interaction term between modality and age?
[ATTACH=full]20815[/ATTACH]
Would that be an important effect to test for given the way the study was set up?
Would the findings in this paper be materially important to how you should design the study?
[ATTACH=full]20826[/ATTACH]
Now go back and look at Table 1 again. How do the p-values for Hepcidin suppression vs serum TT level above compare with TT level in Table 1 of paper (note you'll have to transform the data in Table 1):
[ATTACH=full]20828[/ATTACH]
Kind of important to make sure you are isolating the effect you are trying to test. Is hepcidin suppression a lurking variable that wasn't properly controlled for in this study?
