The researchers tested two saliva collection methods, chewing on a cotton swab (Salivette, Sarstedt) and passive drooling on 30 healthy males with a median age of 27 years (range 19–65 years) and compared their findings against serum levels, which were measured on the same occasion. Here are their results:
For cortisol, we observed a mean positive bias of 65% for passive drooling versus Salivette. In addition, cortisol determined on passive drooling samples correlated much worse with calculated free cortisol than cortisol analyzed on the cotton samples (respectively r=0.34 versus r=0.70). However, for testosterone passive drooling correlated well with the calculated free fraction (r=0.66). A positive bias of 21% for testosterone collected by passive drooling versus Salivette was observed. In contrast to earlier reports, we found no artefactually high testosterone results with the cotton-based collection method.
Obviously, these results seem to indicate that saliva testing for testosterone is a good, while saliva testing for cortisol is a bad idea, BUT these results are in fact intrinsically flawed. By just calculating free serum values instead of using equilibrium dialysis to actually measure them, the scientists' control values and with them their assessment of the "appropriateness" and "inappropriateness" of salivary testosterone, respectively cortisol testing are questionable.
From what I hear from practitioners like Dr Crisler, it is highly advisable to spend the extra cash to have measured (not calculated) serum testosterone levels as reliable baseline and 4x saliva cortisol profiles to see where your levels are at throughout a 24h period. Arguably, this is yet another case in which a questionable study design leads to results which refute common practice and aggravate a confusion around hormonal testing procedures which affects practitioners and patients, alike.