Testing Testosterone Part 2: Apples vs Oranges

Here's a little factoid for you:  those testosterone tests that we run for men aren't necessarily all that accurate for women.  So states a study published this month in JCEM.  More specifically, there's no standardized assay method for testosterone in women.  We use the same technology as we do for men but it's scary to consider the (lack of) accuracy.

First, you have to understand (trust?) that women in most age groups have a normal range of 20-60ng/dL compared to men's 300-1000ng/dL.  Granted you might see slightly different values for either the upper limit of normal or the lower limit of normal but just get a feel for the ballpark difference.  There are certain clinical conditions in which women might have substantially higher testosterone levels than usual but still not into the male range.

Second, there are many different ways to measure testosterone, each with its own reference range.  While your lab might publish a reference range, there are in fact no age & gender normal ranges, nor is there an assay upon which to base this standard.  That's why I mentioned last month that the Endocrine Society and Centers for Disease Control & Prevention are developing a plan to come to a consensus.

But which testing method is most accurate?  The authors took blood samples from 596 women with a condition (polycystic ovarian syndrome or PCOS) that should raise their testosterone levels high enough to minimize as best we can assay difficulties.  These samples were each then divided individually into 3 aliquots.  Two went to two different academic facilities and one went to a commercial facility.  One methodology was used by both an academic facility and its commercial counterpart while a different method was used by the other academic facility.  Got it?

In an ideal world (which we clearly don't live in), all three values would be exactly the same.  For your purposes & mine, getting 90+% correlation would be satisfactory.  However, using the same testing methodology (LCMSMS), the closest the academic and commercial facilities could come to each other was just 83%.  Worse, when compared to a different method (RIA), the academic facility only achieved 79% correlation while the commercial facility was only 67% accurate.  And the lower the testosterone level, the worse the correlation.

Funny then that the study concluded that the LCMSMS & RIA methods are "comparable".  Like an apple to an orange, I think.  At least they admitted there's significant variability between assay methods.  Bottom line for you & me?  Treat the patient, not just the numbers.  And have your blood tests run at the same place using the same technology each time.  So you can compare apples to apples.