EthnicMuse's race and testosterone "meta-analysis"

A commenter asks:

What do you think of this website and his analysis of testosterone levels in different racial groups?

"EthnicMuse" is attempting to aggregate numbers that can't be aggregated, and the results lack face validity. T levels as measured by different techniques and/or at different laboratories are not in general directly intercomparable.
Clinicians are being presented with normal male reference ranges for serum T from these automated platforms that have low end clinical limits down to 170–200 ng/dl (5.9–6.9 nmol/liter) and upper range limits of 700–800 ng/dl (24.3– 27.7 nmol/liter). These stated reference ranges provided by the manufacturer are significantly lower than the 300-1000 ng/dl (10.4–34.7 nmol/liter) reference range referred to in numerous publications over the past 30 yr based on tradi- tional RIA methods with or without the chromatography step as well as some research techniques employed by in- ternal recovery standards to correct for procedural losses (5).

External quality control programs such as that provided by the College of American Pathologists allow laboratories to compare results with other laboratories using the same method or kit reagents. As shown in Table 1, the median value of a quality control sample (Y-04,2002) varied between 215 and 348 ng/dl (7.5 and 12.0 nmol/liter) among methods with coefficients of variation among laboratories using the same method or instrument ranging between 5.1% and 22.7%. The median average for this sample from all methods was 297 ng/dl (10.3 nmol/liter) and results were as low as 160 or as high as 508 ng/dl (5.5 to 17.6 nmol/liter). These results span the hypogonadal to eugonadal range.

[Measurement of Total Serum Testosterone in Adult Men: Comparison of Current Laboratory Methods Versus Liquid Chromatography-Tandem Mass Spectrometry]

Differences that are to be expected between different assays and different laboratories, apart from any other factors, would likely swamp any anticipated racial differences in circulating testosterone levels. Between-study differences in collection times, sample handling, age and health condition of subjects, and so on, add further noise.

I see that EM is at least vaguely aware of these issues, but he rationalizes publishing his "meta-analysis" as follows:

One cannot and should not compare different testosterone studies with different measurement methods. However, for the race-realist purpose of aggregating data, there is nothing inherently wrong with what the PDF file lists. If JP Rushton can use a few studies and make wild claims which are then used by the Internet-o-sphere, using 150 independent peer-reviewed sources with large samples is much more scientific than anything similar from the race realist community. [. . .]

Age differences will affect the results but healthy males should have negligible decreases. Assuming a 0.4% annual decline from 5000 pg/ml after age 40, a man at 80 should have 4275 pg/mL, less than a 15% difference if my spreadsheet math is correct. It would have been better to normalize for age. So while the tabled rankings is flawed, the point is that the entire issue is flawed as there is no standard measuring method in the first place. That race realists routinely use flawed data should be the issue but …

That blindly aggregating data from disparate studies (which in this realm I've never seen anyone other than EM attempt) is nonsensical does not mean all attempts at comparing circulating testosterone levels between races are "flawed". It means that if one wants to attempt such comparisons, one should focus on studies in which a single set of researchers, using standardized methods, publish results for multiple ethnic groups.

EM is aware, for example, of a study (pdf) in which blood samples from Swedes and Koreans "were analyzed in the same laboratory using the same assay". The result (in EM's words): "the Swedes had 25% more T than the Koreans in this study". I've seen other studies showing lower or similar levels of testosterone in East Asians compared to whites (and none showing anything like the 10% higher testosterone in East Asians asserted by EM). But EM apparently did not like where the data pointed (thus his version of "meta-analysis", in which valid data is swamped with garbage).