Reevaluating the Threshold for Low Total Testosterone

Arun AS, Durant TJS, El-Khoury JM, Krumholz HM · Clinical Chemistry 71(5):609–611 · 2025 · DOI: 10.1093/clinchem/hvaf025

PMID: 40238540 · Letter to the Editor (reanalysis of public NHANES data). Yale group (Center for Outcomes Research and Evaluation + Department of Laboratory Medicine).

TL;DR

The widely reported “epidemic” of low testosterone in American men is, on this reanalysis, largely a measurement artifact. Across the five National Health and Nutrition Examination Survey (NHANES) cycles used to establish the US trend, total testosterone (TT) was measured by Roche Elecsys immunoassay in the two earliest cycles (2001–02, 2003–04) and by LC-MS/MS — the modern, lower-reading gold standard — in the three later cycles (2011 onward). Median TT in self-reported-healthy men dropped ~16% precisely across that assay transition, and the prevalence of “low T” (<300 ng/dL) nearly doubled. Re-scoring the later cycles against the LC-MS/MS-appropriate Endocrine Society cutoff of 264 ng/dL returns prevalence to its earlier-era level — i.e. the apparent decline is an artifact of measuring more accurately against an outdated immunoassay-derived threshold.

Design

• Data: NHANES cycles that measured TT — 2001–02, 2003–04, 2011–12, 2013–14, 2015–16 — plus self-reported health status, age, and sex. “Healthy” = response of “very good” or “excellent” to the overall health-status question.
• Population: self-reported-healthy males >18 yr (NHANES complex-survey weighting applied).
• Analysis: survey-weighted quantiles of TT per cycle and the weighted fraction with TT <300 ng/dL (and recomputed at 264 ng/dL). R 4.1.2; code on Zenodo (10.5281/zenodo.14908077). IRB-exempt (public deidentified data).

Key findings (all verified against the PDF)

NHANES cycle	Assay	Median TT (ng/dL)	% with TT <300 ng/dL
2001–02	Roche Elecsys immunoassay	495 (17.2 nmol/L)	11%
2003–04	Roche Elecsys immunoassay	500 (17.3 nmol/L)	12%
2011–12	LC-MS/MS	418 (14.5 nmol/L)	22%
2013–14	LC-MS/MS	409 (14.2 nmol/L)	21%
2015–16	LC-MS/MS	433 (15.0 nmol/L)	18%

• Median TT dropped 16.4% between the 2003–04 and 2011–12 cycles — i.e. exactly across the immunoassay→LC-MS/MS migration — producing a near-100% increase in healthy men classified as “low T” over a decade.
• Applying the Endocrine Society’s 264 ng/dL cutoff to the LC-MS/MS cycles gives prevalences of 13% / 14% / 11% (2011–12 / 2013–14 / 2015–16) — in line with the immunoassay-era 300 ng/dL figures (11–12%). The “epidemic” disappears once the threshold is matched to the method.
• Guideline discordance underscores the threshold problem: AUA 2018 = 300 ng/dL; Endocrine Society 2018 = 264 ng/dL (lowered explicitly because of the assay changes); European Association of Andrology = 349 ng/dL.
• Clinical consequence: using 300 ng/dL with modern LC-MS/MS overestimates low-T prevalence by ≥50% relative to 264 ng/dL, risking over-diagnosis of hypogonadism, unnecessary treatment and its side effects, excess cost, and delayed workup.

Limitations / scope

• It is a Letter to the Editor, not a full original article — concise methods, no formal sensitivity analyses beyond the threshold recomputation.
• NHANES-specific. The argument cleanly addresses the US/NHANES prevalence narrative (and therefore directly bears on NHANES-based studies such as Lokeshwar et al. 2021, Eur Urol Focus, doi:10.1016/j.euf.2020.02.006, which itself flagged “differing assays” as a limitation). It does not speak to non-NHANES cohorts that used a single assay throughout (e.g. the Israeli Maccabi cohort, Danish/Finnish surveys, or the immunoassay-era Massachusetts Male Aging Study) — those secular-decline signals are not dissolved by this particular 2011 assay switch.
• Addresses the prevalence of men classified low and the diagnostic threshold, not whether any individual man’s testosterone falls with age (it does — see testosterone § age-related decline).
• Self-reported (not measured) health status defines the “healthy” subgroup.

Relevance to the wiki

This is the linchpin deflationary argument in the contested secular testosterone-decline debate (see testosterone § The secular (population-level) decline). It reframes a frequently-cited “men are becoming less manly” narrative as, in large part, a laboratory-harmonization problem — a clean worked example of why assay-method changes must be reconciled before a temporal biomarker trend is interpreted as biology.

Footnotes

No external claims beyond the primary source; all data above are from the paper itself (DOI 10.1093/clinchem/hvaf025), read in full.