A source-traced investigation. Compiled 2026-05-28. All cited URLs accessed the same day unless noted.
A modest secular decline in mean serum testosterone among Western males is plausible and is supported by multiple independent population studies. It is not, however, well-established at the magnitudes most commonly cited in popular and clinical media (e.g. "25% in 17 years"), and the strongest single piece of synthetic evidence — a 2025 systematic review of >1 million subjects — finds that the apparent decline is detectable in older immunoassay-based studies but not in the (smaller) sub-set of studies using mass-spectrometric measurement, the current gold standard.
Net classification (per scientific-fact-classification spectrum):
- "Some Western cohorts show a measured downward trend in mean total testosterone over recent decades" — established fact, modest magnitude (traced).
- "There is a real, biologically driven, generation-wide decline in male testosterone of ~25% per generation" — contested; substantial portion of the most-cited estimates is plausibly attributable to (a) assay drift from immunoassay to LC-MS/MS, (b) BMI confounding, (c) sample-selection bias in the source cohorts (traced).
- "Median male testosterone is in free-fall and reaching crisis levels" — not supported by the primary literature; this framing is mostly downstream of one paper (Lokeshwar 2021) whose data window straddles a major assay change at the U.S. Centers for Disease Control and Prevention (CDC) (traced).
Per the pre-registration-honesty discipline, the analytical hypothesis space is not "blank": prior published controversies framed it before any search was run. Provenance of each hypothesis is tagged.
-
H_A — Real decline. A genuine secular decline in serum testosterone across recent decades in Western males, robust to assay change, BMI, and age adjustment. Provenance: training-data familiarity with Travison 2007 and Lokeshwar 2021 framed this hypothesis. Not a free pre-registration.
-
H_B — Methodological artefact. Apparent decline is wholly or substantially produced by (i) assay drift from immunoassay to LC-MS/MS, (ii) BMI/comorbidity confounding, (iii) selection bias in source cohorts, (iv) shifting reference-range conventions linked to testosterone-replacement-therapy (TRT) prescribing interest. Provenance: training-data familiarity with Handelsman critiques framed this hypothesis. Not a free pre-registration.
-
H_C — Stratified / partial. Real in some cohorts/age bands/countries; absent or much smaller in others. The aggregate "Western male" generalisation is too coarse. Provenance: synthesised compromise; common in this literature.
Each row labels the warrant per the discipline: (traced) = fetched in this session via WebFetch/WebSearch with URL stated; (memory — unverified) = recalled, not verified this session; (user-supplied — unverified) = supplied by the requester. No (user-supplied) items in this analysis. All non-(traced) items are explicitly flagged.
-
Travison et al. 2007, Massachusetts Male Aging Study (MMAS). 1,532 American men aged 45–79, three measurement waves 1987–1989, 1995–1997, 2002–2004. Median total testosterone 501 → 435 → 391 ng/dL across waves; estimated age-matched time trend ~1.2 % per year. Decline persisted after adjustment for smoking and obesity. NIH-funded. (traced — PubMed 17062768; healio summary)
-
Andersson et al. 2007, Danish population surveys. 5,350 male samples across four surveys 1982–2001. Significant age-independent decline in total testosterone — but after adjustment for rising BMI the testosterone decline became non-significant; the SHBG decline persisted. (traced — PubMed 17895324)
-
Perheentupa et al. 2013, Finnish surveys 1972/1977/2002. 3,271 men aged 25–74. Significantly lower testosterone in same-age men born in more recent decades. Published European Journal of Endocrinology (not JCEM as sometimes cited). (traced — PubMed 23161753)
-
Mazur, Westerman & Mueller 2013, US Air Force Health Study (Ranch Hand cohort). 991 USAF veterans, six exam cycles 1982–2002, fasting morning bloods. Secular decline 8.9 ng/dL annually, more than double the aging effect (3.9 ng/dL annually). Weight-stable and weight-losing men still showed substantial declines (~117 ng/dL over 20 years among normal-weight men). Authors explicitly conclude that increasing obesity "cannot be a sufficient or primary explanation." No reported funding; no competing interests declared. (traced — PLOS ONE 10.1371/journal.pone.0076178)
-
Chodick, Epstein & Shalev 2020, Israeli health-care data. 102,334 male members of Maccabi Healthcare Services, 2006–2019. At age 21 (testosterone peak), 19.68 → 17.76 nmol/L across the period. Authors filtered to maintain a uniform assay/lab norms across the window. (traced — PMC7063751)
-
Lokeshwar, Patel, Fantus et al. 2021, NHANES 1999–2016, adolescent/young-adult men. 4,045 men. Reported mean total testosterone 605 ng/dL (1999–2000) → 451 ng/dL (2015–2016), a ~25 % drop. The decline reportedly remained significant in normal-BMI men. European Urology Focus 7(4):886–889. Critical caveat: the NHANES data window crosses two assay-platform transitions (see counter-evidence row 9). Senior co-author R. Ramasamy has disclosed industry ties including consultant/investigator/advisory-board roles with Aytu Biosciences (manufacturer of the testosterone product Natesto), Endo Pharmaceuticals, Boston Scientific and Coloplast — all with direct or adjacent financial interest in TRT-related diagnostics and prescribing. (traced abstract — PubMed 32081788; industry ties — Doximity / AUA disclosure, PMC6642974 corrigendum acknowledging Aytu/Natesto investigator role)
-
Santi, Spaggiari et al. 2025, systematic review and meta-regression. 1,256 papers, 1,504 study groups, 1,064,891 subjects, years 1970–2024. Significant negative linear regression of testosterone on year of measurement (coefficient −0.6, p < 0.001 after BMI/age adjustment). LH also declining; FSH and SHBG flat. Decline confirmed in studies using immunoassay, ELISA, or radioimmunoassay (RIA), but not in studies using mass spectrometry (8 MS study-groups pre-2000 — small N, but the cleanest assay sub-set). Authors declare no competing interests; institutional funding only (University of Modena and Reggio Emilia). (traced — PMC12602663; PubMed 40748419)
-
Fraile-Martínez, Ortega, García-Montero 2026, narrative review. Int J Mol Sci 27(2):692. Accepts the decline as real ("approximately 0.5–1.0 % annually") and enumerates proposed mechanisms (obesity, endocrine-disrupting chemicals (EDCs), inactivity, ultraprocessed diet, sleep, stress, medications, SARS-CoV-2, gut dysbiosis). Does not discuss the NHANES assay change. Does not cite the Santi 2025 mass-spec sub-finding. This is a narrative review and does not itself add new data. (traced — PMC12841019)
-
Vesper, Wang, Vidal et al. 2015, NHANES 2011–2012 reference paper. Clin Chem 61(12):1495–1504. Documents that NHANES 2011–2012 used CDC-standardised LC-MS/MS (isotope-dilution HPLC/MS/MS). Critically, the paper explicitly states:
"The unadjusted percentile values for totalT in men were lower than those reported in other population studies… [reflecting] differences in analytical methodologies, such as assay calibration. Previous NHANES cycles used an assay reported to be inaccurate, especially at low testosterone concentrations." This is the CDC laboratory itself attributing a substantial portion of the apparent post-2011 "decline" in NHANES testosterone to the platform switch, not to a biological trend. (traced — PMC5698798)
-
NHANES three-platform sequence (corroboration of 9). Search snippet from urology-trade press confirms NHANES used (a) Biotin-Streptavidin immunoassay 1999–2004, (b) isotope-dilution LC chromatography 2011–2012, (c) HPLC-tandem mass spectrometry from 2013 onward. The Lokeshwar 2021 cycle-by-cycle drops align with these platform changes more tightly than with any inflection in plausible biological exposure. (traced — search snippet from Urology Times coverage; primary paper itself paywalled at Eur Urol Focus)
-
Bhasin, Travison, Vesper et al. 2017, Harmonized Reference Ranges. J Clin Endocrinol Metab 102(4):1161–1173. 9,054 community-dwelling men across the Framingham Heart Study, European Male Aging Study (EMAS), Osteoporotic Fractures in Men Study (MrOS), and a Belgian sibling-study. Established a harmonised reference range for healthy non-obese 19–39-year-olds of 264–916 ng/dL (median 531). The exercise was explicitly motivated by assay non-comparability across cohorts — "variations in assays is an important contributor to variation in testosterone levels in cohorts from different geographic regions" (Bhasin). The paper does not discuss secular decline. (traced — J Clin Endocrinol Metab; press release — Endocrine Society)
-
Vesper & Botelho 2012, CDC Hormone Standardization (HoSt) overview. Documents that CDC's HoSt program drove a ~50 % reduction in mean absolute bias between mass-spectrometry assays and the CDC reference method 2007–2011, that immunoassay testosterone is unreliable especially below 100 ng/dL, and that the Endocrine Society convened a 2010 consensus to standardise testosterone measurement. The temporal coincidence of (a) the bias-reduction program and (b) the apparent population-level "decline" in immunoassay-dominated literature is itself a confounder. (traced — Clinical Laboratory News, June 2012)
-
Handelsman 2022, "The Illusory Case for Treatment of an Invented Disease." Front Endocrinol 12:682620. Argues against both age-related testosterone decline ("late-onset hypogonadism", "andropause", "Low T") as a meaningful disease entity AND against TRT-prescribing for non-pathological men. Author declares no commercial relationships. Handelsman is the senior endocrinologist most consistently associated with the artefact / disease-mongering critique. His position is not industry-independent of the question — he is an academic who has spent two decades opposing the TRT expansion narrative — but it is not commercially conflicted. (traced — PMC8803734; related 2015 commentary — Scientific American)
-
Levine, Jørgensen, Martino-Andrade et al. 2023, sperm-count meta-regression. Hum Reprod Update 29(2):157–176. 223 studies, 57,168 men, 1973–2018. Sperm concentration −51.6 %, total sperm count −62.3 % over the period; post-2000 decline rate (−2.64 %/yr) more than double the post-1972 baseline (−1.16 %/yr). Decline documented across all six inhabited continents. Funded by the Grantham Foundation for the Protection of the Environment — an environmental-cause funder; declared no conflicts. The findings are biologically plausible co-evidence for a declining-male-reproductive-function thesis but are themselves the subject of methodological dispute (selection of "unselected" cohorts, hemocytometer-only inclusion). A 2024 systematic review reportedly finds stable sperm concentration among fertile American men (Fert Steril — not fetched this session; (memory — unverified), included only as a flag for unresolved debate). (Levine et al. traced — Hum Reprod Update)
-
Han & Jin 2025, EDC review. Front Endocrinol 16:1573526. Reports that human populations with significant endocrine-disrupting-chemical exposure show ~28.5 % lower testosterone (95% CI −24.1 to −32.9 %); explicit mechanism via 17β-hydroxysteroid-dehydrogenase suppression, aromatase upregulation, androgen-receptor antagonism. Authors link contemporary EDC environmental load to the documented sperm-count decline. (traced — Frontiers in Endocrinology)
The single most-cited number in popular coverage — a ~25 % decline in adolescent/young-adult American testosterone between 1999 and 2016 (Lokeshwar 2021) — is drawn from NHANES data that spans two complete assay-platform transitions (immunoassay → isotope-dilution LC → HPLC-MS/MS). The CDC laboratory itself, in Vesper et al. 2015, states explicitly that the post-2011 LC-MS/MS values are lower than prior immunoassay values because the immunoassay was inaccurate at the relevant concentrations. The cycle-by-cycle Lokeshwar data points (605 → 567 ng/dL pre-switch, then drop to ~425 ng/dL coinciding with the LC-MS switch, then partial recovery to 451 ng/dL on HPLC-MS/MS) are consistent with a methodological discontinuity at least as well as with a biological trend. The Lokeshwar paper does not adjust for the platform change; the abstract acknowledges "differing assays for TT measurement" as a limitation. (traced)
A 1.06-million-subject meta-regression detected a significant negative time trend in testosterone — but the trend was detectable in immunoassay/ELISA/RIA studies and not in mass-spectrometry studies. The MS sub-set is small (8 study-groups pre-2000) and so the null could be a power problem rather than a true null. But the asymmetry is exactly what the artefact hypothesis predicts: the apparent decline is concentrated in the methodology known to be biased and to have been progressively standardised downward. The decline-thesis literature has not produced a published rebuttal addressing this sub-finding as of the search date. (traced)
The Air Force Health Study cohort showed substantial testosterone decline even in men whose weight was stable or decreased — i.e. independent of the BMI explanation that collapses the Danish Andersson 2007 finding. The cohort used a fixed protocol over 20 years. If the assay was consistent across the 1982–2002 window, the Mazur result is hard to explain away methodologically and is the best evidence that something biological (lifestyle, EDC exposure, sleep, or unmeasured) is contributing. (traced)
Two well-conducted analyses come to opposite conclusions on whether BMI accounts for the trend (Andersson 2007: yes, Mazur 2013: no). Santi 2025 reports the aggregate trend held after BMI adjustment. The honest reading: rising obesity contributes substantially but is not the whole story. (traced)
If male testosterone really were in steep decline, one would expect concurrent decline in spermatogenesis and reproductive output — and indeed Levine 2023 reports a ~50 % drop in sperm concentration with an inflection circa 2000. The EDC-mechanism literature provides a plausible biological pathway. But the sperm-count literature has its own methodological critics (Esteves and others), and at least one 2024 systematic review reports stable counts in fertile American men. The lines of evidence are mutually supportive but not independent: many of the same research groups and funders span both literatures. (traced for Levine and Han & Jin; flagged unresolved for the 2024 counter-review)
A reader of any single source should know who is funding it, prescribing under its conclusions, or campaigning against its conclusions.
| Cluster | Representative figures / studies | Stake in the question |
|---|---|---|
| Urology / men's-health-clinic axis (US) | Lokeshwar, Ramasamy and co-authors; many translational-andrology journals | TRT is a multi-billion-dollar market in the US. Ramasamy has disclosed paid investigator/advisory roles with Aytu Biosciences (Natesto), Endo Pharmaceuticals, Boston Scientific, Coloplast. The clinical-financial interest in "declining T = treatable disease" is direct. (traced for Ramasamy via AUA disclosure; corrigendum) |
| Endocrinologist standardisation axis (CDC / Endocrine Society / Bhasin / Travison / Vesper) | 2017 Harmonized Reference Ranges; CDC HoSt program | Their professional interest is in measurement reliability rather than in "decline" per se. The 2017 reference-range paper, jointly authored by Travison (decline-thesis pioneer) and Vesper (CDC standardisation), pointedly does not invoke secular decline. (traced) |
| Skakkebaek / Scandinavian reproductive-health network | Andersson 2007, Perheentupa 2013, Levine 2023 (Skakkebaek co-authored prior versions) | Two-decade research programme on the "testicular-dysgenesis syndrome" thesis — EDC-driven decline in male reproductive function. Network of overlapping authors. Programmatic prior in favour of the decline thesis. (traced for the papers; programmatic-prior inference is (memory — unverified)) |
| Environmental-cause funders | Grantham Foundation funded Levine 2023 | Environmentalist funding has a programmatic interest in EDC-driven decline. Does not invalidate the data, but the funding chain is not "neutral." (traced) |
| Independent academic skeptics | Handelsman (Sydney); Mazur (Syracuse); Esteves; Cocuzza | Career-long opposition to TRT-marketing expansion (Handelsman) or independent population analyses (Mazur). No commercial conflicts disclosed. (traced for Handelsman / Mazur; Esteves / Cocuzza (memory — unverified)) |
"Independent corroboration" of the decline thesis is weaker than it appears: many of the Western European studies share authors and conceptual framework with the Skakkebaek network; the US adolescent-T-decline result is generated by a sub-specialty (TRT-prescribing urology) whose senior authors are paid by TRT manufacturers. Removing any single cluster does not collapse the decline finding to zero — but the apparent consilience is partly an artefact of looking at three overlapping networks and counting them as three votes.
The verdict above ("modest decline plausible; large decline overstated") would shift toward stronger affirmation of H_A if:
- A new large-N study using exclusively CDC-standardised LC-MS/MS across a 15+ year window in a representative Western population showed a clearly significant downward trend.
- The Santi 2025 systematic review were updated with more mass-spec studies and the decline persisted in that sub-set.
- An independent re-analysis of NHANES restricted to the post-2011 LC-MS/MS cycles (2011–2018) showed a continuing decline within the consistent-assay window.
It would shift toward stronger affirmation of H_B if:
- An independent re-analysis of Lokeshwar 2021 with cycle×platform interaction terms showed the decline collapsed to non-significance.
- A direct head-to-head comparison of stored immunoassay-era samples re-measured by LC-MS/MS demonstrated that the historical "high" readings were assay-driven.
- The 2024 Fertility and Sterility "stable sperm count among fertile American men" review, on independent fetch, were corroborated by additional consistent-methodology data.
Would the same verdict be reached if the politically/socially expected answer were reversed?
The "Western male testosterone is collapsing" thesis sits at an unusual cross-political junction:
- It is amplified by right-coded "demasculinisation / civilizational decline" commentary and by commercial TRT-clinic marketing.
- It is also amplified by left-coded environmentalist / EDC-regulation advocacy.
- The opposing "it's mostly an artefact" position is supported by mainstream academic endocrinology and by professional skepticism of disease-mongering, neither of which has a strong partisan valence.
The reasoning above does not appear to be driven by political alignment; the load-bearing pivot — Santi 2025's mass-spec sub-finding combined with the NHANES assay-change documentation from the CDC itself — would be just as load-bearing if the politics around it pointed the other way. A residual concern: the analysis leans heavily on Handelsman's framing of "invented disease," and Handelsman is a polemicist on this question. The arbitration is placed in the empirical sub-finding (Santi 2025 MS-vs-immunoassay) rather than in Handelsman's authority, which is the right way to handle this.
A second residual concern: a generally available cognitive bias is "boring null reading," in which one preferentially credits the artefact explanation because it is intellectually safer. This is a real risk here. The Mazur 2013 Air Force finding is the strongest single piece of evidence against the artefact-only reading, and it is given full weight in Pivot 3 above. The verdict is not "no decline"; it is "decline plausible at smaller magnitude than commonly cited."
All warranted-(traced) items above were fetched in this session on 2026-05-28 via WebFetch or WebSearch, with URLs stated inline. No load-bearing claim rests on (memory — unverified) or (user-supplied) evidence. Items so labelled appear only as flags for unresolved sub-questions (e.g. the 2024 sperm-count counter-review; Esteves' critiques of Levine), not as inputs to the verdict.
Two source-fetch attempts were blocked by paywall (Springer for Santi 2025 full text; ScienceDirect for Lokeshwar 2021 full text). The Santi 2025 PMC mirror was reachable and supplied the key sub-findings; the Lokeshwar 2021 abstract plus the corroborating CDC paper (Vesper 2015) and a journalistic summary supplied the NHANES-assay-window facts. Both gaps are flagged honestly.
A research-tier verdict (per Rule 7 thresholds) would warrant a dedicated deep-research session — ≥40 sources, including re-analysis of NHANES with assay-platform covariate, and a systematic comparison of within-assay-platform secular trends across all available cohort studies. The present investigation fetched ~17 substantive primary or near-primary sources and used search-result extraction for ~5 more; it is sufficient for a calibrated verdict on the contested headline claim but not a definitive resolution of the underlying biology.