The sperm count decline is real in one dataset and unproven in another

The claim that sperm counts have halved in fifty years comes from one research group's meta-regression, and it is contested in print by other researchers who think the method cannot carry the conclusion. Nobody has shown the decline is not happening; nobody has shown it is. The honest answer is that the largest analysis says yes, two substantial published critiques say the data are not good enough to know, and none of the three disputes the underlying numbers — only what can be inferred from them.

What the big analysis found

The 2022 meta-regression by Levine and colleagues in Human Reproduction Update combined 223 studies yielding 288 estimates, from semen samples collected between 1973 and 2018 [s1]. It updated the same group's 2017 analysis by searching PubMed/MEDLINE and EMBASE for studies published from 2014 to 2019, screening 2,936 abstracts and 868 full articles, and adding 44 estimates from 38 studies [s1].

In the simplest model — a straight line through sample collection year — sperm concentration fell by 0.87 million per millilitre per year (95% CI −0.89 to −0.86) [s1]. In the adjusted meta-regression, the declines varied by group: −1.27 per year among unselected men from North America, Europe and Australia (95% CI −1.78 to −0.77), −0.65 among unselected men from South and Central America, Asia and Africa (95% CI −1.29 to −0.01), and −0.50 among men already known to be fertile in the first group of continents (95% CI −1.00 to −0.01) [s1].

Two headline figures came out of that. Among unselected men from all continents, mean sperm concentration fell 51.6% between 1973 and 2018 [s1]. Total sperm count fell 62.3% over the same period [s1]. And the authors report the rate is getting worse rather than levelling off: restricted to data collected after 2000, the slope steepened to −1.73 per year (95% CI −3.23 to −0.24), and the percentage decline per year roughly doubled, from 1.16% after 1972 to 2.64% after 2000 [s1].

The first critique: what counts as a normal range

A 2021 paper in Human Fertility by Boulicault and colleagues gave the underlying assumption a name — the Sperm Count Decline hypothesis — and argued it is a set of implicit choices about how to measure and interpret sperm counts rather than a finding [s2]. Their alternative, the Sperm Count Biovariability hypothesis, holds that sperm count varies across a wide range, much of which is non-pathological and typical of the species, and that interpreting a shift in a population average requires knowing how individual counts relate to life history and environment [s2].

The paper describes the 2017 analysis as the largest and most methodologically rigorous of its kind to that point, and still identifies what it calls weaknesses and inconsistencies in the framework around it [s2]. The disagreement is not that the arithmetic is wrong. It is that a falling mean in a highly variable biological measure does not automatically mean a population is getting less fertile, and the leap from one to the other is where the public conversation happens.

The second critique: what the source studies can support

The more technical objection came in a 2022 review in Nature Reviews Urology by Auger and colleagues, which examined four decades of studies of semen quality by design type [s3]. Its conclusion about the multicentre retrospective literature — the kind that feeds meta-analyses — is blunt: conclusions drawn from those studies are "questionable, owing to intrinsic limitations in the studies performed" [s3].

The review's positive findings matter as much as its negative ones. Standardised studies comparing specific cities within a country give clear evidence of geographical differences in sperm production, sometimes over short distances inside the same country [s3]. Many single-centre retrospective studies with homogeneous groups of men and appropriate methods do suggest a temporal decrease in the area studied [s3]. And measures of motility and morphology remain limited by how variable the assessment itself is, regardless of study design [s3]. The review's summary is that the available data do not support concluding semen quality is deteriorating worldwide or across the Western world, but that a trend is seen in some specific areas [s3].

Why the disagreement is hard to settle

Semen analysis is not a standardised laboratory measurement in the way a cholesterol assay is. The studies pooled across five decades used different counting methods, different populations — sperm-donor candidates, men attending fertility clinics, military conscripts, partners of pregnant women — and different abstinence periods before collection. The 2022 meta-regression handles this statistically, adjusting for predetermined covariates, testing for effect modification by fertility status and continent group, and running multiple sensitivity analyses that it reports changed the results only minimally [s1]. The critics' position is that no amount of statistical adjustment fixes source studies that were never designed to be compared to each other across decades [s3].

There is also a straightforward power problem in the newer result. The 2022 paper's own novelty is the first reported decline among unselected men outside North America, Europe and Australia — and that estimate's confidence interval runs from −1.29 to −0.01, with a p-value of 0.045 [s1]. An interval whose upper bound sits at effectively zero is a finding that would flip with a modest amount of additional data.

What this leaves a reader with

Three things can be true simultaneously, and appear to be. Sperm concentration has fallen in the studied populations over the studied period, by a large amount, in the largest pooled analysis available [s1]. That analysis rests on a body of source studies whose comparability across decades is genuinely disputed by other researchers publishing in the field [s3]. And the relationship between a population's average sperm concentration and any individual man's chance of conceiving is not established well enough for a falling average to be read as a falling personal probability [s2].

What would resolve it is the thing both critiques ask for: prospective studies using standardised methods in defined populations, followed forward, rather than retrospective pooling of whatever laboratories happened to measure [s3]. Until those exist, a reader encountering the halving figure should know that it is a real published estimate, that it is not a consensus, and that the scientists who dispute it are not fringe.

Sources

• Temporal trends in sperm count: a systematic review and meta-regression analysis of samples collected globally in the 20th and 21st centuries — Human Reproduction Update, 2022-10-19
• The future of sperm: a biovariability framework for understanding global sperm count trends — Human Fertility, 2021-05-10
• Spatiotemporal trends in human semen quality — Nature Reviews Urology, 2022-08-17