Microplastics found in semen and placentas. What does it mean for fertility?

Of 281 men who qualified, 113 took part, a response rate of 40.2%. Semen and urine were both tested for plastic in 100 of them, and 100 of those 100 carried at least 1 kind in each. Most carried 3 to 5 kinds at once: 91 of 100. The plastic bits measured 1.2 to 20 micrometers across.^[1]

Per extra kind of plastic, adjusted for age, BMI, smoking, alcohol and site, only sperm concentration held: beta −0.18 on a log scale (95% CI −0.35 to −0.01), p = 0.036. Total sperm number: beta −0.19 (−0.38 to 0.006), p = 0.067. Forward (progressive) motility: beta −0.08 (−0.23 to 0.07).^[1]

Non-stick coating plastic (PTFE), unadjusted: 28 of 55 men carrying it had too few forward-swimming sperm, against 14 of 45 without it (51% against 31%), p = 0.046. Adjusted, the odds ratio for low forward swimming was 2.29 (0.96 to 5.49) and not significant. The authors say there is no direct evidence tying the PTFE to cookware.^[1]

The paper's abstract prints effect sizes per extra kind of plastic for total count, concentration and forward movement (β −15.4, −7.2 and −8.3), a pattern-analysis odds ratio (3.5) and PTFE concentration averages that appear nowhere in the Results, figures or 12 supplementary tables, and the body reports different numbers. We use the body and supplement. The abstract and text also reverse the PTFE forward-motility averages against Table S3. Our re-check of the PubMed abstract on 2026-10-05 confirms those abstract-only figures are still there.^[1]

Polystyrene turned up in 100 of 100 men tested, so its effect could not be tested. The study recorded which plastics were present, not how much; it had no fertile men for comparison; and fewer than half of eligible men took part. No source was measured. The page's "men carrying the most kinds of plastic" wording follows this: the study counted kinds, not amounts.^[1]

Human samples: 23 testicles from the New Mexico Office of the Medical Investigator, collected in 2016, from males aged 16 to 88 who "typically did not experience a natural death", in the authors' words. The samples had been stored for 7 years and were due to be thrown away. The paper describes them as "Twenty-three human testis samples" and plots each as "a unique human sample ID" with its own age; it does not say whether any two samples came from one person. Range 161.22 to 695.94 µg/g; mean 328.44 µg/g. Dogs: 47 testicles; range 2.36 to 485.77 µg/g; mean 122.63 µg/g. Human levels were higher, p ≤ .001.^[2]

Dog correlations, unadjusted, 36 tests with no correction for multiple testing: testicle weight per kg of body weight with PVC r = −0.30, PET r = −0.33 and SBR r = −0.31; sperm count with PC r = −0.30, ABS r = −0.35 and polyurethane r = +0.34. PVC against sperm count was r = −0.14 and not significant, so the page ties PVC to testicle weight, not to sperm count. No plastic was linked to epididymis weight. Polyethylene, the main plastic, showed no link with any measure. The text mentions the two negative sperm-count results and not the positive one. The authors write that "a decreased testis weight is indicative of reduced spermatogenesis", citing Creasy 2003.^[2]

Adjusted for the dog's age (the numbers are in Supplementary Table 1, which we could not obtain): PVC, PET, PC and PS went with lower dog testicle weight, and age, N66 and ABS with higher. PC and PS are not on the authors' later high-confidence list. For dog epididymis weight, PVC went with lower weight. For dog sperm count, age went with lower counts and PMMA with higher counts; no plastic went with lower counts.^[2]

The paper reports no contamination controls of its own; the authors point to their 2024 placenta paper. The comment by Uppu, Peijnenburg and Hays was funded by the American Chemistry Council; it said there was "no way of knowing if false positives exist" and named 2 plastics as hard to measure this way, both among the 3 tied to the dog result. In reply, the authors named 6 plastics they measure with enough confidence: polyethylene, polypropylene, PVC, nylon, styrene-butadiene rubber and PET. PC, ABS, polyurethane and PMMA are not among them. All 3 plastics tied to lighter dog testicles in the simple comparison are on that list. The same reply calls polyethylene, the top plastic in both species here, "challenging" to measure, because body fats can interfere and create false positives; the authors chose their digestion method to limit this.^{[2, 9, 10]}

6 of the paper's 10 authors also wrote the Nihart 2025 brain study, which uses the same method and tissue from the same medical examiner's office; that study faces the same fat-interference question, covered in our article on plastic, the brain and memory. The university's press release differs from the paper on the human average (329.44 against 328.44), on which plastic went with dog sperm counts and on the donors' ages; we use the paper.^{[2, 11]}

Shen 2026 (Ecotoxicology and Environmental Safety): 1,750 mother-infant pairs from a large cohort in China, births 2022 to 2024, described by the authors as a cross-sectional analysis. Plastic was counted per 10 g of placenta by laser direct infrared imaging, which sees particles above roughly 10 to 20 microns and misses smaller ones; medians PVC 7.0, PP 5.0, PBS 4.0, PET 3.0 particles per 10 g. Total plastic against birth weight: −107.7 g (95% CI −138.87 to −76.49), p < 0.001, in an adjusted model whose covariates the abstract does not list; the abstract also does not state the unit of that estimate (per particle, per quartile or per log unit), so the page prints no grams in the story. Mixture models: birth weight −117.7 g (−155.3 to −80.1) and −100.9 g (−146.2 to −55.5); length −0.89 cm (−1.20 to −0.58) and −0.96 cm (−1.34 to −0.58); head circumference lower; the BKMR model was monotonic for all three; the link was stronger in boys (p for interaction 0.04). Read at abstract level; the journal is open access and the full text is still to be read. PBS, a biodegradable plastic, ranking so high is unusual for human tissue and may be a library-matching issue. The same cohort appears to report other outcomes in separate 2025 and 2026 papers, so those are not independent replications of this one.^[7]

The full paper is paywalled, and no open copy exists. We read the published abstract, the highlights and the figure images; the figure captions were not available. Our re-checks of the PubMed abstract on 2026-10-05 and 2026-10-12 match the figures printed here.^[3]

The abstract says plastic was below the detection limit in the normal pregnancies except for 3 of 30, and the page prints that count as the abstract's. The paper's Figure 4 appears to show a different number, and two later summaries agree with the figure, not the abstract. Without the paywalled full paper, we cannot settle which is right.^[3]

Up to 64% of the bits were smaller than 10 microns; a human hair is about 50 to 100 microns wide. Most were polyethylene and polystyrene, the plastics in shopping bags and foam cups.^[3]

The abstract does not say whether the correlations cover all 43 babies or only the 13, whether results were adjusted for the mother's age, weight, smoking or weeks of pregnancy, or what blanks and detection limits were used. The authors describe the pattern as nonlinear, so for this study the page says "more plastic went with smaller newborns"; the lines "more plastic, smaller babies" and "babies with more plastic in the placenta were born smaller" are used only for Shen 2026, where the dose pattern ran in one direction. Because the groups were split by baby size, the link with size is partly built into the design.^[3]

The 2024 rapid systematic review by Chartres and colleagues rated birth size and pregnancy length "unclassifiable" because the evidence was of low and very low quality; Amereh 2022 is one of its 3 human studies. One Amereh author works for Veolia North America, a water company; the paper declares no competing interests.^{[3, 13]}

2020 averages at 70°C after sterilizing, 5 bottles per product: 1,314,000 to 16,224,000 particles per liter across the 8 bottle bodies, and 69,000 to 2,532,000 for add-on parts. 16.2 million is one product's 5-bottle average. The glass-beaker control averaged 170 per liter. The 10 products tested held about 69% of online sales across 48 countries and regions, and about 8 in 10 bottles sold there were made fully or partly of the plastic tested.^[4]

Temperature test, bottle 1, no sterilizing: 625,000 at 25°C; 3,139,000 at 40°C; 6,130,000 at 70°C; 54,712,000 at 95°C, per liter. Sterilizing raised the counts by 34.7%, 84.2% and 72.1% for 3 products; the manuscript gives no p-value.^[4]

Only 4 spots on each filter, 1.5136 of 227 square millimeters, were counted and scaled up. Recovery of the spiked test particles was 94.9%, and a blind count by Keyence matched the lab's. Neither check could catch additive specks read as plastic.^[4]

In 2022 the same team showed that slip additives such as oleamide and erucic acid match polypropylene closely in the signal band the 2020 paper used. An alcohol rinse cut bottle levels from over 1 million to around 100 to 100,000 per liter. The recount is one sentence; no bottle data tables were published with it. The Münster lab (Gerhard 2022) tested 7 bottles and found 1 above the detection limit, with particles more likely contamination than bottle plastic.^{[5, 6]}

The paper's per-baby daily estimates are a model built on online sales data, made before the recount, and we do not print them. No babies were tested. We read the authors' accepted manuscript and the publisher's open supplement; the paywalled published version was not read.^[4]

Jin 2022 (Nanjing University, Particle and Fibre Toxicology): 7 groups of 15 male mice; polystyrene particles in drinking water at 100 or 1,000 µg/L for 180 days; lower sperm viability, higher sperm abnormality and lower serum testosterone, LH and FSH than mice on clean water. It is the one controlled experiment on this page, so it is the one place the page uses causal verbs. It is cited through the explainer's fact sheet; the same lab's brain study used mice from the same batch, so the two Jin papers are not independent. The particles are one plastic at set doses; the human tissue here held mostly polyethylene.^{[2, 8]}

Zhang 2024: public research funds in Shanghai and China; the authors declare no competing interests. Hu 2024: US National Institutes of Health grants; no conflicts declared. Amereh 2022: funding not stated in the parts we could read; no competing interests declared. Li 2020: Enterprise Ireland, Science Foundation Ireland, a Trinity scholarship and the China Scholarship Council; no competing interests declared. Shen 2026: funding and competing interests are not in the abstract we read.^{[1, 2, 3, 4, 7]}

Read in full: Zhang 2024 and its supplement; Hu 2024, the comment and the reply; the Li 2020 accepted manuscript, supplement and source data; Li 2022. Read at abstract level: Amereh 2022 (paywalled), Gerhard 2022 (paywalled) and Shen 2026 (open access; full text not yet read). No paywall or bot check was bypassed. The key numbers were re-checked against the PubMed records on 2026-10-05.^{[1, 2, 3, 4]}

This version opens with the share of samples that held plastic in the semen and testicle studies, stated as a percentage; the exact counts (100 of 100, 23 of 23) stay on the page in the Short Summary, the body and these sections. The limits of the evidence are stated once, in the boundary block. The mouse experiment has its own section, so the body runs from what was found in people, to the link with sperm counts, to the controlled test in mice, then to placentas and the bottle as the everyday route. The placenta finding now leads with Shen 2026, a study of 1,750 placentas that is free to read (open access), read at abstract level. The 2022 Tehran study (Amereh 2022) follows it, briefly, labelled wherever it appears as read from the abstract only, because the full paper is paywalled; its counts (13 of 13 and, from the abstract, 3 of 30) stay in the Short Summary, the body and these sections. The per-baby exposure model, the three-week bottle test and the kettle result stay out of the story. The 2020 hot-water chart left the story too, because the same team's 2022 recount revised that method; its numbers are in the baby-bottle record above. Every figure that left the story is printed in these sections, from the papers' own text and tables.