Phthalates: what's established and what's extrapolated

Phthalates are the chemicals that make plastic soft and flexible, and they turn up in vinyl flooring, food packaging, medical tubing and fragranced products. Two things about them are well established: nearly everyone is exposed, and the exposure is measurable in urine. A third set of claims — that phthalates cause hundreds of thousands of deaths a year — is not measured but modelled, and the distinction between the two is the whole story.

The established part

Phthalate metabolites are detectable in the urine of most people tested, and the body clears them quickly, which is why exposure is continuous rather than accumulated. Because they clear fast, a change in what a person eats or handles shows up in urine within days. A 2026 randomised trial demonstrated exactly that: a seven-day low-plastic diet in 60 people cut urinary levels of two phthalate metabolites — mono-n-butyl phthalate by 37.5% and monobenzyl phthalate by 53.5% — showing that exposure is real, ongoing and responsive to behaviour [s3].

Also reasonably established is that some phthalates behave as endocrine disruptors in laboratory and animal studies, interfering with hormone signalling, particularly in reproductive development. That biological activity is why regulators restrict certain phthalates in children's toys and why the research attention exists. None of this is in serious dispute.

The proposed biological route in the cardiovascular work is specific: DEHP exposure has been linked to increased oxidative stress and metabolic dysfunction, which are plausible contributors to cardiovascular disease [s1]. That is a coherent mechanism, and it is part of why the association is taken seriously rather than dismissed. It is still a mechanism inferred largely from experimental systems, not one demonstrated to operate at population scale in the people the burden estimates count.

The extrapolated part

The numbers that make headlines come from disease-burden models, and they require care. A 2025 study estimated that in 2018, exposure to di-2-ethylhexyl phthalate — DEHP, a common plasticiser — was attributable to 356,238 cardiovascular deaths worldwide, about 13.497% of cardiovascular deaths among people aged 55 to 64, of which 349,113 were linked specifically to plastics [s1]. It put the associated loss at 10.473 million years of life [s1]. A 2026 companion study estimated that DEHP exposure was attributable to 1.97 million preterm births in 2018, 8.74% of the global total, along with 74,000 deaths [s2].

The preterm-birth estimate extended to a second plasticiser, diisononyl phthalate, or DiNP, with a similar attributable burden — about 1.88 million preterm births and 64,000 deaths, a population attributable fraction of 8.32% [s2]. Both analyses found the burden concentrated in the same regions: the Middle East and South Asia accounted for over 54% of attributable preterm births, followed by Africa at 26%, tracking where plastic production and underlying disease rates are highest [s2].

These are large, specific figures, and they are estimates, not counts. Both studies build a population-attributable-fraction model: take an exposure level from biomonitoring surveys, apply a hazard ratio from prior cohort studies, and multiply against background disease rates from a global burden dataset [s1][s2]. Each of those inputs carries uncertainty, and the studies say so. The preterm-birth analysis, testing effect estimates from four different meta-analyses, found the attributable burden could be several times lower than its central figure at the low end of the uncertainty range [s2]. Where biomonitoring surveys did not exist, exposure was extrapolated from other regions [s2].

Why the gap matters

An attributable estimate answers a hypothetical: if the assumed causal relationship is real and this strong, how much disease would follow from the measured exposure? It does not, on its own, prove the causal relationship — it inherits that assumption from the cohort studies feeding it. If the true hazard ratio is smaller, or partly explained by other factors that track with phthalate exposure, the attributable numbers shrink accordingly.

That is not a reason to dismiss them. The cardiovascular study was explicitly built to inform negotiations over a Global Plastics Treaty, and attributable-burden modelling is a standard tool for weighing environmental exposures at population scale [s1]. It is a reason to read the numbers as what they are: policy-facing projections with stated uncertainty, not tallies of identified victims.

What a reader can take from it

The defensible summary has two halves. Exposure to phthalates is near-universal, biologically active in the lab, and reducible — the trial evidence for that is direct. And the attributable death and preterm-birth figures are model outputs resting on assumed causal strengths, informative for policy but softer than their precision suggests [s1][s2].

The measurement side of phthalates is unusually solid for an environmental exposure. The health-toll side is unusually confident-sounding for what is, underneath, a chain of estimates. Keeping those two apart is how to read the field without either overstating a modelled projection or waving away a genuine and pervasive exposure.

This article is informational and is not medical advice.

Sources

• Phthalate exposure from plastics and cardiovascular disease: global estimates of attributable mortality and years life lost — EBioMedicine, 2025-04-29
• Preterm birth attributable to exposure to chemicals used in plastic materials: a global estimate — eClinicalMedicine, 2026-03-31
• Low-plastic diet and urinary levels of plastic-associated phthalates and bisphenols: the randomized controlled PERTH Trial — Nature Medicine, 2026-04-21