Microplastics in Drinking Water: Studies, Filters & Risks (2026)
What 2024-2025 microplastics studies actually show – bottled vs tap, RO filtration, nanoplastics in brain tissue. Free ZIP-level water check + filter recommendations.
Reviewed by Dr. Marcus Reilly, PhD – PFAS & Drinking Water Scientist
Quick Answer
Microplastics are in tap water, bottled water, and the human body. The 2024 Columbia/Rutgers study (PNAS) found 240,000 plastic fragments per liter of bottled water – roughly 100x prior estimates. Reverse osmosis (NSF/ANSI 58) is the only home filter type proven to remove sub-micron and nano-scale particles. Carbon block at 0.5 micron catches larger microfibers but misses nanoplastics. Standard Brita pitchers are not designed for microplastics.
Key facts
- Plastic fragments per liter of bottled water
- 240,000 (Columbia/Rutgers 2024, PNAS)
- Nanoplastic concentrations in human brain tissue
- 7-30x higher than liver/kidney (Nature Medicine 2025)
- Microplastics in carotid plaque, stroke risk
- 4.5x higher (NEJM 2024 Marfella study)
- Particles released from heated baby bottles
- 16 million per liter (Trinity College 2020, Nature Food)
- Most effective home filter
- Reverse osmosis (NSF/ANSI 58), 0.0001 micron pore
Frequently asked questions
Are microplastics in my tap water?
Almost certainly some. A 2017 Orb Media study tested 159 tap water samples from 14 countries; 83% contained microplastic fibers. US samples averaged 9.6 fibers per 500 mL. Concentrations vary by source water type and pipe material.
What filters remove microplastics?
Reverse osmosis is the only home filter type proven to remove sub-micron and nano-scale plastic particles. NSF/ANSI 58 RO membranes have a pore size around 0.0001 micron – smaller than every plastic particle ever measured.
Are microplastics worse in bottled water than tap?
Yes – substantially. The Columbia/Rutgers 2024 study used Raman microscopy to find an average of 240,000 plastic fragments per liter of bottled water, roughly 100x prior estimates. Tap water concentrations from the same testing approach are 10-100x lower.