Guides / ICP-MS vs XRF for heavy metals
Why an XRF heavy metal screen is not a compliance result
Is XRF good enough for heavy metals testing?
Energy-dispersive X-ray fluorescence has detection limits in the parts per million range for arsenic, cadmium, lead and mercury. ICP-MS reaches parts per billion and below, three to six orders of magnitude lower. Because ICH Q3D permitted daily exposures translate into concentrations at parts per billion for most products, an XRF result of not detected is entirely consistent with a sample that exceeds the limit several times over.
X-ray fluorescence irradiates the sample and reads the characteristic X-rays each element emits as its inner-shell vacancies are filled. It is non-destructive, needs essentially no sample preparation, and returns an answer in minutes. Those are real advantages and they make it a genuinely useful sorting and screening tool.
ICP-MS destroys the sample first. Microwave digestion in concentrated nitric acid decomposes the matrix completely, the digest is nebulised into an argon plasma at six to ten thousand kelvin which atomises and ionises essentially everything, and the ions are separated by mass to charge ratio and counted individually. Counting individual ions is why the sensitivity is what it is.
| ED-XRF | ICP-MS | |
|---|---|---|
| Typical detection limit | Parts per million | Parts per trillion to billion |
| Sample preparation | Minimal or none | Microwave acid digestion |
| Sample destroyed | No | Yes |
| Time per sample | Minutes | Hours including digestion |
| Compendial status | Alternative, requires validation | Named in USP <233> |
| Suited to | Screening, sorting, gross contamination | Compliance against Q3D limits |
ICH Q3D(R2) sets permitted daily exposures rather than concentrations. Lead is 5 micrograms per day by both oral and parenteral routes. Cadmium is 5 micrograms per day orally and 2 parenterally. Arsenic is 15 by both. Mercury is 30 orally and 3 parenterally.
Converting to a concentration requires the daily intake. A supplement taken at 1 gram per day and permitted 5 micrograms of lead per day corresponds to 5 micrograms per gram, which is 5 parts per million and comfortably within reach of XRF. The same limit on a product taken at 100 milligrams per day corresponds to 50 parts per million, easier still.
The problem appears at the other end. A peptide dosed at a few milligrams gives a permitted concentration in the parts per thousand range for that element, which sounds generous, but the meaningful question for an injectable is usually not the permitted daily exposure but whether the material carries contamination at all, and that question lives well below XRF's floor. Speciation, trace-level supply-chain monitoring and any parenteral assessment all sit in territory XRF cannot reach.
USP <233> names ICP-MS and ICP-OES as the compendial procedures for elemental impurities. Alternative procedures are permitted, but only where validated to demonstrate performance equivalent to the compendial method for the elements and levels in question.
That validation is the crux. An alternative method must be shown to be capable at the relevant level. XRF cannot demonstrate equivalence at parts per billion because it does not detect there, so it cannot be validated into a compliance role for those levels no matter how it is documented.
Used as intended, XRF has a proper place: incoming raw material sorting, identifying grossly contaminated lots, and triaging which samples deserve full digestion and ICP-MS. The failure mode is not the technique, it is presenting a screening result in the position where a compliance result belongs.
Can XRF be used for USP <232> and <233> compliance?
Not at typical limits. USP <233> names ICP-MS and ICP-OES; alternatives must be validated as equivalent, and XRF cannot demonstrate equivalence at parts-per-billion levels because it does not detect there.
What does not detected mean on an XRF heavy metals result?
That the element was below roughly parts-per-million. It is compatible with a sample that exceeds a parts-per-billion limit by a wide margin.
Is XRF useless then?
No. It is fast, non-destructive and genuinely useful for raw material sorting and for finding gross contamination. The problem is presenting a screening result as a compliance result.
Does ICP-MS distinguish inorganic from organic arsenic?
Not by default; standard ICP-MS reports total arsenic. Speciation requires a chromatographic separation coupled to the instrument and is a separate determination.
References
- ICH Q3D(R2), Guideline for Elemental Impurities
- USP <232>, Elemental Impurities Limits
- USP <233>, Elemental Impurities Procedures
Sources
Updated 2026-08-24