Uncertainty Propagation in High-Resolution Mass Spectrometry Non-Target Screening for Food Contact Conformity Verification

Spike

Simulant Contact Protocols and Extraction Recoveries

Analytical measurements for food contact materials begin long before eluate reaches a chromatographic column. Non-target screening by liquid chromatography or gas chromatography coupled to high-resolution mass spectrometry evaluates complex chemical mixtures migrating from polymers, adhesives, coatings, and printing inks. The raw measurement signal reflects both analyte concentration in the simulant and the compound’s thermodynamic partitioning and chemical stability during exposure.

Food simulants prescribed under European Union Regulation 10/2011, such as ten percent ethanol, three percent acetic acid, or vegetable oil, extract migrating species under defined time and temperature regimes. A ten-day contact period at sixty degrees Celsius alters labile migrants through hydrolysis, oxidation, or thermal rearrangement. When a polyolefin film containing ester-based antioxidant breakdown products undergoes migration testing in three percent acetic acid, acid-catalyzed hydrolysis degrades parent molecules into smaller fragment species, shifting the apparent molecular weight distribution observed during mass analysis.

Extraction efficiency across diverse chemical structures is the primary source of analytical uncertainty in non-target workflows. Unlike targeted quantitation, where isotopically labeled internal standards correct for recovery losses of known compounds, non-target screening covers hundreds of unknown non-intentionally added substances with uncharacterized polarities, partition coefficients, and pKa values. Solutes with high octanol-water partition coefficients adsorb onto glass collection vials, PTFE tubing, or syringe filter membranes during transfer steps.

Solid-phase extraction protocols used to concentrate migrants from aqueous simulants like ten percent ethanol show variable retention, ranging from under ten percent for highly hydrophilic organic acids to over ninety percent for lipophilic oligomers.

Solid-phase extraction on polymeric sorbents returns recoveries below twenty percent for polar primary aromatic amines extracted from three percent acetic acid at forty degrees Celsius.

Sources of Systematic Extraction Variance

Quantifying recovery bias across unknown migrants requires accounting for physical losses during sample pre-treatment. Liquid-liquid extractions using dichloromethane or hexane selectively enrich non-polar compounds while leaving ionic species in the aqueous phase. Evaporation of extraction solvents under nitrogen streams drives off volatile migrant species, including monomer residues such as vinyl chloride, butadiene derivatives, or short-chain aldehyde decomposition products.

The accumulated error from extraction recovery introduces an uncertainty factor of two to ten in final concentration estimates prior to mass spectrometric detection.

• Adsorptive Surface Depletion losses occur when non-polar migrants bind irreversibly to active silanol sites on glassware during solvent concentration steps.
• Thermal Degradation Artefacts generate synthetic signals when heat-sensitive additives decompose inside hot injector ports or during high-temperature simulant exposures.
• Volatilization Losses deplete low-boiling migrants during solvent evaporation, distorting the relative peak area ratios of light compounds against heavier homologues.
• Differential Sorption Efficiency leaves hydrophilic degradation products behind in aqueous food simulants during liquid-liquid extractions using non-polar organic solvents.

Standard additions using isotopically labeled analogues are impossible when peak identities are unknown during data acquisition. Surrogate standards spiked into food simulants before exposure provide a benchmark for recovery corrections, yet a surrogate only reflects compounds sharing its precise physical property space. Spiking deuterium-labeled bisphenol A measures recovery for structural analogues, but fails to capture the recovery dynamics of cyclic polyester oligomers or slip agent degradation products present in the same migration solution.

Drift

Mass Accuracy and Resolving Power Stability

High-resolution mass spectrometry relies on precise mass-to-charge ratios to assign chemical formulas to observed molecular ions. Architectures like Quadrupole Time-of-Flight and Orbitrap mass analyzers achieve resolving powers exceeding thirty thousand and one hundred thousand, respectively. Mass accuracy degrades over long analytical runs due to ambient laboratory temperature fluctuations, power supply instability, and space-charge accumulation within ion traps or flight tubes.

A calibration drift of just two parts per million expands the set of plausible elemental compositions for an unknown ion from three candidates to dozens, directly broadening identification ambiguity.

Continuous internal mass calibration using lock masses mitigates systemic drift across analytical batches. Lock mass compounds infused continuously or injected alongside chromatographic peaks correct mass scale shifts in real time. However, lock mass signals experience ion suppression when co-eluting high-concentration migrants dominate ionization.

A heavy slip additive like erucamide eluting at high concentration saturates detector channels, depressing the lock mass signal and causing localized mass assignment errors above five parts per million right during the elution window of critical unknown migrants.

Where Does Mass Drift Impact Formula Assignment Accuracy?

Determining whether an observed peak comes from a regulated packaging component or an uncharacterized contaminant hinges on isotopic pattern matching and narrow mass tolerances. At a mass tolerance of one part per million, a monoisotopic mass of 285.1122 yields a single formula, C17H16O4. Expanding the tolerance window to five parts per million because of thermal drift introduces alternate formulas with nitrogen and sulfur, such as C14H17N2O4S.

This shift alters the predicted toxicity profile and regulatory classification, turning a benign ester into a potential structural alert requiring targeted toxicological evaluation.

Chromatographic retention time drift compounds mass accuracy errors during feature alignment across multi-sample studies. Gradient elution high-performance liquid chromatography experiences slight retention shifts caused by mobile phase evaporation, column temperature variation, or matrix-induced stationary phase modification. When aligning features across ten consecutive migration extracts, a retention time drift exceeding three seconds risks misaligning identical chemical entities or merging distinct isomers into a single composite feature, skewing peak abundance calculations.

Signal intensity drift across analytical runs further complicates semi-quantitative screening. Ion source contamination from non-volatile matrix components eluting from food simulants reduces ionization efficiency over time, causing identical migrant concentrations to yield lower peak areas at the end of an analytical sequence than at the start. Regular injection of quality control pools throughout sequence runs allows software algorithms to model and correct intensity decay, provided the decay remains linear.

A calibrated instrument operated in a controlled environment keeps mass error within predictable boundaries over routine sequence durations.

Deconvolution

Algorithm Architecture and Feature Extraction Errors

Raw files generated by mass spectrometers contain millions of data points spanning retention time, mass-to-charge ratio, and ion intensity. Deconvolution software converts these continuous signals into chemical features representing distinct molecular species. Algorithms must distinguish real analyte signals from background noise, electronic spikes, solvent contaminants, and mobile phase adducts.

Setting peak detection thresholds too high omits low-concentration migrants near the regulatory screening threshold of ten parts per billion, while setting them too low introduces thousands of false positives from baseline noise.

Adduct formation, in-source fragmentation, and isotopic distributions convolute feature lists by generating multiple peaks for a single chemical entity. Electrospray ionization yields protonated molecules alongside sodium, potassium, and ammonium adducts. In recycled polyolefin extracts, a single synthetic antioxidant like Irganox 1010 produces protonated ions, sodiated ions, dimer species, and multiple in-source fragment ions.

If deconvolution software fails to group these related signals into one compound feature, it reports five distinct chemical species ~ undercounting the integrated mass of the parent antioxidant while generating phantom non-intentionally added substances in conformity dossiers.

Standardized evaluation under ISO/IEC 17025 requires documented signal-to-noise ratios exceeding ten to one prior to reporting unassigned chromatographic features in compliance dossiers.

Isotopic Pattern Matching and Formula Exclusion

Isotopic distributions provide a secondary filter for validating formula assignments derived from high-resolution mass measurements. Chlorine, bromine, sulfur, and silicon feature distinctive polyisotopic ratios that constrain chemical composition choices. Carbon-13 natural abundance patterns confirm the number of carbon atoms in an unknown molecule.

High intensity uncertainty in small chromatographic peaks distorts observed isotopic ratios, causing deconvolution algorithms to reject correct chemical formulas because of a mismatch with theoretical patterns.

Co-eluting compounds with overlapping mass spectra distort background subtraction routines. When a plasticizer like diethylhexyl phthalate co-elutes with an unknown adhesive degradation product, shared fragment ions mask the spectrum of the lower-abundance migrant. Automatic spectral deconvolution algorithms attempt to unmix composite spectra using ion chromatographic profiles, but tightly correlated elution profiles lead to incomplete spectral extraction and corrupted library searches against databases like NIST or Wiley.

1. Signal Threshold Definition establishes the baseline noise floor to exclude electronic artifacts while retaining trace migrant peaks above ten parts per billion.
2. Chromatographic Peak Detection fits mathematical models to total ion chromatograms to isolate eluting chemical species from mobile phase noise.
3. Adduct and Isotope Grouping clusters sodium, potassium, and protonated species alongside natural isotope patterns into unified molecular features.
4. Background Subtraction Routine eliminates chemical noise contributions originating from migration cell components and extraction solvents.
5. Library Mass Spectral Matching compares deconvoluted fragmentation spectra against certified reference databases to calculate match confidence scores.

Calculating the true confidence level of a deconvoluted feature requires propagating mass accuracy errors, isotopic fit scores, and fragmentation match metrics into a unified identification probability score. Yet bounding the false-discovery rate of unknown migrants remains difficult when spectral libraries contain fewer than fifteen percent of all commercially used plastic additives and transformation products.

Surrogate

Semi-Quantitative Response Factor Variance

Determining the concentration of an unidentified non-intentionally added substance requires converting raw chromatographic peak areas into mass fractions. In targeted analysis, pure reference standards generate compound-specific calibration curves that correlate ion current directly with concentration. Non-target screening lacks reference standards for newly discovered migrants, forcing laboratories to use semi-quantitative approaches that apply the response factor of an added internal standard or surrogate to calculate concentrations for every unknown feature in the extract.

Ionization efficiency in electrospray mass spectrometry varies by up to three orders of magnitude across chemical structures. A highly ionizable tertiary amine produces a signal intensity one thousand times greater than a weakly ionizable saturated fatty acid ester at identical molar concentrations. Assuming a uniform response factor introduces substantial quantitative error, underestimating the concentration of a poorly ionizable migrant by a factor of one hundred or overestimating a highly ionizable contaminant by the same margin.

Uncertainty Propagation Mathematics in Concentration Estimation

Quantifying the combined uncertainty of semi-quantitative non-target screening requires integrating multiple variance terms. The total relative standard uncertainty of a calculated migrant concentration depends on extraction recovery, mass spectrometer peak integration repeatability, internal standard concentration uncertainty, and the probability distribution of response factors across the chemical domain evaluated.

Models using response factor distributions constructed from hundreds of target compounds estimate the probability that an unknown feature’s true concentration lies below a regulatory limit. By fitting log-normal distribution curves to response factor databases, analysts derive a conversion factor corresponding to the eightieth or ninety-fifth percentile of ionization efficiency. Applying an upper-percentile response factor yields a conservative estimate, minimizing false-negative compliance decisions while increasing unnecessary follow-up investigations for benign migrants.

Ionization response factors across unknown migrants span three orders of magnitude in electrospray ionization, making single-standard semi-quantification a screening estimate rather than an absolute measurement.

Propagating uncertainty through a log-normal response factor distribution follows a distinct sequence during data processing.

1. Calculates raw peak area ratios between the unknown non-target feature and the nearest eluting deuterated surrogate standard.
2. Applies the known concentration of the surrogate standard to establish a baseline nominal concentration value.
3. Multiplies nominal concentration by the extraction recovery correction factor derived from matrix validation spikes.
4. Incorporates the log-normal variance parameter of the compound-class response factor model to compute upper confidence bounds.
5. Compares the calculated upper confidence limit concentration directly against toxicological screening thresholds.

Using uncalibrated semi-quantitative screening data to declare food contact compliance creates immediate legal exposure if border control authorities perform targeted testing with authentic reference standards and find restricted migrants exceeding legal limits.

Benchmark

Toxicological Threshold Mapping and Schymanski Levels

Interpreting non-target screening data for regulatory compliance requires combining chemical identification confidence with toxicological risk criteria. The Schymanski communication scale categorizes identification confidence into five tiers. Level 1 represents confirmed structures verified against authentic reference standards with matching mass spectra, retention times, and fragmentation profiles.

Level 2 covers probable structures supported by library spectrum matching or diagnostic fragment interpretation. Level 3 indicates tentative candidate structures sharing specific molecular formulas or sub-structural fragments. Level 4 establishes molecular formulas without structural resolution, and Level 5 defines exact mass features lacking formula assignment.

Regulatory evaluation of unknown migrants relies on the Threshold of Toxicological Concern framework when empirical toxicological data is unavailable. Under European Food Safety Authority guidelines, uncharacterized substances lacking structural alerts for genotoxicity operate under a default exposure threshold of 0.015 micrograms per kilogram body weight per day ~ corresponding to a concentration limit of 0.01 milligrams per kilogram in food or food simulants. If an unknown migrant contains structural alerts for genotoxicity, such as aromatic amines, epoxides, or alkyl hydrazines, the toxicological threshold drops to 0.0025 micrograms per kilogram body weight per day, requiring analytical detection down to 0.15 parts per billion in food simulants.

Integration of QSAR Tools and Cramer Classification

Assigning a Threshold of Toxicological Concern to a Level 2 or Level 3 non-target feature requires running molecular structures through Quantitative Structure-Activity Relationship models like ToXsuite or Derek Nexus. These tools screen candidate structures against databases of known carcinogens, mutagens, and endocrine disruptors. If an analytical feature maps to five candidate structures at Level 3, and even one contains a structural alert for mutagenicity, the analyst must apply the most conservative toxicological threshold across all plausible candidates.

Uncertainty in identification confidence propagates directly into toxicological evaluation. Misinterpreting an isotope peak as a protonated molecular ion leads to an incorrect molecular weight calculation, feeding false structural inputs into screening software. A false negative during alert screening can assign a high toxicological threshold of 0.05 milligrams per kilogram to a potent carcinogen, allowing a hazardous migrant to pass compliance evaluation undetected.

Submitting non-target screening data in compliance dossiers supporting declarations of conformity requires explicit documentation.

• Mass Accuracy Traceability Records document daily instrument calibration routines and internal lock mass stability across analytical batches.
• Schymanski Tier Classifications assign explicit confidence levels to every reported chemical feature exceeding screening thresholds.
• Ionization Mode Complementarity Data demonstrates acquisition using both positive and negative electrospray modes alongside gas chromatography mass spectrometry for non-polar fractions.
• Extraction Blank Profiles identify matrix spikes, solvent impurities, and laboratory system contaminants to prevent false non-intentionally added substance assignments.

Article 16 of Regulation EC 1935/2004 demands that supporting documentation demonstrate compliance through test reports backed by verified analytical methodologies.

Discrepancy

Inter-Laboratory Variability and Legal Defensibility

When two accredited laboratories analyze identical samples of migrating packaging film using high-resolution non-target screening, reported chemical feature lists frequently diverge. Discrepancies arise from differences in sample preparation, instrument front-end geometries, stationary phase selectivity, deconvolution algorithms, and spectral library coverage. Laboratory A, running a liquid chromatography time-of-flight mass spectrometer with a methanol-water gradient, identifies forty distinct migrant features, whereas Laboratory B, operating an Orbitrap system with an acetonitrile-water gradient, reports sixty-five features from the same simulant extract.

Variability between analytical facilities creates significant commercial friction during cross-border trade inspections. A regulatory authority performing official controls may flag an unlisted non-intentionally added substance exceeding ten parts per billion based on an in-house protocol, while the manufacturer’s dossier contains a test report from a contract laboratory showing no signals above background noise. Resolving these conflicting measurements requires standardized workflows for uncertainty estimation that account for instrument-specific variance, ionization suppression, and algorithm settings.

Quantifying measurement uncertainty transforms raw mass spectrometry counts into defensible legal evidence. Combining individual uncertainty components ~ extraction recovery, mass measurement drift, peak deconvolution variance, and response factor spread ~ yields an expanded measurement uncertainty that routinely reaches fifty to two hundred percent of the estimated concentration. A reported non-target concentration of eight parts per billion with a two hundred percent expanded uncertainty spans a true value range from two point six to twenty-four parts per billion, straddling the regulatory action limit of ten parts per billion.

Regulatory authorities faced with data straddling compliance limits evaluate the statistical probability of non-compliance based on confidence intervals. Importers placing packaging materials on the market bear the responsibility of demonstrating that migrant concentrations remain safely below legal limits even at the upper bound of analytical uncertainty. Downstream brand owners increasingly demand that converted films and multi-layer laminates come backed by non-target screening reports that explicitly report lower and upper quantitative confidence intervals rather than single-point semi-quantitative values.

Standardizing non-target screening protocols across enforcement laboratories requires certified reference materials containing known mixtures of non-intentionally added substances at trace concentrations. Synthetic mixtures containing cyclic oligomers, degradation products of secondary antioxidants, and ink photoinitiators allow testing facilities to benchmark feature detection thresholds, formula assignment accuracy, and semi-quantitative response models against standardized ground truth data. Method validation frameworks developed under European Committee for Standardization protocols aim to harmonize reporting formats, signal-to-noise calculations, and identification confidence assignments, ensuring that non-target screening results stand up to judicial review during market enforcement actions.