When exposure isn’t dose: parameter non-identifiability as the central barrier to human microplastic toxicokinetics

Human health risk assessment for microplastics (MPs) rests on an implicit assumption: that ingestion exposure estimates can be translated into internal tissue dose and lifetime accumulation burden. This Critical Review evaluates whether the available human evidence base can support that translation, and argues that it cannot, not because of data scarcity alone, but for a structural reason: the underlying toxicokinetic (TK) systems are structurally non-identifiable under the measurement designs currently in use. Using a scoping review of 22 eligible studies to map empirical baseline data, we executed a synthetic dependency and sensitivity demonstration using a literature-anchored synthetic dataset and probabilistic Monte Carlo simulations to evaluate the structural integrity of commonly used intake–elimination assumptions. Quantitative analyses are restricted to microplastics of 1 μm or larger; nanoplastics (< 1 μm) are treated qualitatively throughout, because the analytical methods that detect them and the toxicokinetic evidence available for them differ substantially from those for microplastics. While this dependency simulation yields a seemingly robust and near-perfect relationship between estimated ingestion and faecal excretion (r = 0.949; no significance test is reported, as the variables are dependent by construction), we demonstrate that this correlation is a purely mathematical artifact of the model’s structural architecture rather than evidence of biological coupling. Individual clinical biomonitoring correlations (such as in tissue biopsies or fluids) are fundamentally uncoupled from ingestion paths due to unmeasured multi-route confounding. Synthesizing 22 studies across human biomonitoring, ex vivo organoid, animal, and in vitro mechanistic evidence, we show that no dataset co-measures intake and elimination in the same individuals and that tissue detections aggregate unattributed multi-route (ingestion, inhalation, dermal) burdens; fixed-fraction retention models are therefore fitted to structurally under-determined systems. We formalize this with a convolution model of multi-route tissue burden and demonstrate, via probabilistic simulation of a literature-anchored synthetic intake–excretion dataset, that the near-perfect intake–excretion correlations that arise when excretion is defined as a fixed fraction of intake (Pearson r ≈ 0.88–0.98 across Monte Carlo conditions) are a consequence of model construction, not evidence of biological coupling. We then derive the minimum identifiability conditions, longitudinal, same-subject, route-resolved intake–elimination–retention measurement that any valid human MP-TK model must satisfy and translate these into concrete study-design and reporting requirements. Reframing the field’s central problem from “insufficient data” to “structurally uninformative data” reorients microplastic risk assessment toward the specific measurements that would make quantitative body-burden modelling possible.