Linear-No-Threshold Default Assumptions for Noncancer and Nongenotoxic Cancer Risks: A Mathematical and Biological Critique

Quantitative risk analysis of treatment plans for patients with tumor by mining historical similar patients from electronic health records using federated learning

The determination of a treatment plan for a target patient with tumor is a difficult problem due to the existence of heterogeneity in patients' responses, incomplete information about tumor states, and asymmetric knowledge between doctors and patients, and so on. In this paper, a method for quantitative risk analysis of treatment plans for patients with tumor is proposed. To reduce the impacts of the heterogeneity in patients' responses on analysis results, the method conducts risk analysis by mining historical similar patients from Electronic Health Records (EHRs) in multiple hospitals using federated learning (FL). For this, the Recursive Feature Elimination based on the Support Vector Machine (SVM) and Deep Learning Important FeaTures (DeepLIFT) are extended into the FL framework to select key features and determine key feature weights for identifying historical similar patients. Then, in the database of each collaborative hospital, the similarities between the target patient and all historical patients are calculated, and the historical similar patients are determined. According to the statistics of tumor states and treatment outcomes of historical similar patients in all collaborative hospitals, the related data (including the probabilities of different tumor states and possible outcomes of different treatment plans) for risk analysis of the alternative treatment plans can be obtained, which can eliminate the asymmetric knowledge between doctors and patients. The related data are valuable for the doctor and patient to make their decisions. Experimental studies have been conducted to verify the feasibility and effectiveness of the proposed method.

Ultrasensitive dose-response for asbestos cancer risk implied by new inflammation-mutation model

Alterations in complex cellular phenotype each typically involve multistep activation of an ultrasensitive molecular switch (e.g., to adaptively initiate an apoptosis, inflammasome, Nrf2-ARE anti-oxidant, or heat-shock activation pathway) that triggers expression of a suite of target genes while efficiently limiting false-positive switching from a baseline state. Such switches exhibit nonlinear signal-activation relationships. In contrast, a linear no-threshold (LNT) dose-response relationship is expected for damage that accumulates in proportion to dose, as hypothesized for increased risk of cancer in relation to genotoxic dose according to the multistage somatic mutation/clonal-expansion theory of cancer, e.g., as represented in the Moolgavkar-Venzon-Knudsen (MVK) cancer model by a doubly stochastic nonhomogeneous Poisson process. Mesothelioma and lung cancer induced by exposure to carcinogenic (e.g., certain asbestos) fibers in humans and experimental animals are thought to involve modes of action driven by mutations, cytotoxicity-associated inflammation, or both, rendering ambiguous expectations concerning the nature of model-implied shape of the low-dose response for above-background increase in risk of incurring these endpoints. A recent Inflammation Somatic Mutation (ISM) theory of cancer posits instead that tissue-damage-associated inflammation that epigenetically recruits, activates and orchestrates stem cells to engage in tissue repair does not merely promote cancer, but rather is a requisite co-initiator (acting together with as few as two somatic mutations) of the most efficient pathway to any type of cancer in any reparable tissue (Dose-Response 2019; 17(2):1-12). This theory is reviewed, implications of this theory are discussed in relation to mesothelioma and lung cancer associated with chronic asbestos inhalation, one of the two types of ISM-required mutations is here hypothesized to block or impede inflammation resolution (e.g., by doing so for GPCR-mediated signal transduction by one or more endogenous autacoid specialized pro-resolving mediators or SPMs), and supporting evidence for this hypothesis is discussed.

Thresholds for carcinogens

Current regulatory cancer risk assessment principles and practices assume a linear dose-response relationship-the linear no-threshold (LNT) model-that theoretically estimates cancer risks occurring following low doses of carcinogens by linearly extrapolating downward from experimentally determined risks at high doses. The two-year rodent bioassays serve as experimental vehicles to determine the high-dose cancer risks in animals and then to predict, by extrapolation, the number of carcinogen-induced tumors (tumor incidence) that will arise during the lifespans of humans who are exposed to environmental carcinogens at doses typically orders of magnitude below those applied in the rodent assays. An integrated toxicological analysis is conducted herein to reconsider an alternative and once-promising approach, tumor latency, for estimating carcinogen-induced cancer risks at low doses. Tumor latency measures time-to-tumor following exposure to a carcinogen, instead of tumor incidence. Evidence for and against the concept of carcinogen-induced tumor latency is presented, discussed, and then examined with respect to its relationship to dose, dose rates, and the dose-related concepts of initiation, tumor promotion, tumor regression, tumor incidence, and hormesis. Considerable experimental evidence indicates: (1) tumor latency (time-to-tumor) is inversely related to the dose of carcinogens and (2) lower doses of carcinogens display quantifiably discrete latency thresholds below which the promotion and, consequently, the progression and growth of tumors are delayed or prevented during a normal lifespan. Besides reconciling well with the concept of tumor promotion, such latency thresholds also reconcile favorably with the existence of thresholds for tumor incidence, the stochastic processes of tumor initiation, and the compensatory repair mechanisms of hormesis. Most importantly, this analysis and the arguments presented herein provide sound theoretical, experimental, and mechanistic rationales for rethinking the foundational premises of low-dose linearity and updating the current practices of cancer risk assessment to include the concept of carcinogen thresholds.

Implications of nonlinearity, confounding, and interactions for estimating exposure concentration-response functions in quantitative risk analysis

Recent advances in understanding of biological mechanisms and adverse outcome pathways for many exposure-related diseases show that certain common mechanisms involve thresholds and nonlinearities in biological exposure concentration-response (C-R) functions. These range from ultrasensitive molecular switches in signaling pathways, to assembly and activation of inflammasomes, to rupture of lysosomes and pyroptosis of cells. Realistic dose-response modeling and risk analysis must confront the reality of nonlinear C-R functions. This paper reviews several challenges for traditional statistical regression modeling of C-R functions with thresholds and nonlinearities, together with methods for overcoming them. Statistically significantly positive exposure-response regression coefficients can arise from many non-causal sources such as model specification errors, incompletely controlled confounding, exposure estimation errors, attribution of interactions to factors, associations among explanatory variables, or coincident historical trends. If so, the unadjusted regression coefficients do not necessarily predict how or whether reducing exposure would reduce risk. We discuss statistical options for controlling for such threats, and advocate causal Bayesian networks and dynamic simulation models as potentially valuable complements to nonparametric regression modeling for assessing causally interpretable nonlinear C-R functions and understanding how time patterns of exposures affect risk. We conclude that these approaches are promising for extending the great advances made in statistical C-R modeling methods in recent decades to clarify how to design regulations that are more causally effective in protecting human health.

A global environmental health perspective and optimisation of stress

The phrase "what doesn't kill us makes us stronger" suggests the possibility that living systems have evolved a spectrum of adaptive mechanisms resulting in a biological stress response strategy that enhances resilience in a targeted quantifiable manner for amplitude and duration. If so, what are its evolutionary foundations and impact on biological diversity? Substantial research demonstrates that numerous agents enhance biological performance and resilience at low doses in a manner described by the hormetic dose response, being inhibitory and/or harmful at higher doses. This Review assesses how environmental changes impact the spectrum and intensity of biological stresses, how they affect health, and how such knowledge may improve strategies in confronting global environmental change.

Commentary: EPA's proposed expansion of dose-response analysis is a positive step towards improving its ecological risk assessment

The United States Environmental Protection Agency (US EPA) has recently proposed changes to strengthen the transparency of its pivotal regulatory science policy and procedures. In this context, the US EPA aims to enhance the transparency of dose-response data and models, proposing to consider for the first time non-linear biphasic dose-response models. While the proposed changes have the potential to lead to markedly improved ecological risk assessment compared to past and current approaches, we believe there remain open issues for improving the quality of ecological risk assessment, such as the consideration of adaptive, dynamic and interactive effects. Improved risk assessment including adaptive and dynamic non-linear models (beyond classic threshold models) can enhance the quality of regulatory decisions and the protection of ecological health. We suggest that other countries consider adopting a similar scientific-regulatory posture with respect to dose-response modeling via the inclusion of non-linear biphasic models, that incorporate the dynamic potential of biological systems to adapt (i.e., enhancing positive biological endpoints) or maladapt to low levels of stressor agents.

Linear-No-Threshold Default Assumptions are Unwarranted for Cytotoxic Endpoints Independently Triggered by Ultrasensitive Molecular Switches

Crump's response in this issue to my critique of linear-no-threshold (LNT) default assumptions for noncancer and nongenotoxic cancer risks (Risk Analysis 2016; 36(3):589-604) is rebutted herein. Crump maintains that distinguishing between a low-dose linear dose response and a threshold dose response on the basis of dose-response data is impossible even for endpoints involving increased cytotoxicity. My rebuttal relies on descriptions and specific illustrations of two well-characterized ultrasensitive molecular switches that govern two key cytoprotective responses to cellular stress-heat shock response and antioxidant response element activation, respectively-each of which serve to suppress stress-induced apoptotic cell death unless overwhelmed. Because detailed dose-response data for each endpoint is shown to be J- or inverted-J-shaped with high confidence, and because independent pathways can explain background rates of apoptosis, LNT assumptions for this cytotoxic endpoint are unwarranted, at least in some cases and perhaps generally.

Bogen's Critique of Linear-No-Threshold Default Assumptions

In an article recently published in this journal, Bogen⁽¹⁾ concluded that an NRC committee's recommendations that default linear, nonthreshold (LNT) assumptions be applied to dose- response assessment for noncarcinogens and nonlinear mode of action carcinogens are not justified. Bogen criticized two arguments used by the committee for LNT: when any new dose adds to a background dose that explains background levels of risk (additivity to background or AB), or when there is substantial interindividual heterogeneity in susceptibility (SIH) in the exposed human population. Bogen showed by examples that SIH can be false. Herein is outlined a general proof that confirms Bogen's claim. However, it is also noted that SIH leads to a nonthreshold population distribution even if individual distributions all have thresholds, and that small changes to SIH assumptions can result in LNT. Bogen criticizes AB because it only applies when there is additivity to background, but offers no help in deciding when or how often AB holds. Bogen does not contradict the fact that AB can lead to LNT but notes that, even if low-dose linearity results, the response at higher doses may not be useful in predicting the amount of low-dose linearity. Although this is theoretically true, it seems reasonable to assume that generally there is some quantitative relationship between the low-dose slope and the slope suggested at higher doses. Several incorrect or misleading statements by Bogen are noted.

Low-Dose Dose-Response for In Vitro Nrf2-ARE Activation in Human HepG2 Cells

Kelch-like ECH-associated protein 1 (Keap1), nuclear factor erythroid 2-like factor 2-related factor 2 (Nrf2), and the antioxidant response element (ARE) are interacting components of a master regulatory signaling pathway that coordinates redox homeostasis, cytoprotective responses, and shifts in stem cell state. This study reexamined detailed dose–response (DR) data reported for in vitro Nrf2-ARE activation in human hepatoblastoma HepG2 cell lines containing either a ARE-bla or ARE-luc reporter at 12 different concentrations of each of 15 chemicals. The normalized study data were combined among chemicals exhibiting a positive response, yielding n = 531 (179) DR data for 9 (7) chemicals using the ARE-bla (ARE-luc) assay. Three-parameter linear/kth-power regression fits obtained to each combined set of ARE-bla- or ARE-luc-assay response data provided good fits (R² = .99 or .91, respectively, P_fit > .99) that each incorporate a highly significant negative initial linear slope (P = 4 × 10⁻⁵ or .00025) and an overall J-shaped DR pattern. Results from this reanalysis of high-resolution ARE response data support the hypothesis that nonlinear ARE-mediated adaptive cellular responses to oxidative stress are governed by an ultrasensitive molecular switch.

Low-dose dose-response for reduced cell viability after exposure of human keratinocyte (HEK001) cells to arsenite

The in vitro arsenite (AsIII) cytotoxicity dose-response (DR) of human keratinocytes (HEK001) was examined at greater statistical resolution than ever previously reported using the MTT assay to determine cell viability. Fifty-four 96-well plates were treated with AsIII concentrations of 0.25, 0.5, 1, 2, 3, 4, 5, 7, 10, 15, 20, 25, or 30 μM. Because of unexpected variation in viability response patterns, a two-stage DR analysis was used in which data on plate-specific viability (%), estimated as 100% times the ratio of measured viability in exposed to unexposed cells, were fit initially to a generalized lognormal response function positing that HEK001 cells studied consisted of: a proportion P of relatively highly sensitive (HS) cells, a proportion Po of relatively resistant cells, and a remaining (1-P-Po) fraction of typical-sensitivity (TS) cells exhibiting the intermediate level of AsIII sensitivity characteristic of most cells in each assay. The estimated fractions P and Po were used to adjust data from all 54 plates (and from the 28 plates yielding the best fits) to reflect the condition that P = Po = 0 to provide detailed DR analysis specifically for TS cells. Four DR models fit to the combined adjusted data were each very predictive (R2 > 0.97) overall but were inconsistent with at least one of the data set examined (p < 10-5). Adjusted mean responses at ≤3 μM were approximately equal (p > 0.30) and exceeded 100% significance (p ≤ 10-6). A low-dose hormetic model provided the best fit to the combined adjusted data for TS cells (R2 = 0.995). Marked variability in estimates of P (the proportion of apparent HS cells) was unexpected, not readily explained, and warrants further study using additional cell lines and assay methods, and in vivo.

The Influence of Human Interindividual Variability on the Low-Dose Region of Dose-Response Curve Induced by 2,3,7,8-Tetrachlorodibenzo-p-Dioxin in Primary B Cells

The influence of interindividual variability is not typically assessed in traditional toxicological studies. Given that chemical exposures occur in heterogeneous populations, this knowledge gap has the potential to cause undue harm within the realms of public health and industrial and municipal finances. A recent report from the National Research Council (NRC) suggests that when accounting for interindividual variation in responses, traditionally assumed nonlinear dose-response relationships (DRRs) for noncancer-causing endpoints would better be explained with a linear relationship within the low-dose region. To address this knowledge gap and directly test the NRC's assumption, this study focused on assessing the DRR between 2,3,7,8-tetracholorodibenzo-p-dioxin (TCDD) exposure and immune suppression in a cohort of unique human donors. Human B cells were isolated from 51 individual donors and treated with logarithmically increasing concentrations of TCDD (0-30 nM TCDD). Two endpoints sensitive to TCDD were assessed: (1) number of IgM-secreting B cells and (2) quantity of IgM secreted. The results show that TCDD significantly suppressed both the number of IgM-secreting B cells and the quantity of IgM secreted (P < .05). Statistical model comparisons indicate that the low-dose region of the two DRRs is best explained with a nonlinear relationship. Rather than assuming low-dose linearity for all noncancer-causing DRRs, our study suggests the need to consider the specific mode of action of toxicants and pharmaceuticals during risk-management decision making.