Overview

toxval is an R package providing a unified, flexible framework for estimating ecotoxicological toxicity metrics from concentration–response data. It currently supports extraction of effect-based metrics (ECx) and no-effect metrics — including the No-Significant-Effect Concentration (NSEC) — from fitted frequentist (drc) and Bayesian (bayesnec) model objects.

For the theoretical basis of the NSEC, see Fisher & Fox (2023). For model-averaged N(S)EC estimation across smooth and threshold model classes, see Fisher et al. (2024).

Motivation

Contemporary ecotoxicology increasingly relies on model-based inference to derive toxicity metrics from concentration–response data. However, the statistical tools used to estimate these metrics remain fragmented, inconsistently defined, and often tied to specific modelling frameworks. These inconsistencies hinder transparency, reproducibility, and comparability in regulatory decision-making.

A core aim of toxval is to operationalise NSEC estimation across both frequentist and Bayesian concentration–response models and to generalise this capability to a broader suite of toxicity metrics. In doing so, the package highlights a critical and often overlooked issue in ecotoxicology: ambiguity in the definition of “effect.” ECx values — while widely used — depend on how effect is defined relative to controls, model structure, and response scale, leading to inconsistencies that are not always transparent to users (see issue #19).

toxval addresses these challenges by providing:

• a consistent interface for extracting ECx, NOEC, and NSEC values from fitted models;
• support for multiple model classes and inference frameworks;
• explicit, transparent definitions of effect and no-effect that are independent of model class; and
• improved reproducibility and comparability of toxicity estimates across modelling approaches.

The NSEC

The No-Significant-Effect Concentration (NSEC) is a model-based alternative to the traditional No-Observed-Effect Concentration (NOEC). Unlike the NOEC, which is tied to the experimental design and is estimated via hypothesis testing, the NSEC:

• decouples the no-effect estimate from the set of concentrations actually tested;
• enables proper statistical inference with uncertainty quantification; and
• retains the intuitive concept of “no observable effect.”

For threshold-based models (NEC models), toxval also supports extraction of the NEC parameter directly, and is designed to support model-averaged N(S)EC estimates that combine NSEC values from smooth models with NEC values from threshold models — providing a more statistically defensible approach to no-effect estimation (see Fisher et al. 2024 and the bayesnec package).

Installation

The current development version can be downloaded from GitHub via

if (!requireNamespace("remotes")) {
  install.packages("remotes")
}
remotes::install_github("open-aims/toxval", ref = "main")

Usage

Usage and further information about toxval can be seen on the project page and the reference page.

Planned Development

toxval is under active development. The following enhancements are planned over the coming months, drawing on open issues in the repository:

Expanded model class support (#21)

Currently toxval supports drc and bayesnec model fits. Support will be extended to a wider range of commonly used R model classes, including glm, glmer, gam, and gamm. The refactoring strategy is to shift the bulk of computation to operate on a prediction matrix, so that adding support for a new model class requires only a thin wrapper around that class’s existing prediction method.

Support for increasing responses (#20)

nsec() will be generalised to handle both increasing and decreasing concentration–response curves. Currently only nsec_multi() supports both directions; this capability will be folded into the main nsec() function, with nsec_multi() becoming a deprecated special case for multivariate responses. ecx() may also be extended to support stimulatory (hormetic) effects where a biologically meaningful upper bound exists.

NOEC estimation (#17)

A method for deriving the NOEC will be added for models fitted with a factor predictor. The control treatment will be inferred from the lowest concentration level or specified by the user, with support for one-tailed tests and a user-controlled significance threshold consistent with the NSEC workflow.

Model-averaged N(S)EC (#18)

Support will be added for estimating a model-averaged N(S)EC — combining NSEC estimates from smooth (ECx-type) models with NEC values from threshold models. Extraction from bayesnec multi-model fits will be straightforward; extension to other fitting frameworks (e.g., drc or custom threshold models) will follow.

ECx definition clarity and consistency (#19)

Ongoing work will address the implicit ambiguity in ECx definitions across model classes and implementations. Planned improvements include:

• clearer documentation of the type argument ("relative", "absolute", "direct") and when each is appropriate;
• explicit warnings when ECx definitions are poorly defined for a given model family (e.g., unbounded Gaussian responses);
• transparent exposure of the reference points (e.g., top, bottom, control) used in the calculation.

Model fitting wrappers (#22)

There is an open question about whether toxval should provide a thin abstraction layer over existing concentration–response fitting frameworks (e.g., drc, bayesnec). Such wrappers would standardise input data structures, model specification, and output objects, while making assumptions about bounds, link functions, and ECx/NEC definitions explicit and consistent. Community input is welcome — see issue #22.

Further Information

toxval is provided by the Australian Institute of Marine Science and Poisson Consulting under the GPL-2 License (GPL-2).