New perspectives in assessing environmental risks for birds: a simple TKTD framework to link growth and reproduction energy budget to chemical stress

This paper introduces BIRDkiss, a new open-source R-based modeling framework that integrates simplified energy budget and toxicokinetic-toxicodynamic principles to predict how pesticides and food availability interact to affect bird growth and reproduction, thereby enhancing the ecological relevance of environmental risk assessments.

The Gist

Imagine you are trying to predict how a bird will do in the real world. In a lab, scientists feed birds a specific amount of food and give them a specific dose of pesticide. But in nature, things are messy. Birds might find less food one day, more the next, and they might be exposed to a cocktail of different chemicals, not just one.

This paper introduces a new digital tool called BIRDkiss (which stands for Bird - Impact on Reproduction via Diet, keep it simple and suitable). Think of BIRDkiss as a high-tech "bird simulator" that helps scientists understand how pesticides and food shortages affect a bird's ability to grow and lay eggs.

Here is how it works, broken down into simple concepts:

1. The Bird's Wallet: Energy Budgeting

Imagine a bird's body is like a person with a monthly budget.

• Income: The bird earns "energy currency" by eating food.
• Expenses: It has to pay bills for maintenance (keeping its heart beating, staying warm) and growth (getting bigger).
• Savings: Whatever is left over goes into a "savings account" for reproduction (laying eggs).

In a healthy world, the bird spends its energy wisely. But what happens when a pesticide enters the picture?

2. The Pesticide as a "Tax" or "Leak"

The paper explains that chemicals don't just kill birds instantly; they act like a leak in the wallet or a heavy tax on the bird's energy.

• When a bird eats contaminated food, the chemical causes "damage" inside its body.
• The bird has to spend extra energy to fix this damage (detoxification).
• Because energy is limited, if the bird spends more on fixing the chemical damage, it has less money left in the savings account for making eggs.

3. The "Keep It Simple" Approach

Older models were like complex, 500-page instruction manuals that were hard to read and required data we often don't have.

• BIRDkiss is like a smartphone app: it's streamlined, easy to use, and focuses on the most important numbers (how much the bird weighs and how many eggs it lays).
• It uses a "Bayesian" method, which is like a smart guesser. It takes the limited data we have from lab tests, makes a best guess about how the bird works, and then updates that guess as it learns more, giving us a range of likely outcomes rather than just one rigid number.

4. The Big Discoveries: Food vs. Poison

The researchers used this simulator to test some "What if?" scenarios that are hard to test in real life:

• The "Too Much Food" Trap: You might think giving a bird more food would make it lay more eggs. The model shows that there is a limit. Once a bird has enough food to be healthy, extra food doesn't help it lay more eggs. It's like having a full tank of gas; driving faster doesn't help if the engine is already running at max speed.
• The "Starvation" Danger: If food is scarce, the bird stops laying eggs to save energy for its own survival. It's a "survival mode."
• The Double Whammy (The Most Important Finding): This is the paper's big reveal. When you combine low food with chemicals, the damage is much worse than just adding the two problems together.
  • Analogy: Imagine you are trying to run a marathon (survival) while carrying a heavy backpack (chemicals). If you are well-fed, you can handle the backpack. But if you are also starving, the backpack crushes you. The bird has to choose between fixing the chemical damage and staying alive, leaving zero energy for eggs.

5. Mixing Chemicals: The "Team" Effect

In the real world, birds are rarely exposed to just one pesticide. They face mixtures.

• The model can simulate two ways chemicals work together:
  1. Concentration Addition (CA): Like a team of workers doing the same job; their strength adds up.
  2. Independent Action (IA): Like workers doing different jobs; they don't help each other much.
• The model found that the "Team" approach (CA) usually predicts a worse outcome (more damage) than the "Independent" approach. For safety regulations, it's better to assume the chemicals work together to cause maximum harm.

Why Does This Matter?

Currently, risk assessments for birds are based on clean lab tests where birds are well-fed and only see one chemical. This paper argues that this is too optimistic.

By using BIRDkiss, regulators can see that chemicals are much more dangerous when food is scarce. This tool allows scientists to predict real-world scenarios, potentially saving birds from unexpected population crashes and helping us write better, safer rules for using pesticides.

In short: BIRDkiss is a simple but powerful calculator that tells us: If you starve a bird and poison it at the same time, the result is a disaster for the next generation of birds.

Technical Summary

1. Problem Statement

Environmental Risk Assessment (ERA) for pesticides traditionally relies on single-species laboratory tests to determine the toxicity of individual active substances. However, this approach faces significant limitations:

• Extrapolation Gap: Statistical data-driven models struggle to extrapolate static laboratory results to dynamic, real-world field scenarios involving time-varying exposures.
• Complexity of Real-World Stressors: Real-world environments involve multiple chemical compounds (mixtures) and varying abiotic factors (e.g., food availability), which are rarely tested in standard OECD guidelines.
• Sub-lethal Effects: Standard assessments often focus on mortality, whereas sub-lethal effects on growth and reproduction are critical for population sustainability but harder to model mechanistically.
• Data Scarcity: There is a lack of specific mechanistic models for birds compared to aquatic organisms, and existing Dynamic Energy Budget (DEB) models are often too complex for routine regulatory use due to high parameter requirements.

2. Methodology

The authors propose BIRDkiss (Bird - Impact on Reproduction via Diet, keep it simple and suitable), a novel modeling framework that integrates a simplified Dynamic Energy Budget (DEB) model with Toxicokinetic-Toxicodynamic (TKTD) principles.

A. Model Architecture

The model is fully embedded in an open-source R package using Stan for Bayesian inference. It consists of three main modules:

1. DEBkiss Component (Energy Budget):
   • Based on the simplified DEBkiss theory (Jager et al.), which removes the reserve compartment found in standard DEB models to reduce complexity.
   • State Variables: Structural body mass ($W_V$) and reproduction buffer mass ($W_R$).
   • Mechanism: Food assimilation ($J_A$) is split via the $\kappa$-rule: a fraction $\kappa$ goes to somatic maintenance and growth, while $(1-\kappa)$ goes to the reproduction buffer.
   • Reproduction Logic: Egg production is driven by the reproduction buffer, subject to a carrying capacity to prevent simultaneous egg laying, and a stress function that modulates the rate.
2. Toxicokinetic (TK) Component:
   • Models the internal scaled damage ($D(t)$) resulting from contaminated food ingestion.
   • Uses a single dominant rate constant ($k_d$) representing the balance between accumulation and elimination.
   • Incorporates bioavailability ($y_{BU}$) to account for the fraction of the chemical entering systemic circulation.
3. Toxicodynamic (TD) Component:
   • Converts internal damage into a stress function applied to egg production.
   • Assumes an Individual Tolerance hypothesis where toxicity thresholds follow a log-logistic distribution defined by a median ($\alpha$) and shape parameter ($\beta$).
   • The stress function ranges from 0 (100% reduction in reproduction) to 1 (no stress).

B. Mixture Toxicity

The framework implements two classical mixture paradigms:

• Concentration Addition (CA): Assumes chemicals share a mode of action. Concentrations are scaled by their respective EC50 values and summed.
• Independent Action (IA): Assumes chemicals act on different pathways. The combined effect is calculated by multiplying the probabilities of non-effect for each compound.

C. Statistical Framework

• Bayesian Inference: Parameters are estimated using non-informative priors (Beta, Exponential, or Inverse Gamma distributions).
• Stochasticity: The model accounts for observation error in body weight and egg counts using Gamma distributions.
• Calibration Data: The model was calibrated using standard OECD Guideline 206 avian reproduction test data for Bobwhite quail and Mallard ducks exposed to various single chemicals (e.g., Atrazine, Abamectin, Deltamethrin).

3. Key Contributions

• Simplification for Regulation: By adopting the DEBkiss approach, the model reduces the number of parameters to identify (only 8 parameters total) while maintaining biological realism, making it suitable for regulatory application.
• Integration of Multiple Stressors: Unlike standard models, BIRDkiss explicitly decouples food intake from chemical toxicity, allowing for the simulation of scenarios involving food limitation (starvation) combined with chemical exposure.
• Mixture Assessment: It provides a mechanistic implementation of CA and IA hypotheses for bird reproduction, a capability rarely found in avian risk assessment tools.
• Open-Source Tool: The release of a dedicated R package facilitates reproducibility and allows regulators and researchers to perform calibration, validation, and prediction without proprietary software.

4. Results

• Calibration Performance: The model successfully calibrated against observed data for single compounds. Posterior Predictive Checks (PPC) showed high coverage (mostly >95%) for both body weight and cumulative egg production across multiple species and chemicals.
• Food Availability Effects:
  • Saturation: Increasing food beyond a certain level did not increase egg production, indicating a physiological saturation limit.
  • Starvation: Significant food reduction drastically decreased egg production as energy was reallocated to maintenance.
• Synergistic Stress: The model demonstrated that chemical stress and nutritional stress are not additive but synergistic. When food is scarce, the negative impact of chemical exposure on reproduction is amplified because the organism lacks the energy reserves to support both maintenance/detoxification and reproduction.
• Mixture Simulations:
  • The IA model was simpler to implement (no scaling required) but tended to underestimate the combined effects compared to the CA model.
  • The CA model was identified as the more conservative (regulatory-safe) approach, as it accounts for the contribution of sub-threshold concentrations.
• EC50 Estimation: The model generated robust EC50 values for egg production, revealing species-specific sensitivities (e.g., Bobwhite quail was more sensitive to Azoxystrobin than Mallard ducks).

5. Significance and Implications

• Regulatory Relevance: BIRDkiss offers a "ready-to-use" mechanistic tool that can be integrated into regulatory frameworks (e.g., as an analysis method for OECD 206). It moves risk assessment from static endpoints to dynamic, scenario-based predictions.
• Ecological Realism: By accounting for food limitation, the model addresses a critical gap in current ERA. It suggests that laboratory-only assessments may underestimate risks for wild populations facing simultaneous starvation and contamination.
• Animal Welfare: The model supports a "de-risking" strategy by potentially reducing the need for extensive animal testing through robust in silico predictions and the ability to design more targeted experiments.
• Future Directions: While the model is robust for single compounds, the authors note that validation for chemical mixtures still requires empirical mixture data. However, the framework provides a solid theoretical foundation for advancing mixture toxicity assessment in birds.

In conclusion, BIRDkiss represents a significant step forward in avian ecotoxicology by bridging the gap between complex biological energy budgets and practical regulatory risk assessment, specifically highlighting the critical interaction between nutritional status and chemical toxicity.