---
author: Christoph Helma
date: October 12, 2022
title: Variability of chronic rodent bioassays
---

# Content

Rodent Carcinogenicity

:   E Gottmann, S Kramer, B Pfahringer and C Helma\
    *Data quality in predictive toxicology: reproducibility of rodent
    carcinogenicity experiments*\
    Environ Health Perspect 109:509--514 (2001)\
    <https://doi.org/10.1289/ehp.01109509>

Lowest observed adverse effect level (LOAEL)

:   C Helma, D Vorgrimmler, D Gebele, M Gütlein, B Engeli, J Zarn,
    B Schilter and E Lo Piparo\
    *Modeling Chronic Toxicity: A Comparison of Experimental Variability
    With (Q)SAR/Read-Across Predictions*\
    Front Pharmacol 9 (2018)\
    <https://doi.org/10.3389/fphar.2018.00413>

# Carcinogenicity Data

-   Carcinogenic Potency Database(CPDB, Gold 1997)
-   1,289 unique compounds
-   2 Subsets
    -   National Toxicology Program (NTP)
    -   General literature
-   121 common compounds in both subsets

# Carcinogenicity Classification

-   **57%** concordant classifications (69/121 compounds, 39
    carcinogens, 30 non-carcinogens)

Rats
:   62% concordant classifications

Mice
:   49% concordant classifications

Multi species carcinogens
:   58% concordant classifications

Multi organ carcinogens:
:   52% concordant classifications

-   poor reproducibility of sex, species and organ specific effects

# Carcinogenicity TD50's

![](fig2.png)

# Carcinogenicity caveats

-   low sample size
-   no standardized protocols for literature data

Gold et al. (1987)

:   -   38 compounds from the literature
    -   93% reproducibility for rats
    -   76% for mice
    -   34 studies were published by the same authors (!)

# LOAEL Data

Chronic (\>180 days) lowest observed effect levels (LOAEL) for rats
(Rattus norvegicus) after oral (gavage, diet, drinking water)
administration

Nestlé Database

:   567 LOAEL values for 445 unique chemical structures from the
    literature (Mazzatorta et al., 2008)

Swiss Food Safety and Veterinary Office (FSVO) Database

:   493 rat LOAEL values for 381 unique chemical structures from
    pesticide evaluations (Zarn et al., 2011, 2013)

    -   European Food Safety Authority (EFSA) (EFSA, 2014)
    -   Joint FAO/WHO Meeting on Pesticide Residues (JMPR) (WHO, 2011)
    -   US EPA (US EPA, 2011)

Combined dataset

:   -   compounds that occur in both databases
    -   375 LOAEL values for 155 unique chemical structures

# LOAEL Variability

**Both** datasets contain substances with multiple measurements

![](fphar-09-00413-g003.jpg)

All datasets have almost the same experimental variability (standard
deviations: 0.56 mg/kg_bw/day (Nestlé), 0.57 mg/kg_bw/day (FSVO), 0.56
mg/kg_bw/day (combined))

# LOAEL Correlation

![r\^2: 0.52, RMSE: 0.59, p-value \< 2.2e-16](fphar-09-00413-g004.jpg)

As both databases contain duplicates medians were used for the
correlation plot and statistics

# LOAEL Experiments vs Predictions

![](fphar-09-00413-g005.jpg)

# Conclusions

-   Carcinogenicity classifications seem to be poorly reproducible (57%
    concordant classifications for repeated experiments)

-   Experimental LOAEL values have a variablity of approximately 1.5 log
    units (orders of magnitude)

-   Variability in chronic *in vivo* bioassays might be caused by

    -   biological complexity
    -   long term experimental conditions
    -   evaluation complexity
    -   statistical limitations (low number of animals/treatment)

-   Good *in-silico* models have the same accuracy as biological
    experiments (*in-vivo* and *in-vitro*) for **compounds in their
    applicability domain**

\
\
<https://in-silico.ch/presentations/epa-nam-2022/>