Search

In order to automate the counting of ovarian follicles required in multigeneration reproductive studies performed in the rat according to Organization for Economic Co-operation and Development guidelines 443 and 416, the application of deep neural networks was tested. The manual evaluation of the differential ovarian follicle count is a tedious and time-consuming task that requires highly trained personnel. In this regard, deep learning outputs provide overlay pictures for a more detailed documentation, together with an increased reproducibility of the counts. To facilitate the planned good laboratory practice (GLP) validation a workflow was set up using MLFlow to make all steps from generating of scans, training of the neural network, uploading of study images to the neural network, generation and storage of the results in a compliant manner controllable and reproducible. PyTorch was used as main framework to build the Faster region-based convolutional neural network for the training. We compared the performances of different depths of ResNet models with specific regard to the sensitivity, specificity, accuracy of the models. In this paper, we describe all steps from data labeling, training of networks, and the performance metrics chosen to evaluate different network architectures. We also make recommendation on steps, which should be taken into consideration when GLP validation is aimed for.

Pineal gland tumors are very rare brain lesions in rats as well as in other species including humans. A total of 8 (out of 1,360 examined) Wistar rats from 3 different combined chronic toxicity/carcinogenicity or mere carcinogenicity studies revealed pineal gland tumors. The tumors were regarded to be spontaneous and unrelated to treatment. The morphology and immunohistochemical evaluation led to the diagnosis malignant pinealoma. The main characteristics that were variably developed within the tumors were the following: cellular atypia, high mitotic index, giant cells, necrosis, Homer Wright rosettes, Flexner-Wintersteiner rosettes and pseudorosettes, positive immunohistochemical reaction for synaptophysin, and neuron-specific enolase. The pineal gland is not a protocol organ for histopathological examination in carcinogenicity studies. Nevertheless, the pineal gland can occasionally be encountered on the routine brain section or if it is the origin of a tumor protruding into the brain, the finding will be recorded. Therefore, although known to be a rare tumor in rats, pineal neoplasms should be included in the list of possible differential diagnoses for brain tumors, especially when the tumor is located in the region of the pineal body., (© 2015 by The Author(s).)

The Bovine Corneal Opacity and Permeability (BCOP) test is commonly used for the identification of severe ocular irritants (GHS Category 1), but it is not recommended for the identification of ocular irritants (GHS Category 2). The incorporation of human reconstructed tissue model-based tests into a tiered test strategy to identify ocular non-irritants and replace the Draize rabbit eye irritation test has been suggested (OECD TG 405). The value of the EpiOcular™ Eye Irritation Test (EIT) for the prediction of ocular non-irritants (GHS No Category) has been demonstrated, and an OECD Test Guideline (TG) was drafted in 2014. The purpose of this study was to evaluate whether the BCOP test, in conjunction with corneal histopathology (as suggested for the evaluation of the depth of the injury( and/or the EpiOcular-EIT, could be used to predict the eye irritation potential of agrochemical formulations according to the UN GHS, US EPA and Brazil ANVISA classification schemes. We have assessed opacity, permeability and histopathology in the BCOP assay, and relative tissue viability in the EpiOcular-EIT, for 97 agrochemical formulations with available in vivo eye irritation data. By using the OECD TG 437 protocol for liquids, the BCOP test did not result in sufficient correct predictions of severe ocular irritants for any of the three classification schemes. The lack of sensitivity could be improved somewhat by the inclusion of corneal histopathology, but the relative viability in the EpiOcular-EIT clearly outperformed the BCOP test for all three classification schemes. The predictive capacity of the EpiOcular-EIT for ocular non-irritants (UN GHS No Category) for the 97 agrochemical formulations tested (91% sensitivity, 72% specificity and 82% accuracy for UN GHS classification) was comparable to that obtained in the formal validation exercise underlying the OECD draft TG. We therefore conclude that the EpiOcular-EIT is currently the best in vitro method for the prediction of the eye irritation potential of liquid agrochemical formulations., (2015 FRAME.)

In 2009, the Bovine Corneal Opacity and Permeability (BCOP) test was accepted by the regulatory bodies for the identification of corrosive and severe ocular irritants (Global Harmonised System [GHS] Category 1). However, no in vitro test is currently accepted for the differentiation of ocular irritants (GHS Category 2) and non-irritants (GHS No Category). Human reconstructed tissue models have been suggested for incorporation into a tiered testing strategy to ultimately replace the Draize rabbit eye irritation test (OECD TG 405). The purpose of this study was to evaluate whether the EpiOcular(TM) reconstructed cornea-like tissue model and the COLIPA pre-validated EpiOcular Eye Irritation Test (EpiOcular-EIT) could be used as suitable components of this testing strategy. The in-house validation of the EpiOcular-EIT was performed by using 60 test substances, including a broad variety of chemicals and formulations for which in vivo data (from the Draize rabbit eye irritation test) were available. The test substances fell into the following categories: 18 severe irritants/corrosives (Category 1), 21 irritants (Category 2), and 21 non-irritants (No Category). Test substances that decreased tissue viability to ≤ 60% (compared to the negative control tissue) were considered to be eye irritants (Category 1/2). Test substances resulting in tissue viability of > 60% were considered to be non-irritants (No Category). For the assessed dataset and the classification cut-off of 60% viability, the EpiOcular-EIT provided 98% and 84% sensitivity, 64% and 90% specificity, and 85% and 86% overall accuracy for the literature reference and BASF proprietary substances, respectively. Applying a 50% tissue viability cut-off to distinguish between irritants and non-irritants resulted in 93% and 82% sensitivity, 68% and 100% specificity, and 84% and 88% accuracy for the literature reference and BASF proprietary substances, respectively. Further, in the EpiOcular-EIT (60% cut-off), 100% of severely irritating substances under-predicted by the BCOP assay were classified as Category 1/2. The results obtained in this study, based on 60 test substances, indicate that the EpiOcular-EIT and the BCOP assay can be combined in a testing strategy to identify strong/severe eye irritants (Category 1), moderate and mild eye irritants (Category 2), and non-irritants (No Category) in routine testing. In particular, when the bottom-up strategy with the 60% viability cut-off was employed, none of the severely irritating substances (Category 1) were under-predicted to be non-irritant. Sensitivity for Category 1/2 substances was 100% for literature reference substances and 89% for BASF SE proprietary substances., (2011 FRAME.)

Acute oral toxicity testing is still required for the classification and labelling of chemicals, agrochemicals and related formulations. There have been increasing efforts over the last two decades to reduce the number of animals needed for this testing, according to the Three Rs concept. To evaluate the utility of an in vitro cytotoxicity test in our routine testing for acute oral toxicity, we have implemented in our laboratory the neutral red uptake (NRU) method, with Balb/c 3T3 fibroblasts after a 48-hour exposure, which was recommended in ICCVAM Report 07-4519, 2006. Initially, we tested 16 substances that had existing in vivo and in vitro data available, to prove our technical proficiency with the in vitro test. Then, testing was performed with 187 test substances, including a broad variety of chemicals, agrochemicals and formulations. The starting dose for acute oral systemic toxicity assays in rats (LD50) was estimated by using the prediction model presented in the ICCVAM validation study, and subsequently compared to the results obtained by in vivo testing performed according to, or similar to, OECD Test Guideline 423. Comparison of all of the 203 predicted LD50 values that were deduced from the in vitro IC50 values, with the in vivo results from oral toxicity studies in rats, resulted in a low overall concordance of 35%. The in vitro cytotoxicity assay achieved a good concordance of 74%, only for the weakly toxic substances (EU-GHS Cat. 4). However, it must be noted that 71% of the substances tested (i.e. 145/203) were classified as being weakly toxic in vitro. We further analysed the utility of the in vitro test for predicting the starting dose for an in vivo study, and the potential reduction in animal usage that this would engender. In this regard, the prediction by the cytotoxicity test was useful for 59% of the substances. However, the use of a standard starting dose of 300 mg/kg bw by default (without previous cytotoxicity testing) would have been almost as useful (50%). In contrast, the prediction by an experienced toxicologist was correct for 95% of the substances. However, this was only performed for 40% of the substances, mainly those of no to low toxicity. Calculating the theoretical animal numbers needed in several scenarios supported these results. The additional analysis, considering some physicochemical data (solubility, molecular weight, log POW), substance class and mode of action, revealed no specific applicability domains. In summary, the use of the 3T3 NRU cytotoxicity data alone did not sufficiently contribute to refinement and reduction in the acute oral toxicity testing of the substance portfolio tested routinely in our laboratory., (2011 FRAME.)

Data on eye irritation are generally needed for the hazard identification of chemicals. As the Bovine Corneal Opacity and Permeability (BCOP) test has been accepted by many regulatory agencies for the identification of corrosive and severe ocular irritants since September 2009 (OECD Test Guideline 437, TG 437), we evaluated this alternative method for routine testing at BASF. We demonstrated our technical proficiency by testing the reference standards recommended in TG 437, and 21 additional materials with published BCOP and in vivo data. Our results matched the published in vitro data very well, but with some intentionally selected false negatives (FNs) and false positives (FPs), the concordance was 77% (24/31), with FN and FP rates of 20% (2/10) and 24% (5/21), respectively. In addition, we tested 21 in-house materials, demonstrating the utility of the BCOP assay for our own test material panel. Histopathological assessment of the corneas by light microscopy was also conducted, as this was suggested as a means of improving the identification of FNs. The histopathology corrected the classification of some FNs, but also increased the number of FPs. Parallel to the test method evaluation, we compared three new opacitometer models with the current standard device. We recommend the use of an opacitometer developed in our BASF laboratory, which has certified components and electronic data storage, resulting in what we consider to be excellent sensitivity, stability and reproducibility., (2011 FRAME.)

Data on eye irritation are generally needed for the hazard identification of chemicals. For the routine testing of a broad variety of chemicals and formulations, we have used the Hen's Egg Test-Chorioallantoic Membrane (HET-CAM) method. In the course of a tiered-testing strategy, and due to the lack of global regulatory acceptance of the HET-CAM method, we have also performed the Rabbit Eye Irritation test according to OECD Test Guideline 405. Of the 145 substances tested, 76% were classified as non-irritant/mild irritant and 13% were identified as irritant in vivo, according to the EU classification system (61% and 28%, respectively, with the GHS classification). The remaining 11% were severe irritants in vivo, based on the irreversibility of the effects and not due to sufficiently high irritation scores in the three days following application. The retrospective analysis revealed that the overall accuracy of the HET-CAM assay was 65% and the overall rates of false-negatives (FN) and false-positives (FP) were 50% and 33%, respectively. The HET-CAM assay was sufficiently specific (few FP) for water-soluble substances, but failed to identify nearly all the severe irritants within this group. In contrast, it was highly sensitive (no FN) for non-soluble and oil-soluble substances, but the specificity for this group was rather low. Therefore, we conclude that the HET-CAM assay is not useful in our tiered-testing strategy for eye irritation testing. However, for water-insoluble substances, it might be applicable in combination with another in vitro method, provided that regulatory acceptance is gained., (2010 FRAME.)