Summary of the meeting:
The afternoon workshop on the 6th of April about “Adversity in safety studies: How to set a NOAEL?” is hosted by the NVT risk assessment section together with the section Toxicological Pathology. We were welcomed to AkzoNobel by Kim Doornebosch. Theo Vermeire, chair of the NVT risk assessment section, introduced the program and the speakers.
This meeting addresses the contemporary view on “Adversity in animal studies” and its consequences for deriving a NOAEL. The reason for this meeting is the proposal by the Society of Toxicological Pathology (STP) to denote a test article effect as being “adverse” if it is harmful to the test animal resulting in impaired (organ) function/wellbeing of the test species. As recently supported by the Society of Toxicological Pathology (STP) in one of their recent publications. To underline the actuality of this topic, the EFSA recently published a Guidance of Biological Relevance which is open for comments (hyperlink).
In the first part of the meeting, the STP position on “adversity” was highlighted and an alternative view on “adverse” was presented. In a concluding presentation a comparison was made of the potential consequences of the new concept of adversity on both the NOAEL and Bench Mark Dose approach.
In the 2nd part of the meeting, the new concept of adversity has been dealt with in an interactive workshop in which the attendees discussed in small groups the setting of a NOAEL for a number of case studies.
The first speaker, Eveline de Rijk (CRL), gave an introduction to the papers written by the American (STP) as well as European Society on Pathology (ESTP) on what is adverse in animal safety studies?
The STP publication provides guidance on the declaration of adversity of an effect and how this is translated into a NOAEL. In addition, guidance is given on how to report and communicate a NOAEL and adversity of effects. The STP publication is a procedural publication based on historical issues and misinterpretations. There are many definitions of what is considered an adverse effect; statistical and biological differences are of importance, or effects on well-being, or highest level of exposure that does not lead to toxicity.
The STP publication recommends considering adversity as an effect indicating harm to the animal. In a study it should be considered if:
1. An effect is related to test item?
2. If yes: is the effect harmful?
A. If no: provide an unambiguous justification for non-adverse interpretation
B. If yes: incorporate in the decision of defining NOAEL
It is very important to keep in mind that adversity should be applied only to the test species and under the conditions of the study. Toxic effects should be assessed on their own merits. For example, effects considered as adaptive might actually be adverse. If there are sub reports, such as pathology reports, these should be consistent with the overall study report. For an overall hazard assessment NOAELs should be communicated in an overview document based on multiple studies including explanatory text and not just numbers. All the experts should be involved in assigning NOAELs and all data should be use.
In the ESTP publication a more practical approach is presented based on examples. The workshop was organized as follow up on the paper. In addition to the STP paper, the ESTP paper mentions that the NOAEL should take into consideration if there is an impairment of the capacity to respond to an additional challenge. The paper also defines at what levels adverse effects should be reported: pathological nature of effects (mortality, malignant neoplasms, neuronal necrosis), lesions severity (and incidence), exacerbation of spontaneous findings (in case of clear dose response, test item related), direct versus indirect effects (adversity depends on effect itself like severe depression of body weight), adaptive responses, MOA-exaggerated pharmacology, reversibility (which does not automatically imply a non-adverse effect), extrapolation to longer/higher exposure and translation to humans in overview documents.
In the end there should be no difference in the conclusion on adversity in studies for pharmaceutical, biocides or industrial chemicals. This should be independent of the application of a substance.
Next Frieke Kuper (TNO) presented a model that can be used to look at adversity in another way. In the current practice any negative impact on ‘good health’ is based on expert judgement. These experts generally perform evaluations per group of parameters and per expert (pathologist, toxicologist etc.). In the ‘other way’ model the data are interpreted at an individual animal level. Four studies have been performed with this model. Correlations and patterns are visualized. A principal component analysis has been performed on the dataset collected in a study. The model tries to identify for a 2D space in 26 dimensional space and calculates the best fit and at the same time tries to separate the animals as much as possible. In this way you can identify which parameters contribute the most to the effect. You also assess which animals stand out of the population. Keep a close eye on any outliers, they tell a lot also about variation in a population. In complex systems you can compare homeostasis with hills and valleys. It takes a lot of energy to reach another state (resistance up the hill); once you are over the hill there is high risk of transition/low resilience. The model helps to recognize recognition patterns. This benefits identification important parameters and group (population) responses. More research and evaluation is needed.
The third speaker is Bas Bokkers (RIVM). Bas gave an overview of adversity of effects, consequences for deriving a point of departure. He started with presenting results of a personal small survey at RIVM he performed among his colleagues on how risk assessors identify NOAEL. He identified a number of different approaches.
When statistical analysis is applied you look at the null hypothesis versus the alternative hypothesis; what is the borderline between zero effect and non-zero effect? What does non- zero mean? This distinction has no practical meaning. A frequent error is the interpretation of non-significant as no effect. The outcome should be that it is inconclusive and the NOAEL should not be based on statistical significance.
The benchmark dose (BMD) approach is a more sophisticated approach to derive a NOAEL. Using this analysis you can define a “small” effect size called the benchmark response (BMR). What is the acceptable additional risk for continuous data (individual level) and for quantal data (population level)? A pragmatic approach is to use a BMR of 5% for continuous and 10% for quantal data as a default. In the US a standard deviation approach is also used (to make sure you are outside the sampling error range). Drawback of this approach is that this highly depends on study and lab quality.
In a recent study (Slob, 2016) a new theory is presented on effect size, involving meta-analysis on a lot of toxicological endpoints and compares severities between endpoints.
The last part of the afternoon Anja Slikkerveer and Liesbeth Heijink (Astellas Pharma) facilitated an interactive workshop on adversity cases (from the STP workshop and EMA workshop in 2016). Within this workshop the audience was divided in small groups to debate on a total of 4 cases. The questions were simple: what is the NOAEL? And what are key points supporting your assignment of the NOAEL? However, during the discussion it became clear that the studies give sufficient reason for debate. It was concluded that it is realistic that people disagree on a NOAEL for a study. It is more important to be transparent about the rationale behind determination of the NOAEL.
Thanks to the contributors and the enthusiastic participants we look back on an interactive afternoon and good discussions.