Potential issues associated with the introduction of virtual control groups into non-clinical toxicology studies

Abstract

In recent years, introducing virtual control groups (VCGs) into toxicology studies is increasingly discussed because of the 3Rs and non-human primate (NHP) supply issues. Evaluating toxicology study results using historical control data is not new; however, introducing a VCG means replacing the concurrent control group in a toxicology study with a VCG, thereby reducing the number of animals used by approximately 30%. While it may be possible to conduct a toxicology study of a developmental compound in which the concurrent control group is replaced with a VCG, the scientific appropriateness of introducing a VCG and its regulatory acceptability needs to be considered. Therefore, we identified the following five issues that may arise when implementing a VCG: 1) regulatory requirements, 2) common issues when introducing a VCG, 3) issues related to histopathological examinations when introducing a VCG, 4) statistical analysis when introducing a VCG, and 5) facility monitor (sentinel) animals. Current regulatory guidelines require a concurrent control group for a pivotal toxicology study, whose results, if do not meet the requirements of these guidelines, cannot be used for new drug approval applications. Even if the use of VCGs is justified from animal welfare and scientific points of view, it is critical that the industry work with health authorities to ensure that data from these studies are accepted. The Japan Pharmaceutical Manufacturers Association will continue to hold necessary discussions with key stakeholders to accelerate efficient and effective new drug development pertaining to the use of VCGs.

Highlight

The Japan Pharmaceutical Manufacturers Association (JPMA) Taskforce Team has continuously discussed concerns about introducing virtual control groups (VCGs) into non-clinical toxicology studies from different perspectives at team meetings. To ensure that the issues identified in this study ultimately lead to efficient and effective new drug development, the intent was to expand discussions involving key stakeholders, including regulatory authorities.

Introduction

Animal studies are a critical component of drug development. However, efforts are being made to reduce the number of laboratory animals used. Over 50 years ago, the principles of the 3Rs (Replacement, Reduction, and Refinement) were developed to promote ethical animal research [1]. In December 2022, a U.S. law was enacted, which eliminated the requirement that drugs under development be tested in animals prior to being administered to humans in clinical trials [2]. Although this new law was groundbreaking, it did not ban animal experimentation. In reality, there is no guarantee that drug development is possible without animal experimentation, and there is still a strong belief that nothing will change. It is expected that, for the foreseeable future, animal experiments will continue to be performed to support the development of new drugs using rodent (mice and rats) and non-rodent (dogs and non-human primates [NHPs]) animal species.

NHPs have been widely used as non-rodent species in nonclinical toxicology studies [3,4,5]. One reason for using NHPs instead of other large animals such as dogs is that NHPs are genetically similar to humans. Furthermore, when evaluating certain classes of drug candidates, such as biologics, NHPs are considered the only pharmacologically appropriate species. However, in February 2022, the US Food and Drug Administration (FDA) issued guidance restricting the use of NHPs in toxicology studies [6]. The primary reason for this guidance was to reserve NHPs for testing investigational COVID-19 treatments and vaccine candidates as the supply of NHPs was affected by the COVID-19 pandemic. Reducing the number of NHPs used is of great concern worldwide because it affects the feasibility of non-clinical pharmacological and toxicological studies, and has developed into a discussion involving various consortiums and regulatory authorities [7, 8]. This guidance was withdrawn in 2023 with the expiration of the public health emergency [9]. However, issues remain regarding the stable supply of NHPs and the impact of a shortage of NHPs on new drug development has not been completely resolved.

Efforts to reduce the number of animals, not only NHPs, but also other species, such as rodents, are required in consideration of the 3Rs mentioned above. Although this is not a complete solution, the concept of virtual control groups (VCGs) was introduced by Steger-Hartmann et al. in 2020 [10] to partially contribute to animal reductions. According to their study, VCGs were constructed from control-group animal data from previously conducted studies that have been accumulated in a repository. Examples of conditions that are considered to be used for selection criteria in selecting animals for VCGs by their definition may include, but are not limited to, animal species/strain, sex, breeder, and age, although this depends on the evaluation endpoints. By applying a VCG created from historical control data, approximately 30% of the total number of animals could be reduced in repeated-dose toxicology studies consisting of a control group and three dose groups. Therefore, topics on VCGs, including, but not limited to, how to construct one, as well as potential issues, are now being discussed at conferences [11, 12] and international consortia [13,14,15]. The VCG concept was also discussed as a non-clinical topic in PHUSE CSS 2023 [16]. In addition, when the Japan Pharmaceutical Manufacturers Association (JPMA) presented Taskforce Team activities at the 2022 Fall Clinical Data Interchange Standards Consortium (CDISC) face-to-face meeting [17], questions from participants focused on the topic of VCGs, which demonstrated a high level of global interest.

The ultimate goal of pharmaceutical companies is the regulatory acceptance of toxicology studies applying VCGs while achieving animal population reductions with scientific validity. However, major issues need to be resolved before VCGs are implemented. For example, can the effect of the compound under development be appropriately evaluated scientifically, even if the concurrent control group in the study is replaced with a VCG? Does the application of VCG pose any problems from a statistical perspective? Will regulatory agencies accept studies that include no real control animals? We believe that discussing each of these issues carefully is important. Therefore, the JPMA Taskforce Team, which deals with issues on safety evaluation in the Non-Clinical Evaluation Expert Committee, Drug Evaluation Committee, and JPMA, identified various issues that may arise when implementing VCGs and discussed these issues in detail in this paper.

In this study, VCG refers to a control group virtually created based on study data collected and accumulated in the past. The contents of the VCG are similar to those of the historical control data in a test facility; however, the VCG is a subset of the historical control data. To generate a VCG from historical control data, the required number of animals was randomly selected from animals that matched conditions such as species, strain, sex, age, vehicle, and route of administration. Replacing the control group with existing data has already been partially implemented in the field of clinical trials and is being applied to trials in which it is difficult to establish a control group owing to recruitment issues such as rare diseases or ethical considerations [18, 19].

Materials and Methods

Concerns regarding the introduction of VCGs into non-clinical toxicology studies

The JPMA Taskforce Team on safety evaluation systems (including VCG-related topics) consists of 16 members with expertise in general toxicology, pathology, reproductive and developmental toxicology, genotoxicity, immunotoxicology, statistics, and regulatory agencies. The Taskforce Team discussed concerns about introducing VCGs to non-clinical toxicology studies from different perspectives at FY2022 to 2023 team meetings. The remainder of this paper is organized as follows.

1) Regulatory requirements: description of the creation of control groups and dose levels in guidelines (confirmation of currently valid guidelines).

2) Common issues when introducing VCGs: filtering conditions (conditions that should or may be matched with treated animals).

3) Issues related to histopathological examinations when introducing VCGs (lot-to-lot differences in normal tissue images).

4) Statistical analysis when introducing VCGs (differences in the importance of statistical analysis depending on study type).

5) Facility monitor (sentinel) animals (necessity of monitor animals).

Guidelines examined

The guidelines in Table 1 were examined for the creation of the control groups.

Table 1. Guidelines examined

Organization	Guideline	Reference
ICH	S4A guideline “Duration of chronic toxicity testing in animals (rodent and non rodent toxicity testing)”	[20]
OECD	TG 407 “Repeated dose 28-day oral toxicity study in rodents”	[21]
	TG 408 “Repeated dose 90-day oral toxicity study in rodents”	[22]
	TG 409 “Repeated dose 90-day oral toxicity study in non-rodents”	[23]
	TG 452 “Chronic toxicity studies”	[24]
EMA	Guideline on repeated dose toxicity	[25]
MHLW	The partial revision of the guidelines for repeated dose toxicity studies (Notification No. 655)	[26]

EMA: the European Medicines Agency; ICH: the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use; MHLW: the Japanese Ministry of Health, Labour and Welfare; OECD: the Organisation for Economic Co-operation and Development.

Investigation of the distribution of test values in the control group in toxicology studies

To investigate the distribution of test values in the control group in toxicology studies, a series of histograms of the distribution of test values in the control groups of toxicology studies were generated using R version 4.1.0 and R Studio version 1.4.1717. The results of 10 repeated trials, in which 10 samples were randomly selected from a normally distributed population, are shown.

Results

Regulatory requirements: description of the creation of control groups and dose levels in guidelines

The ICH S4A guideline “Duration of chronic toxicity testing in animals (rodent and non rodent toxicity testing)” [20] specifies the duration of treatment for chronic toxicology studies in rodents and non-rodents (6 and 9 mo, respectively), along with a discussion of the duration of treatment in non-rodents (6 vs. 12 mo). However, details regarding the design of the toxicology studies are not provided in the ICH guidelines. In contrast, the OECD repeated-dose toxicology study guidelines (repeated dose 28-day oral toxicity study in rodents: TG 407 [21]; repeated dose 90-day oral toxicity study in rodents: TG 408 [22]; repeated dose 90-day oral toxicity study in non-rodents: TG 409 [23]; chronic toxicity studies: TG 452 [24]) state that a control group should be used and the vehicle used to administer the test substance should be administered to control groups. The EMA guideline “guideline on repeated dose toxicity” [25] also states that “the treatment should include appropriate control group (s)”. A partial revision of the guidelines for repeated dose toxicity studies (Notification No. 655), issued by the MHLW (ex. the Ministry of Health and Welfare [MHW]) in 1999 [26], similarly states that “a control group should be set for which the test substance is not administered (vehicle administration)”. Thus, the current regulatory guidelines on repeated-dose toxicology studies (OECD, EMA, and MHLW), other than the ICH, describe the use of a control group to which a vehicle is administered.

Although this is not directly related to introducing VCGs to non-clinical toxicology studies, there have been some discussions in the Taskforce Team regarding dose levels from the perspective of reducing the number of animals. TG 408 of the OECD states “if a test at one dose level equivalent to at least 1,000 mg/kg body weight/day produces no observed adverse effects and if toxicity would not be expected based upon data from structurally-related compounds, then a full study using three dose levels may not be considered necessary”. TG 407 and 409 have similar statements. The dose levels section of the EMA guideline states “Dose levels may need to be adjusted, in unexpected toxic responses or lack of responses occurs during the study”. This means that there are possibilities of both increasing or decreasing the three dose levels, suggesting that flexibility should be considered depending on the situation. The Japanese MHLW guidelines do not refer to dosing flexibility.

In some cases, three doses may not be necessary. For example, highly selective biologics can be administered up to the maximum feasible dose to normal animals without any signs of toxicity. The EMA and OECD guidelines permit flexibility to adjust such cases, but conversely, a fourth dose-group may be set as necessary [21,22,23, 25]. The ability to flexibly increase or decrease the dose levels with scientific appropriateness would lead to a reduction in the total number of animals. Thus, there is content in the guideline documents that offers flexibility as long as the number of dose groups goes, and this sort of flexibility may, with proper justification and health authority agreement, be extended to involve the utilization of VCGs.

Common issues when introducing VCGs: filtering conditions

When trying to set a VCG in a toxicology study, if the conditions for animal selection are too strict, there may be insufficient data to use the VCG. Conversely, if the selection criteria are overly general, inappropriate data may be included. The Taskforce Team listed the possible filter conditions and used an asterisk (*) to indicate those that should be matched (i.e., critical ones) in Table 2. The basis for setting these conditions were set to minimize the variation in the normal values in the control group. Our idea is similar to that reported by Aulbach et al. [27] regarding factors that influence clinical pathology data when attempting to distinguish between test article- and non-test article-related effects in non-clinical safety studies; the listed factors overlapped with each other. According to this report, red blood cell parameters (erythrocyte count, hemoglobin concentration, and hematocrit percentage) and white blood cell composition (e.g., lymphocytes and neutrophils) change with aging in rats and dogs. In terms of the dosing route, there is a possibility that parameters such as AST and ALT may be slightly elevated owing to tissue damage caused by intravenous administration. Thus, these values cannot be combined across studies as control group values unless certain conditions are unified. Furthermore, some parameters may vary depending on the animal supplier and/or breeding conditions [28], including the frequency of pathological changes in corneal mineralization in rats [29]. However, cases of such variations have rarely been reported.

Table 2. Virtual control groups (VCG) filter conditions

• Species and strain*

• Sex*

• Age (with a certain range)*

• Test facility

• Breeder

• Origin of the animal

• Breeding conditions (including food and drinking water)

• Dosing route

• Vehicle

• Clinical laboratory equipment

• Possible genetic variation (e.g., conducted within 5 years for rodents)

*Conditions that the Taskforce Team think should be matched (i.e., critical ones).

Issues related to histopathological examinations when introducing VCGs

Technologies, such as the digitization of pathological slides and machine learning of normal/abnormal tissue images for artificial intelligence (AI) pathological diagnosis, are advancing globally [30]. The JPMA also launched another Taskforce Team on AI pathology (hereafter referred to as the AI pathology Taskforce Team) in FY2023. This AI pathology Taskforce Team initiated activities with the aim of building a specific system in cooperation with related organizations, rather than simply collecting information. Given these circumstances, it is highly likely that abnormal histopathological findings caused by compound administration can be diagnosed using a database of background control data. However, compared to the clinical pathology data of the control groups, histopathological data are image data and could be large in size, and a one-slide image may contain both normal and different types of abnormal regions. Furthermore, dyeability may differ depending on the facility, making it difficult to handle the data.

Upon discussion within the Taskforce Team, it was noted that pathologists maintained that reading the slides of control animals within a particular study is valuable. This is especially true for small animals, as there may be lot-to-lot differences in normal tissue images in the control group, such as frequency of spontaneous corneal and lens opacity in rats [31, 32]. At this time, no conclusion can be drawn regarding this point, although pathologists are conducting histopathological analysis. However, advances in this field may occur with the creation of an AI pathological diagnosis system that considers lot-to-lot differences. If the issues mentioned above are resolved, the situation in which histopathological examination becomes the rate-limiting factor for introducing VCGs to non-clinical toxicology studies will be avoided.

Statistical analysis when introducing VCGs

It is assumed that multiple 4-week repeated-dose toxicology studies (at least three animals per sex per group for large animals and 10 animals per sex per group for small animals), including concurrent controls, even if repeated under exactly the same conditions, would likely result in different statistical outcomes each time. It is possible that the replacement of control groups with VCGs may also result in different statistical outcomes because the groups to be compared are different. Additionally, as the number of animals in the virtual control group increased, the number of endpoints that were statistically significant also tended to increase. Therefore, the statistical significance of these findings must be considered.

This is related to the interpretation of statistical analysis results in the evaluation of short-term repeated-dose toxicology study results. Although appropriate statistical methods are required as per the OECD and EMA guidelines [21,22,23,24,25], we believe that it is useful to consider all study data in addition to statistical significance when evaluating a study. A statistically significant difference suggests the effects of the test compound, and comprehensive conclusions should be drawn based on the results of other endpoints, including the histopathological examination. This is different from clinical trials or carcinogenicity among non-clinical studies, which emphasize statistical results. Conversely, even if there were no statistically significant differences, the findings were judged as toxicologically significant. Therefore, the objective of this study can be achieved if a VCG that can detect potential toxic findings is used with a reduced number of animals.

In a typical 4-week repeated-dose toxicology study in rodents, the number of animals per group is 10 per sex, as is the number of animals in the control group. As discussed in “Investigation on the distribution of test values in the control group in toxicology studies” described later, even clinical pathology data (e.g., glucose in this case) are normally distributed in the population, but in 10 samples, although the mean values were similar, the distributions differed depending on the trials. In addition, the standard deviations varied considerably depending on the study. If this statistically natural variation also occurs in toxicology studies, the statistically significant difference shown by the 10 control group values cannot be considered definitive. As mentioned above, toxicology assessments should be performed comprehensively with other parameters while referring to statistical differences as a component of the analysis.

Facility monitor (sentinel) animals

One of the roles of concurrent control groups in repeated-dose toxicology studies is to determine the influence of factors other than the effects of the test compound (e.g., breeding environment). The introduction of the VCG may affect the ability to identify these external factors. Particularly in long-term studies, it is possible to use “facility monitor animals” for this purpose in addition to study animals. The OECD TG 452 [24] states: “In chronic toxicity studies, additional groups of sentinel animals may also be included for the monitoring of disease status, if necessary, during the study”. The use of facility monitor animal is not always essential. However, the use of such animals in the animal room should be considered to eliminate suspicion of infectious diseases not caused by the administration of the test compound in long-term rearing, even if VCG is introduced in the future.

Investigation on the distribution of test values in the control group in toxicology studies

Some human clinical test results were normally distributed. According to Okubo [33], normally distributed parameters include serum sodium, calcium, chloride, inorganic phosphorus, total protein, albumin, cholinesterase, uric acid, and glucose.

Since we know from collective experience that serum sodium in rats has a very narrow distribution of values, we selected glucose as an example, which has a relatively wider distribution of values. A simplistic model was created with the assumption that blood glucose levels in rats were normally distributed, as in humans. The R programming language was used to generate 100,000 random numbers with a mean of 250 and a standard deviation of 30 (Fig. 1). Ten samples were then selected from these random numbers, which were repeated 10 times, and the distribution was examined (Fig. 2). Sampling of 10 was intended to mimic the number of animals per sex per group in short-term (e.g., 4-week) repeated dose toxicology studies in rodents, as mentioned above. Means and standard deviations for each trial were calculated (Table 3).

Fig. 1.

Distribution of 100,000 generated glucose values using R programming (mean 250, standard deviation 30).

Fig. 2.

Distribution of 10 samples collected from the population (10 trials).

Table 3. Mean and standard deviation (SD) values for 10 trials

Trial no.	1	2	3	4	5	6	7	8	9	10
Mean	258.6	254.4	243.3	244.5	248.6	231.9	256.6	253.9	251.7	243.1
SD	31.5	25.5	27.6	27.9	32.8	31.5	27.1	25.1	25.8	36.7

The R code used was as follows:

When n=10 sampling was repeated 10 times, the average value was generally similar to that of the population, but the standard deviation was different each time, and the distribution was biased such that it could not be said to be a normal distribution. There are simplified examination results using R programming, but we can obtain results similar to those of the simulation if we repeat this process of randomly extracting 10 samples from the accumulated glucose data (data not shown). The similarity between the created glucose model and the actual glucose data supports the use of VCGs. Despite the use of VCG, it should be noted that appropriate toxicology studies should take the historical control data of the facility, and the results of histopathological examinations should be considered.

Discussion

Kawaguchi et al. [12] retrospectively introduced VCG into a 4-week repeated-dose toxicology study in rats that had already been evaluated and examined. Candidate animals for VCG were selected from previous studies with matching conditions, such as study animals (e.g., species, strain, sex, and age), experimental conditions, and measurement equipment. Body weight was selected as the selection criterion for retrospective VCG analysis. Body weights at the time of grouping should be similar so that changes can be evaluated. That is, the VCG is set as within (a) the mean ± SD or (b) the mean ± 2SD relative to the original control group (i.e., concurrent control group). The number of animals in the VCG group was 10, which was the same as that in the compound treatment groups. A statistical significance test was performed between the original control group or VCG group and the treatment groups to detect the effects of the compound. Results of this statistical analysis in the evaluation of body weight changes indicated that when (a) mean ± SD was selected for the VCG, a statistically significant difference was shown, similar to when using concurrent controls as expected. Similar results were also obtained when (a) mean ± SD or (b) mean ± 2SD based on the body weight at grouping was selected for the VCG in evaluating hematology and blood chemistry tests. In other words, while some of the results when using the concurrent control were reproduced, some items showed new statistically significant differences when compared with the VCG. Conversely, some items lost statistical significance when compared with the VCG. However, these results should not completely preclude the use of VCGs in toxicology studies, as there could be variations between the actual control groups in the two different studies conducted in the same manner. The use of VCGs should continue to be explored, especially in certain types of studies, such as dose-range-finding or short-term repeated dose toxicology studies. However, some endpoints such as histopathological evaluations may have specific concerns.

Another important point that may be raised when discussing the introduction of VCGs is revisiting current evaluation methods for toxicology studies (especially short-term repeated-dose toxicology studies). Many toxicologists evaluate toxicology study results regarding their toxicological significance by referring to statistically significant differences in body weight changes and clinical pathology results after confirming related histopathological changes, and then comprehensively draw final conclusions. There is a concern that completely consistent statistical results may not be obtained by replacing a concurrent control group with VCG. However, this does not imply that the statistical power will change. This is because, as shown in the simulation of glucose values, if there are 10 animals of each sex per group, the distribution of values in the concurrent control group may be biased by chance (the mean value of Trial #6 is low). By performing a comprehensive evaluation, including histopathological changes, the impact of the introduction of VCGs should be minimized in terms of toxicological significance.

To address this point, Wright et al. [34] conducted a larger investigation by retrospectively analyzing the impact of replacing concurrent control groups with VCGs on the treatment relatedness of histopathological findings in the eTOX database. The results suggested that VCGs could be used most successfully when only incorporating historical control data that had strong similarity to concurrent control groups, that is, the same species, strain, sex, administration route, and vehicle. These results support our discussion (Table 2) and are helpful in the selection of the VCG filter conditions.

Specifically related to historical control data, Kluxen et al. [35] performed a very detailed review of statistical tests for the use of historical control data in the evaluation of toxicology study results. According to their study, statistical approaches can be used to set historical control limits or relevance thresholds based on the historical control data. Statistical approaches can be used to compare concurrent responses to derived relevance thresholds. If a value exceeds a relevance threshold, the response may be abnormal or considered biologically irrelevant. However, they stated that this should be considered as one piece of evidence contributing to, but not replacing, toxicological assessments.

General toxicology studies, particularly short-term repeated-dose toxicology studies, are discussed in this manuscript. However, the use of VCGs may also be applicable to other types of studies. It is necessary to consider the usefulness of VCGs in terms of evaluating endpoints, such as electrocardiogram (ECG) and respiratory data in safety pharmacology studies, which can be modeled in a standard format by implementing the latest CDISC Standard for Exchange of Nonclinical Data (SEND) implementation guide (IG) Version 3.1.1 [36]. The Taskforce Team believed that these possibilities require further discussion.

In the near future, the accumulation of such evidence will support the appropriate use of VCGs. However, if VCGs are to be used, acceptance by regulatory authorities is of utmost importance. Although this is not directly related to the introduction of VCGs in non-clinical toxicology studies, there have been some discussions regarding dose levels from the perspective of reducing the number of animals. As mentioned in the Results section “Regulatory requirements: description of the creation of control groups and dose levels in guidelines”, the currently effective regulatory guidelines on repeated dose toxicology studies (OECD, EMA, MHLW), other than the ICH, describe the use of a control group to which the vehicle is administered. In addition, as for the dose level, 3 doses are basically set, which can be adjusted flexibly in the EMA and OECD guidelines. As discussions on the introduction of VCGs progress and from the perspective of animal welfare, it will be important to update the guidelines accordingly. The active participation of regulatory authorities in VCG discussions is extremely valuable and strongly encouraged.

Assuming that VCGs are scientifically justified and regulatory authorities accept toxicology studies with study designs that include VCGs, sponsors of non-clinical studies may need to consider their interactions with CROs regarding whether to include a VCG in the study design. Consequently, CROs may be required to enrich their facilities’ historical control data under various conditions to create an IT environment that facilitates the use of VCGs. To achieve this, it may be useful to use data accumulated in the SEND format, which is a non-clinical CDISC standard, and extract the appropriate data. However, this approach needs to be discussed separately and extensively, as the SEND data may contain information on other sponsors (i.e., information on the original study).

Conclusions

In recent years, the concept of VCGs has spread rapidly, and is now being discussed globally. In this study, we identified issues that may arise when implementing VCG and explored them. The issues discussed in this paper, namely regulatory requirements, common issues when introducing a VCG, issues related to histopathological examination when introducing a VCG, statistical analysis when introducing a VCG, and facility monitor (sentinel) animals, are examples that will arise with the introduction of VCGs, and unexpected issues are anticipated. Additionally, the introduction of the VCG concept may impact the current evaluation methods for toxicology studies. Although the introduction of VCGs stems from a reduction in the number of animals from the viewpoint of the 3Rs and the problem of NHP supply, with the increasing utilization of data in standard formats, the VCG concept is even more likely to gain traction. Even if the use of VCGs is justified from animal welfare and scientific perspectives, it is critical that the industry works with health authorities to ensure that data from studies utilizing VCGs are accepted. The JPMA will continue to hold the necessary discussions with key stakeholders to accelerate efficient and effective new drug development pertaining to the use of VCGs. Furthermore, the JPMA would also be involved in discussions on the international harmonization of study designs implementing VCGs (e.g., how to choose appropriate animals for VCGs and the animal number of VCGs).

Conflict of Interest

The authors (Gen Sato, Mikio Nakajima, Kuniyoshi Sakai, Yuko Togashi, Masakatsu Yamamoto, Yuki Inoue, Takeshi Oshima, Tetsuyoshi Soh, Mayumi Watanabe, Izumi Matsumoto, Toshinobu Yamamoto, Takashi Tanaharu, Akio Kawakami, Keiko Motoyama, Kiyohiro Hashimoto, and Mutsumi Suzuki) are employees of companies (Eisai Co., Ltd., Asahi Kasei Pharma Corporation, ASKA Pharmaceutical Co., Ltd., EA Pharma Co., Ltd., MSD K.K., Otsuka Pharmaceutical Co., Ltd., Kyowa Kirin Co., Ltd., Shionogi & Co., Ltd., Daiichi Sankyo Co., Ltd., Sumitomo Pharma Co., Ltd., Mitsubishi Tanabe Pharma Corporation, Bristol-Myers Squibb K.K., Merck Biopharma Co., Ltd., Janssen Pharmaceutical K.K., Takeda Pharmaceutical Co. Limited, and Kyowa Kirin Co., Ltd.) who are funding the research.

Acknowledgements

We are grateful to the former Taskforce Team member Kyohei Nishimura (Shionogi & Co., Ltd.), who was replaced during the research period, for his valuable contributions to our research work. We would like to thank Shun Kawaguchi (Mediford Corporation) for his multiple discussions of VCGs based on their implementation using a retrospective approach. We would also like to thank all the reviewers of the Taskforce Team member companies for their thorough reviews and useful suggestions in the preparation of this manuscript. Furthermore, we wish to acknowledge Thomas Visalli (Eisai Inc.) for providing valuable feedback during the preparation of this manuscript and for improving its language quality.

References

• 1.   Gorzalczany,  S. B. and  Rodriguez Basso,  A. G. 2021. Strategies to apply 3Rs in preclinical testing. Pharmacol. Res. Perspect. 9: e00863.
• 2.   Wadman,  M. 2023. FDA no longer needs to require animal tests before human drug trials. Science Insider.
• 3.  EU: European Union 2009. Non-human primates in research and safety testing. https://ec.europa.eu/health/scientific_committees/opinions_layman/en/non-human-primates/index.htm [accessed August 23, 2023].
• 4.   Prior,  H.,  Haworth,  R.,  Labram,  B.,  Roberts,  R.,  Wolfreys,  A. and  Sewell,  F. 2020. Justification for species selection for pharmaceutical toxicity studies. Toxicol. Res. (Camb.) 9: 758–770.
• 5.   Monticello,  T. M.,  Jones,  T. W.,  Dambach,  D. M.,  Potter,  D. M.,  Bolt,  M. W.,  Liu,  M.,  Keller,  D. A.,  Hart,  T. K. and  Kadambi,  V. J. 2017. Current nonclinical testing paradigm enables safe entry to First-In-Human clinical trials: the IQ consortium nonclinical to clinical translational database. Toxicol. Appl. Pharmacol. 334: 100–109.
• 6.  FDA: US Food and Drug Administration. 2022. Nonclinical Considerations for Mitigating Nonhuman Primate Supply Constraints Arising from the COVID-19 Pandemic Guidance for Industry. Federal Register 87(37), 10373–10375. https://www.govinfo.gov/content/pkg/FR-2022-02-24/pdf/2022-03915.pdf [accessed August 23, 2023].
• 7.  NC3Rs. 2022. Webinar: Minimising non-human primate use in drug development. https://www.nc3rs.org.uk/events/minimising-NHP-use [accessed August 23, 2023].
• 8.   Ackley,  D.,  Birkebak,  J.,  Blumel,  J.,  Bourcier,  T.,  de Zafra,  C.,  Goodwin,  A.,  Halpern,  W.,  Herzyk,  D.,  Kronenberg,  S.,  Mauthe,  R.,  Shenton,  J.,  Shuey,  D. and  Wange,  R. L. 2023. FDA and industry collaboration: identifying opportunities to further reduce reliance on nonhuman primates for nonclinical safety evaluations. Regul. Toxicol. Pharmacol. 138: 105327.
• 9.   Brown,  P. C. and  Wange,  R. L. 2023. Considerations regarding the use of nonhuman primates in assessing safety endpoints for pharmaceuticals. Regul. Toxicol. Pharmacol. 143: 105449.
• 10.   Steger-Hartmann,  T.,  Kreuchwig,  A.,  Vaas,  L.,  Wichard,  J.,  Bringezu,  F.,  Amberg,  A.,  Muster,  W.,  Pognan,  F. and  Barber,  C. 2020. Introducing the concept of virtual control groups into preclinical toxicology testing. Altern. Anim. Exp. 37: 343–349.
• 11.   Andaya,  R. V.,  Pourmohamad,  T.,  Sullivan,  R.,  Hayes,  M. and  Anger,  L. 2023. Virtual Control Groups in Preclinical Safety: Decreasing Animal Use while Maintaining In Vivo Study Interpretability. The 62nd Annual Meeting of the Society of Toxicology.
• 12.   Kawaguchi,  S.,  Takahashi,  K. and  Yamamoto,  D. 2023. Introduction of Virtual Control Group in Nonclinical safety studies. The 50th Annual Meeting of the Japanese Society of Toxicology.
• 13.  eTRANSAFE. 2022. eTRANSAFE announces the launch of the “eTRANSAFE ViCoG Initiative” https://etransafe.eu/etransafe-announces-the-launch-of-the-etransafe-vicog-initiative/ [accessed August 23, 2023].
• 14.   Sanz,  F.,  Pognan,  F.,  Steger-Hartmann,  T.,  Díaz,  C.,  Asakura,  S.,  Amberg,  A.,  Bécourt-Lhote,  N.,  Blomberg,  N.,  Bosc,  N.,  Briggs,  K.,  Bringezu,  F.,  Brulle-Wohlhueter,  C.,  Brunak,  S.,  Bueters,  R.,  Callegaro,  G.,  Capella-Gutierrez,  S.,  Centeno,  E.,  Corvi,  J.,  Cronin,  M. T. D.,  Drew,  P.,  Duchateau-Nguyen,  G.,  Ecker,  G. F.,  Escher,  S.,  Felix,  E.,  Ferreiro,  M.,  Frericks,  M.,  Furlong,  L. I.,  Geiger,  R.,  George,  C.,  Grandits,  M.,  Ivanov-Draganov,  D.,  Kilgour-Christie,  J.,  Kiziloren,  T.,  Kors,  J. A.,  Koyama,  N.,  Kreuchwig,  A.,  Leach,  A. R.,  Mayer,  M. A.,  Monecke,  P.,  Muster,  W.,  Nakazawa,  C. M.,  Nicholson,  G.,  Parry,  R.,  Pastor,  M.,  Piñero,  J.,  Oberhauser,  N.,  Ramírez-Anguita,  J. M.,  Rodrigo,  A.,  Smajic,  A.,  Schaefer,  M.,  Schieferdecker,  S.,  Soininen,  I.,  Terricabras,  E.,  Trairatphisan,  P.,  Turner,  S. C.,  Valencia,  A.,  van de Water,  B.,  van der Lei,  J. L.,  van Mulligen,  E. M.,  Vock,  E. and  Wilkinson,  D. 2023. eTRANSAFE: data science to empower translational safety assessment. Nat. Rev. Drug Discov. 22: 605–606Comment.
• 15.  TransCelerate 2023. The use of virtual controls in preclinical research: promise and predicament. https://www.transceleratebiopharmainc.com/events/the-use-of-virtual-controls-in-preclinical-research-promise-and-predicament/ [accessed August 23, 2023].
• 16.  PHUSE CSS 2023, Nonclinical topics. 2023. https://www.phuse-events.org/attend/frontend/reg/tOtherPage.csp?pageID=15403&ef_sel_menu=1694&eventID=26 [accessed August 23, 2023].
• 17.   Sato,  G. 2022. Updates from the JPMA SEND team. 2022 Fall CDISC Face-to-Face Meeting.
• 18.   Nakatsuka,  Y. 2019. Utilization of clinical trial control groups in clinical development. Office of Pharmaceutical Industry Research News 58 https://www.jpma.or.jp/opir/news/058/pdf/pdf-13–01.pdf [accessed August 23, 2023] (in Japanese).
• 19.   Gökbuget,  N.,  Kelsh,  M.,  Chia,  V.,  Advani,  A.,  Bassan,  R.,  Dombret,  H.,  Doubek,  M.,  Fielding,  A. K.,  Giebel,  S.,  Haddad,  V.,  Hoelzer,  D.,  Holland,  C.,  Ifrah,  N.,  Katz,  A.,  Maniar,  T.,  Martinelli,  G.,  Morgades,  M.,  O’Brien,  S.,  Ribera,  J. M.,  Rowe,  J. M.,  Stein,  A.,  Topp,  M.,  Wadleigh,  M. and  Kantarjian,  H. 2016. Blinatumomab vs historical standard therapy of adult relapsed/refractory acute lymphoblastic leukemia. Blood Cancer J. 6: e473.
• 20.  ICH: The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use S4A: Duration of Chronic Toxicity Testing in Animals (Rodent and Non Rodent Toxicity Testing) https://database.ich.org/sites/default/files/S4_Guideline.pdf [accessed August 23, 2023].
• 21.  OECD: Organisation for Economic Co-operation and Development 2008. Test No. 407: Repeated Dose 28-day Oral Toxicity Study in Rodents. OECD Guidelines for the Testing of Chemicals, Section 4. OECD Publishing, Paris. https://www.oecd-ilibrary.org/environment/test-no-407-repeated-dose-28-day-oral-toxicity-study-in-rodents_9789264070684-en [accessed August 23, 2023].
• 22.  OECD 2018. Test No. 408: Repeated Dose 90-Day Oral Toxicity Study in Rodents. OECD Guidelines for the Testing of Chemicals, Section 4. OECD Publishing, Paris. https://www.oecd-ilibrary.org/environment/test-no-408-repeated-dose-90-day-oral-toxicity-study-in-rodents_9789264070707-en [accessed August 23, 2023].
• 23.  OECD 1998. Test No. 409: Repeated Dose 90-Day Oral Toxicity Study in Non-Rodents. OECD Guidelines for the Testing of Chemicals, Section 4. OECD Publishing, Paris. https://www.oecd-ilibrary.org/environment/test-no-409-repeated-dose-90-day-oral-toxicity-study-in-non-rodents_9789264070721-en [accessed August 23, 2023].
• 24.  OECD 2018. Test No. 452: Chronic Toxicity Studies. OECD Guidelines for the Testing of Chemicals, Section 4. OECD Publishing, Paris. https://www.oecd-ilibrary.org/environment/test-no-452-chronic-toxicity-studies_9789264071209-en [accessed August 23, 2023].
• 25.  EMA: European Medicines Agency 2010. Guideline on repeated dose toxicity. CPMP/SWP/1042/99 Rev 1 Corr. https://www.ema.europa.eu/en/repeated-dose-toxicity-scientific-guideline [accessed August 23, 2023].
• 26.  MHLW: Japanese Ministry of Health, Labour and Welfare. 1999. Partial Revision of the Guidelines for Repeated Dose Toxicity Studies (Notification No. 655) https://www.pmda.go.jp/files/000156632.pdf [accessed August 23, 2023] (in Japanese).
• 27.   Aulbach,  A.,  Provencher,  A. and  Tripathi,  N. 2017. Influence of study design variables on clinical pathology data. Toxicol. Pathol. 45: 288–295.
• 28.   Shimomura,  K. 2018. 5.5 Developmental and reproductive toxicology in Toxicology, 3rd ed. Asakura Publishing Co., Ltd., Tokyo, Japan. pp.156–165 (in Japanese).
• 29.   Hashimoto,  S.,  Doi,  T.,  Wako,  Y.,  Sato,  J.,  Wada,  S. and  Tsuchitani,  M. 2013. Corneal mineralization in wistar hannover rats. J. Toxicol. Pathol. 26: 275–281.
• 30.  BIGPICTURE 2023. https://bigpicture.eu/ [accessed August 17, 2023].
• 31.   Onodera,  H.,  Sasaki,  S.,  Otake,  S.,  Tomohiro,  M.,  Shibuya,  K. and  Nomura,  M. 2013. Scientific viewpoints in ocular toxicity assessment: departure from conventional practice. Anim Eye Res. 32: 3–13(in Japanese).
• 32.   Onodera,  H.,  Sasaki,  S.,  Otake,  S.,  Tomohiro,  M.,  Shibuya,  K. and  Nomura,  M. 2015. General considerations in ocular toxicity risk assessment from the toxicologists’ viewpoints. J. Toxicol. Sci. 40: 295–307.
• 33.   Okubo,  A. 1993. 2. How to read normal and abnormal values. J Jpn Soc Int Med 82: 485–489(in Japanese).
• 34.   Wright,  P. S. R.,  Smith,  G. F.,  Briggs,  K. A.,  Thomas,  R.,  Maglennon,  G.,  Mikulskis,  P.,  Chapman,  M.,  Greene,  N.,  Phillips,  B. U. and  Bender,  A. 2023. Retrospective analysis of the potential use of virtual control groups in preclinical toxicity assessment using the eTOX database. Regul. Toxicol. Pharmacol. 138: 105309.
• 35.   Kluxen,  F. M.,  Weber,  K.,  Strupp,  C.,  Jensen,  S. M.,  Hothorn,  L. A.,  Garcin,  J. C. and  Hofmann,  T. 2021. Using historical control data in bioassays for regulatory toxicology. Regul. Toxicol. Pharmacol. 125: 105024.
• 36.  CDISC 2021. Standard for Exchange of Nonclinical Data Implementation Guide (SENDIG): Nonclinical Studies. Version 3.1.1 (Final). https://www.cdisc.org/standards/foundational/send/sendig-v3-1-1 [accessed November 16, 2023].