The Road to Virtual Control Groups and the Importance of Proper Body Weight Selection

Virtual control groups (VCGs) created from historical control data (HCD) can reduce the number of concurrent control group animals needed in regulatory toxicity studies by up to 25%. This study investigates the performance of VCGs on statistical outcomes of body weight development between treatment and control groups in legacy studies. The objective is to reproduce the statistical outcomes of 28-day sub-chronic studies (legacy studies) after replacing the concurrent control group with virtual ones. In rodent toxicity studies initial body weight is used as surrogate for the age of animals. For the assessment of VCG-sampling methods three different approaches are explored: (i) sampling VCGs from the entire HCD ignoring initial body weight information of the legacy study, (ii) sampling from HCD matching the legacy study's initial body weights, and (iii) sampling from HCD with assigned statistical weights derived from legacy study initial body weight information. It is shown that the ability to reproduce statistical outcomes by virtual controls is mainly determined by the congruence between the legacy study and the HCD weight distribution: regardless of the chosen approach, the ability to reproduce statistical outcomes was well for VCGs when the legacy study's initial-body-weight distribution was similar to the HCD's. When the initial body weight range of the legacy study was at the extreme ends of the HCD's distribution, the weighted-sampling approach was superior. This article highlights the importance of proper HCD-matching by the legacy study's initial body weight and discusses required conditions to accurately reproduce body weight development.


1.2
Selection of legacy studies All animal studies were conducted internally and recorded using Pristima's laboratory information management system (LIMS) 1 .Harmonization of the study data was performed according to the standard for exchange of nonclinical data (SEND)2 controlled terminology for enabling data analysis using Amazon Warehouse System (AWS) S3 solution.The selection criteria for the legacy studies used for the qualification procedure were based on the following study design parameters: -Dosing duration of 4 weeks.
-Studies were initiated between 2017 and 2021.
-A control group was used along with three dose groups: low dose (LD), mid dose (MD), and high dose (HD).
-Studies were conducted in the laboratory of Bayer Pharmaceuticals AG, Wuppertal, Germany.
-Study was performed with Wistar HAN rats.
-Route of administration was oral by gavage.
-Animals were supplied by the same breeder.
-Food and water supply during the study was ad libitum.
-Animals were housed in group cages with 2-3 animals per cage.
-Body weight was measured on day 1-28.
Of a total of 1602 rat studies, 14 studies fulfilled all criteria.The attrition of studies is illustrated in Figure S1.

2.1
Mean body weight gain of selected studies Table S2 shows age at study initiation, initial body weight, and mean weight gain of male and female animals for Study-01 to Study-14.The selected studies are sorted by ascending initial body weight of male control group animals.For calculation of the mean weight gain across all studies, body weight data from days 22-26 were removed to avoid bias due to urine measurements frequently taken during that time.Of note is the difference in mean initial body weights between males and females (e.g., 16 g in Study-01 and 72 g in .The subsequent figures show the mean body weight per study over time (Fig. S2 for males and Fig. S3 for females).Four studies are highlighted where urine measurements were taken for males resulting in substantial transient body weight declines.Historical control data should be examined for such transient declines, and data should be flagged accordingly.
Tab. S2: Age, mean initial body weight ± SD, and mean weight gain ± SD with relative weight gain [%] ± SD of male and female concurrent control group (CCG) animals for Study-01 to Study-14 Mean weight gain was cleaned of days on which urine measurements were usually taken, i.e., day 22 to 26 were excluded for calculating weight gains.

Mean body weight gain of legacy study control groups compared to mean body weight gain of historical control data
The following 28 figures (Fig. S4-S31) show the mean body weight gain of the legacy studies Study-01 to Study-14 of concurrent control group animals, first for males (Fig. S4-S17), then for females (Fig. S18-S31).The solid lines represent the mean body weight with respect to the study day and the dashed lines show the 2× standard deviation area.Of note is the mean weight gain of male animals in Study-01 in males (Fig. S4).The mean initial body weight (on day 1) is very low compared to the historical control data, while the weight gain of the animals is high compared to the weight gain of the historical control data animals.Generating virtual control groups and replacing the concurrent controls therefore resulted in poor reproducibility of subsequent statistical outcomes.Only the body weight values of the last week (day 22 and onwards) were reproduced well by historical control data.As a result, statistical outcomes were reproduced only in 15% of all VCG iterations.

Fig. S6: Mean body weight growth of male control group animals of Study-03 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S7: Mean body weight growth of male control group animals of Study-04 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S8: Mean body weight growth of male control group animals of Study-05 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S9: Mean body weight growth of male control group animals of Study-06 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S10: Mean body weight growth of male control group animals of Study-07 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S11: Mean body weight growth of male control group animals of Study-08 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S24: Mean body weight growth of female control group animals of Study-07 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S25: Mean body weight growth of female control group animals of Study-08 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S26: Mean body weight growth of female control group animals of Study-09 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S27: Mean body weight growth of female control group animals of Study-10 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S28: Mean body weight growth of female control group animals of Study-11 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S29: Mean body weight growth of female control group animals of Study-12 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S30: Mean body weight growth of female control group animals of Study-13 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S31: Mean body weight growth of female control group animals of Study-14 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28
The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

HCD histograms, exemplary VCG selection, and initial body weight box plots of legacy studies
In the following 28 figures, initial body weights (body weights measured on day 1 prior to the first substance application) of historical control data (HCD) are displayed as histograms in comparison to the initial body weight of the legacy study.Objective is to replace the concurrent control with a virtual control.on the location parameters of the legacy study's groups, the VCG is either close to or far away from the concurrent control.The latter would result in poor statistical reproducibility of the body weight differences between control and dose groups.The studies (Study-01 to Study-14) are sorted by increasing initial body weight of male animals, with Study-01 having the lightest animals (compared to the HCD) and Study-14 having the heaviest (see Fig. S32-S45).Note, that the same study nomenclature has been used for females as well, which however is not sorted by increasing initial body weight of females (Fig. S46-S59).
The figures share the same format: On the left, the histogram shows the probability density distribution of historical control data (HCD) used to generate virtual control groups (VCGs).Adjacent to the histogram, box plots show one example iteration of a VCG in red, the legacy study's concurrent control group (CCG) in grey, and dose groups (low dose (LD), mid dose (MD), and high dose (HD)) in white.The black dashed lines represent the ranges (min.and max.body weight) of the dose groups.In Panel A, VCG was sampled from HCD not matched to legacy study weight distribution.Panel B shows VCG sampled from HCD matched to legacy study dose groups' initial body weight.Panel C shows VCG generated by weighted sampling with weights assigned to HCD to match the probability density of legacy study dose groups (black solid line on the histogram).

Fig
Fig. S1: Rat study selection and attrition

Fig. S2 :
Fig. S2: Mean body weight gain of control group animals for all 14 studies (males) Four studies are highlighted where a transient decrease in body weight was observed.These transient decreases were caused by urine measurements.

Fig. S3 :
Fig. S3: Mean body weight gain of control group animals for all 14 studies (females) The same studies are highlighted as for the males in Fig. S1, in which a transient decrease in body weight was observed.Only two studies show the same transient decrease in body weight.

Fig. S4 :
Fig. S4: Mean body weight growth of male control group animals of Study-01 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S5 :
Fig. S5: Mean body weight growth of male control group animals of Study-02 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S12 :
Fig. S12: Mean body weight growth of male control group animals of Study-09 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S13 :
Fig. S13: Mean body weight growth of male control group animals of Study-10 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S14 :
Fig. S14: Mean body weight growth of male control group animals of Study-11 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S15 :
Fig. S15: Mean body weight growth of male control group animals of Study-12 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S16 :
Fig. S16: Mean body weight growth of male control group animals of Study-13 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S17 :
Fig. S17: Mean body weight growth of male control group animals of Study-14 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S18 :
Fig. S18: Mean body weight growth of female control group animals of Study-01 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S19 :
Fig. S19: Mean body weight growth of female control group animals of Study-02 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S20 :
Fig. S20: Mean body weight growth of female control group animals of Study-03 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S21 :
Fig. S21: Mean body weight growth of female control group animals of Study-04 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S22 :
Fig. S22: Mean body weight growth of female control group animals of Study-05 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.

Fig. S23 :
Fig. S23: Mean body weight growth of female control group animals of Study-06 (orange line) compared to the mean growth of historical data control groups (grey line) from study day 1-28 The dashed lines with the painted area represent the 2×SD range of the legacy study and historical control data respectively.