Ethical Perspectives Influence the Formation of Researchers’ Perceptions of Experimental Conduct in a Preclinical Animal Experimentation Training Program

Abstract

Animal experimentation is conducted under regulatory and institutional frameworks that require ethical standards, safety management, scientific validity, and compliance with relevant guidelines. Accordingly, institutions that conduct preclinical research are responsible for providing structured training programs that prepare researchers to conduct animal experiments appropriately. However, empirical evaluations of such training programs and their outcomes remain limited. This study evaluated a comprehensive preclinical animal experimentation training program covering ethics, regulatory requirements, safety, and technical knowledge, using survey data collected from researchers who completed the program at Samsung Medical Center. The Kirkpatrick training evaluation model was applied as an analytical framework, and partial least squares structural equation modeling (PLS-SEM) was used to explore researchers’ perceptions of animal experimentation conduct. The results demonstrated that the training program was associated with improved researcher perceptions, supported by acceptable reliability and validity of the measurement instrument. In particular, perceptions related to animal research ethics were significantly associated with experimental performance. An exploratory structural model was developed to identify key determinants of researcher perceptions and to support future refinement of training evaluation approaches. These findings suggest that well-designed training programs can foster ethically grounded and practically oriented attitudes among researchers, thereby strengthening accountability in animal experimentation and supporting the implementation of the Replacement, Reduction, and Refinement (3Rs) principles. This study provides evidence-based insights for institutions, training coordinators, and oversight bodies seeking to improve laboratory animal research training programs.

Introduction

Animal experimentation is a highly regulated and technically demanding activity that requires strict adherence to ethical standards, animal welfare principles, safety procedures, and legal requirements. There is growing emphasis on animal researchers’ responsibility for animal welfare, Replacement, Reduction, and Refinement (3Rs) principles, and the application of alternative methods. As a result, many jurisdictions, including the United States, European Union member states, Australia, South Korea, and others, mandate formal education and training for researchers before their participation in animal experiments. In South Korea, animal experimentation is regulated under the Laboratory Animal Act, which aims to enhance the ethical standards and scientific reliability of animal research and requires training for personnel involved in animal experimentation.¹

Institutional animal experimentation training programs play a critical role in ensuring regulatory compliance, promoting ethical research conduct, and supporting the successful execution of animal-based studies. Well-designed training programs can reduce unnecessary repetition of experiments, improve technical proficiency, enhance perception of humane endpoints, and ultimately contribute to the responsible use of laboratory animals. From an institutional perspective, such programs are also essential components of animal care and use oversight systems including IACUC operations.

Despite their importance, relatively few studies have systematically evaluated the outcomes of animal experimentation training programs, particularly with respect to how training influences researchers’ perceptions and attitudes toward animal research practices. Although prior studies have examined training effectiveness in various contexts,^2–6 many have relied on descriptive analyses and lack structured analytical approaches to understanding how specific training components influence researchers’ perceptions and attitudes.

The purpose of this study was to evaluate researchers’ perceptions following participation in a Preclinical Animal Experimentation Training Program (PAETP) conducted at a large medical research institution. The PAETP is a comprehensive training program that covers the ethics, relevant legal and regulatory frameworks, safety management, and technical knowledge of animal experimentation.

This study sought to address two research questions.

• (1)How does participation in an animal experimentation training program influence researchers’ perceptions related to the conduct of animal experiments?
• (2)Which topics of the training program are most strongly associated with perceptions toward animal experimentation?

Survey data collected from researchers at the Samsung Medical Center (SMC) in South Korea were analyzed to explore relationships among training components and researcher perceptions. An exploratory analytical approach was employed to examine how perceptions related to ethics, safety, and experimental practice are interconnected within the context of institutional training.

The findings of this study highlight the central role of ethical training, particularly education related to the 3R principles and humane experimental endpoints, in shaping researchers’ perceptions toward animal experimentation. These results suggest that animal experimentation training programs can function not only as compliance mechanisms but also as meaningful educational interventions that promote responsible research conduct and reinforce researchers’ accountability for animal welfare.

Although this study was conducted at a single institution, the results provide practical insights applicable to other research institutions that offer animal experimentation training programs. The findings may inform the development, refinement, and evaluation of training curricula aimed at improving animal welfare practices and supporting high-quality animal research.

Ethical review.

No animal experiments were conducted in this study; therefore, approval from an IACUC was not required. Institutional Review Board (IRB) approval was obtained for the survey conducted in this research (SMC IRB No. 2025-12-020). All participants provided informed consent, and the survey was conducted anonymously.

Materials and Methods

Study design.

The content of the PAETP was developed in accordance with national training guidelines for animal experimentation researchers issued by the Ministry of Food and Drug Safety (MFDS), South Korea. The program covers key domains including animal research ethics, relevant legal and regulatory requirements, safety management, and technical procedures. Participants in this study were researchers affiliated with SMC, including members of the Sungkyunkwan University School of Medicine, who were preparing to conduct animal experiments.

Survey data were collected immediately before and after completion of the training program. To reduce potential response bias, the survey was conducted anonymously, and participants were assured that their responses would be used solely for research purposes. In addition, questionnaire items were carefully designed using clear and neutral wording to minimize the likelihood of socially desirable responses. Following data cleaning, descriptive statistical analyses were performed, and the validity and reliability of the measurement instrument were evaluated.

Given the limited prior research on outcomes of animal experimentation training programs, this study emphasized an exploratory analytical approach. Relationships among researchers’ perceptions were examined using partial least squares structural equation modeling (PLS-SEM), which is suitable for exploratory analyses involving complex latent constructs. Model evaluation included assessments of reliability, validity, and explanatory power to determine whether the proposed model provided meaningful insights into the effects of training on researcher perceptions. An overview of the study procedure is presented (Figure 1).

Figure 1.

Research Procedure.

Training program and participants.

In South Korea, the Ministry of Food and Drug Safety (MFDS) requires all researchers involved in animal experimentation to complete a designated animal experimentation training program. Individual research institutions, including SMC, are authorized to implement MFDS-aligned training programs, which are formally recognized as fulfilling this national training requirement.

The Preclinical Animal Experimentation Training Program (PAETP) implemented at SMC is structured around 4 core content domains:

• (1)Animal experimentation ethics, including the 3R principles and the role and IACUC
• (2)Structure and key provisions of relevant laws and regulations governing animal research
• (3)Safety management for researchers and facilities, including accident prevention
• (4)Fundamental knowledge of experimental techniques commonly used in animal research

Each domain consists of multiple detailed training components designed to support ethical, scientific, and safe conduct of animal experiments. Eligibility for participation in the PAETP is limited to researchers who plan to directly handle laboratory animals. Researchers conducting preclinical animal research are required to complete both a formal training session and an on-site laboratory orientation before initiating animal experiments.

The PAETP is delivered as a one-day, 5-hour training program and is conducted by veterinarians, senior research staff with doctoral degrees, and experienced personnel with 5-10 years of professional experience working in an accredited animal facility (PRC).

During the study period, the PAETP was administered 9 times, and a total of 97 participants completed the program. No duplicate participants were included in the analysis. In accordance with institutional policy, 13 participants completed the program as recertification trainees, reflecting the requirement that researchers renew their PAETP training every 3 years to maintain eligibility for animal experimentation. Both first-time trainees and recertification trainees were included in the study sample. All participants completed the same standardized PAETP during the study period, and the analyses focused on within-subject changes in perception before and after the training program.

Theoretical framework of the survey.

The Kirkpatrick model is a widely used evaluation framework for training programs, originally proposed in 1959.^7,8 This model categorizes training evaluation into 4 levels: level 1 (“reaction”), which assesses participants’ interest, needs, and immediate responses to training; level 2 (“learning”), which evaluates knowledge acquisition, skill development, and attitudinal change; level 3 (“behavior”), which examines behavioral changes in real-world practice; and level 4 (“results”), which focuses on productivity gains, performance outcomes, and organizational contributions. These 4 levels are conceptually interconnected, and the selection of appropriate evaluation levels depends on the research purpose and contextual constraints.^9–11

In this study, perceptions regarding the need for animal experimentation ethics training among PAETP participants were derived from both level 1 and level 2. In contrast, the formation of ethical perception and the application of ethical perception to animal experiments were derived from level 2. Evaluation at level 3 requires measurement of behavioral changes during or after actual animal experimentation, which was beyond the scope of the present study. Nevertheless, this framework provides a theoretical foundation for future follow-up studies examining longitudinal changes in participants’ perceptions. Level 4 evaluation, which concerns organizational performance and outcome achievement, was also outside the scope of this study.

A review of previous studies in the field of animal experimentation training revealed no cases of the Kirkpatrick model for training evaluation, suggesting the methodological relevance of applying this framework in the present study. Within the questionnaire design, survey items functioned as measured variables, while the 4 core training themes of the PAETP were modeled as latent variables. A reflective measurement model was adopted, in which measured variables are assumed to reflect the underlying latent constructs. This reflective design is appropriate for perception-based research and serves as a prerequisite for structural model analysis of researchers’ perceptions. Furthermore, the reflective research model allows for expandability and analytical flexibility, as additional variables may be incorporated in future studies.

Surveys were conducted before and after participation in the PAETP. The pretraining survey assessed participants’ baseline perceptions, whereas the posttraining survey captured perceptions following completion of the program. All measurements used a Likert interval scale. The 4 latent variables to be used as independent variables were measured on a 5-point scale (12 items in total), while the dependent latent variable was measured using a single item on a 7-point scale. The use of a 7-point single-item measure for the dependent variable was intended to capture participants’ responses with greater precision. This measure was designed to reflect respondents’ evaluations and ensure the accuracy of their responses.^12,13 Differences in scale ranges between independent and dependent variables are acceptable in surveys.¹⁴

Data acquisition, normality, and descriptive statistics.

Data were collected between August 12, 2024, and June 13, 2025. During this period, the PAETP was conducted 9 times, and a total of 97 participants completed the program. After data cleaning, a valid sample size of n = 90 (92.7%) was retained for analysis. Participants with 2 or more missing values in the measurement variables and those with values exceeding ±3 SD were treated as outliers and excluded from the analysis. Normality was assessed using commonly accepted criteria for skewness (absolute value <3) and kurtosis (absolute value <7), which were satisfied for all measured variables.¹⁵

The final sample size of n = 90 was considered appropriate for structural model analysis in an exploratory research context. This sample size was adequate for applying PLS-SEM for several reasons. First, according to the 10-times rule, a minimum sample size of n = 40 was required to analyze structural relationships involving 4 exogenous latent variables (independent variables) predicting one endogenous latent variable (dependent variable).^16–18 Second, based on the Cohen power analysis guidelines,¹⁹ the minimum required sample size was 41, assuming 4 exogenous latent variables, a significance level of 5%, and a minimum explanatory power (R²) of 0.25. Third, sample size estimation using G*Power software²⁰ indicated that a minimum of n = 85 was required to achieve a statistical power of 0.80 with a significance level of 5%, a medium effect size (f² = 0.15), and 4 predictor variables. Although G*Power is not specifically designed for SEM-based sample size estimation, it provides a useful reference for evaluating statistical adequacy.

In this study, SmartPLS 4 software²¹ was used for PLS-SEM analysis with a secured sample size of n = 90. Bootstrapping with 5,000 resamples was conducted to enhance the stability of parameter estimates and the reliability of the results.

Demographic characteristics and descriptive statistics were examined for the final sample (n = 90). Demographics were categorized by gender, age, researcher type, education level, animal experimentation experience, training attendance type, and expected experimental duration. Descriptive statistics were calculated for all measured variables both before and after participation in the PAETP. These data were used to examine changes in perceptions using a paired-samples t test. IBM SPSS Statistics 28 was used for the analysis.

Validity and reliability of survey instrument.

In this study, the variables were categorized into latent variables and measured variables (Table 1). The operational definitions of all variables are presented in the Results section (Tables 2 and 3). The latent variables were organized into 4 training themes of the PAETP: ethical perception (EP), legal perception (LP), safety perception (SP), and experimental knowledge perception (EK). These latent variables served as independent variables in the structural model. Each latent variable was measured using 3 indicators, reflecting a parsimonious measurement design. Limiting the number of indicators per construct helps reduce model complexity and the risk of overfitting, while maintaining sufficient construct representation.²² The 3 indicators, necessity, formation, and application of perception, were theoretically derived from levels 1 (reaction) and 2 (learning) of the Kirkpatrick training evaluation model.

Table 1.

Construction, Validity, and Reliability of the Measurement Instrument

Latent variables	Measurement variables	Abbreviation	Factor loadings	Communality	Eigen value	Explained variance percent (cumulative variance percent)	Cronbach α	Types of variables
Ethical perception (EP)	(1) I need a perception of animal experimentation ethics.	EPN	0.749	0.834	2.704	22.425 (44.958)	0.917	IVa
	(2) I have developed an ethical perception regarding animal experimentation.	EPF	0.837	0.896
	(3) I can apply perception of ethics to animal experiments.	EPA	0.762	0.867
Legal perception (LP)	(1) I need a perception of laws related to animal experimentation.	LPN	0.512	0.700	2.691	20.053 (86.777)	0.896	IV
	(2) I have developed a legal perception regarding animal experimentation.	LPF	0.883	0.945
	(3) I can apply perception of laws to animal experiments.	LPA	0.900	0.965
Safety perception (SP)	(1) I need a perception of animal experimentation safety.	SPN	0.887	0.914	2.612	22.533 (22.533)	0.895	IV
	(2) I have developed a safety perception regarding animal experimentation	SPF	0.773	0.873
	(3) I can apply perception of safety to animal experiments	SPA	0.654	0.751
Experimental knowledge perception (EK)	(1) I need knowledge of animal experimentation.	EKN	0.622	0.770	2.406	21.767 (66.725)	0.915	IV
	(2) I have developed knowledge regarding animal experimentation.	EKF	0.880	0.961
	(3) I can apply knowledge of experiments to animal experiments.	EKA	0.878	0.938
Level of contribution in conducting animal experimentation (LCCAE)	I believe the training program will contribute to conducting animal experimentation.	LCCAE	N/A	N/A	N/A	N/A	N/A	DVb

KMO = 0.839; Bartlett χ² = 1,200.48 (P < 0.001).

Abbreviations: EPN, Ethical perception necessity; EPF, Ethical perception formation; EPA, Ethical perception application; LPN, Legal perception necessity; LPF, Legal perception formation; LPA, Legal perception application; SPN, Safety perception necessity; SPF, Safety perception formation; SPA, Safety perception application; EKN, Experimental knowledge perception necessity; EKF, Experimental knowledge perception formation; EKA, Experimental knowledge perception application; LCCAE, Level of contribution in conducting animal experimentation.

^a Independent variable in PLS-SEM.

^b Dependent variable in PLS-SEM.

Table 2.

Contents of the Preclinical Animal Experimentation Training Program

Topics	Contents
Animal experimentation ethics: ethical perception (EP)	(1) Animal welfare, animal freedom, definition of laboratory animals, and trends in laboratory animal use
	(2) 3R principles, IACUC functions, support, and oversight
	(3) Animal experiment procedures, protocol review, evaluation for humane termination of experiments, implementation, and euthanasia
Animal experimentation legal systems: legal perception (LP)	(1) Details of laws and government systems related to animal experiments
	(2) Components of animal experimentation facilities, legal registration, and inspection
	(3) Government agency supervision of animal experimentation facilities based on law
Animal experimentation safety: safety perception (SP)	(1) Biosafety, LMO research, and researcher safety
	(2) Infection experimentation (ABSL 2), disaster prevention
	(3) Animal experimentation accident response, procedures, and support process
Animal experimental (technical) knowledge: experimental knowledge perception (EK)	(1) Rodent physiology, laboratory animal classification, breeding, and care
	(2) Mouse/rat handling, anesthesia, surgery, care, and molecular imaging experiments
	(3) Substance administration routes, techniques, and results acquisition

Abbreviations: 3Rs, Replacement, Reduction, Refinement; ABSL 2, Animal Biosafety Level 2; LMO, living modified organisms.

Table 3.

Demographics for the Sample

Demographics	Frequency	Ratio
Total frequency	90	100.0%
Gender
Male	37	41.1%
Female	53	58.9%
Age
20s	28	31.1%
30s	47	52.2%
40s	12	13.3%
50s	3	3.3%
Position
Researcher	84	93.3%
Principal investigator	6	6.7%
Education level
Bachelor’s	16	17.8%
Master’s	36	40.0%
Doctorate	38	42.2%
Animal experimentation experience
Less than 1 y	47	52.2%
1-3 y	13	14.4%
3-5 y	4	4.4%
5 y or more	26	28.9%
Types of training program attendance
First registration	77	85.6%
Reregistration	13	14.4%
Expected experimentation period
1-2 mo	12	13.3%
3-6 mo	16	17.8%
6-11 mo	15	16.7%
Continued for 1 y	10	11.1%
1-2 y and more	37	41.1%

The dependent variable was defined as the level of contribution in conducting animal experimentation (LCCAE). A single-item measurement approach was adopted for this construct. The use of a single-item measure is considered academically and practically acceptable when the construct is concrete, unidimensional, and easily understood by respondents.^23,24

As a preliminary step in instrument validation, exploratory factor analysis (EFA) was conducted to examine the underlying factor structure of the measured variables. EFA was performed using IBM SPSS Statistics 28, employing principal component analysis with varimax orthogonal rotation. Factor loadings and communalities were evaluated against commonly accepted thresholds (≥0.40) to confirm construct validity.²⁵

Following this preliminary validation, the measurement model was further evaluated within the PLS-SEM framework. Reliability analysis of the latent constructs was conducted using IBM SPSS Statistics 28, and internal consistency was assessed using the Cronbach α. The obtained values were compared with established reference criteria to confirm acceptable reliability.²⁶ Assessments of measurement model validity and reliability using PLS-SEM-specific criteria are reported in the subsequent section.

PLS-SEM analysis and hypothesis testing.

The causal relationships among the latent variables, as explained by their corresponding measured variables, were specified as the research model (Figure 2). Based on this model, a hypothesis was formulated to examine whether researchers’ perceptions formed through participation in the PAETP influence their perceived contribution to conducting animal experimentation.

Figure 2.

Proposal of an Exploratory Research Model.

Research hypothesis.

Researchers’ perceptions of PAETP topics (ethics, legal systems, safety, and experimental knowledge) positively influence their perceived level of contribution in conducting animal experimentation.

This hypothesis addresses the research questions presented in the Introduction. Before PLS-SEM analysis, a paired-samples t test (α = 0.05) was conducted to assess changes in researchers’ perceptions before and after participation in the PAETP.

PLS-SEM analysis was conducted after confirming significant improvements in researchers’ perceptions across all measured variables. PLS-SEM is suitable for exploratory research when theoretical development is at an early stage, prior empirical evidence is limited, or the objective is to examine relationships among latent constructs rather than test a fully established theory.^27,28

The analysis followed a 2-stage procedure: measurement model assessment and structural model assessment. Evaluation of the measurement model is a prerequisite for interpreting structural relationships, as violations of validity and reliability criteria may bias structural estimates.²⁹ Accordingly, structural model assessment was performed only after the measurement model met established thresholds. Measurement model assessment included indicator reliability, internal consistency reliability, convergent validity, and discriminant validity using the Fornell-Larcker criterion and the heterotrait–monotrait (HTMT) ratio.

Common method bias (CMB) was assessed using variance inflation factor (VIF) values within the PLS-SEM model.^30,31

Structural model assessment involved estimating relationships between latent variables using a bootstrapping procedure in SmartPLS 4. The significance of standardized path coefficients was evaluated, and effect sizes (f²) were calculated. Model explanatory power was evaluated using the coefficient of determination (R²) and adjusted R².

Predictive relevance and predictive power were also evaluated using the Stone-Geisser Q² and average loss difference. All results were interpreted based on established methodological criteria.

Results

Contents of PAETP and general characteristics of participants.

The PAETP is a 5-hour, one-day training program. The detailed content of the program, consisting of 4 training topics, is presented (Table 2). The program covers ethics, laws and regulations, safety, and experimental knowledge relevant to the conduct of animal experimentation. The animal experimentation ethics module includes animal welfare and freedoms, definitions and usage trends of laboratory animals, the purpose and function of IACUC, review of animal experimental protocols, the 3R principles, humane endpoints, and euthanasia.

The demographic characteristics of the participants (n = 90) are summarized (Table 3). Although demographic variables can be incorporated into an analytical model depending on the research objective, this study focused on describing overall participant characteristics and did not include demographic factors as variables in the model.

Among researchers who directly handled laboratory animals, most participants were female (58.9%), and 83.3% were in their 20s or 30s. Practical researchers accounted for the majority of participants (93.3%), whereas principal investigators comprised 6.7%. In terms of education level, 82.2% held a master’s or doctoral degree.

With respect to animal experimentation experience, 52.2% of participants had less than one year of experience. Most participants (85.6%) attended the PAETP for the first time, while 14.4% were repeat participants due to the expiration of their previous training. In addition, 58.9% of participants planned animal experiments with an expected duration of one year or less.

Validity and reliability of measurement instrument.

The measurement instrument consisted of measured variables and latent variables. The list and operational definitions of these variables are presented (Table 1). Each measured variable assessed participants’ perceptions regarding the necessity, formation, and application of the PAETP.

EFA was conducted to evaluate the construct validity (Table 1). All factor loadings and communality values exceeded the minimum acceptable threshold of 0.4.²⁵ The Kaiser-Meyer-Olkin (KMO) was 0.839, and the Bartlett test of sphericity was significant (χ² = 1,200.48; P < 0.001), indicating that the data were suitable for factor analysis.

EFA is appropriate for validating measurement instruments in exploratory research where theoretical foundations are limited.^32,33 Internal consistency reliability was assessed using the Cronbach α, and all latent variables exceeded the recommended threshold of 0.6, indicating acceptable reliability.²⁶

Descriptive statistics and changes in perception.

Descriptive statistics, normality assessments, and changes in researchers’ perceptions before and after the PAETP are presented (Table S1 ^{(26.5KB, docx)} ). Mean values increased across all measured variables following completion of the PAETP. Skewness and kurtosis values variables satisfied accepted normality criteria (|skewness| < 3, |kurtosis| < 7), supporting the use of parametric analyses. A paired-samples t test was conducted to assess whether these changes were statistically significant.

The results showed significant increases across all measured variables after the PAETP (P < 0.001), indicating researcher perceptions following training. These findings support the inclusion of these variables in the subsequent PLS-SEM analysis (Figure 2).

Measurement model assessment in PLS-SEM.

Before evaluating the structural model, the PLS-SEM measurement model was assessed to confirm the reliability and validity. Established criteria widely accepted in the literature were applied.

The results for indicator reliability, internal consistency reliability, and convergent validity are presented (Table S2 ^{(25.1KB, docx)} ). Indicator reliability was confirmed using outer loadings (L_i) and squared loadings (L_i2), exceeding the thresholds of ≥0.70 and ≥0.50, respectively.³⁴ Internal consistency reliability (Cronbach α, rho_A, and composite reliability [rho_C]) exceeded the recommended threshold of 0.70.³⁵ Convergent validity was supported by average variance extracted (AVE) values greater than 0.50.³⁶

Discriminant validity was confirmed using both the Fornell-Larcker criterion and the heterotrait–monotrait (HTMT) ratio. The square root values of AVE exceeded interconstruct correlations, satisfying the Fornell-Larcker criterion. All HTMT values were below 0.85, confirming discriminant validity.³⁷

Common method bias (CMB) was assessed using variance inflation factor (VIF) values.³¹ Both full collinearity assessment and random variable approaches indicated VIF values below 3.3, suggesting that CMB was unlikely to affect the results.

Overall, the measurement model demonstrated adequate reliability and validity, supporting subsequent structural model analysis.

Structural model assessment in PLS-SEM.

Structural model analysis was conducted to examine whether researchers’ perceptions formed through the PAETP influence their perceived LCCAE. The sample size (n = 90) met established requirements for exploratory PLS-SEM analysis.¹⁹ Bootstrapping with 5,000 resamples was applied to estimate path coefficients and their statistical significance.

The standardized path coefficients and corresponding significance levels are presented (Figure 3). EP showed a significant positive effect on LCCAE (β = 0.579; P < 0.001). In contrast, LP (β = 0.169), SP (β = 0.112), and EK (β = −0.208) were not statistically significant. The effect size of EP on LCCAE (f² = 0.219) corresponds to a medium effect according to established criteria.³⁸

Figure 3.

PLS-SEM Structural Model Analysis Results.

Paired-samples t test results indicated that researchers’ perceptions significantly improved across all PAETP topics following the training program. However, only ethical perception showed a significant effect on LCCAE, suggesting that improvements in perception do not uniformly translate into perceived experimental contribution. This discrepancy is further discussed in the Discussion section.

The R² and adjusted R² values for LCCAE were 0.436 and 0.409, respectively, indicating a moderate explanatory power in PLS-SEM.¹⁷ Predictive relevance was supported by a Q² value of 0.359.³⁹ The average loss difference was −0.132 (P = 0.004), indicating adequate predictive performance.⁴⁰

Overall, the structural model demonstrates acceptable explanatory power, predictive relevance, and robustness for exploratory analysis. These findings support the proposed hypothesis partially.

Discussion

The researchers’ perceptions of animal experiments were analyzed using the Kirkpatrick evaluation framework and exploratory PLS-SEM approach. Following the PAETP, participants demonstrated significant improvements across all perception domains, indicating that the training program effectively enhanced awareness related to animal experimentation. The measurement and structural model assessments supported the adequacy of the proposed analytical framework. Notably, among the PAETP topics, only ethical perception exerted a significant influence on researchers’ perceived contribution to conducting animal experiments. This finding suggests that ethical training plays a foundational role in shaping researchers’ readiness to engage in animal experimentation. EP may function as a core construct that integrates LP, SP, and EK.

The absence of significant direct effects from LP, SP, and EK does not imply that these topics are unimportant. Rather, the results indicate that their influence may be indirect, potentially mediated, or moderated through ethical perception. Previous SEM studies have reported similar interaction effects among latent perceptual constructs in complex behavioral models.^27,41,42 Establishing the present baseline model allows future studies to examine extended pathways, such as LP → EP → LCCAE, SP → EP → LCCAE, and EK → EP → LCCAE.

This study differs from prior research on animal experimentation training in several important aspects. Previous studies primarily demonstrated perception changes following ethics education using descriptive or cross-sectional approaches, occasionally supplemented by follow-up surveys or qualitative comparisons.^2–6

A few studies have attempted qualitative approaches to compare programs across countries on animal experimentation ethics training⁴³ and emphasize ethical responsibility and improve welfare in animal experimentation.⁴⁴

In contrast, the present study applied a theory-based evaluation framework and modeled perception as a system of interacting latent variables, enabling causal interpretation of how ethical perception contributes to experimental engagement.

Training-induced perceptions may be a prerequisite for ethical and humane animal experimentation. Ethics topics can reduce psychologic barriers, alleviate anxiety related to animal use, and facilitate the internalization of the 3R principles. From a practical perspective, the findings suggest that emphasizing ethical perception within training programs may enhance researchers’ preparedness to conduct animal experiments responsibly. This approach may contribute to improved animal welfare, reduced animal use, and increased consideration of alternative methods in preclinical research environments.

This study is consistent with previous studies published in Europe, Africa, and Asia,^2–4,6 which have shown the importance of ethical perceptions in animal experiments.

Limitations.

This study was not a multi-institutional or multinational survey; therefore, the findings may have limited generalizability across regions and institutional contexts. Although the Kirkpatrick model was applied, alternative models could also be considered for training evaluation. Perceptions were assessed before conducting animal experiments and were not longitudinally tracked during or after experimentation.

In addition, subgroup analyses based on demographic characteristics were not conducted due to the relatively limited sample size. Therefore, potential differences across researcher characteristics (for example, experience level or training background) could not be examined.

Furthermore, the inclusion of recertification trainees may have introduced variability in baseline perceptions due to prior training experience. Although all participants completed the same standardized training and the analysis focused on within-subject changes, potential differences between first-time and recertification trainees were not examined due to the limited sample size.

Finally, as this study relied on self-reported survey data collected immediately after a mandatory institutional training program, the results may be subject to social desirability bias. Although procedural measures such as anonymous data collection and the use of clear and neutral questionnaire items were implemented to mitigate this risk, such bias cannot be fully excluded. Moreover, although common method bias was assessed using VIF, this approach does not fully account for social desirability bias.

Further study.

Future studies may extend the present SEM framework by incorporating additional variables to further examine the causal pathways involving ethical perception. In particular, mediation or moderation models may help clarify the indirect roles of LP, SP, and EK.

With a larger and more diverse sample, multigroup PLS-SEM based on demographic characteristics would enable subgroup analyses, providing deeper insights into how ethical perceptions of animal experimentation differ across researcher groups.

Future research may also incorporate additional methodological approaches, such as validated social desirability scales or behavioral outcome measures, to further address potential response bias in self-reported data. In addition, future studies should examine differences between first-time and recertification trainees using larger samples to better understand the impact of prior training experience.

Supplementary Materials

Table S1. ^{(26.5KB, docx)} Descriptive statistics, normality, and paired-samples t tests before and after PAETP.

Table S2. ^{(25.1KB, docx)} PLS-SEM measurement model validity.

Conflict of Interest

The authors have no conflicts of interest to declare.

Funding

This study was supported by the internal budget of the Preclinical Resource Center.

Data availability

The survey raw data may be provided according to the review of the corresponding author.