An exploratory within-cohort paired comparison study of learner-centered and traditional teaching in medical endocrinology education

Abstract

Objective

This exploratory within-cohort paired comparison study aimed to compare the outcomes of a learner-centered learning (LCL) approach and Traditional Teaching (TT) in medical endocrinology education, using an unbalanced self-controlled design adapted to real-world teaching contexts.

Results

Sixteen undergraduates from the same cohort at Wenzhou Medical University were taught two endocrine topics (hyperuricemia and hyperthyroidism) by two associate chief physicians of equal qualification. Each topic was delivered using either LCL or TT. Although the LCL group achieved higher mean scores on knowledge tests and across all satisfaction survey dimensions, these differences were not statistically significant (p > 0.05). Qualitative analysis of student feedback revealed a strong preference for LCL elements, with students reporting enhanced deep understanding and clinical thinking. The knowledge test exhibited a ceiling effect, and the small sample size limited the analysis of structural validity; content validity was established, and internal consistency estimates were high (Cronbach’s α > 0.9).

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13104-026-07794-7.

Introduction

The shift from teacher-centered to learner-centered teaching/learning approach (LCL) is a cornerstone of contemporary medical education reform [1, 2]. LCL encompasses strategies such as case-based learning and flipped classrooms that actively engage students in constructing knowledge [3, 4]. Evidence demonstrates its efficacy in enhancing engagement, knowledge retention, and clinical reasoning compared to passive lecture-based formats [5–7].

Despite this evidence, a gap persists between the theoretical advocacy for LCL and its rigorous evaluation within real medical school curricula [8, 9]. Most validation studies are conducted under idealized conditions, often using randomized controlled trials (RCTs) with highly controlled interventions. Yet in institutional teaching management—governed by fixed schedules, existing faculty assignments, and cohort-based structures—pure RCTs are frequently impractical [2]. This creates a critical challenge: how can educators robustly assess innovative pedagogies when they cannot randomize students or fully standardize instructors and content across groups? Consequently, the evidence base lacks studies that both compare LCL to traditional methods and explicitly navigate these real-world constraints [10].

This pilot study explored this gap using a pragmatic, exploratory within-cohort paired comparison design within a live undergraduate endocrinology course. By using the same cohort of students as their own controls and matching instructor expertise and content difficulty, this design mitigated inter-individual learner differences while accepting the inherent “non-equivalence” of real-world teaching assignments [11]. Our research aimed to answer two questions: First, in a real-world setting where instructor and content differ across teaching formats, what comparative patterns emerge in students’ knowledge acquisition and learning experiences under LCL versus Traditional Teaching (TT)? Second, can this pragmatic design serve as a feasible model for evaluating educational innovations in authentic contexts? Thus, this study contributes to the discourse on LCL and offers a methodological reflection on conducting valid educational research within medical school constraints.

Study population and methods

Study population and teaching context

This was an exploratory within-cohort paired comparison study conducted in a real-world undergraduate medical education setting. The same cohort of students experienced both teaching formats, with instructor and topic differing across conditions; therefore, the study estimated comparative patterns rather than isolated pedagogy effects.

The initial study population comprised 29 undergraduate students from Wenzhou Medical University. After data cleaning (see Fig. 1), 16 students who completed both course questionnaires with valid data were included in the comparative analysis. Missingness was primarily attributable to logistical factors (student absence, time constraints, or duplicate submissions) and was treated as plausibly random. All analyses comparing the two teaching methods were conducted on the complete-case sample (n = 16; 9 males, 7 females; mean age 21.7 ± 0.7 years).

Fig. 1.

Flowchart of participant selection

This study was conducted in accordance with the Declaration of Helsinki. Ethical approval was granted by the Ethics Committee of Shanghai Pudong Hospital (Fudan University Pudong Medical Center) (Approval No.: 2025-KY-057-E01). Informed consent was obtained from all participants. The teaching activities were part of the university’s regular curriculum.

A total of 29 undergraduate students were enrolled and provided informed consent. The two courses were conducted three weeks apart. For the hyperthyroidism (TT) course, 21 questionnaires were received; 1 duplicate submission was excluded, yielding 20 valid questionnaires. For the hyperuricemia (LCL) course, 21 questionnaires were received; 1 duplicate submission and 1 invalid questionnaire (fabricated student ID) were excluded, yielding 19 valid questionnaires. After matching students across both courses, 16 participants with complete data for both teaching formats were included in the paired comparative analysis. The remaining students with data for only one course (TT-only: n = 4; LCL-only: n = 3) were excluded from paired comparisons.

Teaching implementation

To ensure comparability between the two teaching implementations, the following standardization measures were applied:

Instructor matching: Both instructors were associate chief physicians in endocrinology, matched for age and years of clinical and teaching experience to minimize variability attributable to individual expertise and teaching style.

Topic selection: Both “Hyperthyroidism” and “Hyperuricemia” are core curriculum components, with cognitive load and knowledge complexity assessed as comparable by subject matter experts.

TT Group (Hyperthyroidism): Instructor A employed lecture-based instruction supplemented by chalkboard notes, PowerPoint slides, and knowledge summary tables.

LCL Group (Hyperuricemia): Instructor B employed a case-driven active learning model, featuring core components such as typical case analysis, the “metabolic factory” analogy, interactive discussions, and visual charts.

Intervention Sequence: The two teaching interventions were conducted three weeks apart, with data collected separately after each intervention.

Evaluation tools and indicators

Data collection employed the Questionnaire Star platform [12], with surveys distributed immediately after class sessions and a 30-minute time limit. The questionnaire comprised three sections (Additional Files 1 and 2):

1. Knowledge Assessment: Closed-book tests administered after each chapter covered core knowledge points. The first 10 questions were single-choice (1 point each), while questions 11–15 were multiple-choice (2 points each), totaling 20 points.

Validity Analysis: The validity of the knowledge test was primarily ensured through content validity. The test content was based on the Wenzhou Medical University syllabus and reviewed by two senior endocrinology experts and one medical education expert. All experts agreed that the questions effectively covered and represented the core knowledge points of “hyperthyroidism” and “hyperuricemia,” ensuring strong content validity (Supplementary Table 1.1 and 2.1).

Due to the ceiling effect observed in knowledge scores (scores clustered near the maximum, as shown in Table 1), the data lacked sufficient variability to meet the statistical prerequisites for structural validity analyses (e.g., exploratory factor analysis)[13–14].

Table 1.

Comparison of knowledge test scores between two teaching methods (x ± s, n = 16)

Score	Teaching model		Mean difference (95% CI)	d = t/√n	t	P
	Hyperuricemia	Hyperthyroidism
	(LCL)	(TT)
Total Score	19.81 ± 0.75	19.31 ± 2.09	0.50 (− 0.72, 1.72)	0.22	0.877	0.394
Single-choice question	9.94 ± 0.25	9.81 ± 0.54	0.13 (− 0.21, 0.46)	0.2	0.808	0.432
Multiple-choice questions	9.88 ± 0.5	9.63 ± 1.09	0.25 (− 0.41, 0.91)	0.2	0.808	0.432

P < 0.05 was considered statistically significant; all p-values are two-tailed

LCL Learner-centered learning, TT Traditional teaching, SD standard deviation, CI confidence interval

Reliability Analysis: Internal consistency testing of questionnaire data yielded a Cronbach’s α of 0.991 for the hyperuricemia questionnaire and 1.000 for the hyperthyroidism questionnaire. Internal consistency estimates were extremely high (Cronbach’s α > 0.9), but this is likely inflated by the restricted variance and ceiling effects observed in this pilot sample. Therefore, these values should be interpreted with caution. Content validity was supported by expert review, but structural validity could not be meaningfully examined in this pilot sample.

• 2.Instructional Satisfaction Questionnaire: A self-designed 10-item Likert 5-point scale.

Validity: Content validity was assessed as good based on Wenzhou Medical University’s teaching syllabus and review by three medical education experts (Supplementary Table 1.2 and 2.2). Structural validity was not calculated due to insufficient sample size (N = 16) for factor analysis.

Reliability: The Cronbach’s alpha coefficient was 1.0 for both satisfaction questionnaires. While this indicates a perfect statistical correlation among all items, it is more likely attributable to the highly restricted range of student responses, which were overwhelmingly positive and exhibited minimal variance (i.e., a ceiling effect). Under such conditions of limited response variability, alpha coefficients can become artificially inflated and may not accurately reflect the instrument’s reliability in a more diverse sample. Therefore, these reliability estimates should be interpreted with caution.

• 3.Analysis of Qualitative (Open-ended) Feedback.

Textual data from two open-ended questions—exploring students’ perceptions of the most effective teaching components and their suggestions for improvement—were analyzed using thematic analysis following the approach outlined by Braun and Clarke [15, 16]. This process involved a systematic, iterative procedure to ensure the trustworthiness of the findings:

1. Familiarization: Researchers repeatedly read all responses to gain a comprehensive understanding.

2. Initial Coding: Significant phrases and concepts were generated to describe key features of the data.

3. Searching for Themes: The initial codes were collated into potential overarching themes that captured shared patterns of meaning across the dataset.

4. Reviewing and Defining Themes: The candidate themes were refined, checked against the coded data and the entire dataset to ensure coherence and distinctness.

5. Analysis and Reporting: The essence of each theme was defined, and vivid, illustrative excerpts from the students’ responses were selected for presentation in the results.

To enhance the reliability of the coding process, two researchers independently analyzed a portion of the transcripts and discussed the coding framework to reach a consensus. This rigorous approach allowed for a rich, detailed, and nuanced account of the student experience.

Statistical analysis

SPSS 26.0 software was employed. Quantitative data are expressed as mean ± standard deviation. Pairwise comparisons were performed using the paired t-test. P < 0.05 was considered statistically significant. For within-person comparisons (n = 16 students with complete data for both course types), we calculated mean differences with 95% confidence intervals. Effect sizes were calculated using the formula d = t/√n for paired samples [17], which provides a standardized measure of effect that directly reflects the relationship between the t-statistic and sample size. This approach was adopted given the pilot study’s limited power, to complement null-hypothesis significance testing. Thematic analysis was used to summarize open-ended feedback.

Results

Comparison of knowledge test scores

For the 16 students with complete data for both courses, paired comparisons revealed no statistically significant differences between the LCL and TT formats across all outcomes (Table 1). The mean knowledge score was slightly higher in the LCL group (19.81 ± 0.75) compared to the TT group (19.31 ± 2.09), with a mean difference of 0.50 (95% CI [− 0.72, 1.72], t(15) = 0.877, p = 0.394). Effect sizes calculated for paired samples were small for all comparisons (d = 0.22 for knowledge scores, 0.20 for single-choice scores, and 0.20 for multiple-choice scores), consistent with the non-significant p-values.

Comparison of teaching satisfaction

To more accurately evaluate the effectiveness of different teaching models, this study employed two distinct satisfaction questionnaires with varying focuses. Both questionnaires share five core evaluation dimensions for directly comparing common experiences across the two models. These dimensions include: clarity of course content logic, comprehensibility of instructor explanations, rationality of classroom time allocation, alignment of post-class assessment questions with teaching objectives, and overall course design.

Additionally, each questionnaire incorporated unique dimensions tailored to its respective teaching model:

The LCL group questionnaire added dimensions such as the effectiveness of case studies, the inspirational quality of interactive segments, and the supportive role of multimedia/visual tools to evaluate its distinctive teaching components.

The TT group questionnaire incorporates dimensions such as the logical coherence of board work/PPT presentations, the emphasis on key knowledge points, and the completeness of lecture notes to evaluate the critical elements of lecture-based instruction.

Although the satisfaction items differed slightly between the LCL and TT courses, they were designed to measure common underlying dimensions. We created an item mapping table (Additional File 3) to align items by conceptual dimension for comparative interpretation. However, we acknowledge that differences in item wording and response anchors may introduce measurement non-equivalence, and thus cross-method comparisons should be treated as descriptive rather than direct statistical contrasts.

This design ensured the content validity of the evaluation tools, enabling both a horizontal comparison of core teaching experiences and a deeper exploration of the unique strengths and weaknesses of each teaching model.

Satisfaction ratings were generally high in both groups across all five core dimensions, with mean scores above 4.6 on 5-point scales (Table 2). Scores in the LCL group were consistently near the maximum (4.94 ± 0.25 for all dimensions), while TT group scores showed slightly more variability, ranging from 4.69 ± 0.60 to 4.75 ± 0.58. Due to differences in item wording and response anchors between the two questionnaires (agreement scale for LCL vs. satisfaction scale for TT), these data were presented descriptively without statistical comparison, consistent with the exploratory nature of this pilot study.

Table 2.

Descriptive comparison of core satisfaction dimensions between LCL and TT formats (n = 16)

Shared core dimensions	LCL (mean ± SD)	TT (mean ± SD)
Clarity of course content logic	4.938 ± 0.25	4.688 ± 0.602
Ease of understanding in instructor explanations	4.938 ± 0.25	4.688 ± 0.602
Rational allocation of class time	4.938 ± 0.25	4.75 ± 0.577
Relevance of post-lesson assessment questions	4.938 ± 0.25	4.688 ± 0.704
Overall course design	4.938 ± 0.25	4.688 ± 0.602

LCL Learner-centered learning, TT Traditional Teaching

The LCL group received exceptionally high ratings on its distinctive dimensions (Effectiveness of case analysis: 4.938 ± 0.25; Inspirational nature of interactive segments: 4.938 ± 0.25).

The TT group also demonstrated stable performance on its key dimensions (Emphasis on core knowledge: 4.875 ± 0.342; Completeness of class notes: 4.875 ± 0.342).

The dimensions compared are conceptually aligned across the two questionnaires, but the specific item wording and response anchors differed. Therefore, these comparisons are descriptive rather than direct psychometric equivalents.

Thematic analysis of open-ended feedback

Thematic analysis of open-ended feedback (Table 3) revealed significant differences in the focus areas of the two student groups regarding the teaching model. Feedback on the LCL approach centered on internalizing knowledge, developing clinical reasoning, and enhancing learning experiences. In contrast, feedback on the TT method focused on memorization tools, structured knowledge organization, and expectations for deeper teaching tools. This qualitative analysis confirmed how the two teaching models emphasize different levels of learning objectives.

Table 3.

Comparative analysis of open-ended feedback themes across two instructional models

Teaching module	Core theme	Theme essence	representative student feedback (selected)
LCL (hyperuricemia)	1. Vividness and depth in mechanism understanding	Emphasizes transforming abstract pathophysiological mechanisms into easily internalized mental models through metaphors and visualization tools	“The doctor used the ‘metabolic factory’ analogy to explain the uric acid elevation process… It instantly made me understand why some people have metabolic issues while others have excretion problems.” (Student P)
	2. Development of Clinical Reasoning and Decision-Making Skills	Case analysis sessions were deemed effective in simulating real clinical scenarios, fostering logical reasoning from symptoms to diagnosis and treatment	“The clinical case-based explanations helped me grasp the clinical manifestations and treatment approaches for hyperuricemia more profoundly.” (Student I)
	3. Stimulating Learning Interest and Engagement	Interactive segments and video-based instruction enhanced classroom appeal and student participation	“Video teaching sparks interest.” (Student J)
	4. Affirmation of Teaching Tools and Methods	Students highly endorsed specific pedagogical approaches like case studies, flowcharts, and mind maps	“Visualization of pathological mechanisms and teaching simulations (enabling intuitive understanding of uric acid metabolism’s complex pathways).” (Student G)
TT (hyperthyroidism)	1. Reliance on Knowledge Structuring Tools	Strong affirmation of comparison tables, flowcharts, etc., for summarizing, organizing, and memorizing systematic knowledge	“Comparison tables are better.” (Student M) “Comparison tables directly present key information… crystal clear, easier to understand and memorize.” (Student P)
	2. Dialectical Reflection on the Merits of Mnemonics	Acknowledges the aid of mnemonics for memorization, but critically examines their limitations—confusion and instability during application	“Mnemonics are smooth to recite, but they fail when you actually need them… Comparison tables work better.” (Student P)
	3. Desire for Enhanced Teaching Tools	While affirming traditional tools, explicitly calls for more intuitive, clinically relevant resources	“Hope to add medication treatment flowcharts.” (Student F) “Could include clinical patient images and case analysis questions.” (Student A)

LCL Learner-centered learning, TT Traditional teaching

Discussion

The paradigm shift toward learner-centered learning (LCL) represents a cornerstone of contemporary medical education reform, championed for its potential to cultivate the self-directed, critically-thinking practitioners demanded by modern healthcare [2, 18]. This reform is underpinned by educational theories such as constructivism, which posits that learners actively build knowledge, and social learning theory, which emphasizes the importance of interaction and observation [19, 20]. Consequently, pedagogical strategies like case-based learning and flipped classrooms have proliferated, with a robust body of evidence demonstrating their efficacy in enhancing engagement and clinical reasoning skills in controlled settings [21, 22]. However, the translation of these evidence-based practices into the complex, resource-constrained reality of institutional teaching often remains challenging, creating a critical gap between theoretical efficacy and proven effectiveness in authentic contexts. This exploratory within-cohort paired comparison study directly explored this gap by implementing an “unbalanced self-controlled”design to evaluate LCL against traditional lecturing under rigorously matched real-world conditions. Our findings reveal a nuanced landscape of functional equivalence in quantitative outcomes, juxtaposed with a qualitative divergence in the learning experience, offering critical insights for curriculum design and educational research methodology.

Functional equivalence and the re-evaluated role of direct instruction

Our primary quantitative finding—the absence of statistically significant differences in knowledge acquisition across all assessment formats—challenges a simplistic narrative of LCL’s inherent superiority. This result, often termed a “negative” finding in a binary framework, instead provides a powerful affirmation of the contextual efficacy of established methods.

The enduring value of structured knowledge transmission

The high performance of the traditional lecture group underscores that well-designed direct instruction remains a profoundly efficient vehicle for foundational knowledge consolidation [23, 24]. From the perspective of cognitive load theory, the explicit guidance and structured presentation of information in a lecture can effectively manage intrinsic cognitive load for novice learners, allowing them to efficiently build the essential schemas upon which clinical reasoning depends [25, 26]. Our results align with meta-analytic findings that the “direct guidance” aspect of instruction is a more reliable predictor of student achievement than unguided discovery-based methods [27]. Therefore, the strategic use of lectures should not be viewed as a pedagogical failure, but as a deliberate and effective choice for achieving specific, foundational learning objectives [28].

The evolved value proposition of LCL

The failure of LCL to produce superior test scores does not equate to a failure of the method itself. Its value proposition appears to reside in domains that are crucial yet elusive to standardised testing. The rich qualitative feedback from students, filled with active, process-oriented language describing “understanding mechanisms” and “clinical thinking,” provides compelling evidence of LCL’s success in fostering perceived relevance and metacognitive engagement [1, 29]. This aligns with the goals of transformative learning theory, which seeks to shift learners’ perspectives and meaning structures [30]. Furthermore, the observed trend towards more consistent performance (smaller standard deviations) in the LCL group on complex multiple-choice items, while not statistically significant, may indicate a possible “floor-raising” effect. However, this speculative pattern requires confirmation in larger samples before any conclusions can be drawn about its impact on uniform competency attainment [31]. This combination of qualitative depth and performance consistency highlights LCL’s unique role in cultivating the affective and cognitive dimensions of clinical preparedness that are central to competency-based education but difficult to quantify on a multiple-choice exam [32].

Students in the LCL group reported positive experiences with case-based learning and interactive sessions, while TT students valued the clarity of lecture organization. These descriptive patterns suggest that each approach may offer distinct experiential benefits, though direct comparisons should be interpreted cautiously given the measurement differences noted above.

Methodological contributions and the science of improvement

A secondary but significant contribution of this work is its methodological approach. The “unbalanced self-controlled” design represents a pragmatic and rigorous response to the common challenge of conducting experimental research in real educational ecosystems where randomization is often logistically or ethically impossible [33, 34].

By using students as their own controls, this design powerfully controlled for the major confounding variable of inter-individual differences in prior knowledge and academic ability, a feature that strengthens internal validity over many between-group cohort studies [35]. This approach is aligned with the principles of “improvement science” in education, which prioritizes practical, context-specific inquiry that can inform local change while contributing to generalizable knowledge [34, 36]. We openly acknowledge the design’s inherent limitation—the inability to fully disentangle the effect of pedagogy from the unique influence of each instructor’s persona. However, by transparently defining, implementing, and critiquing this model, we provide a valuable case study for other educators and researchers operating under similar constraints. This demonstrates a pragmatic approach to educational evaluation in real-world settings, while acknowledging that such designs estimate comparative patterns rather than isolated causal effects [37].

Limitations

Several limitations of this feasibility study must be considered when interpreting the findings. First, the modest sample size (n = 16 for the comparative analysis), though justified in this exploratory, controlled-comparison study, limited the statistical power to detect small-to-moderate effect sizes and precluded more complex multivariate analyses [38]. Second, the knowledge test exhibited a ceiling effect, with scores clustering near the maximum. This indicates that the measure lacked discriminative sensitivity in this cohort, reducing statistical power to detect potential differences between the two teaching formats. Future studies should consider more challenging items or a wider difficulty range to better differentiate student performance [39]. Third, the satisfaction survey was likely compromised by social desirability bias, a well-documented phenomenon in course evaluations that can artificially compress score variance [40]. Finally, the short-term nature of the assessment leaves the critical question of long-term knowledge retention and transfer to clinical practice unanswered [41].

Future research should therefore build upon this foundation by: (1) conducting large-scale, multi-institutional studies to enhance generalizability and statistical power ; (2) developing and validating more sensitive and multidimensional assessment tools, including direct observation of clinical skills, structured clinical reasoning exams (e.g., Script Concordance Tests), and longitudinal follow-up to measure the decay or consolidation of knowledge and skills ; and (3) exploring the optimal design of blended learning models that systematically integrate the efficiency of direct instruction for knowledge transmission with the engagement of LCL for clinical application and reasoning, potentially informed by learning analytics to personalize the pathway.

Conclusion

In conclusion, this pilot investigation moves the discourse beyond the simplistic dichotomy of “traditional versus learner-centered” teaching. Our findings demonstrate that in the hands of skilled educators, both pedagogies can effectively achieve core knowledge outcomes, a state of functional equivalence. The true differentiator of LCL lies in its capacity to enrich the learning experience, foster a sense of clinical relevance, and promote a more uniform understanding of complex concepts. The implication for medical educators is not to choose one method over the other, but to become adept at strategic pedagogical design—knowing when to leverage the efficient clarity of a lecture and when to create space for the messy, engaging, and transformative work of learner-centered inquiry. By doing so, we can cultivate a generation of physicians who are not only knowledgeable but also adaptable, engaged, and equipped with the clinical mindset for a lifetime of practice.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

LCL

Learner-centered learning

TT

Traditional teaching

RCTs

Randomized controlled trials

Author contributions

FY conducted the research and drafted the initial manuscript. CW participated in teaching and assisted with statistical analysis. JR contributed to data collection and assisted with statistical analysis. DX and WPparticipated in teaching. XN conceived and designed the study, critically revised key content of the manuscript, and approved the final version for publication. All authors reviewed and approved the final version of the manuscript.

Funding

Project: Fudan University Undergraduate Teaching Reform Project: learner-centered teaching Design and Implementation: Endocrine and Metabolic Diseases (FD2025A162 ).

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to an agreement with the subjects but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in full compliance with the principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Ethics Committee of the Shanghai Pudong Hospital (Fudan University Pudong Medical Center) (Approval No.: 2025-KY-057-E01). Written informed consent was obtained from all individual participants included in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.