Nationwide Pathology Proficiency Testing in China (2025): An Analysis of Participation Patterns, Institutional Characteristics, and Diagnostic Performance

Abstract

Background: Proficiency testing (PT) is central to external quality assessment in pathology, providing a standardized approach to evaluate diagnostic performance and inter-institutional consistency. However, the national-scale implementation patterns and performance landscape of PT programs in China remain insufficiently characterized.

Materials and Methods: We conducted a nationwide analysis of PT program in China in 2025, encompassing eight assessment categories (breast, lung, colorectal, gastric, thyroid, urothelial, lymphoma, and cytology). We evaluated the geographic and institutional participation patterns by region, hospital grade, ownership type, county-level status, and assessment categories. Participation breadth and diagnostic performance were analyzed using descriptive statistics and comparative approaches.

Results: Registered participation expanded rapidly from 2022 to 2025, with a compound annual growth rate of 86.6%. In 2025, participation remained geographically uneven (Gini coefficient = 0.405), with institutions concentrated in economically developed regions. High-tier public hospitals predominated, particularly Tertiary Grade A institutions (64.8%). Category-level participation was high (72.3%–90.1%), and 59.7% of institutions participated in all eight assessment categories. Participation breadth was weakly but significantly associated with hospital grade (Spearman ρ = 0.21; P < 0.001), while county-level hospitals demonstrated significantly lower participation breadth, particularly in lymphoma (FDR < 0.001).

Discussion: The findings show that national-scale pathology PT has rapidly expanded in China. However, participation remains shaped by regional and institutional capacity. The concentration of high-tier hospitals and lower engagement among county-level institutions suggest that high overall coverage and favorable PT performance should be interpreted in the context of structural participation patterns.

Conclusions: Nationwide pathology PT is feasible and scalable in China. However, persistent geographic and institutional disparities indicate that observed PT performance may not fully represent the overall landscape of pathology quality across China. Targeted strategies are needed to enhance participation in underrepresented regions and low-tier institutions, thereby promoting more equitable and comprehensive development of pathology quality nationwide.

Graphical abstract

Keywords

pathology / proficiency testing / external quality assessment / diagnostic performance / geographic disparities

Introduction

Proficiency testing (PT) is a key component of external quality assessment in pathology, providing a standardized framework to evaluate diagnostic performance and inter-institutional consistency[1,2]. As pathological diagnosis plays a central role in clinical decision-making, biomarker evaluation, and clinical research, ensuring its accuracy and reproducibility is essential for both patient care and the generation of reliable medical evidence[2–5]. Recent national PT initiatives have demonstrated that such programs can effectively establish baseline diagnostic performance, while also identifying systematic biases and areas requiring improvement in routine pathology practice[1].

In China, PT programs have expanded rapidly in recent years. The initial nationwide PT in 2022 was designed to establish a baseline of diagnostic performance, revealing a characteristic pattern of relatively strong immunohistochemical (IHC) interpretation but weaker performance in morphological assessment, particularly in tumor grading[1]. Building upon this foundation, the subsequent PT in 2023 advanced toward more granular evaluation, identifying emerging challenges in the accurate classification of HER2-low expression and uncovering the presence of systematic diagnostic biases across institutions[2]. During this process, the whole slide imaging-based centralized digital platform was constructed, which further transformed PT implementation by enabling standardized workflows, scalable participation, and thereby strengthening the consistency of nationwide quality evaluation[6,7]. Collectively, these efforts have established PT as a feasible and credible tool for large-scale diagnostic quality evaluation in China.

Despite these advances, key aspects of PT implementation at the national level remain insufficiently characterized. In particular, the geographic distribution of participating institutions, variation across hospital grades and institutional types, and the overall landscape of diagnostic performance across assessment categories have not been systematically evaluated. Therefore, in this study, we conducted a nationwide analysis of the 2025 PT program in China to address these gaps and provide a comprehensive overview of current landscape of pathology quality assurance in China.

Methods

Program design and implementation

This study was based on a nationwide pathology PT program conducted in 2025 under the coordination of the National Quality Control Center for Cancer. The PT provider is the pathology department of the National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital. The program was designed in accordance with the international standard for PT (ISO/IEC 17043:2023) to establish a standardized external quality assessment framework[8]. It consists of program initiation, case selection, expert-assigned value determination, participant assessment, performance evaluation and scoring, followed by result publication and certificate issuance, as described by our previous study[2]. For the present analysis, institutions that completed the PT assessment and submitted valid results within the designated assessment period were included. Registered institutions that did not submit assessment results were excluded.

The PT program comprised eight assessment categories, including breast, lung, colorectal, gastric, thyroid, urothelial, lymphoma, and cytology. These categories represented major domains of oncologic pathology practice and encompassed a broad spectrum of diagnostic challenges in routine clinical practice. Collectively, they served as a comprehensive framework for evaluating institutional diagnostic capability.

Candidate cases for each cancer type and the corresponding assigned values were determined through an expert-driven process to ensure diagnostic representativeness, evaluation value, and clinical validity. Cases were initially screened by specialized working groups and independently evaluated by experienced pathologists within each cancer-specific panel. Their assessment results were aggregated to generate preliminary consensus interpretations. Subsequently, senior expert panels for each cancer type reviewed both the selected cases and the preliminary consensus results, finalized the case set, and established the assigned values as the reference standards in accordance with current diagnostic guidelines. Pathology slides from all selected cases were digitized into whole-slide images (WSIs) using a KF-PRO-040 Magscanner system (KFBIO, China) at 20× magnification. For data protection, the original pathological identification numbers on the glass slides were masked prior to scanning. The resulting WSIs were subjected to rigorous quality-control procedures and subsequently reviewed by the expert panel to ensure image quality, diagnostic accuracy, and inter-institutional consistency.

Then, participating institutions accessed digital slides via the centralized platform and performed diagnostic evaluations according to predefined criteria. Although assessment components varied across assessment categories, they generally included histological subtype diagnosis, invasion and growth pattern evaluation, diagnostic biomarker interpretation, and therapeutic biomarkers assessment. All results were submitted electronically using standardized reporting templates within predefined timeframes to ensure data consistency and completeness.

Institutional diagnostic performance was evaluated by comparing submitted results with assigned values. Overall PT scores were calculated as the proportion of correctly classified items and scaled to a 0–100 range. Performance grading was defined as excellent (≥ 85), good (75–84), pass (60–74), and unsatisfactory (< 60). In addition, mean and median scores were calculated for each cancer type to characterize performance distributions.

Statistical analysis

All statistical analyses were performed using R (version 4.3.1; R Foundation for Statistical Computing), with ggplot2, dplyr, and sf packages used for data processing, visualization, and spatial analyses. Continuous variables were summarized as mean with 95% confidence intervals (CIs) or median with range, as appropriate, and categorical variables were presented as counts and proportions.

Participation breadth was defined as the number of participated assessment categories. Differences in participation breadth across institutional characteristics were evaluated using the Wilcoxon rank-sum test for two-group comparisons. Associations between hospital grade and participation breadth were assessed using Spearman’s rank correlation analysis.

Participation rates across assessment categories were compared between groups defined by hospital grade, ownership, independent pathology laboratory status, and county-level status. Fisher’s exact test was used for these analyses because of imbalances in group sizes[9]. To account for multiple comparisons across assessment categories, false discovery rate (FDR) correction was performed using the Benjamini–Hochberg method, and adjusted P values are reported.

For geographic inequality analysis, the Lorenz curve was constructed to assess the cumulative distribution of participating institutions across provinces, and the Gini coefficient with 95% CI was calculated to quantify spatial inequality.

Temporal growth of participating institutions was summarized descriptively, and compound annual growth rates (CAGR) were calculated to characterize expansion trends over time.

All statistical tests were two-sided, and a P value < 0.05 or FDR-adjusted P value < 0.05 was considered statistically significant.

Results

Rapid national expansion of participating institutions in 2025 with geographic imbalance and structural concentration

By 2025, the total number of registered institutions participating in the PT program had reached 657, reflecting rapid nationwide expansion (Fig. 1A). Using breast cancer as an example, both registered and reporting institutions showed marked growth from 2022 to 2025, with compound annual growth rates (CAGR) of 86.6% and 83.4%, respectively (Fig. 1B). Despite this expansion, the Lorenz curve demonstrated moderate spatial inequality (Gini coefficient = 0.405; 95% CI: 0.297–0.488), suggesting that institutional resources were concentrated in a limited number of provinces (Fig. 1C). Guangdong had the largest number of participating institutions (n = 85), followed by Sichuan (n = 54), Shandong (n = 49), Jiangsu (n = 41), and Zhejiang (n = 39), whereas participation remained relatively limited in central and northeastern China, as well as in Inner Mongolia (Fig. 1A).

At the institutional level, participation was strongly skewed toward high-tier hospitals. Tertiary Grade A hospitals accounted for the majority (64.8%), followed by other tertiary (15.2%) and secondary Grade A hospitals (9.3%), while low-tier institutions were underrepresented (Fig. 1D). In terms of ownership, public hospitals predominated (92.2%), whereas participation of private and other types of institutions was relatively limited (Fig. 1E). By hospital type, general hospitals represented the largest group (77.6%), followed by specialty hospitals (14.0%), while independent pathology laboratories and other facility types accounted for only a small proportion of participants (Fig. 1F).

Collectively, these findings indicate that although the program has achieved rapid nationwide expansion, participation remains geographically and structurally uneven, with a predominant concentration in economically developed regions and high-tier public general hospitals.

Assessment items and participation across assessment categories in 2025

Across the eight assessment categories assessed in 2025, the number of evaluation items differed by cancer type (Fig. 2A). Colorectal cancer included the largest number of items (n = 10), followed by lung adenocarcinoma (n = 7) and breast cancer (n = 6), whereas thyroid fine-needle aspiration (FNA) cytology comprised two items (Fig. 2A).

Across assessment categories, morphology/cytology-based diagnosis was included in all assessments. Other commonly assessed components included therapeutic biomarker interpretation, invasion or growth pattern, and diagnostic biomarkers, with varying combinations across assessment categories (Fig. 2B). For example, the colorectal cancer assessment included histological diagnosis, lymphovascular invasion, perineural invasion, high-grade tumor budding, and poorly differentiated clusters (PDCs), as well as elastic fiber staining for the evaluation of serosal invasion and extramural venous invasion, together with mismatch repair (MMR) markers (MSH2, MSH6, MLH1, and PMS2). In contrast, the lymphoma assessment included histological diagnosis and IHC markers CD3, CD20, and CD30.

Therapeutic biomarker inclusion also varied across assessment categories. Lung adenocarcinoma included PD-L1, ALK, and c-MET, whereas breast cancer included HER2 and ER/PR (Fig. 2C).

Participation rates were high across all assessment categories, ranging from 72.3% for lymphoma to 90.1% for breast cancer (Fig. 2D). In terms of participation breadth, 59.7% of institutions participated in all eight cancer type assessments, 12.0% participated in seven, and 6.7% participated in only one (Fig. 2E).

Participation breadth across hospital grades and geographic distribution of high-tier institutions

Participation breadth differed across hospital grades. Tertiary Grade A hospitals participated in the largest number of assessment categories (mean 6.77, 95% CI: 6.55–6.98), significantly exceeding other tertiary hospitals (5.98, 5.49–6.47) and secondary Grade A hospitals (5.74, 5.18–6.29), whereas other hospital types showed lower or more variable participation (Fig. 3A).

A weak but statistically significant positive association was observed between hospital grade and the number of participated assessment categories (Spearman ρ = 0.21; P < 0.001), indicating that higher grade institutions tended to participate in a broader range of assessment categories (Fig. 3B).

At the provincial level, Tertiary Grade A hospitals were unevenly distributed across provinces, with higher proportions observed in economically developed regions, particularly in eastern and southern China, as well as in Sichuan (Fig. 3C).

When stratified by hospital grade, similar participation patterns were observed across assessment categories. Participation rates were generally highest for breast, colorectal, and thyroid cancer assessments, whereas lymphoma consistently exhibited the lowest participation rates across nearly all hospital grade groups, particularly among Secondary Grade A hospitals (Fig. 3D).

Participation patterns of independent pathology laboratories and institutions by ownership

Independent pathology laboratories accounted for a small proportion overall and were unevenly distributed across provinces, with relatively higher representation in a limited number of regions (Fig. 4A). The highest proportion was observed in Guangdong (0.8%, n = 5), followed by Anhui (0.5%, n = 3), with additional representation in Jiangxi, Henan, Hubei, Sichuan, and Gansu (each 0.3%, n = 2).

In terms of participation breadth, the number of participated assessment categories was comparable between independent pathology laboratories and other institutions (P = 0.889) (Fig. 4B). Similarly, participation rates across assessment categories did not differ significantly between the two groups, with no statistically significant differences observed after multiple testing correction (Fig. 4C). Independent pathology laboratories also showed a similar level of participation breadth compared with Tertiary Grade A hospitals (Fig. 4D).

With respect to ownership, no significant differences were observed between public and private hospitals in the number of participated assessment categories (P = 0.204) (Fig. 4E). Participation rates across assessment categories were likewise comparable between ownership types, with no statistically significant differences detected after multiple comparison correction (Fig. 4F).

Collectively, despite differences in geographic distribution, independent pathology laboratories and private institutions demonstrated participation breadth comparable to that of other institutional types.

Participation differences by county-level hospital status

County-level hospitals showed an uneven geographic distribution, with relatively higher proportions in selected central and southwestern regions (Fig. 5A). The highest proportion was observed in Sichuan (2.1%, n = 14), followed by Yunnan (1.2%, n = 8), with additional representation in Anhui and Shandong (each 0.9%, n = 6). No county-level hospitals were represented among participants in Guangdong and Jiangsu, where participation by Tertiary Grade A institutions was high.

In terms of participation breadth, county-level hospitals participated in fewer assessment categories compared with non-county hospitals (P < 0.001) (Fig. 5B). Participation rates across assessment categories were broadly comparable between the two groups for most categories. However, in lymphoma, county-level hospitals showed a significantly lower participation rate than non-county hospitals (FDR < 0.001) (Fig. 5C).

Collectively, these findings indicate that county-level hospitals have reduced participation breadth relative to non-county institutions, particularly for lymphoma.

Performance distribution and institutional composition across assessment categories

PT performance varied across assessment categories. Lymphoma showed the highest central values, with mean and median scores approaching 95, followed by colorectal cancer and urothelial carcinoma (approximately 85), whereas cytology had the lowest mean and median values (approximately 68) (Fig. 6A).

The institutional composition was broadly consistent across assessment categories. Tertiary Grade A hospitals constituted the majority of participants in all categories (approximately 66%–72%), with smaller but stable contributions from other tertiary and secondary Grade A hospitals, and limited representation from low-tier institutions (Fig. 6B).

Notably, the relatively higher performance observed in lymphoma coincided with a higher proportion of Tertiary Grade A hospitals among participating institutions compared with other assessment categories, which may have contributed to the overall score distribution.

Discussion

In this nationwide analysis of a pathology PT program conducted in China in 2025, we systematically characterized participation patterns, institutional composition, and performance distributions. Three principal findings emerge. First, PT participation expanded rapidly nationwide but remained with substantial geographic and structural imbalance. Second, county-level institutions demonstrated significantly narrower participation breadth across assessment categories compared with non-county institutions, indicating persistent gaps in diagnostic capacity at the primary care level. Third, lymphoma exhibited the highest diagnostic performance among the eight assessment categories, which may be partly attributable to to the relatively high proportion of Tertiary Grade A hospitals among participating institutions.

We observed a marked increase in PT participation over time, reflected by substantial growth in both registered and reporting institutions. This rapid expansion demonstrates the scalability of PT and its broad acceptance within the pathology community as a national quality assurance mechanism[2]. The growth of the program was further facilitated by the improved accessibility and implementation efficiency enabled by WSI and online platforms[6].

However, this rapid expansion was accompanied by moderate geographic inequality in PT participation, as demonstrated by the Lorenz curve and Gini coefficient analyses. Participating institutions were disproportionately concentrated in economically developed regions, particularly in eastern and southern China[10], suggesting that regional healthcare capacity and resource availability remain key determinants of engagement in national quality initiatives. These findings highlight the challenge for large-scale quality assurance systems to achieve rapid expansion while maintaining equitable participation[11,12].

At the institutional level, participation was predominantly concentrated in high-tier hospitals. Tertiary Grade A hospitals accounted for approximately 65% of participating institutions and demonstrated broader engagement across multiple assessment categories. The observed association between hospital grade and participation breadth suggests that institutional capacity, including subspecialty expertise, availability of ancillary testing, and diagnostic infrastructure, is a critical determinant of engagement in complex, multi-domain assessments.

In contrast, county-level hospitals showed reduced participation breadth, particularly in lymphoma, where diagnostic requirements are more specialized. This pattern likely reflects persistent disparities in diagnostic capacity at the primary care level, including limitations in specialized expertise, infrastructure, and access to advanced diagnostic resources[13,14]. Nevertheless, these challenges may be partially mitigated through established referral and consultation networks. In routine practice, many county-level hospitals refer lymphoma cases to nearby high-tier hospitals or independent pathology laboratories for expert consultation and diagnostic support, thereby improving access to specialized pathology services and enhancing diagnostic accuracy. Meanwhile, pathology consensus recommendations tailored to county-level settings have been developed, with particular emphasis on the accessibility and feasibility of diagnostic approaches in resource-limited healthcare environments[15]. Together, these initiatives may help narrow existing diagnostic disparity in China.

Despite heterogeneity in participation, overall PT performance remained consistently high across assessment categories. This finding, however, is likely influenced by the overrepresentation of high-tier institutions, which may inflate overall performance estimates. Lymphoma showed the highest performance, and also had the greatest proportion of Tertiary Grade A hospitals, suggesting that institutional composition may partly account for differences in performance across assessment categories. Cytology demonstrated comparatively lower overall performance than the other assessment categories. Although the present study was not designed to investigate the determinants of category-specific performance, this finding may partly reflect the inherent interpretive challenges of cytologic diagnosis and the relatively limited availability of ancillary diagnostic information. Therefore, PT results should be interpreted in light of both the diagnostic complexity of each assessment category and the institutional composition of participating centers.

These findings have several important implications for the future development of PT systems. The adoption of digital platforms has greatly improved the standardization, accessibility, and scalability of PT, laying the foundation for nationwide quality assurance in pathology. However, the uneven distribution of participating institutions indicates that additional efforts are still needed to improve participation among underrepresented regions and low-tier hospitals, particularly in diagnostically challenging subspecialties. In addition, the current evaluation framework remains largely based on binary diagnostic accuracy and overall error rates, which may not fully capture the complexity of diagnostic discrepancies encountered in routine practice. Future refinement of PT systems may therefore require more detailed and structured analyses of diagnostic discordance to provide more informative feedback and support targeted educational interventions. Encouragingly, the 2025 lymphoma PT assessment conducted a preliminary exploration of artificial intelligence-assisted discrepancy analysis, highlighting the potential approaches to further refine PT methodologies.

Several limitations should be acknowledged. First, participation in the program was voluntary, which may have introduced selection bias by overrepresenting high-tier institutions. Institutions with greater diagnostic expertise, more established pathology infrastructure, and stronger commitment to quality improvement may have been more inclined to participate. As a result, higher-tier hospitals were likely overrepresented, which may have led to an overestimation of national diagnostic performance and limit the generalizability of the findings to all healthcare institutions in China. Second, the analysis was based on institution-level aggregated data. Although this facilitates comprehensive system-level assessment of participation patterns and diagnostic performance, it precludes more detailed investigation of diagnostic workflows, error patterns, and their underlying causes. Furthermore, although the eight assessment categories included in this study cover major domains of oncologic pathology and provide a broad framework for evaluating institutional diagnostic capability, they do not fully capture the breadth and complexity of pathology practice. Several diagnostically demanding subspecialties, such as soft tissue tumors and gynecologic malignancies, were not included in the current PT program. Future expansion of PT assessments to encompass these highly specialized areas may provide a more comprehensive evaluation of diagnostic performance and further strengthen pathology quality assurance efforts.

Conclusion

In conclusion, this nationwide PT program demonstrates that large-scale pathology quality assessment is both feasible and scalable in China. However, persistent geographic and structural imbalances highlight the need for targeted strategies to promote more equitable participation and to strengthen diagnostic capacity across all levels of the healthcare system.

References

[1] Xue X, Guo L, Guo C, et al. Proficiency testing of diagnosis in histopathology and immunohistochemistry of breast pathology in China: results from a pilot work of national single disease quality control program for breast cancer. BMC Cancer. 2024;24(1):23.

[2] Xue X, Guo L, Guo C, et al. Challenges and improvements in HER2 scoring and histologic evaluation: insights from a national proficiency testing scheme for breast cancer diagnosis in China. Breast Cancer Res. 2024;26(1):128.

[3] Rosanu NM, Gervaso L, Lobrano R, et al. Diagnostic and therapeutic challenges related to HER2 heterogeneity in gastric cancer: bridging molecular pathology and clinical decision-making. Int J Mol Sci. 2026;27(3):1542.

[4] Kim WG, Cummings MC, Lakhani SR. Pitfalls and controversies in pathology impacting breast cancer management. Expert Rev Anticancer Ther. 2020;20(3):205-219.

[5] Quinn C, Maguire A, Rakha E. Pitfalls in breast pathology. Histopathology. 2023;82(1):140-161.

[6] Masjoodi S, Anbardar MH, Shokripour M, Omidifar N. Whole Slide Imaging (WSI) in pathology: emerging trends and future applications in clinical diagnostics, medical education, and pathology. Iran J Pathol. 2025;20(3):257-265.

[7] Copland RR, Hanke S, Rogers A, et al. The digital platform and its emerging role in decentralized clinical trials. J Med Internet Res. 2024;26:e47882.

[8] International Organization for S, International Electrotechnical C. ISO/IEC 17043:2023 Conformity assessment -General requirements for the competence of proficiency testing providers. Geneva, Switzerland: International Organization for Standardization, 2023.

[9] Kim HY. Statistical notes for clinical researchers: chi-squared test and fisher’s exact test. Restor Dent Endod. 2017;42(2):152-155.

[10] Wan S, Chen Y, Xiao Y, Zhao Q, Li M, Wu S. Spatial analysis and evaluation of medical resource allocation in China based on geographic big data. BMC Health Serv Res. 2021;21(1):1084.

[11] Kobeissi MM, Hickey JV. An infrastructure to provide safer, higher-quality, and more equitable telehealth. Jt Comm J Qual Patient Saf. 2023;49(4):213-222.

[12] Patel J, More S, Sohani P, et al. Reshaping the equitable and inclusive access to healthcare: a qualitative study. Clin Epidemiol Glob Health. 2024;26.

[13] Wang Y, Sun T, He Z, et al. Current situation and capacity of cancer diagnosis and treatment in county hospitals in China: a multicenter cross-sectional survey. Holistic Integrative Oncology. 2024;3(1):51.

[14] Zhu Y, Zhao Y, Dou L, et al. The hospital management practices in Chinese county hospitals and its association with quality of care, efficiency and finance. BMC Health Serv Res. 2021;21(1):449.

[15] Lung Cancer Study Group of Tumor Pathology Committee of Chinese Anti-Cancer A, Consensus Expert Group on Standardized Pathologic Diagnosis of Lung Cancer in C, County H. Expert consensus on pathological diagnosis of lung cancer in county-level and prefectural medical institutions (2025 edition). Chin J Oncol (Zhonghua Zhong Liu Za Zhi). 2025;47(12):1137-1151 (in Chinese).

RIGHTS & PERMISSIONS

The Author(s) 2026. This article is published by Higher Education Press at journal.hep.com.cn.