Enhancing laboratory test consistency through linear transformation: A multi-center study

Abstract

Objectives

China is promoting the mutual recognition of clinical laboratory test results to reduce redundant testing, provide more convenient medical services, and lower economic burdens. This study aimed to enhance the consistency of test results across laboratories using a linear transformation method, focusing on five representative biochemical parameters: ALP, CA, TBIL, TC, and TG.

Methods

Five ISO 15189 accredited laboratories participated in this study. We established inter-laboratory and intra-laboratory conversion relationships using patient sample comparisons and daily quality control (QC) data. These relationships were used to develop a web-based tool enabling real-time conversion and mutual recognition of laboratory test results.

Results

The study found that the linear transformation method effectively improved the consistency of test results. After three stages of conversion, most test results showed deviations within ±1/2 TEa when compared to a reference laboratory. However, some parameters in the low-value range exhibited less significant conversion effects, likely due to the sensitivity of percentage deviation measurements in this range.

Conclusions

The developed approach and web-based tool show potential for enhancing result consistency and facilitating mutual recognition across laboratories. Despite its effectiveness, the study's limitations, such as a small sample size and a narrow focus on five biochemical parameters, indicate the need for further research and broader application.

Highlights

• •Enhancement of result mutual recognition.
• •Application of Deming regression models.
• •Cloud-based data processing platform.
• •Validation using patient samples.
• •Potential impact on clinical practice.

1. Introduction

China is now vigorously promoting the mutual recognition of clinical laboratory test results, aiming to avoid waste of resources caused by repeated testing, provide patients with more convenient medical services and reduce the economic burden on patients and society [1,2]. The General Office of China's Ministry of Health requires clinical laboratories of medical institutions at the same level to achieve the recognition of their test results [1].

A variety of laboratory quality and personnel competency assessment and laboratory accreditation programs exist to ensure the consistency of test results among laboratories as much as possible. At present, more than 550 laboratories in China were ISO15189 accredited, and 97 laboratories have obtained CAP accreditation. These accreditation programs ensure laboratory service quality and personnel safety through on-site inspections once every 2–3 years and annual self-inspections, and then ensure mutual recognition of results between laboratories through external quality assessment (EQA) or proficiency tests (PT) 2–3 times a year [[3], [4], [5], [6]]. Daily quality control (QC) is required to ensure stable and reliable results within the laboratory [7]. The actual situation is that the test results among laboratories have large deviations, and the EQA results can only represent the comparability of peer group instruments at that time, and it does not mean that the results on any day of the year are comparable, which has certain risks and hidden dangers. In addition, EQA specimens still exhibit matrix effects, and establishing inter-system conversion relationships through comparison with patient specimens can improve comparability between systems [8]. However, because of the additional expenditure and manpower consumption, the feasibility of this measure to increase the frequency of EQA or patient specimens is limited.

QC is a quality control activity that clinical laboratories do every day [9]. If the QC results and data can be reasonably used to make up for the lack of EQA frequency, the real-time mutual recognition of test results among laboratories can theoretically be achieved. The consistency of results can be improved to another level, providing better and more reliable services for patients and clinics. The main problems facing the realization of this theory include: whether an effective conversion relationship can be established between QC data at different times within the laboratory; whether the conversion relationship between QC data is applicable to patient data; Real-time sharing of QC data and conversion relationships between rooms.

This study plans to include five of 15189-certified laboratories, establish inter-laboratory conversion relationships through patient-specimen comparison, establish intra-laboratory conversion relationships through the daily quality control data of each laboratory, and then establish the real-time conversion relationships among different laboratories through the conversion of these two relationships. Then, this study also verified the feasibility of real-time comparison of transformed results using QC data in purpose of improving the consistency of laboratory results.

2. Methods

2.1. Research objects

Five laboratories accredited by ISO 15189 participated in this study, namely Hosp 1–5. Five analytic items were selected as the research objects of this study, including Total cholesterol (TC), Triglyceride (TG), Total bilirubin (TBIL), Alkaline phosphatase (ALP), and Calcium (CA).

These 5 items are required to have continuous quality control data (Quality control material was provided by Qualab Biotech Co., Ltd. Shanghai, China) and to pass a linear range verification within 180 days form the date of data gathering for this study. All experimental data fall within the linear range of their respective detection systems.

All experimental procedures were approved by the Institutional Review Board of the First Hospital of China Medical University (approval number: ScientificResearchEthicsReview[2020]No.314). Written informed consent was obtained from all participants. This study adhered to all applicable ethical guidelines, including those of the Declaration of Helsinki.

2.2. Experimental process

From March 15 to April 2, 2021, each laboratory tested the issued quality control material (batch number: 376B7A7, 376DEB7, manufactured by Qualab Biotech Co., Ltd. Shanghai, China) according to the manufacturer's manuals and collected quality control data under the premise of ensuring indoor quality control and collected quality control data.

From March 15 to March 19, 2021, a total of 15 serum samples were collected and tested for the 5 analytic items. After the samples are collected and aliquoted, store at −40 °C for distribution. In this study, serum samples were distributed to various laboratories in two batches and tested at the same time: (1) The first batch of 10 serum samples were distributed on 22 March and the testing was completed on 23 March. At this time, the status of the testing system in all laboratories was recorded as condition “a”; (2) The second batch of 5 serum samples were distributed on 1 April and the testing was completed on 2 April. At this time, the status of the testing system in all laboratories was recorded as condition “b”.

All serum samples were tested repeatedly twice. On the day of testing, in addition to the daily quality control process (Quality control material were provided by Qualab Biotech Co., Ltd. Shanghai, China), 2 additional Qualab controls (batch numbers: 376B7A7, 376DEB7) need to be performed while testing the serum samples.

The specific procedures and requirements were shown in Fig. 1.

• 1.Laboratory quality control monitoring.

Fig. 1.

The Experiment Procedure including the establishment of equations and the verification process.

The quality control data of each laboratory from March 15 to April 2 were monitored to ensure that the daily control was under control and the CV% was less than 1/3 total error allowable (TEa).

• 2.Establishment of mathematical relationships between laboratories.

Laboratory A stands for the Department of Clinical Laboratory, First Hospital of China Medical University, and Laboratory B stands for other laboratories. Under the condition “a”, laboratories tested the 10 serum samples simultaneously. Then, the test results of laboratory A were assigned as the dependent variable y and those of laboratory B were assigned as the independent variable x. The mathematical relationship between results of laboratories was constructed by using the Deming linear regression model and this relationship was assigned as R_B- > A. Under the condition “a”, all laboratories also tested 2 levels of Qualab quality controls and results were recorded as L1a and L2a.

• 3.Establishment of the intra-laboratory mathematical relationships.

In the condition “b”, all laboratories tested 2 levels of Qualab controls and the results were recorded as L1b and L2b. The quality control results L1b and L2b under the condition “b” were assigned as the independent variables x, and the quality control results L1a and L2a under the condition “a” were assigned as the dependent variables y. The Intra-laboratory mathematical relationship was then constructed by using the Deming regression equation and such relationships were assigned as R_b->a for Hospital 2–5 from condition “b” to condition “a”, and similarly, R_a- > b for Hospital 1 from condition “a” to condition “b”.

• 4.Conversion of the testing results under different conditions between laboratories.

According to the previous procedures, three equations has been established: (1) R_b->a to convert the results tested by Laboratories from condition “b” to condition “a”; (2) R_B- > A to convert the results from laboratory B to laboratory An under the same condition “a”; (3) R_a- > b to convert the results tested by or transformed to Laboratory A from condition “a” to condition “b”. By applying the equations described above, we can convert any result tested by laboratory B under condition “b” to the estimated result tested by laboratory An under condition “b”.

• 5.Comparability verification of the laboratory results after conversion.

Under the condition “b”, 5 fresh blood samples were issued to each enrolled laboratory for verification purpose. The results of the five samples detected by laboratory B (Hosp 2–5) under condition “b” were recorded as Bb_1-5, which were then converted to the estimated values of laboratory B under condition “a” by applying Rb->a(B) and recorded as Ba_1-5. Then, by applying R_B- > A, Ba_1-5 were converted to the estimated Aa_1-5. Then, the estimated Aa_1-5 were converted to the estimated Ab_1-5 by applying R_b->a(A). Finally, we compared the estimated Ab_1-5 to the actual measured Ab_1-5. If the error was less than 1/2TEa, we considered the results to be comparable.

• 6.Cloud platform for data analysis.

Cloud platform was applied to realize the data upload, storage, conversion, and the following statistical analyses.

Data processing and statistics.

The quality control data were uploaded through Qualab software. Serum samples and follow-up quality control test data were carefully recorded for further statistical analyses. All statistical calculations and transformations were done using Microsoft Excel and figures were generated using GraphPad Prism 10.1.0.

3. Results

• 1Laboratory quality control monitoring.

The quality control data of each laboratory from March 15 to April 2 were all under control and the CV was less than 1/3 total error (TEa).

• 2Establishment of mathematical relationships between laboratories.

The mathematical relationships between results of laboratories were constructed by using the Deming linear regression equations (Table 1) and the comparison of results before and after the transformation were shown in Fig. 2, Fig. 3.

• 3Establishment of the intra-laboratory mathematical relationships.

Table 1.

The parameters of the regression equations R_B- > A representing the result transformation from Hospital 2–5 to Hospital 1 under condition “a”, respectively.

Analyte	Parameter	Hosp 2	Hosp 3	Hosp 4	Hosp 5
ALP	Slope	1.093105	0.930065	1.089898	0.932155
	Intercept	0.307212	0.295363	0.111376	1.748229
CA	Slope	0.994567	1.014085	1.028238	1.022166
	Intercept	−0.01329	−0.00067	−0.04836	−0.06656
TBIL	Slope	0.928203	0.959368	0.909143	0.843898
	Intercept	1.404589	5.420425	1.404886	6.467648
TC	Slope	1.045222	1.088898	1.018837	0.997386
	Intercept	−0.06946	0.014233	−0.03863	0.244501
TG	Slope	1.071787	1.117122	1.067187	1.020343
	Intercept	−0.23395	−0.2356	−0.18035	−0.20451

Fig. 2.

Comparison of results before and after the inter-laboratory transformations.

(a) to (e) represent for ALP, CA, TBIL. TC, and TG, respectively.

Fig. 3.

Percentage deviation graphs before and after the inter-lab transformation.

Hosp2′-5′ represent for the results after the inter-lab transformation.

The quality control results L1b and L2b under the condition “a” and “b” were shown in supplementary material 1. The Intra-laboratory mathematical relationship was then constructed by using the Deming linear regression equations and such relationship were assigned as R_b->a and R_a- > b. The parameters of the regression equations R_b->a, representing the result transformation from condition “b” to condition “a” for Hospital 2–5, were shown in Table 2. The parameters of the regression equations R_a- > b, representing the result transformation from condition “a” to condition “b” for Hospital 1, were shown in Table 3.

• 4Conversion of the testing results under different conditions between laboratories.

Table 2.

The parameters of the regression equations R_b->a representing the result transformation from condition “b” to condition “a” for Hospital 2–5, respectively.

Analyte	Parameter	Hosp 2	Hosp 3	Hosp 4	Hosp 5
ALP	Slope	1.00495	0.98863636	1.029145	1.028902
	Intercept	−0.44059	−1.0340909	−6.01137	−8.45665
CA	Slope	1	1	1.011236	1.068966
	Intercept	−0.02	−0.06	−0.03573	−0.19103
TBIL	Slope	0.974747	1.0255102	1.058347	1.246296
	Intercept	0.363636	0.54540816	0.67812	0.967593
TC	Slope	0.940639	1.0245098	0.986047	0.976303
	Intercept	0.404292	−0.1279902	0.024	0.067204
TG	Slope	1.015957	0.98369565	1.005076	1.048257
	Intercept	0.002926	−0.0125543	−0.03563	−0.0526

Table 3.

The parameters of the regression equations R_a- > b representing the result transformation from condition “a” to condition “b” for Hospital 1, respectively.

Analyte	Parameter	Hosp 1
ALP	Slope	0.975903614
	Intercept	4.879518072
CA	Slope	1.023255814
	Intercept	−0.011395349
TBIL	Slope	0.989001397
	Intercept	−1.043032472
TC	Slope	1.022222222
	Intercept	−0.106
TG	Slope	1.019607843
	Intercept	−0.01

According to the previous procedures, three equations has been established: (1) R_b->a to convert the results tested by Laboratories from condition “b” and condition “a” (Table 2); (2) R_B- > A to convert the results from laboratory B and laboratory An under the same condition “a” (Table 1); (3) R_a- > b to convert the results tested by or transformed to Laboratory A from condition “a” and condition “b” (Table 3). By applying the equations described above, we can convert any result tested by laboratory B under condition “b” to the estimated result tested by laboratory An under condition “b”.

• 5Comparability verification of the laboratory results after conversion.

Under the condition “b”, 5 fresh blood samples were issued to each enrolled laboratory for verification purpose. The original data and estimated data were shown in Supplementary material 2. The comparison of results before and after the transformation were shown in Fig. 4, Fig. 5. If the error was less than 1/2TEa, we considered the results to be comparable.

• 6Cloud platform for data analysis.

Fig. 4.

The comparison of results before and after the three-step transformation.

(a) to (e) represent for ALP, CA, TBIL. TC, and TG, respectively.

Fig. 5.

Percentage deviation graphs for the five analytes before and after the three-step transformation.

Hosp2′-5′ represent for the results after the three-step transformation.

Cloud platform was developed and applied to realize the data upload, storage, conversion, and the following statistical analyses. Sample figures (screenshots) of the developed cloud platform were provided in supplementary material 2.

4. Discussion

In this study, we sought to enhance the consistency of test results across different laboratories and time points by applying a linear transformation method. The objective was to assess the feasibility of real-time mutual recognition of laboratory test results based on this approach. We selected five representative biochemical parameters with significant inter-system variability, including ALP, CA, TBIL, TC, and TG, to conduct our experiments. Additionally, we attempted to utilize internal quality control data to establish the relationship between patient sample results obtained at different times within a single testing system. The final results demonstrated that our method effectively improved the consistency of laboratory test results, thereby facilitating the advancement of result mutual recognition projects.

It is important to emphasize that linear validation was the foundational premise of this study. All data obtained were within the linear range of the corresponding testing systems. Consequently, we employed a linear model to establish the mathematical relationships between systems [10]. Furthermore, we believe that establishing relationships between patient sample results at different times within a single testing system using internal quality control data is subject to specific conditions. We categorized influencing factors into two major groups: internal factors within the reaction complex (referring to factors that affect the physical and/or chemical properties of the reaction complex, composed of substances such as samples and reagents during the testing process) and external factors outside the complex. For internal factors within the complex (e.g., changes in reagent batch numbers, variations in quality control material batch numbers, fluctuations in temperature and humidity), differential impacts on quality control materials and patient samples could potentially invalidate the originally established mathematical relationship. For external factors outside the complex, even when changes occur, their effects on quality control materials and patient samples are equivalent, thereby preserving the integrity of the established mathematical relationship. In our study, reagent and quality control material batch numbers were kept consistent, and the influence of internal factors within the complex was minimized under ISO 15189 standards. When internal factors within the complex are present, the pre- and post-changes can be considered as distinct testing systems, and a mathematical relationship can be established using the aforementioned method for multi-level conversion. This necessitates that laboratories establish a mathematical relationship between time points using patient sample comparisons when changes in reagent or quality control material batch numbers occur. Similar recommendations are outlined in the Medical laboratories—Requirements for quality and competence and Quality management in clinical laboratories [5,11].

Following three stages of conversion, the overall consistency of test results improved. The deviation of most test results relative to Laboratory A was within ±1/2 TEa. However, it was observed that the conversion efficacy for certain parameters in the low-value range was less pronounced. This can largely be attributed to the sensitivity of percentage deviation measurements in the low-value range. On a local scale, this phenomenon manifested as deviations in the results of individual laboratories for ALP and TG, suggesting that significant random deviations might have occurred in the low-value quality control results on that day, while high-value quality control results remained stable. Thus, it may be necessary to increase the frequency of quality control within the analysis batch and use the mean for conversion. In the case of TBIL, low-value samples may have been near the linear low point, making it difficult to represent the inter-laboratory relationship near this point using data from other concentrations. This indicates that local evaluation may be needed near the endpoints of linear validation. As for Ca, its high consistency before conversion resulted in minimal improvement post-conversion.

Based on the theoretical framework developed in this study, we designed a web-based tool and implemented practical applications within the materials and methods framework. This tool allows laboratories to view the conversion of their test results to values from other laboratories in real-time, thereby facilitating initial mutual recognition between laboratories. We believe that the results of this study and the established theoretical foundation will have a positive impact on clinicians and patients. Such impacts include, but are not limited to, aiding in the enhancement of result consistency and mutual recognition initiatives, improving patient convenience, reducing redundant testing, lowering testing costs, and alleviating patients' medical burdens. Additionally, the findings of this study could help provide logical explanations for deviations in test results between laboratories involved in mutual recognition and assist in delivering more reliable and user-friendly test results to clinicians and patients.

Nonetheless, this study has certain limitations, including a small sample size, the experimental design encompassing only five biochemical parameters, and experiments conducted solely among five ISO 15189 accredited laboratories. Moreover, external factors, such as human errors or instrument malfunctions, were not considered during the analysis process. The dissemination and application of this study's findings rely on a cloud platform for data sharing and automated analysis comparisons, which may limit its applicability to laboratories capable of participating in cloud-based data sharing.

5. Conclusion

This study successfully demonstrated the potential of using a linear transformation method to enhance the consistency of laboratory test results across different facilities. By establishing both inter-laboratory and intra-laboratory conversion relationships, we developed a framework that significantly improves the comparability of results, thereby supporting the initiative of real-time mutual recognition of clinical laboratory results. The web-based tool developed as part of this study provides a practical means for laboratories to apply these conversions in real-time, which could lead to reduced redundancy in testing, increased efficiency, and better patient care. However, the study's limitations, such as the small sample size and focus on a limited number of biochemical parameters, suggest that further research is necessary to validate and extend the applicability of this approach. Future studies should explore the scalability of this method across a broader range of tests and in diverse clinical settings to fully realize its benefits.

CRediT authorship contribution statement

Shitong Cheng: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. Dongliang Man: Conceptualization, Data curation, Formal analysis, Investigation, Validation, Writing – review & editing. Zhiwei Zhou: Conceptualization, Formal analysis, Software, Validation, Visualization. Hui Kang: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing.

Funding

This study was supported by the CAMS Innovation Fund for Medical Sciences (2019-I2M-5–027).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.