Sample size calculation method and system for quantitative consistency evaluation

Abstract

The invention discloses a sample size calculation method and system for quantitative consistency evaluation. The method comprises the following steps: firstly, determining MDL and a clinical consistency boundary value, then collecting pre-test data, and determining a required sample size through iterative calculation. Then, verification samples are extracted from the overall data for consistency analysis, and the robustness of a calculation result of the verification sample size is simulated through Bootstrap; according to the method, MDL and least square regression are fused for the first time, a sample size formula for quantitative consistency evaluation is constructed, the reliability and effectiveness of clinical test data are ensured by providing a new sample size calculation method, more accurate sample size estimation is provided for clinical research, and therefore the quality of clinical decisions is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for calculating sample size for quantitative consistency assessment. Background Technology

[0002] In the field of medical testing, with the continuous emergence of new testing methods, the demand for quantitative consistency assessment is increasing. Quantitative consistency assessment is a crucial means of ensuring the comparability and reliability of measurement results across different testing systems or methods. Accurate sample size calculation is a key aspect of quantitative consistency assessment, directly impacting the reliability and scientific validity of the assessment results.

[0003] Currently, sample size calculation methods primarily rely on two-tailed one-term tests of regression coefficients and constant terms. While this method can provide sample size estimates to some extent, it has several limitations. First, it lacks consideration of expected bias at the Medical Decision Level (MDL). The MDL is a crucial concept in medical testing, referring to the numerical values ​​at which test results are significant for clinical decision-making within a specific medical context. At the MDL, the accuracy of test results is particularly important because these results directly influence clinical decisions and treatment choices. However, traditional sample size calculation methods fail to adequately account for expected bias at the MDL, potentially leading to inaccurate sample size estimates and thus affecting the conclusions of quantitative consistency assessments.

[0004] Secondly, existing sample size calculation methods primarily focus on the goodness of fit of the overall data, neglecting in-depth analysis of specific data points. This can lead to the omission of crucial data information, affecting the accuracy and application value of research results in patient populations. Furthermore, when clinical consistency thresholds are missing or not predetermined, traditional sample size calculation methods cannot assess consistency, limiting their widespread practical application. Summary of the Invention

[0005] Based on this, the embodiments of this application provide a sample size calculation method and system for quantitative consistency assessment, which can improve the accuracy and reliability of sample size calculation by introducing the expected bias at MDL, enhance the scientificity and practicality of quantitative consistency assessment, and provide more reliable sample size estimation for medical testing.

[0006] Firstly, a method for calculating sample size for quantitative consistency assessment is provided, the method comprising:

[0007] S1 determines the level of medical decision-making based on preset medical testing indicators and sets clinical consistency thresholds;

[0008] S2, performing a single test on a set of samples using at least two different detection methods, and collecting preliminary test data; wherein the detection methods include at least different analytical instruments, reagent kits or detection platforms;

[0009] S3, Substitute the pre-test data into the MDL-based sample size estimation formula, and calculate the sample size required to achieve the consistency requirements through multiple iterations;

[0010] S4. Based on the calculated sample size, extract a corresponding number of validation samples from the overall experimental data and perform consistency analysis on the validation samples.

[0011] S5. After the consistency analysis is passed, a bootstrap simulation is performed to verify the robustness of the sample size calculation results. The bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power.

[0012] S6, when the test power meets the preset statistical threshold, the sample size estimate is determined as the final sample size calculation result.

[0013] Optionally, the method further includes:

[0014] If the test power does not meet the preset statistical threshold, adjust the parameters in the sample size estimation formula, such as increasing the sample size or adjusting the clinical consistency threshold, and then repeat S3 to S5 until the test power meets the preset statistical power.

[0015] Optionally, the pre-experimental data is substituted into the MDL-based sample size estimation formula, and the sample size required to achieve the consistency requirements is calculated through multiple iterations, specifically including:

[0016] According to the formula

[0017]

[0018] The sample size was obtained by processing the preliminary experimental data, where n is the sample size, α is the significance level, and α = 0.05 for a two-tailed test. n-2,1-α / 2 and t n-2,1-β / 2 , b0 and b1 represent the two-sided t-distribution critical values ​​at confidence levels of 1-α and 1-β, respectively; b0 represents the regression constant of the expected regression model; b1 represents the regression coefficient of the expected regression model; S yx The remaining standard deviation, s represents the average value of the sample measurements in the experimental method. x The standard deviation of the sample measurements of the experimental method is represented by Δ, which is the clinical consistency threshold. cThe value of the independent variable at the medical decision level.

[0019] Optionally, consistency analysis of the validation samples may be performed using the MDL method, least squares regression method, and PB regression method.

[0020] Secondly, a quality control method based on quantitative consistency assessment is provided, the method comprising:

[0021] The sample size calculation method according to any one of claims 1-4 is used to extract a corresponding number of samples from the target dataset;

[0022] At least two different detection methods or systems are applied to the extracted samples to collect clinical trial data;

[0023] Perform a consistency analysis on the collected clinical trial data to assess the consistency between different testing methods;

[0024] If the consistency analysis results meet the preset consistency criteria, the test results of the extracted samples will be accepted for use in subsequent clinical trial studies.

[0025] If the consistency analysis results do not meet the preset consistency criteria, the testing method shall be adjusted according to the analysis results, or the sample shall be re-extracted and tested until the consistency criteria are met.

[0026] Thirdly, a sample size calculation system for quantitative consistency assessment is provided, the system comprising:

[0027] The setting module is used to determine the level of medical decision based on preset medical test indicators and to set clinical consistency thresholds;

[0028] A detection module is used to collect pre-test data; wherein, the collection of pre-test data includes performing a single test on a set of samples using at least two different detection methods or obtaining the data directly from a database; wherein, the detection methods include at least different analytical instruments, reagent kits, or detection platforms;

[0029] The calculation module is used to substitute the pre-experiment data into the MDL-based sample size estimation formula and calculate the sample size required to achieve the consistency requirements through multiple iterations.

[0030] The consistency analysis module is used to extract a corresponding number of validation samples from the overall experimental data based on the calculated sample size, and to perform consistency analysis on the validation samples.

[0031] The validation module is used to perform bootstrap simulation after the consistency analysis is passed to verify the robustness of the sample size calculation results. The bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power.

[0032] The determination module is used to determine the sample size estimate as the final sample size calculation result when the test power meets the preset statistical threshold.

[0033] Fourthly, a quality control system based on quantitative consistency assessment is provided, implemented in the sample size calculation system as described in the third aspect, the system comprising:

[0034] The extraction module is used to extract a corresponding number of samples from the target dataset based on the calculated results of the determined sample size.

[0035] The processing module is used to apply at least two different detection methods or systems to the extracted samples to collect clinical trial data; and to perform consistency analysis on the collected clinical trial data to evaluate the consistency between different detection methods.

[0036] The judgment module is used to accept the test results of the extracted samples for subsequent clinical trials if the consistency analysis results meet the preset consistency criteria; if the consistency analysis results do not meet the preset consistency criteria, the detection method is adjusted according to the analysis results, or the samples are extracted and tested again until the consistency criteria are met.

[0037] Fifthly, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the sample size calculation method described in any of the first aspects above.

[0038] In a sixth aspect, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the sample size calculation method described in any of the first aspects above.

[0039] In a seventh aspect, a computer program product is provided, on which a computer program is stored, and when the computer program is executed by a processor, it implements the sample size calculation method described in any of the first aspects above.

[0040] The beneficial effects of the technical solutions provided in this application include at least the following:

[0041] (1) This application provides a novel method for calculating sample size by combining medical decision level and least squares regression principles. This method can more accurately assess the consistency of two detection systems at a specific concentration, thereby improving the accuracy and reliability of sample size calculation.

[0042] (2) By introducing anticipated bias at the MDL, the method of this application improves the scientific rigor and practicality of quantitative consistency assessment. It provides a more reliable sample size estimate for medical testing, helping to ensure the validity of clinical trials and research results. Attached Figure Description

[0043] To more clearly illustrate the embodiments of this application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0044] Figure 1 A flowchart illustrating the steps of a sample size calculation method for quantitative consistency assessment provided in this application embodiment;

[0045] Figure 2 The expected bias plot at the medical decision level provided in the embodiments of this application;

[0046] Figure 3 This is a schematic diagram of the decomposition of Type I and Type II errors in a two-tailed test;

[0047] Figure 4 The system deviation diagrams for ALT1 and ALT2 are shown. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] In the description of this application, the terms "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations conceived in this application.

[0050] The medical and healthcare industry is entering a period of rapid development, with domestically developed diagnostic methods and the introduction of advanced international technologies continuously enriching the Chinese market. This progress has driven the demand for quantitative consistency evaluation methods. Sample size estimation is the core issue addressed in this application. However, the current methodological foundation for sample size calculation is weak, leading to discrepancies in guideline protocols and challenges in clinical practice. Therefore, improving the construction of sample size formulas for quantitative consistency evaluation is particularly important.

[0051] In recent years, the concept of Medicine Decide Level (MDL) has played a crucial role in various fields such as laboratory medicine and biochemistry. It aims to provide important data for disease diagnosis, grading, classification, and prognostic assessment by setting specific numerical limits, thereby guiding clinical decision-making and treatment strategy selection. Some scholars suggest that when selecting a regression model, emphasis should be placed on MDL, critical values, and reference limits, ensuring that the regression line passes through or approaches these clinically significant data points. This allows for more accurate model estimates at these specific concentrations, and thus a more accurate assessment of the consistency of testing instruments at these concentrations. If the 95% confidence interval of the expected bias at the MDL is within the pre-defined allowable error range, the two testing systems can be considered consistent.

[0052] This application will use the least squares regression principle, combined with the expected bias at the MDL, to derive the sample size for quantitative consistency evaluation. The accuracy of the formula will be verified through examples, and the results will be compared with those calculated by the sample size formula based on the regression coefficient and intercept, providing theoretical support for the application and promotion of this method.

[0053] The purpose of this application is to construct and validate a sample size calculation formula based on expected bias at the Medical Decision Level (MDL) for use in quantitative consistency assessment. Methods: Applying the principle of least squares regression, this application obtains a sample size calculation formula based on expected bias at the MDL. Simultaneously, the accuracy and robustness of the formula are validated using example data and Bootstrap simulation, and compared with the results calculated by a quantitative consistency sample size formula based on regression coefficients and intercepts. Results show that the sample size formula derived based on MDL is approximately consistent with the sample size calculated by the regression method. Validation using the results of these two formulas (121 / 139 samples) for both MDL and regression analyses yields conclusions satisfying quantitative consistency, and Bootstrap simulation validation achieves over 90% statistical power. Conclusion: This application is the first to integrate MDL and least squares regression to construct a sample size formula for quantitative consistency assessment, providing a new sample size calculation method for this field and strengthening its theoretical foundation.

[0054] Please refer to Figure 1 The document illustrates a flowchart of a sample size calculation method for quantitative consistency assessment provided in an embodiment of this application. This method may include the following steps:

[0055] S1 determines the level of medical decision-making based on preset medical testing indicators and sets clinical consistency thresholds.

[0056] S2, collect preliminary test data.

[0057] The collection of pre-test data includes performing a single test on a set of samples using at least two different detection methods or obtaining the data directly from a database; wherein the detection methods include at least different analytical instruments, reagent kits, or detection platforms.

[0058] The preliminary experimental data in this application include the mean of the sample measurements, the standard deviation of the sample measurements, and the residual standard deviation.

[0059] S3. Substitute the pre-experimental data into the MDL-based sample size estimation formula, and calculate the sample size required to achieve the consistency requirements through multiple iterations.

[0060] Specifically, according to the formula

[0061]

[0062] The sample size was obtained by processing the preliminary experimental data, where n is the sample size, α is the significance level, and α = 0.05 for a two-tailed test. n-2,1-α / 2 and t n-2,1-β / 2 , b0 and b1 represent the two-sided t-distribution critical values ​​at confidence levels of 1-α and 1-β, respectively; b0 represents the regression constant of the expected regression model; b1 represents the regression coefficient of the expected regression model; S yx The remaining standard deviation, s represents the average value of the sample measurements in the experimental method. x The standard deviation of the sample measurements of the experimental method is represented by Δ, which is the clinical consistency threshold. c The value of the independent variable at the medical decision level.

[0063] S4. Based on the calculated sample size, extract a corresponding number of validation samples from the overall experimental data and perform consistency analysis on the validation samples.

[0064] Consistency analysis of the validation samples includes the use of MDL method, least squares regression method, and PB regression method.

[0065] S5. After the consistency analysis is passed, a bootstrap simulation is performed to verify the robustness of the sample size calculation results.

[0066] The Bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power.

[0067] S6, when the test power meets the preset statistical threshold, the sample size estimate is determined as the final sample size calculation result.

[0068] In addition, if the test power does not meet the preset statistical threshold, the parameters in the sample size estimation formula are adjusted, such as increasing the sample size or adjusting the clinical consistency threshold. Then S3 to S5 are repeated until the test power meets the preset statistical power.

[0069] In addition, this application includes a quality control method based on quantitative consistency assessment, comprising:

[0070] Based on the determined sample size, a corresponding number of samples are extracted from the target dataset.

[0071] At least two different detection methods or systems are applied to the extracted samples to collect clinical trial data;

[0072] Perform a consistency analysis on the collected clinical trial data to assess the consistency between different testing methods;

[0073] If the consistency analysis results meet the preset consistency criteria, the test results of the extracted samples will be accepted for use in subsequent clinical trial studies.

[0074] If the consistency analysis results do not meet the preset consistency criteria, the testing method shall be adjusted according to the analysis results, or the sample shall be re-extracted and tested until the consistency criteria are met.

[0075] The following are specific embodiments of this application:

[0076] 1.1 Sample size calculation based on medical decision level.

[0077] In traditional regression analysis practice, researchers typically rely on two-sided one-term tests of the regression coefficients and the constant term. This application focuses on evaluating the confidence interval band of the population regression line. Figure 2 As shown, the confidence band resembles a "double flare," meaning it is narrower in the center and wider at both ends. Considering that MDL often exhibits extreme values ​​in patients' physiological and biochemical data, the confidence band is wider here. By comparing this confidence band width with a predetermined expected bias standard, if the confidence band width falls within the predetermined bias range, it can be more conservatively concluded that the two detection methods show good agreement.

[0078] According to the bias prediction algorithm in E09-A3, the difference between a given data point and the regression line is called the residual at that point. For point (x... i ,y i The residuals can be calculated using the following formula:

[0079]

[0080] At a given medical decision level x c At that time, expected bias B c The estimated value It can be represented in the following form:

[0081]

[0082] The 95% confidence interval is given by the following formula:

[0083]

[0084] In the formula, t represents the average value of the experimental method. (n-2,1-α / 2) Let represent the critical value of the two-tailed t-distribution when the degrees of freedom are n-2 and the confidence level is 1-α. Let the total significance level for the global hypothesis test in the consistency evaluation be α, and the total Type II error level be β. According to the principles of consistency evaluation, the total Type I error α can be decomposed into upper and lower tails, with both the two-tailed and one-tailed significance levels being α / 2. The total Type II error level β can also be decomposed into... The data consists of two parts: the lower limit of the 95% confidence interval for Type II error (β1) and the upper limit of the 95% confidence interval for Type II error (β2).

[0085] like Figure 3 The diagram shown illustrates the decomposition of Type I and Type II errors in a two-tailed test. Based on the t-distribution and the diagram, the following can be derived:

[0086]

[0087] Therefore, the formula for calculating Type II errors can be transformed into:

[0088]

[0089] Combining the two, the Type II error level is calculated as follows:

[0090]

[0091] Based on the properties of least squares regression:

[0092]

[0093] Among them, Syx To find the remaining standard deviation, further substituting formula (9) into formula (8), and simultaneously substituting... S is the sum of squares of the total deviations from the mean of the experimental method, where S x 2 Let be the sample variance of the measured values ​​of the experimental method. Δ is the clinical consistency threshold. Simplifying and rearranging this formula, we obtain the sample size estimation formula based on MDL (MDL-based quantitative consistency sample size formula):

[0094]

[0095] In the formula, α is the significance level, usually taken as α = 0.05 for two-tailed tests; 1 - β is the statistical power, not less than 80%; t n-2,1-α / 2 and t n-2,1-β / 2 These are the two-sided t-distribution critical values ​​for confidence levels of 1-α and 1-β, respectively; b0 represents the regression constant of the expected regression model, usually taken as 0; b1 represents the regression coefficient of the expected regression model, usually taken as 1; S yx The remaining standard deviation; S represents the average value of the sample measurements in the experimental method. x The standard deviation of the sample measurements represents the experimental method; Δ is the clinical consensus threshold. If a clinically accepted consensus threshold is difficult to obtain, Δ can also be taken as x. c The t-distribution threshold is 5%-10%. The t-distribution threshold is related to the sample size and can be determined using an iterative method. In the first iteration, the standard normal distribution quantile Z is used to replace the t-distribution threshold to obtain an initial value of n. The iteration stops when the result of n does not change between the two iterations. The final result is the sample size.

[0096] 1.2 Formula for calculating sample size using regression coefficients and intercepts.

[0097] Based on the principle of quantitative consistency evaluation using regression, and combined with the basic properties and statistical inference of the least squares method, the applicant obtained the sample size calculation formula (sample size formula for quantitative consistency using regression) n0 and n1 for quantitative consistency studies based on regression coefficients and regression constants.

[0098]

[0099] In the formula, α, 1-β, b0, b1, S yx , S x Same meaning as 1.1; t n-2,1-α / 4 and t n-2,1-β / 4These represent the two-sided t-distribution thresholds at confidence levels of 1-α and 1-β, respectively; Δ0 is the threshold of the clinically acceptable concordance bound (-Δ0, +Δ0) for the regression constant b0, and Δ1 is the threshold of the clinically acceptable concordance bound (1-Δ1, 1+Δ1) for the regression coefficient b1. When calculating the sample size, an iterative method is used to determine the t-distribution threshold, as it is related to the sample size. This method involves two formulas, one based on the regression constant and the other on the regression coefficient, to satisfy the confidence intervals of b0 and b1 and the clinically accepted concordance bounds. In practice, the sample size is calculated independently for each formula, and the larger of the two values ​​is taken as the accurately required sample size.

[0100] The following is an example analysis of this application:

[0101] 2.1 Calculate the sample size according to the formula to verify consistency.

[0102] A company recently developed a fully automated biochemical analyzer and established its own testing system. To compare the consistency of this system with other commonly used clinical testing systems in terms of measurement results, the company used two methods for single-sample testing. To ensure the study could achieve the expected testing power at a lower cost, a pilot test was first conducted on 30 volunteers, using two measurements of alanine aminotransferase (ALT) as an example to determine the optimal sample size.

[0103] To compare the differences between the formula derived in this application (MDL method formula) and the sample size formula derived based on regression coefficients and intercepts (tentatively referred to as the regression method formula), the sample size was calculated using both the MDL method and the regression method formulas for the example data. A corresponding number of samples were then drawn from the population, and MDL analysis and least squares regression analysis were performed respectively. Figure 4 The differences between the two detection systems are evident: in the low concentration range, this difference remains relatively constant; however, in the high concentration range, the difference increases with increasing concentration. This indicates the presence of mixed-type variance bias, and Passing-Bablok regression (PB regression) analysis should be performed on the data to determine whether the consistency conclusion holds. Therefore, PB regression is supplemented in the analysis to demonstrate the stability of the results. The specific steps for the comparison are as follows:

[0104] 1) Set the overall parameters: Set the Type I error level α = 0.05, the test power to 90%, MDL = 60 (U / L) [11,12], set the clinical consistency threshold to 5% of MDL, which is 3 (U / L), and set the clinical consistency threshold of the regression coefficient Δ0 = 3 (U / L) and the clinical consistency threshold of the regression constant Δ1 = 0.1 (U / L);

[0105] 2) Using 30 preliminary test data, calculate the sample size: Substitute the parameters obtained from the preliminary test into the MDL method formula and the regression method formula, and after multiple iterations, calculate the sample size required to achieve the consistency requirement, which are 121 cases and 139 cases respectively.

[0106] 3) Extracting validation samples of the expected sample size: Extracting a validation sample of size from the total experimental data, and using the MDL method, least squares regression method and PB regression method respectively to determine whether the consistency conclusion is valid.

[0107] The results showed that neither the MDL method nor the regression method could yield a consensus conclusion for the 30 cases in the preliminary trial. The MDL method calculated a 95% confidence interval (CI) of -0.10 to 1.88 for 121 samples in the MDL analysis, which was within the clinical consensus margin (5% x 100%). c The values ​​were in the range of (-3, +3) U / L, meeting the consistency requirement. In the least squares regression analysis, the 95% CI of b0 was (-0.87, 1.65), between (-3, +3), and the 95% CI of b1 was (0.98, 1.04), between (0.90, 1.10), meeting the consistency requirement. In the PB regression analysis, the 95% CI of b0 was (-2.09, 0.07), between (-3, +3), and the 95% CI of b1 was (0.99, 1.08), between (0.90, 1.10), meeting the consistency requirement. The regression method calculated 139 samples. In the MDL analysis, B... c The 95% CI for b0 is (-0.5, 1.51), falling within the range of (-3, +3) U / L, thus meeting the consistency requirement. In the least squares regression analysis, the 95% CI for b0 is (-1.19, 1.47), falling within the range of (-3, +3), and the 95% CI for b1 is (0.97, 1.03), falling within the range of (0.90, 1.10), thus meeting the consistency requirement. In the PB regression analysis, the 95% CI for b0 is (-1.72, 1.16), falling within the range of (-3, +3), and the 95% CI for b1 is (0.98, 1.08), falling within the range of (0.90, 1.10), thus meeting the consistency requirement. In the above examples, verification experiments were conducted based on the sample size calculation results of the MDL method and the regression method, and both analytical methods yielded the conclusion that consistency was valid. The MDL method calculates a sample size that is slightly smaller than that calculated by the regression method. However, given that the sample size calculated by the regression method is satisfied, both consistency analysis methods can ultimately achieve the expected consistency conclusions.

[0108] The specific results are shown in Table 1: Estimation results of ALT measurement consistency evaluation parameters.

[0109] Table 1. Estimation Results of ALT Measurement Consistency Evaluation Parameters

[0110]

[0111] 2.2 Bootstrap simulation to verify robustness.

[0112] To verify the robustness of the formula in this application, bootstrap simulations were performed. This process involves extensive repeated sample extraction and model estimation to evaluate the consistency and reliability of the model predictions under specific conditions.

[0113] 2.2.1 Bootstrap simulation verification of the MDL formula.

[0114] Set the number of iterations (1000), and draw 121 / 139 samples in each iteration. Perform Bootstrap sampling and model fitting: in each iteration, randomly draw samples from the data, allowing repeated sampling. Use these samples to fit a linear regression model, calculate and check confidence intervals: calculate in x c Predicted value at location Calculate S yx and as well as Use these statistics to calculate x c At 95% CI, output B C (95% CI). Check if the upper and lower limits of these confidence intervals are between (-3, +3). Calculate the test power: The test power is calculated by dividing the number of times the confidence interval conditions are met by the total number of iterations, and then outputting this ratio is the test power. (See Table 1)

[0115] 2.2.2 Bootstrap simulation verification of the regression formula.

[0116] Set the number of iterations (1000), and sample 121 / 139 samples in each iteration. Perform Bootstrap sampling and model fitting: in each iteration, randomly sample samples with replacement from the dataset. Fit a linear regression model using these samples. Calculate 95% confidence intervals for the fitted model, obtaining confidence intervals for the intercept (b0) and regression coefficients (b1). Determine if the confidence interval for the intercept is between (-3, +3) and the confidence interval for the regression coefficients is between (0.9, 1.1). If both conditions are met, consistency is achieved. Calculate the proportion of iterations that meet the consistency criteria out of the total number of iterations, as the test power. (See Table 1)

[0117] The results show that both the 121 samples calculated by the MDL method and the 139 samples calculated by the regression method have a statistical power of over 90% when performing Bootstrap simulation validation. Although the test power decreased after changing the least squares regression to PB regression, it still reached over 80%, proving that the sample size results obtained by the formula in this application are very robust.

[0118] It is evident that sample size estimation is crucial in quantitative consistency assessment. However, current guidelines only broadly specify minimum sample size requirements for different categories of in vitro diagnostic reagents, without detailing specific sample size calculation methods. Therefore, in practice, researchers often estimate sample size based on experience. This approach lacks scientific basis and may reduce the statistical power, thus overlooking potentially clinically significant results. Therefore, developing a reasonable sample size formula for quantitative consistency assessment is urgently needed.

[0119] The applicant previously developed a formula for calculating the sample size based on the regression coefficients and intercepts, and verified the formula's correctness using three example datasets. However, this formula still has certain limitations:

[0120] 1) It involves two sample size formulas and selects the larger sample size as the final result, which is complex and inefficient.

[0121] 2) If clinical consistency thresholds are missing or not predetermined, consistency assessment cannot be performed;

[0122] 3) Previous methods primarily focused on the goodness of fit of the overall data, neglecting in-depth analysis of specific data points. This can lead to the omission of crucial data information, affecting the accuracy and applicability of research results in patient populations.

[0123] In this application, a formula for calculating the corresponding sample size is derived based on the expected bias at the MDL (Mean Differential Length). Through application to real data and Bootstrap simulation, this application verifies the accuracy and robustness of the formula and compares it with previous sample size calculation methods based on regression coefficients and intercepts. The introduction of MDL enhances the clinical application value of this formula; using x... C The (5%-10%) alternative clinical conformity limits provide a new solution for conformity studies where it is difficult to obtain clinical conformity limits.

[0124] In the above example, although the preliminary test of 30 samples could not verify consistency, b0 was contained in (-3, +3) and b1 was contained in (0.90, 1.10). This indicates that the inability to determine consistency was due to the small sample size of 30, resulting in an excessively wide confidence interval. The 121 samples calculated using the MDL formula and the 139 samples calculated using the regression method were subjected to MDL analysis, least squares regression analysis, and PB regression analysis for verification, respectively, and all yielded consistent results. The sample size obtained by the MDL method was slightly smaller than that of the regression method, and the bootstrap simulation using the MDL formula yielded the highest test power, suggesting that the MDL formula has higher computational power.

[0125] The Bland-Altman (BA) plot method is also one of the most commonly used methods in domestic and international consistency evaluation studies. The standard BA method has certain applicable conditions: the difference follows a normal distribution, the difference does not change with the measured value, and the variance of the difference does not change with the measured value. In practice, Fu Yang et al. found some problems: a large number of in vitro diagnostic reagent registration application materials containing the Bland-Altman method lack pre-determined consistency evaluation criteria; Lu Mengjie et al. found that researchers often fail to judge the conditions of the BA method when applying it, blindly applying the standard method, leading to doubts about the correctness of the conclusions. The example data used in this application shows a right-skewed distribution, and the difference increases with the increase of the measured value (e.g., Figure 4 As shown in the figure, this data is not suitable for BA plot analysis. Furthermore, since the derivation of the MDL method and regression method formulas is based on least squares regression theory, which assumes that the data follows a normal distribution, we incorporated PB regression, used for nonparametric regression analysis, when validating the sample size. The results show that PB regression analysis can still verify consistency and achieve a certain level of power (80%). This suggests that the MDL method's quantitative consistency sample size formula and the regression method's quantitative consistency sample size formula have a wide range of applications and can accurately estimate the sample size of some skewed data.

[0126] In practical applications, considering that many physiological and biochemical indicators in clinical practice have multiple MDLs, using the MDL formula for sample size calculation can lead to Type II error inflation. We can use the Bonferroni correction method to correct for Type II error, where the Type II error should be β / 2E (where E is the number of selected MDLs). Furthermore, in this application, the calculation results before and after iteration are approximately consistent. With a large sample size, the t-distribution threshold is close to the Z-critical value of the standard normal distribution. Therefore, in practical applications, the Z-critical value can be used instead of the t-critical value.

[0127] In summary, this application integrates medical decision level with least squares regression to construct a sample size formula for quantitative consistency evaluation, providing a new sample size calculation method for this field and enhancing the theoretical foundation of research in this field.

[0128] This application provides a sample size calculation system for quantitative consistency assessment, which may include:

[0129] The setting module is used to determine the level of medical decision based on preset medical test indicators and to set clinical consistency thresholds;

[0130] A detection module is used to perform a single test on a set of samples using at least two different detection methods and collect preliminary test data; wherein the detection methods include at least different analytical instruments, reagent kits, or detection platforms;

[0131] The calculation module is used to substitute the pre-experiment data into the MDL-based sample size estimation formula and calculate the sample size required to achieve the consistency requirements through multiple iterations.

[0132] The consistency analysis module is used to extract a corresponding number of validation samples from the overall experimental data based on the calculated sample size, and to perform consistency analysis on the validation samples.

[0133] The validation module is used to perform bootstrap simulation after the consistency analysis is passed to verify the robustness of the sample size calculation results. The bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power.

[0134] The determination module is used to determine the sample size estimate as the final sample size calculation result when the test power meets the preset statistical threshold.

[0135] In one embodiment, an electronic device, which may be a computer, is provided. The electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for sample size calculation data for quantitative consistency assessment. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a sample size calculation method for quantitative consistency assessment.

[0136] In one embodiment of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the sample size calculation method described above for quantitative consistency assessment.

[0137] In one embodiment of this application, a computer program product is provided, including a computer program / instructions, which, when executed by a processor, implements the steps of the sample size calculation method described above for quantitative consistency assessment.

[0138] The computer-readable storage medium and computer program product provided in this embodiment are similar in implementation principle and technical effect to the above method embodiments, and will not be repeated here.

[0139] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in M ​​forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0140] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0141] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for calculating sample size for quantitative consistency assessment, characterized in that, The method includes: S1 determines the level of medical decision-making based on preset medical testing indicators and sets clinical consistency thresholds; S2, collecting pre-test data; wherein, the collection of pre-test data includes performing a single test on a set of samples using at least two different detection methods or obtaining the data directly from a database; wherein, the detection methods include at least different analytical instruments, reagent kits, or detection platforms; S3, Substitute the pre-test data into the MDL-based sample size estimation formula, and calculate the sample size required to achieve the consistency requirements through multiple iterations; S4. Based on the calculated sample size, extract a corresponding number of validation samples from the overall experimental data and perform consistency analysis on the validation samples. S5. After the consistency analysis is passed, a bootstrap simulation is performed to verify the robustness of the sample size calculation results. The bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power. S6, when the test power meets the preset statistical threshold, the sample size estimate is determined as the final sample size calculation result.

2. The sample size calculation method according to claim 1, characterized in that, The method further includes: If the test power does not meet the preset statistical threshold, adjust the parameters in the sample size estimation formula, such as increasing the sample size or adjusting the clinical consistency threshold, and then repeat S3 to S5 until the test power meets the preset statistical power.

3. The sample size calculation method according to claim 1, characterized in that, Substituting the pre-experimental data into the MDL-based sample size estimation formula, the required sample size to achieve the consistency requirement is calculated through multiple iterations, specifically including: According to the formula The sample size was obtained by processing the preliminary experimental data, where n is the sample size, α is the significance level, and α = 0.05 for a two-tailed test. n-2,1-α / 2 and t n-2,1-β / 2 , b0 and b1 represent the two-sided t-distribution critical values ​​at confidence levels of 1-α and 1-β, respectively; b0 represents the regression constant of the expected regression model; b1 represents the regression coefficient of the expected regression model; S yx The remaining standard deviation, s represents the average value of the sample measurements in the experimental method. x The standard deviation of the sample measurements of the experimental method is represented by Δ, which is the clinical consistency threshold. c The value of the independent variable at the medical decision level.

4. The sample size calculation method according to claim 1, characterized in that, Consistency analysis of validation samples includes the use of MDL method, least squares regression method, and PB regression method.

5. A quality control method based on quantitative consistency assessment, characterized in that, The method includes: Using the sample size calculation result determined by the sample size calculation method according to any one of claims 1-4, a corresponding number of samples are extracted from the target dataset; At least two different detection methods or systems are applied to the extracted samples to collect clinical trial data; Perform a consistency analysis on the collected clinical trial data to assess the consistency between different testing methods; If the consistency analysis results meet the preset consistency criteria, the test results of the extracted samples will be accepted for use in subsequent clinical trial studies. If the consistency analysis results do not meet the preset consistency criteria, the testing method shall be adjusted according to the analysis results, or the sample shall be re-extracted and tested until the consistency criteria are met.

6. A sample size calculation system for quantitative consistency assessment, characterized in that, The system includes: The setting module is used to determine the level of medical decision based on preset medical test indicators and to set clinical consistency thresholds; A detection module is used to collect pre-test data; wherein, the collection of pre-test data includes performing a single test on a set of samples using at least two different detection methods or obtaining the data directly from a database; wherein, the detection methods include at least different analytical instruments, reagent kits, or detection platforms; The calculation module is used to substitute the pre-experiment data into the MDL-based sample size estimation formula and calculate the sample size required to achieve the consistency requirements through multiple iterations. The consistency analysis module is used to extract a corresponding number of validation samples from the overall experimental data based on the calculated sample size, and to perform consistency analysis on the validation samples. The validation module is used to perform bootstrap simulation after the consistency analysis is passed to verify the robustness of the sample size calculation results. The bootstrap simulation includes setting the number of iterations, performing sampling and model fitting, calculating the 95% confidence interval of the expected bias at MDL or the 95% confidence interval of the regression constant and regression coefficient, and statistically analyzing the proportion of iterations that meet the consistency requirements as a measure of test power. The determination module is used to determine the sample size estimate as the final sample size calculation result when the test power meets the preset statistical threshold.

7. A quality control system based on quantitative consistency assessment, implemented in the sample size calculation system as described in claim 6, characterized in that, The system includes: The extraction module is used to extract a corresponding number of samples from the target dataset based on the calculated results of the determined sample size. The processing module is used to apply at least two different detection methods or systems to the extracted samples to collect clinical trial data; and to perform consistency analysis on the collected clinical trial data to evaluate the consistency between different detection methods. The judgment module is used to accept the test results of the extracted samples for subsequent clinical trials if the consistency analysis results meet the preset consistency criteria; if the consistency analysis results do not meet the preset consistency criteria, the detection method is adjusted according to the analysis results, or the samples are extracted and tested again until the consistency criteria are met.

8. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the sample size calculation method as described in any one of claims 1 to 4.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the sample size calculation method as described in any one of claims 1 to 4.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the sample size calculation method according to any one of claims 1 to 4.

Images