Method for Determining the Required Sample Size
This study is observational and involves no intervention. Therefore, the sample size was estimated using the following formula:
where N is the sample size, Uα is the value of υ corresponding to the test level α, S is the overall standard deviation, and δ is the allowable error.
Since the specific value of δ could not be known in advance, the study referred to the literature and identified three widely accepted methods for estimating δ. Method ①: Conduct a pre-experiment before the study begins, using the inter-group difference as the δ value directly. Method ②: Consult relevant authorities in advance to determine a professionally meaningful δ value. Method ③: In the absence of pre-experiment results and expert opinions, it is permissible to use 0.25 or 0.50 times the standard deviation as δ. According to the allowable error δ reference flowchart by Ni Yanyan and Zhang Jinxin, this study used method ③ to determine δ. Therefore, at α = 0.05, the minimum sample size N can range from 16 to 62 cases. Considering a loss-to-follow-up rate of 15%-20%, the sample size can be expanded to 19-78 cases. Finally, considering that Logistic regression analysis will be used in statistical analysis with 15 quantitative indicators as independent variables, the estimated sample size should be at least 150 cases. Taking into account factors such as economy and time, the sample size for this study is determined to be 150-200 cases.
Research Content
The collection of glioma patients is divided into two parts: the first part is a retrospective study based on already collected glioma patients, and the second part is a prospective collection of glioma patients; detailed as follows:
First Part: Retrospectively collect glioma patients who visited our hospital from May 1, 2014, to December 31, 2024, according to the inclusion criteria.
Second Part: Prospectively collect glioma patients who visited our hospital from January 1, 2021, to December 31, 2022, according to the inclusion criteria. Furthermore, track, obtain, and analyze tumor specimens from the enrolled glioma patients, conducting tumor grade analysis and detecting major molecular gene mutation status related to prognosis (including IDH, 1p19q, TERT, ATRX, EGFR amplification, 7+/10-, etc.). By using quantitative, qualitative, radiomics, and machine learning methods to analyze multiparametric MRI data of glioma patients, statistical analysis and modeling will be performed to establish a model based on multiparametric MRI for predicting glioma grade and prognosis-related molecular mutation status.
Survey Content The outcome indicators are glioma grade and prognosis-related molecular mutation status, which will be tested by the pathology department personnel. The exposure factor is the quantitative analysis data from multiparametric MRI examinations, and potential confounding factors are the age and sex of the subjects; these indicators will be searched, measured, and recorded by specialized personnel in the medical system.
Data Management and Statistical Analysis Plan Data Management: Use EXCEL software to establish an electronic spreadsheet, with specialized personnel manually inputting data, such as general information (replaced by patient numbers), age, sex, etc., multiparametric MRI examination data such as Ktrans values obtained after DCE-MRI post-processing, and pathological data such as tumor grade and classification, different molecular mutation statuses, etc., saving and archiving data in real-time.
Statistical Analysis: Use SPSS 25.0 statistical analysis software for data analysis. For quantitative data such as Ktrans values obtained after DCE-MRI post-processing, perform normality tests with S-W and Levene's tests. Data that is normally distributed is expressed as mean ± standard deviation, and non-normally distributed data is expressed as M (P25, P75). Prognosis-related molecular mutation status, such as IDH mutation, is dichotomous data, with mutations, deletions, or amplifications
Bias Control Bias that is prone to occur in this observational study is information bias. The main research steps where bias is likely to occur and the control methods are as follows: ① Acquisition process of multiparametric MRI quantitative data. The control method involves establishing uniform measurement standards before analysis and adhering strictly to these standards; measuring multiple tumor areas and taking the average result as the final result; and ensuring that the measurements are conducted by the same person throughout. ② Data entry process. The control method involves timely review of each data item during data collection, and randomly selecting some data for double entry to check the quality of data entry. ③ Pathology. The control method involves classifying according to the 2016 WHO new classification standards for glioma; and attempting to control the detection of molecular mutation status to be done by the same person using the same batch of test reagents at the same time. In addition to this, factors such as age and sex may be confounding variables that can cause confounding bias. The control method involves collecting patients' age, sex, etc., during the data collection process, and using multivariate statistical analysis methods, stratified analysis methods, etc., to control bias during statistical analysis.
Quality Management The entire research process is carried out according to the experimental plan, with particular attention to controlling bias. The quality of the study is evaluated based on the quality evaluation criteria recommended by the Agency for Healthcare Research and Quality (AHRQ). The criteria include 11 items, which are answered with "yes," "no," or "unclear": (1) Is the source of the data (survey, literature review) clearly stated? (2) Are the inclusion and exclusion criteria for the exposed and non-exposed groups (cases and controls) listed, or are previous publications referenced? (3) Is the time frame for identifying patients provided? (4) If not from a population source, is the study population continuous? (5) Do the subjective factors of the evaluator obscure other aspects of the study subjects? (6) Is any assessment described that was done to ensure quality (e.g., testing/retesting of primary outcome indicators)? (7) Is the reason for excluding any patients from the analysis explained? (8) Are the measures for evaluating and/or controlling confounding factors described? (9) If possible, is the handling of missing data in the analysis explained? (10) Is the patient response rate and the completeness of data collection summarized? (11) If there is follow-up, is the percentage of expected incomplete data or follow-up results identified. The observational study report is written in the format of the STROBE (the Strengthening the Reporting of Safety Evaluation The main content of this study involves the secondary use of medical records, imaging data, and biological specimens, posing almost no risk to the participants.
Ethical Review and Informed Consent This study complies with medical ethical standards and has applied for ethical approval from the Ethics Committee of the First Affiliated Hospital of Sun Yat-sen University. For the prospective study part, informed consent will be obtained from the participants, and an informed consent form will be signed. The study will strictly adhere to the rules of patient information confidentiality, not recording patient names, and distinguishing cases by image number or examination number. The participants' multiparametric MRI data examination and analysis are non-invasive, involving only the extraction and analysis of image data, with the participants' personal information kept confidential throughout. Subsequent tumor sample analysis is based on specimens already obtained from surgical resection or biopsy, hence there is no additional risk in sample collection. During the study, the main molecular mutation status related to prognosis in the participants' tumor samples will be analyzed. This information will be beneficial for the clinical diagnosis, treatment, and prognosis assessment of cancer, from which the participants will benefit.
