Introduction
This document serves as a formal correction to the research article titled "Radiomics on multi-modalities MR sequences can subtype patients with non-metastatic nasopharyngeal carcinoma (NPC) into distinct survival subgroups," originally published in European Radiology in March 2019. The purpose of this correction is to rectify errors in the descriptive statistics of the patient cohorts presented in the original publication.
Background and Objective
In the field of clinical oncology, the ability to accurately subtype patients is paramount for personalized treatment planning and the prediction of survival outcomes. The original study explored the utility of radiomics—the high-throughput extraction of quantitative features from medical imaging—to identify distinct survival subgroups among patients with non-metastatic nasopharyngeal carcinoma (NPC). By leveraging multi-modality MRI sequences, the researchers aimed to develop a predictive model that transcends traditional clinical staging.
Corrected Patient Characteristics
The integrity of a radiomics model relies heavily on the accuracy of the baseline demographic and clinical data of the cohorts used for training and validation. This correction provides the updated Table 1, which details the characteristics of the 658 patients involved in the study.
Cohort Distribution
The study utilized two distinct groups:
- Training Cohort (N = 424): This group was used to build and refine the radiomic subtyping model.
- Validation Cohort (N = 234): This independent group was used to test the model's generalizability and predictive accuracy.
As shown in the corrected table, the cohorts are well-balanced. The median age was 45 years in the training cohort and 44 years in the validation cohort. Both groups showed a similar sex distribution, with approximately 71-74% male patients, reflecting the typical demographics of NPC in Southern China.
Clinical Staging and Follow-up
The cohorts included a comprehensive range of T (tumor) and N (node) categories. Notably, a significant portion of the patients presented with advanced non-metastatic disease, with Stage III (42.9% in training, 32.5% in validation) and Stage IV-a (26.9% in training, 26.5% in validation) accounting for the majority of cases. The median follow-up time was approximately 63 months for both cohorts, providing a robust longitudinal window for assessing survival outcomes.
Conclusion
While the core methodology and findings of the original radiomics study remain supported, this correction ensures that the foundational clinical data is accurately represented in the scientific record. Maintaining data integrity is essential in high-dimensional imaging studies to ensure the reliability and reproducibility of predictive models in oncology.
Summary updated