Development and Validation of a Biopsy Decision Model for Nasopharyngeal Carcinoma Based on Serological Markers and Endoscopic Features

Participants selected from existing clinical records who underwent nasopharyngeal endoscopy and clinically indicated nasopharyngeal biopsy for NPC risk assessment or suspected NPC. EBV-related blood biomarker results, nasopharyngeal endoscopic scores, and pathological diagnoses will be analyzed for model development.

EBV-related blood biomarkers, including P85-Ab, EBNA1-IgA, and VCA-IgA, among others, will be evaluated together with nasopharyngeal endoscopic scores to support biopsy decisions for nasopharyngeal carcinoma. Biopsy and treatment decisions will follow routine clinical practice.

Participants from the retrospective study population who are not included in the training cohort. Their EBV-related blood biomarker results, nasopharyngeal endoscopic scores, and pathological diagnoses will be used for internal validation of the developed models.

EBV-related blood biomarkers, including P85-Ab, EBNA1-IgA, and VCA-IgA, among others, will be evaluated together with nasopharyngeal endoscopic scores to support biopsy decisions for nasopharyngeal carcinoma. Biopsy and treatment decisions will follow routine clinical practice.

Participants undergoing nasopharyngeal endoscopy and clinically indicated nasopharyngeal biopsy at participating centers will be enrolled prospectively. EBV-related blood biomarker results, nasopharyngeal endoscopic scores, and pathological diagnoses from clinically indicated biopsies will be collected for external validation of the developed models. Clinical decisions will follow routine practice.

EBV-related blood biomarkers, including P85-Ab, EBNA1-IgA, and VCA-IgA, among others, will be evaluated together with nasopharyngeal endoscopic scores to support biopsy decisions for nasopharyngeal carcinoma. Biopsy and treatment decisions will follow routine clinical practice.

Sensitivity and corresponding 95% confidence intervals will be assessed and compared across models based on EBV-related blood biomarkers alone, nasopharyngeal endoscopic scores alone, and their combination at prespecified risk thresholds, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

Specificity and corresponding 95% confidence intervals will be assessed and compared across models based on EBV-related blood biomarkers alone, nasopharyngeal endoscopic scores alone, and their combination at prespecified risk thresholds, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

AUCs and corresponding 95% confidence intervals will be assessed and compared across models based on EBV-related blood biomarkers alone, nasopharyngeal endoscopic scores alone, and their combination, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

The calibration intercept will be calculated for each model using predicted probabilities and histopathological diagnoses, separately in the internal validation cohort and the prospective external validation cohort. An intercept of 0 indicates ideal calibration in this respect.

The calibration slope will be calculated for each model using predicted probabilities and histopathological diagnoses, separately in the internal validation cohort and the prospective external validation cohort. A slope of 1 indicates ideal calibration in this respect.

The Brier score will be calculated for each model as the mean squared difference between predicted probabilities and histopathological outcomes, coded as 1 for nasopharyngeal carcinoma and 0 otherwise, separately in the internal validation cohort and the prospective external validation cohort. Scores range from 0 to 1, with lower scores indicating better overall predictive accuracy.

Positive predictive values and corresponding 95% confidence intervals will be calculated for each model at prespecified risk thresholds, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

Negative predictive values and corresponding 95% confidence intervals will be calculated for each model at prespecified risk thresholds, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

Decision curve analysis will be used to assess and compare the net benefit of biopsy strategies based on serological, endoscopic, and combined models with biopsy-all and biopsy-none strategies across risk thresholds, separately in the internal validation cohort and the prospective external validation cohort, using nasopharyngeal biopsy histopathology as the reference standard.

For each serological, endoscopic, and combined model, the potential percentage of biopsies avoided will be calculated as the percentage of all evaluated participants undergoing biopsy whose predicted risk is below the prespecified threshold. This represents the potential reduction in biopsies relative to a biopsy-all strategy and will be assessed separately in the internal validation cohort and the prospective external validation cohort.