# Diagnostic value of MRI for posttreatment surveillance of early-stage (I–II) glottic larynx cancer

## TL;DR

This study evaluates how well MRI can detect cancer recurrence in patients with early-stage glottic larynx cancer after treatment.

## Contribution

The study provides insights into the diagnostic accuracy of MRI for posttreatment surveillance of early-stage glottic larynx cancer.

## Key findings

- MRI showed high specificity (99%) but limited positive predictive value for detecting recurrences.
- Recurrences were identified with 75% sensitivity using MRI.
- A notable proportion of MRIs were inconclusive, raising concerns about MRI as a standalone surveillance tool.

## Abstract

There is no uniformity across various guidelines in defining the modality and frequency of the follow-up, particularly regarding radiological imaging. The objective is to assess the diagnostic performance of magnetic resonance imaging (MRI)-based posttreatment surveillance for early-stage (I–II) glottic squamous cell carcinoma of the larynx.

The follow-up of patients diagnosed with glottic squamous cell carcinoma of the larynx, treated with radiotherapy or surgery in curative intent, was analyzed over a period of 2 years posttreatment. MRI diagnostic performance metrics were calculated using clinical and post-MRI endoscopic examinations as the reference standard. MRI sequences included both anatomical and functional imaging, including diffusion-weighted imaging.

In total, 171 eligible MRIs were analyzed in the follow-up. Recurrences were identified with a sensitivity of 75% and a specificity of 99%. However, the positive predictive value of MRI surveillance reflects considerable uncertainty in the diagnosis of recurrences based solely on MRI findings, dropping as low as 21% in sensitivity analyses. Moreover, a notable proportion of MRIs were inconclusive.

MRI demonstrates high specificity and acceptable sensitivity; however, the limited positive predictive value raises concerns regarding its utility as a stand-alone surveillance tool.

The online version of this article (10.1007/s00066-025-02460-6) contains supplementary material, which is available to authorized users.