EP083/#954 Diagnostic accuracy of pretreatment imaging regarding lymph node metastasis in early stage cervical cancer

Objectives

Imaging modalities are increasingly used to assess lymph node involvement in early-stage cervical cancer patients even though there is a paucity of evidence on the diagnostic accuracy. This retrospective study aims to evaluate the accuracy of MRI, CT, and PET-CT in detecting nodal metastases.

Methods

Women diagnosed between 2009–2017 with a pretreatment CT, MRI, or PET-CT were selected from the Netherlands Cancer Registry. The pelvic ± para-aortic nodal status was registered according to radiological judgement (i.e. negative, inconclusive or positive); inconclusive nodes were categorized as positive. Pathological results were considered the gold standard in calculating accuracy indices for FIGO 2009 IA-IIA. If pathological results were missing, multiple imputation was applied to limit verification bias risk.

Results

Of 1,955 patients included, the gold standard was available in 84%. Nodal evaluation was assessed in 1,431, 685, and 288 patients by MRI, CT, and PET-CT, respectively. Analyses of original and imputed data are presented in table 1. With an AUC of 0.746, PET-CT was more accurate than MRI (0.717) and CT (0.677) in detecting nodal metastases, but not regarding specificity: 64%, 91%, and 90%, respectively.

Conclusions

In early-stage cervical cancer, the accuracy of PET-CT was superior to MRI and CT in the detection of nodal metastases. Although, this might be related to its use as a second verification modality. Limiting verification bias by multiple imputation positively influences the diagnostic accuracy of pretreatment imaging and increases the incidence of nodal metastases, given that data on positive-imaging with pathologic evaluation in early-stages are often missing.