Carcinoma of the uterine cervix: aspects on preoperative staging and assessment of treatment effect using magnetic resonance imaging

Abstract

Background: Uterine CC mainly affects young women as it is caused by persistent HPV infection most often acquired during adolescence. Early-stage disease is treated with surgery, and locally advanced disease with chemoradiotherapy. Both treatment methods are associated with infertility and morbidity. Combined treatment modalities (both surgery and CRT) should be avoided as this has negative impact on morbidity. The choice of treatment depends on the tumour stage at diagnosis, and accurate initial staging is therefore essential. MRI has been an integral part of the routine diagnostic work-up for CC patients at Karolinska since 2003. MRI is also an important tool in the evaluation of treatment, which traditionally is performed after the treatment is completed. However, identification of imaging parameters for prediction of the treatment effect early during the therapy, would provide more individualized treatment. This would enable early change of the treatment strategy for those predicted to experience treatment failure and may contribute to less morbidity in others.

Aim: The overall aim of this thesis was to identify an optimal examination protocol for MRI of the pelvis in patients with biopsy verified CC scheduled for surgery, improve identification of risk factors for recurrence, and identify MRI standards for therapy monitoring, in particular the timing of MRI in relation to treatment.

In Study I: 57 patients with early-stage disease treated with surgery were evaluated with three different sets of MR protocols. We found that magnetic resonance tumour (mrT), magnetic resonance lymph node (mrN) and magnetic resonance distant metastases (mrM) stages were the same for a basic standard protocol including transaxial and sagittal T2-weighted images, and transaxial T1-weighted sequences, as for protocols with the addition of oblique and/or contrast-enhanced sequences. The inter-observer agreement was “good” between readers for all three protocols. The agreement among readers was negatively affected by prior conization of the patient.

Study II: included 102 patients comprising the patients in study I with the addition of patients receiving brachytherapy prior to surgery. Two main groups were compared regarding 10-year outcome, those with visible and non-visible tumours on pre-treatment MRI. Tumour recurrence was seen in 17.9% of patients with visible tumour, and in 17.6% of patients with non-visible tumours. Recurrence free survival (RFS) was longer for patients having undergone conization prior to MRI than for those who had not.

Study III: The inter-observer agreement among experienced and less experienced observers of MRI and TVS was investigated in 60 patients with all stages of CC for this study. For all MRI observers, the inter-observer agreement was “good” for assessment of stromal- and parametrial invasion (PMI). Only for tumour detection was inter-observer agreement lower for the less experienced observers. For TVS observers, the agreement was “moderate” for assessment of tumour detection, stromal- and parametrial invasion. The agreement was significantly higher among experienced TVS observers regarding PMI.

Study IV: was designed as a pilot-study including 15 patients with stage IB2-IIIB scheduled for concomitant chemoradiotherapy (CRT). MRI was performed at baseline, 3 weeks, 5 weeks, and 12 weeks after treatment start. During follow-up, 7 patients relapsed, (“poor prognosis group”, PP), 8 patients did not relapse (“good prognosis group”, GP). We compared tumour size, change in size (Δsize), ADC and change in ADC (ΔADC), and tumour visibility on MRI at all four time points between the PP and GP group. By combining tumour size at baseline with tumour visibility on DWI at 5 weeks, the area under the curve (AUC) in receiver operating characteristics (ROC) analysis reached 0.83.

The findings of this thesis confirm the value of MRI for CC staging and therapy follow-up. In early-stage disease, unequivocally without parametrial invasion, evaluation consisting of a basic standard protocol including transaxial and sagittal T2-weighted images, and transaxial T1-weighted sequences is not improved by addition of oblique and contrast-enhanced sequences. Interpretation of the images is affected by prior conization but the clinical importance of detecting small tumours on MRI can be questioned as tumour visibility in early-stage disease does not affect long term outcome. Interobserver agreement is higher for MR than for TVS. A reasonable level of inter-observer agreement can be achieved for both experienced and less experienced observers of MRI and TVS, after attending a short basic training session on evaluation of cervical tumours. Prediction of disease recurrence seems feasible by combining visibility on high b-value diffusion-weighted imaging (DWI) at 5 weeks after treatment start with tumour size at baseline MRI. For all four studies in this thesis, the limited number of patients must be considered, and results need to be confirmed in larger cohorts.