Consecutive magnetic resonance imaging during brachytherapy for cervical carcinoma: predictive value of volume measurements with respect to persistent disease and prognosis

Cervical cancer is associated with a yearly mortality of 270.000 patients [1‐3]. The presence of a persistent cervical carcinoma after radiotherapy is a significant predictor of patient survival. In a select group of patients with persistent tumors after radiotherapy, surgical resection can be beneficial [4, 5].

The detection of residual tumors by gynecological examination, MRI, and PET-CT is complicated by the occurrence of false positive results and interobserver dependency, which are mainly due to post-radiation induced fibrosis. Diagnoses based on histological biopsies are limited by the occurrence of false negative results [6‐9]. It has been reported that tumor volume assessment on repeated MRIs during radiotherapy treatment can be used to predict tumor response to radiotherapy [10‐13]. In the past decade a new concept was clinically introduced by the GEC-ESTRO working group taking the tumor regression during external beam radiation into account for treatment planning with brachytherapy (BCT) [14]. Applying this concept, i.e., MR-image guided adaptive brachytherapy (IGABT), assessing the residual tumor volume and/or target volume at high risk at the time of BCT has been demonstrated to improve outcome significantly in locally advanced cervix carcinoma [14‐16]. For optimal assessment of residual tumor burden at the time of BCT T2-weighted MR images are obtained with the BCT applicator in place. This imaging sequence results in the best quality to discriminate tumor from cervical stroma. Applicator based distortions and artefacts are acceptable for T2W sequences using MR-compatible applicators [17].

Until now, volumetric analysis for predicting treatment response on MRI used in this setting, with the BCT applicator in place, has never been used. If MR-based tumor volume assessment at the time of BCT is a reliable option, it could be used to predict tumor response both local and for overall outcome.

The aim of our study was to assess if absolute tumor volume and tumor regression at the time of BCT can discriminate between patients who eventually achieve a complete response from those who do not. The second objective was to evaluate whether local volume regression can predict overall treatment failure.

A total of 48 consecutive patients were identified, of which 35 met the inclusion criteria. Their baseline characteristics are presented in Table 1.

Our study shows that absolute volume measurements as well as volume regression determination at the time of the final BCT are accurate predictors for treatment outcome. Performing volume measurements, i.e., the absolute tumor volume and tumor volume regression (Δvolume) at the time of the final BCT treatment predicted residual tumor after treatment and overall treatment failure during follow-up. Volume analysis at first BCT gave promising results however volumetry during the final BCT was superior for predicting residual disease. Indicative for the impact of BCT treatment for treatment outcome in these tumors. As the results of absolute volume analysis were comparable to those of volume regression analysis, absolute volume measurements for the prediction of residual tumor after BCT could be a better option in general practice because it is less time-consuming.

The presence of persistent disease after radiation treatment is a significant predictor of patient survival, early detection of tumors in this group of patients is of clinical importance. For patients with residual disease situated in cervical or adjacent tissue, surgical resection after radiation treatment can be beneficial [4, 5]. To date, most patients receive expectant management until disease recurrence, which might lead to more extensive surgery with increased morbidity or to a situation where the patient might not even be operable because of large tumor size, local involvement of bladder or bowel, distant metastasis, or extensive radiation induced fibrosis. As radiation-induced fibrosis in the target area is lower after completion of EBRT, surgery is an easier option at that time.

Outcome in locally advanced cervix carcinoma has significantly improved through dose escalation enabled by IGABT [14]. The introduction of IGABT resulted in an increase in overall local control (91–95 %) and a reduction of late toxicity by 50–60 % [14]. Our results show that volumetric analysis during radiation treatment can predict treatment failure for individual patients. In these patients therapy could be further tailored to for example surgery and/or adjuvant systemic treatment.

Several groups have performed semi-objective volumetric tumor measurements for the assessment of treatment responses [10‐13, 20, 21]. Most of these studies were performed retrospectively with small study populations, the imaging protocols employed differed from study to study, and volume evaluations were obtained using ellipsoid based volume measurements, which are inferior to those obtained by manual and freehand delineation of regions of interest [18].

Mayr et al. reported the results of MRI-based volume analysis during treatment at a single institution [10‐13]. Wang and Mayr et al. prospectively assessed reductions in tumor volumes measured by sequential MRI during treatment and found a correlation between the volume reduction ratio and 5-year local control rates [13]. Volume regression after EBRT (40–50 Gy) and not initial tumor volume was the most accurate predictor for local control and disease free survival. They found that a tumor with an initial volume >40 cm³ and a residual volume >20 % after EBRT on an MRI scan is a predictor of local treatment failure (Sensitivity 89 % and Specificity 89 %) [13].

In contrast to our study, Mayr et al. reported performing volumetric analysis without BCT applicators in place, necessitating an additional MRI scan. This consequently increases costs and requires the patient to come in the MR unit at a separate time.

The quality of the MRI performed during brachytherapy with the applicator in situ has shown to be sufficient for brachytherapy treatment planning. System based distortion due to the use of a magnetic field are generally under 1-2 mm and are at their lowest in the center of the image. Additional geometric distortions caused by adding a MR-compatible applicator have shown to be nearly undetectable and are at their lowest in the center of the image [22‐24]. In general MRI and applicator induced reconstruction uncertainties are smaller than the uncertainties induced by tumor contouring and bowell movement.

As the use of MRI-guided BCT for a cervical carcinoma larger than 5 cm is expected to increase significantly in future as more institutions perform 3D treatment planning instead of 2D X-ray, volumetric analyses based on routine MRI could prove to be more cost effective and convenient [14, 16, 25, 26].

The prevalence of recurrent/persistent local tumors in our dataset (11.4 %) is rather low compared to those previously reported. The low recurrence rate could be due to the lower initial tumor volumes, which in 80 % of our population were <40 cm³, or to several other reasons. First, because of the exclusion criteria, patients who did not finish imaging or the treatment protocol because of early progressive disease were excluded. Second, all patients received together with radiation therapy, concurrent chemotherapy or hyperthermia treatment, which resulted in a higher cure rate than that of radiotherapy alone [27, 28]. Third, all patients received MRI-based high-dose-rate intracavitary and interstitial BCT according to the GEC-ESTROO guidelines, which is proven to be superior to 2D-based radiotherapy, especially for cervical carcinomas sized >5 cm [26]. Moreover, we included tumors that were relatively smaller than those reported in other studies.

Our study had some limitations. First, the absence of residual disease was not proven histopathologically in all patients. This was mainly because some patients did not undergo surgery. However, we believe that long-term follow-up in the absence of recurrence is at least as good as performing surgico-pathologic analysis. Especially because in those cases the time interval between pathologic analysis with random biopsies or adjuvant surgery is shorter and it could be that because of this reason microscopic tumors are still existent. Possible this would not have been the case with longer waiting time, because of delayed radiotherapy effects. A second limitation of this study was the small number of patients. MRI-based BCT became the standard protocol in our center in October 2008, which was the main reason for the exclusion of 13 patients who underwent only one MRI scan during BCT.

Volume measurement at the time of BCT is a potential promising predictor of local tumor response in individual patients, especially when performed during the final BCT. As absolute volumetric analysis yielded results comparable to those of volume regression analysis, a single tumor volume measurement at the time of the last BCT, could already be sufficient to reduce measurement time. We also observed an association between overall treatment failure and local volume response, the potential of assessing local response for identifying individual patients at high risk for overall treatment failure is promising. External validation however in a larger and independent study population is warranted.