Prospective intra/inter-observer evaluation of pre-brachytherapy cervical cancer tumor width measured in TRUS and MR imaging

Cervical cancer is the fourth most common cancer in women worldwide and the eight overall. A large majority (around 85%) of the global burden occurs in the less developed regions [1]. For decades radiotherapy (RT) has been the standard of care for locally advanced disease and brachytherapy (BT) is an essential part of the treatment [2, 3]. In the last decade, 3D treatment planning was introduced for cervical cancer BT [4‐9] with outstanding clinical results [10‐14]. The goal of 3D adaptive BT is to tightly shape the radiation dose to the individual patient’s anatomy and tumor topography for each BT fraction, with the intent to deliver 85–90 Gy_EQD2 (radiobiologically equieffective dose of 2 Gy per fraction) to the tumor while minimizing the dose to organs at risk (OAR). The precondition for safe BT treatment individualization is the precise identification of target volumes. Magnetic Resonance (MR) has clear advantages in terms of image quality [15] as it allows the optimal definition of normal peri-cervical soft tissues, tumor extension within the cervix, parametrial infiltration and topography. Additionally, MR enables 4D volume adaptation following tumor regression during external beam radiotherapy (EBRT) [16, 17]. Unfortunately, due to its costs and limited availability, the majority of patients worldwide are precluded from receiving MR-based BT treatment [18, 19]. Computer tomography (CT) imaging alone is not an alternative to MR because it’s poor soft tissue contrast is inadequate to precisely define cervical tumors [20] and burdens of parametrial infiltration [21, 22]. Approaches with less intense MR routine like hybrid MR/CT protocols have been investigated with promising results [23] but still rely on some MR imaging. TRUS has excellent soft tissue resolution, is affordable, and has been used extensively in cervical cancer diagnosis [24]. Moreover, TRUS has been used to aid in proper BT applicator insertion and guidance, and for correct placement of parametrial needles, since it is, among all ultrasound (US) modality, the one that better depicts parametrial infiltration. For all these reasons TRUS is under investigation as a potential alternative to MR for Image-Guided Adaptive Brachytherapy (IGABT) planning [25, 26]. Some shortcomings may however limit TRUS use, such as the intrinsic operator dependency, the inadequate visualization of tumor regression areas in the parametria and lastly, the difficulty in evaluating the relationship of extensive tumors to the pelvic sidewall when the edge of infiltration is beyond the range of the probe. The aim of the present study is a blinded multi-observers comparison of TRUS and MR, assessing tumor maximum width before first BT application (without applicator in place) in a large cohort of cervical cancer patients undergoing IGABT.

One hundred ten consecutive cervical cancer patients were analyzed. HRCTV_MR and HRCTV_TRUS average measures ± standard deviations (SD) are given in Table 2.

In the last few years US-based BT dose adaptation has been increasingly investigated for cervical cancer IGABT [33]. US has been proven as an excellent diagnostic imaging modality in gynecological oncology [34] and it has been extensively used during BT application to guide tandem and needles insertion [35].

One of the largest prospective studies comparing the diagnostic accuracy of TRUS and MR in the local staging of cervical cancer was published by Fischerova et al. [36] in 2008 and included 95 patients with early-stage disease. The study showed a significantly higher accuracy of TRUS when compared to MR in tumor identification (considering also residual tumor after previous biopsy (93.7 vs. 83.2%, p ≤ 0.006), or small tumors ≤1 cm³ (90.5 vs. 81%, p ≤ 0.049)). Similar results were demonstrated by a European multicenter prospective study, which included 182 patients with histologically confirmed early-stage cancer. The diagnostic agreement between ultrasound and pathology was significantly better at detecting residual tumor and parametrial invasion than MR (p < 0.001). A surprising finding was the maintenance of diagnostic accuracy of ultrasound in the detection of residual tumor after cone biopsy, where it is difficult to distinguish post-inflammatory and reparative changes after the procedure from the presence of residual tumor [37].

Pinkakova et al. in a prospective study on a cohort of 42 FIGO IB1-IIB cervical cancer patients (with limited parametrial involvement) demonstrated the non- inferiority of TRUS with respect to MR in assessing tumor regression during neo-adjuvant chemotherapy [38].

The potential of TRUS imaging in tracking the tumor changes and regression during EBRT is a fact of critical relevance if TRUS is used to guide the BT insertion (normally scheduled after 3–4 weeks of radio-chemotherapy) and, potentially, dose adaptation.

The clinical use of US-based BT dose adaptation in cervical cancer BT has been pioneered at Peter MacCallum Cancer Center [39] and promising results have been reported [40]. The method suggested is based on transabdominal US measurements of the uterus taken along the tandem axis in the sagittal plan and has shown a robust correlation with MR measurement. This approach is useful to conform the dose distribution according uterus silhouette in the anteroposterior diameter thus reducing bladder and rectal dose. Nevertheless it seems unfit to target volume delineation at the parametria level, because the limitations of transabdominal US in detecting parametrial invasion and the unfeasibility of true volumetric image acquisition. For all these reasons, Kirisits et al. stated, in an interesting editorial, “this method may be useful, mainly in limited size and well-responding tumors, which are confined to the cervix at the time of BT. However, this clinical scenario does not represent most patients in advanced stage as seen in the countries with high patient numbers and limited resources” [33].

Conversely, TRUS (which is already extensively used for prostate BT imaging and dose optimization) allows a true volumetric image acquisition and a detailed imaging of cervical tumor and eventual extension beyond uterine cervix within the parametrial space [41]. Parametrial infiltration is a well known prognostic parameter for cervical cancer [42] and probably the most relevant factor to take into account at the time of IGABT insertion preplanning [43], to select between intracavitary or intracavitary/interstitial technique [44]. For this reason, the correct assessment of tumor width is a critical point in cervical cancer BT. Validation of TRUS as potentially useful tool for cervical BT imaging and dose adaptation starts from the assessment of the robustness of TRUS measuring tumor width at the time of BT.

Researchers at the Medical University of Vienna in two different studies [25, 45] demonstrated an excellent agreement between MR and TRUS in assessing tumor maximum width after EBRT in respectively 16 and 19 patients diagnosed with cervical cancers (FIGO I-IV). In both studies the mean difference between MR and TRUS measurement were in the same range (− 0.3 ± 3.2 mm and − 1.1 ± 3.2 mm respectively). These data strikingly compare with our results (1.3 ± 3.2 mm; 1.1 ± 4.6 mm and 0.7 ± 3 mm for the 3 observers). We were also able to demonstrate that the magnitude of uncertainty in the vast majority of cases (≈80%) is very small (< 3 mm), but increases in large tumors. Additionally, we demonstrated that in large IIIB/IVA tumors TRUS may underestimate tumor width compared to MR. Several reasons might be claimed to explain such data.

First of all, the higher degree of uncertainty associated to tumor maximum width measurement in large tumors that spread to the pelvic wall. In fact, even if the maximum tumor width in this study is taken perpendicularly to the cervical canal and uterine axis (in order to improve reproducibility of the measurements between MR and TRUS), the irregular shape of such tumors may impair the identification of a reproducible axis of measurement (Additional file 1: Figure S4). This type of error may be at the basis of the progressively lower TRUS concordance among observers in advanced FIGO stage tumors. On the other hand, such type of uncertainty should be stochastic, while our findings show that TRUS underestimates tumor width (> 5 mm) in 20% of stage IV and in 7–14% of stage III tumors (Additional file 1: Figure S2.2).

A second possible explanation could be an intrinsic limited resolution of TRUS imaging in depicting the limit of tumor infiltration into pelvic wall, because poor image quality (due to patient preparation, or because TRUS shortcoming in evaluating the relationship of extensive tumors to the pelvic sidewall when the edge of infiltration is close to the range of the probe). During this investigation, a strict bowel preparation protocol was followed in order to prevent the interference on TRUS image quality. Nevertheless, TRUS performance in large tumors was inferior to MR, as shown by the inter-observer analysis, where the measurement agreement as expressed by ICC values was extremely high for HRCTV_MR width, independently from FIGO stage. Conversely, ICC for HRCTV_TRUS width resulted comparable to MR just for FIGO I/II tumors and progressively declined for larger tumors, although remaining fairly good (Table 4).

A further reason might depend on the shape and type of tumor growth pattern. Very advanced tumor stages (FIGO IVA) show more likely an infiltrative growth with irregular shapes, and thin digitations that deeply infiltrate the parametrial space in comparison to bulky expansive cervical tumors, which are generally more clearly visible in TRUS imaging (Figs. 2 and 3).

It deserves to be mention that a limitation of our study design and a potential bias is that TRUS images were acquired by one investigator only and then analyzed independently by the three investigators. To obtain three sets of TRUS images acquired independently by each observer would be probably more correct. Nevertheless, such study design would be impossible at our Institution because the organization of the workflow. To minimize uncertainties the TRUS acquisition, a protocol (previously described) was thoroughly followed.

Our investigation does not include the assessment of tumor thickness and height. Schmid et al. [25, 45] analyzed tumor thickness as measured by TRUS or MR and found a statistically significant underestimation with TRUS. Such underestimation, in their opinion, is mainly due to the compression of the cervix by the TRUS probe at the moment of image acquisition. We agree with this point. We believe that the insertion of rectal probe (and the angle needed to have it parallel to the uterus) may intrinsically bias any comparison with images taken with relaxed rectum/pelvic floor. On the other hand, the exact burden of tumor cranial infiltration into myometrium (especially after EBRT) is challenging on TRUS imaging (but also on MR). For this reason, in the clinic is generally recommended to load the tandem up to the fundus uteri, making the exact tumor height measurement in MR or TRUS not a critical point.

Taken together our results suggest that TRUS is comparable to MR in assessing preBT tumor maximum width in cervical cancer FIGO stage I/II. In more advance stages TRUS seems to be slightly inferior to MR although maintaining a good agreement to gold standard imaging.

Given the limited cost of TRUS compared to MR and the potential of improved patient accessibility, especially in low income countries, TRUS based IGABT is a major field of research in radiation oncology. Further studies are however still needed to define the technical modality of integration of TRUS in cervical cancer IGABT and TRUS-based dose adaptation.