KEY POINTS
- This prospective imaging study from The Royal Marsden and Institute of Cancer Research, UK, evaluated three-dimensional transabdominal ultrasound for selecting a plan of the day (PotD) during radiotherapy for locally advanced cervical cancer. Fifteen patients contributed 149 evaluable fractions, with 172 ultrasound acquisitions initially performed and 23 excluded because of acquisition, export, or registration errors. Ultrasound images were analyzed offline and did not influence actual clinical treatment decisions.
- Each patient had a four-plan VMAT library, comprising three plans covering different ranges of anticipated uterocervix motion and one robust fallback plan. The standard treatment pathway consisted of 25 fractions of external-beam radiotherapy followed by HDR brachytherapy, with concurrent cisplatin or carboplatin when appropriate. Ultrasound was obtained immediately before CBCT in the treatment position, requiring approximately three additional minutes per fraction. Five independent observers compared CBCT alone with co-registered ultrasound–CBCT images, using a consensus selection by three experts as the reference standard.
- Adding ultrasound did not significantly improve correct plan selection, the primary endpoint. Mean concordance with the expert-selected optimal plan was 67.2% with ultrasound–CBCT versus 66.2% with CBCT alone, an absolute improvement of 1.0 percentage point (p=0.555). Selection of any clinically acceptable plan was similarly unaffected (85.2% vs 84.0%; p=0.448). Thus, better additional soft-tissue imaging did not translate into a measurable improvement in the decision that actually determined treatment delivery.
- Uterocervix visibility improved modestly, but observer confidence did not meet the study's corrected significance threshold. Mean visibility increased from 2.06 to 2.18 on a four-point scale (difference +0.12; p=0.001). Confidence increased from 2.06 to 2.13 (difference +0.07; p=0.048), but the authors prespecified p<0.01 to account for multiple comparisons. Therefore, the visibility improvement was statistically significant, whereas the overall confidence improvement was not under the study's threshold.
- The benefit of additional ultrasound was greater when CBCT visibility was poor, but accuracy still did not improve significantly. Among 69 fractions with limited CBCT uterocervix visibility, ultrasound increased optimal plan-selection accuracy by 3.4 percentage points (p=0.225), visibility score by 0.25 (p<0.001), and confidence by 0.33 (p<0.001). Exploratory analyses suggested greater visibility benefits in patients with lower BMI and larger bladder volumes, as well as among observers with less experience interpreting pelvic CBCT.
- Ultrasound alone performed substantially worse than CBCT. In an exploratory assessment by two ultrasound physicists, optimal plan-selection accuracy was 50.3% using ultrasound versus 69.9% using CBCT, a 19.6-percentage-point reduction (p<0.001). The proportion of acceptable selections was 70.3% versus 82.4% (p<0.001). These results do not support replacing volumetric CBCT with transabdominal ultrasound for cervical plan-of-the-day decisions using the tested system.
- Decision ambiguity and technical limitations affected the assessment. Multiple plans adequately covered the target in 52% of fractions, while no available plan provided adequate coverage in 4%; the reference experts initially disagreed in 38% of fractions. Overall optimal-plan concordance was only approximately 66%, and not all study observers had formal PotD training. The small single-center cohort, offline assessment, operator-dependent ultrasound acquisition, potential probe-induced anatomical displacement, and lack of delivered-dose or clinical outcome evaluation limit generalizability.
CLINICAL TAKEAWAY
Adding transabdominal ultrasound to CBCT slightly improved visualization of the uterocervix but did not significantly improve plan-of-the-day selection accuracy during cervical cancer radiotherapy. The technology may have exploratory value in poorly visualized cases or staff training, but routine integration is not supported by the current evidence. Ultrasound alone performed worse than CBCT and cannot replace it in this workflow.