Search

Brachytherapy is a critical component of the standard-of-care curative radiotherapy regimen for women with locally advanced cervical cancer (LACC). However, existing literature suggests that many patients will not receive the brachytherapy boost. We used machine learning (ML) and explainable artificial intelligence to characterize this disparity. Patients with LACC diagnosed from 2004 to 2020 who received definitive radiation were identified in the National Cancer Database. Five ML models were trained to predict if a patient received a brachytherapy boost. The best-performing model was explained using SHapley Additive exPlanation (SHAP) values. To identify trends that may be attributable to the coronavirus disease 2019 (COVID-19) pandemic, the previous analysis was repeated and limited to 2019 to 2020. A total of 37,564 patients with LACC were identified; 5799 were diagnosed from 2019 to 2020 (COVID cohort). Of these patients, 59.3% received a brachytherapy boost, with 76.4% of patients diagnosed in 2019 to 2020 receiving a boost. The random forest model achieved the best performance for both the overall and COVID cohorts. In the overall cohort, the most important predictive features were the year of diagnosis, stage, age, and insurance status. In the COVID cohort, the most important predictive features were FIGO stage, age, insurance status, and hospital type. Of the 26 patients who tested positive for COVID-19 during their course of radiotherapy, 19 (73.1%) received a brachytherapy boost. A gradual increase in brachytherapy boost utilization has been noted, which did not seem to be significantly impacted by the onset of the COVID-19 pandemic. ML could be considered to identify patient populations where brachytherapy is underutilized, which can provide actionable feedback for improving access. [ABSTRACT FROM AUTHOR]

Simple Summary: Brachytherapy is a key component of radiation treatment in the curative treatment of locally advanced cervical cancer. To improve the delivery of brachytherapy, recent studies have explored the use of imaging, such as magnetic resonance imaging (MRI), in treatment planning. In this review, we reviewed the current evidence for image-guided brachytherapy and discussed future directions for research. Purpose: The standard of care for patients with locally advanced cervical cancer is definitive chemoradiation followed by a brachytherapy boost. This review describes the current status and future directions of image-guided adaptive brachytherapy for locally advanced cervical cancer. Methods: A systematic search of the PubMed and Clinicaltrials.gov databases was performed, focusing on studies published within the last 10 years. The search queried "cervical cancer [AND] image-guided brachytherapy [OR] magnetic resonance imaging (MRI) [OR] adaptive brachytherapy". Discussion: The retroEMBRACE and EMBRACE-I trials have established the use of MRI as the standard imaging modality for brachytherapy application and planning. Quantitative imaging and radiomics have the potential to improve outcomes, with three ongoing prospective studies examining the use of radiomics to further risk-stratify patients and personalize brachytherapy. Another active area of investigation includes utilizing the superior soft tissue contrast provided by MRI to increase the dose per fraction and decrease the number of fractions needed for brachytherapy, with several retrospective studies demonstrating the safety and feasibility of three-fraction courses. For developing countries with limited access to MRI, trans-rectal ultrasound (TRUS) appears to be an effective alternative, with several retrospective studies demonstrating improved target delineation with the use of TRUS in conjunction with CT guidance. Conclusions: Further investigation is needed to continue improving outcomes for patients with locally advanced cervical cancer treated with image-guided brachytherapy. [ABSTRACT FROM AUTHOR]

In select cases of locally advanced cervical cancer, a hybrid brachytherapy (HBT) approach consisting of a combined intracavitary (IC)/insterstitial (IS) implant can yield improved target coverage and/or decreased organ at risk dose compared to IC techniques while limiting invasiveness compared to IS techniques. The technique involves placement of transvaginal and/or perineal needles in addition to the tandem and ring/ovoids using either a specialized applicator or free-hand placement. Following applicator and needle placement, brachytherapy may then be planned using principles similar to IC or IS techniques. During treatment planning, it can be helpful to obtain both MRI and CT imaging, as plastic MRI-compatible needles do not show up well on MRI. In patients where acceptable target coverage cannot be achieved using IC alone or doses to nearby OAR are too high, HBT should be evaluated. HBT can improve both dose to target and OAR while sparing patients the morbidity of perineal template-based interstitial brachytherapy. Specific scenarios where HBT may be preferred include bulky residual primary tumor especially with poor response to EBRT, extension into the lateral parametrium, vaginal extension of tumor, and an asymmetric target. Use of HBT can typically permit extension of dose coverage by an additional 1–2 cm beyond what can be achieved with an IC alone technique. HBT allows for improved therapeutic ratio by improving target volume coverage and/or lowering doses to OARs. Brachytherapists should be trained on the practical aspects of administering HBT to be able to offer a less invasive and impactful treatment option when appropriate. [ABSTRACT FROM AUTHOR]

Brachytherapy (BT) is a critical component of the standard-of-care curative regimen for women with locally advanced cervical cancer (LACC). However, existing literature suggests a sizeable proportion of patients will not receive the BT boost. Although predictive models such as logistic regression have been able to help identify characteristics of patients who are most at-risk for omission of BT boost, its predictive performance could be improved. Therefore, we utilized machine learning (ML) combined with explainable artificial intelligence to characterize this treatment disparity, with an emphasis on more recent trends that might be attributable to the COVID-19 pandemic. Patients with LACC diagnosed from 2004-2020 were identified in the National Cancer Database. The query was subsequently limited to patients who received definitive radiation. Five ML models were trained to predict if a patient received a BT boost. Model performance, measured using area under the curve (AUC), was compared to multivariable logistic regression. The best-performing model was explained using SHapley Additive exPlanation (SHAP) values. To permit more traditional statistical understanding, SHAP values were converted into odds ratios (OR) and used to identify how model features influenced predictions. To identify trends that may be attributable to the COVID-19 pandemic, the previous analysis was repeated and limited to the years 2019-2020. A total of 54,896 patients with LACC were identified; 8298 were diagnosed from 2019-2020. Of these patients, 54.8% received a BT boost, with 70.5% of patients diagnosed in 2019-2020 receiving a boost. The majority of the increase occurred from 2018 onward, with rates every year prior being a maximum of 57.0%. The Extreme Gradient Boosting model outperformed all other models with an AUC of 0.61 (95%CI: 0.57-0.65) compared to 0.56 (95%CI: 0.52-0.60) for logistic regression. SHAP values > zero represent greater probability of BT boost. SHAP summary plots of the ML models from 2004-2020 and 2019-2020 are shown in Figure 1A and 1F, respectively. SHAP dependence plots of the ML models from 2004-2020 and 2019-2020 are shown in Figure 1B-E and 1G-J, respectively. Overall, a more recent year of diagnosis (OR: 1.08; 95%CI: 1.07-1.08; p<0.001) and lower Charlson Deyo score (2+ vs 0; OR: 1.27; 95%CI: 1.17-1.38; p<0.001) were strong predictors of receiving a BT boost. Other factors associated with BT boost receipt included younger age, private insurance, and white (vs black) race. When examining the years 2019-2020, year of diagnosis (OR: 0.98; 95%CI: 0.92-1.01; p=0.31) and Charlson Deyo score (2+ vs 0; OR: 1.07; 95%CI: 0.99-1.22; p=0.19) were no longer significant. It does not appear that a COVID-19 diagnosis during radiation adversely impacted rates of BT utilization, with 24 of 34 (70.5%) patients who tested positive for COVID-19 during their radiation course receiving a boost. There were slight delays reflected in total treatment duration relative to national recommendations of 8 weeks, lasting a median of 10.7 weeks. This study illustrates how ML can be used to examine practice patterns and identify areas of underutilization. ML approaches yielded improved performance, relative to logistic regression, at predicting receipt of BT boost. To our knowledge, this is the first time SHAP values have been converted to OR, CI, and p-values to describe such models. An increase in BT boost utilization has been noted since 2018, which fortunately did not seem to be significantly impacted by the onset of the COVID-19 pandemic. ML should be considered to better identify health disparities across patient populations where BT is underutilized, which can provide actionable feedback for improving access. [ABSTRACT FROM AUTHOR]