Predictors of acute hematologic toxicity in women receiving extended-field chemoradiation for cervical cancer: Do known pelvic radiation bone marrow constraints apply?

We identified women with locally advanced cervical cancers treated at two urban, academic institutions from 2012 to 2018. This study was approved by the institutional review boards of both institutions. Only patients treated with extended-field chemoradiation using IMRT with platinum-based chemotherapy were included in this study, as IMRT has been associated with reduced HT toxicity compared to a 3D-CRT technique [28]. Patient demographics and clinical characteristics are summarized in Table 1. BMI, race (as self-reported in the medical record), Charlson Comorbidity Index (CCI), chemotherapy cycles delivered, and nadirs for white blood cell (WBC) count, absolute neutrophil count (ANC), hemoglobin (Hgb) and platelet (Plt) count were obtained. Stage was defined using the International Federation of Gynecology and Obstetrics (FIGO) 2009 cervical cancer staging system.

Patients underwent individualized CT-based planning before the beginning of treatment with immobilization in alpha cradles with arms placed above their head. BM contours were standardized across all patients and included vertebral bodies within the radiation treatment field, sacrum, coccyx, pelvic bones, and femurs from top of the femoral head to the inferior border of the ischial tuberosities (Figure 1). The extended-field RT clinical target volume (CTV) included treatment to the uterus, cervix, primary mass, paracervical, parametrial, uterosacral region and at least upper half of the vagina. The nodal CTV treatment volumes included the external iliac, hypogastric, obturator and para-aortic lymph nodes with a 0.7-0.8 cm margin around the vessels minus anatomic subtraction as clinically indicated. The superior extent of the para-aortic field was defined by the renal vessels in all patients with the exception of 1 patient who had high para-aortic nodal disease (Supplement, Table B). The nodal PTV was uniformly expanded by 0.7 cm to produce the planning target volume (PTV). A cervix internal target volume (ITV) was created using bladder-empty and bladder-full scans; this was expanded by 1.5-2.5 cm to account for organ motion. Similarly, a vagina and parametrial ITV was created and expanded by 0.5-1.0 cm. In cases of urgent vaginal bleeding, radiation treatments utilized 3D-CRT for the first fractions to expedite treatment. These plans were subsequently incorporated into the composite dose calculations for IMRT treatment planning. Volumetric-modulated arc therapy (VMAT) or static-field IMRT were used to treat the volume to a total dose of 45 Gy in 1.8-Gy daily fractions, which was followed by a boost to gross nodal disease to a total dose of 55-60 Gy using either a sequential or simultaneous integrated boost technique. Organs at risk including the rectum, bowel, and bladder were contoured for radiation planning. BM was contoured, but not utilized as avoidance structures for IMRT planning until 2017. Patients were reassessed towards the end of external beam radiation treatment and were given a parametrial boost if indicated and high-dose-rate intracavitary or interstitial brachytherapy. The boost treatment utilized iridium-192 with a dose of 28-30 Gy to the high-risk CTV or Point A in 4-5 treatments. Radiation therapy and brachytherapy were generally held if absolute neutrophil count was less than 0.5 × 10⁹/L or if the platelet count was less than 50 × 10⁹/L.

Patients received concurrent weekly chemotherapy with cisplatin (40 mg/m²). Doses were based on patient weight and surface area. Patients were planned to receive 5 to 6 cycles of chemotherapy concurrently with radiation. Cisplatin was generally held if an absolute neutrophil count was less than 1 × 10⁹/L and/or platelet count was less than 100 × 10⁹/L, although this was at the discretion of the treating gynecologic oncologist. Any cisplatin dose-reduction was also at the discretion of the treating physician.

Contours and dose summation were completed on the Eclipse treatment planning system version 15.5 and the Pinnacle treatment planning system version 9.8. A cumulative plan was created from all external beam portions of the patient's treatment (if applicable) and a dose-volume histogram was generated for each contoured BM region for each patient. Brachytherapy dose was not included given its minimal contribution to bony structures; however, record of brachytherapy was notated. BM metrics collected included: Mean BM dose, V5Gy, V10Gy, V15Gy, V20Gy, V25Gy, V30Gy, V35Gy, V40Gy and V45Gy (where VxGy is defined as percentage of BM volume receiving x Gy). Specific dosimetric predictors of HT previously published in the literature for pelvic (non-extended-field) radiation were also evaluated in this patient cohort. These included V10Gy ≥ 90%, V20Gy ≥ 75%, and V40Gy > 37% [19, 27, 31]. A previously published dosimetric predictor in the extended-field setting was also evaluated in this patient cohort: mean BM dose > 30.3 Gy [23].

Toxicity grading was based on the Cooperative Group Common Toxicity Criteria. HT endpoints included 1) any grade 2 or higher hematologic acute toxicity, not including lymphopenia, and 2) grade 2 or higher leukopenia. These endpoints were chosen as a measure of toxicity as these may result in modifications to systemic or radiation therapy and were previously utilized as HT endpoints in RTOG 0418 [31]. Previously published HT endpoints were utilized for examination of their respective published BM dosimetric constraints in this patient cohort, which included HT2+ [31], leukopenia 2+ [19, 31], neutropenia 2+ (grade 2-4 neutropenia) [19], HT3+ (grade 3-4 HT) [23], and leukopenia 3+ (grade 3-4 leukopenia) [27].

All statistical analyses were performed using STATA version 17.0 (StataCorp LLC, College Station, Texas). Univariate analysis (UVA) was performed using the chi-squared test and logistic regression. The chi-squared test or Fisher's exact test were used to compare rates of hematologic adverse events (AEs) for patients with a volume of BM irradiation from 5 to 45 Gy (V5Gy to V45Gy) dichotomized at the median as well as to compare rates of hematologic AEs for patients according to dichotomized patient characteristics (African American race vs other, CCI), as appropriate. Tests for normality were performed for age and BMI with the Shapiro-Wilk statistic and these variables were transformed using the natural logarithm to eliminate skew. Logistic regression was used to test for associations between natural log-transformed age or natural log-transformed BMI and hematologic endpoints. Variables with p-value < 0.1 from univariate logistic models were included in the multivariate model. Multivariate analysis (MVA) was performed using logistic regression to correlate HT2+ as well as leukopenia 2+ with patient or dosimetric characteristics. Firth's logistic regression was utilized in cases of rare events [33]. Previously published BM constraints from both the non-extended field setting (V10Gy ≥ 90%, V20Gy ≥ 75%, V40Gy ≥ 37%) [20, 27, 31] as well as the extended-field setting were evaluated (BM mean > V30.3Gy) [23] as described in each respective publication including covariates. Associations of hematologic endpoints with V5Gy ≥ 98% and V20Gy ≥ 70% were performed using Fisher's exact test.

Fifty-two locally advanced cervical cancer patients treated with EF-CRT with concurrent cisplatin were identified (Table 1). All patients were treated using an IMRT technique: 21 (40.4%) with VMAT and 31 (59.6%) with static-field IMRT. Two patients (3.9%) started treatment urgently with 3D-CRT (9 Gy in 5 fractions and 18 Gy in 10 fractions) due to vaginal bleeding and converted to IMRT plans. Fifty-one (98.1%) patients were treated with brachytherapy: 17 (32.7%) were treated with interstitial and 35 (67.3%) were treated with intracavitary. One (1.9%) patient declined brachytherapy. The median mean BM dose was 28.40 Gy (IQR: 26.4-30.2). Table 1 further details radiation dosimetric characteristics.

All patients in the study received concurrent cisplatin, and 17 (33%) patients required dose reduction. The median number of cycles of cisplatin was 5 (IQR: 4-5). Forty-eight (92.31%) of the patients experienced HT2+ and 28 (53.85%) of the patients experienced HT3+ (Table 2).

Supplemental Table A demonstrates univariate analysis of the association between clinical or dosimetric variables with any grade 2+ HT and grade 2+ leukopenia. African American race trended towards association with HT2+ (OR 14.84, 95% C.I.: 0.75 – 292.62, p=0.08) and leukopenia 2+ (OR 5.25, 95% C.I.: 0.79 – 57.25, p=0.06 on univariate analysis. There was no significant association between any of the HT endpoints and age, BMI or CCI (all p>0.05). Amongst the BM metrics that were analyzed, V5Gy $\ge$ 98% (volume of BM receiving 5 Gy) was associated with leukopenia 2+ (OR 5.89, 95% C.I.: 1.10 – 472.72, p=0.02; 96% vs 72%) and trended towards association with HT 2+ (OR 11.51, 95% C.I.: 0.59 – 225.67, p = 0.11; 100% vs 84%). V20Gy ≥ 70% was associated with increased rates of leukopenia 2+ (OR 9.21, 95% C.I.: 1.002 – 431.25, p=0.02; 96% vs 73%,).

When accounting for race on MVA, V5Gy ≥ 98% was associated with leukopenia 2+ (OR 17.73, 95% C.I.: 1.71 – 183.56, p =0.02) and HT2+ (OR 21.76, 95% C.I.: 0.99 – 473.43, p=0.05), while V20Gy ≥70% was associated with leukopenia 2+ (OR 9.43, 95% C.I.: 1.01 – 87.72, p<0.05) (Table 3).

While previously described radiation metrics for pelvic radiation in the literature (V10Gy ≥ 90%, V20Gy ≥ 75%, or V40Gy > 37%) and their respective HT endpoints were not exactly found to be associated in our extended field patient cohort, the V20Gy metric was very similar to our findings [19, 27, 31] (Table 4).

Aside from the retrospective nature of our study, we were principally limited by a small dataset. However, our study is the largest and only IMRT-exclusive cohort to describe HT in EF-CRT. In addition, our dataset was comprised of patients from multiple treatment centers which may bias data by differences in practice patterns of oncologists or of the patient cohort. However, radiation treatment planning was performed utilizing consistent extended-field radiation treatment fields and dosing schemes. Moreover, the multi-institutional nature of this study may improve generalizability. It is also unclear whether varying dose metrics would apply to different anatomic bone marrow regions. Perhaps the biggest limitation of the study was the lack of PET or other functional imaging methods to predict areas of active BM to be used as avoidance structures [47-49]. INTERTECC-2 specifically attempted to address this question in a prospective cohort and found that the use of functional image-guided radiotherapy (IG-IMRT) to guide BM sparing techniques was associated with lower rates of neutropenia compared to standard IMRT for patients treated with non-extended field chemoradiation [28, 29]. On the contrary, Yan et al. did not find an added benefit to the use of functional imaging in identifying dosimetric predictors of HT in the extended field setting [23]. PET-defined active BM was not included in our study due to the challenges in deformable registration of the PET to the treatment planning CT scans due to the large fields involved. These technical factors may have contributed to the lack of benefit of functional imaging as observed by Yan et al. Moreover, areas of active BM are common amongst patients and primarily appears to be in the sacrum and thoracolumbar spine [50], which may be difficult to meaningfully avoid with low-dose isodose lines when treating para-aortic disease burden. Overall, more work must be done to further explore methods to decrease HT for extended field radiation therapy, including exploring the potential utility and feasibility of integrating functional imaging for BM sparing techniques, such as development of a standardized atlas that is inclusive of the para-aortic region.