Pelvic bone marrow sparing intensity modulated radiotherapy reduces the incidence of the hematologic toxicity of patients with cervical cancer receiving concurrent chemoradiotherapy: a single-center prospective randomized controlled trial

One hundred sixty-four eligible female patients with cervical carcinoma who were undergoing treatment with concurrent cisplatin and IMRT ± brachytherapy were prospectively recruited in this single-center prospective RCT study from March 2018 to March 2019.

Major inclusion criteria were: (1) Diagnosis of clinical Stage Ib2–IIIb cervical carcinoma with no previous history of chemotherapy or pelvic irradiation (according to the FIGO clinical stage of 2015 NCCN guidelines). Clinical stage was determined by gynecological examination, endometrial curettage, hysteroscopy, pelvic MRI, lung X-ray, neck and abdominal color Doppler ultrasound and other laboratory examinations. (2) Patients between 20 and 70 with ECOG score < 3 (as defined by Zubrod-ECOG-WHO) [27]. Exclusion criteria were: (1) Patients with blood-related diseases or occupational exposures such as paint, decoration, nail industry, etc.; (2) Patients who had been treated with extended-field radiation therapy; (3) Patients who were participating in other clinical trials at the same time. The study was approved by the institutional review board of the local institution. All patients were informed about the design and potential risk of the trial before the intervention and signed informed consents.

Eligible patients were enrolled on the basis of the CONSORT Statement Extension for Randomized Controlled Trials with allocation concealment, as shown in Fig. 1. A computer-generated randomization table was used to allocate the enrolled patients into the PBMS group (82 cases) or the control group (82 cases). The group-allocation information was blinded to both patients and doctors who participated in the study.

Patients were immobilized in the supine position and underwent a contrast-enhanced CT scan with a 5 mm slices from the L3 - L4 junction to 2 cm below the perineum. The image datasets were transferred to the Raystation planning system (Raysearch Radiation Oncology Systems). The clinical target volume (CTV) was divided into CTV-U, CTV-C, CTV-V, CTV-P and CTV-N determined by CT and MRI, which were defined as the uterus, cervix, vagina, paracancer tissue and the common iliac, internal iliac, external iliac, obturator and sacral lymph node region, respectively. The planning target volume (PTV) was defined by a uniform three-dimensional expansion using 10 to 20 mm around CTV-U and CTV-C, 10–15 mm around CTV-V and CTV-P, 7 mm around CTV-N. All patients had bladder filling and rectal emptying before radiotherapy.

IMRT plans were generated with 7–9 coplanar fields using a 6-MV X ray. The planned IMRT dose was 50.4 Gy in 1.8 Gy daily fractions. PTV planning constraints were to ensure that the PTV should receive 95% of the prescribed dose and that 107% of the prescribed dose should be restricted to ≤5% of the PTV. The bladder, rectum, and small bowel were contoured as the Organs at risk for all patients.

Patients treated definitively received 3D image-based brachytherapy with 4–5 fractions of 6–7 Gy. Treatment volumes and dose prescriptions based on GEC-ESTRO recommendations [28]. The mean dose to 90% of the PTV was required to achieve the prescribed dose. Brachytherapy was initiated no sooner than the fourth week of treatment and insertions were separated by a minimum of 48 h.

For each patient, the external contour of all bones within the pelvis was used as a surrogate for pelvic bone (PB) which was delineated on the window of a planning CT scan (window width: 800 Hu–2000 Hu; window level: 250 Hu–500 Hu). PB was further divided into the hip bone (HIP) and lumbosacral spine (LSS). HIP was the area from the iliac crests extending to the proximal femora, which consisted of the pubes, ischia, acetabula, and proximal femora. LSS was the area extending from the most superior vertebral body contained in the planning treatment volume (usually L5) inferiorly to the entire sacrum. Bone marrow was defined as the low-density regions within the corresponding bones. Pelvic bone marrow (PBM) was also divided into two subsites, namely HIP marrow and LSS marrow (Figs. 1s and 2s).

According to various retrospective studies, the volumes receiving above 10 Gy (V10) and the mean dose of LSS, as well as the volumes receiving above 10, 20, or 40 Gy (V10, V20, V40) and the mean dose of PB were correlated with acute HT in cervical cancer patients with CRT [15‐20]. Therefore, PBMS-IMRT was applied in the PBMS group in order to maximize the constraints of the pelvic bone dose parameters in: LSS V10 < 85%, LSS mean < 30 Gy, PB V10 < 80%, PB V20 < 70%, PB V40 < 30%, and PB mean < 30 Gy. The control group patients were treated with IMRT alone.

After the completion of planning, dose volume histograms (DVHs) were then generated, and the parameters of PB and marrow were recorded (Figs. 3s and 4s). The treatment plan for all patients was established by two physicians and the radiotherapy plan was developed by the same medical physicist.

Chemotherapy consisted of weekly cisplatin (40 mg/m², maximal dose 70 mg) concurrently with IMRT ± brachytherapy. Cisplatin was held and appropriate symptomatic treatment was administered under the following conditions: white blood cell count (WBC) < 2 × 10⁹/L, absolute neutrophil count (ANC) < 1 × 10⁹/L, platelet (PLT) count < 50 × 10⁹/L, or creatinine clearance < 50 mL/min.

The primary endpoint was grade 2 or higher hematological toxicity (HT2+) during the treatment. The median follow-up time was the end of treatment. The blood counts monitoring and gastrointestinal symptoms of all patients were recorded weekly from the beginning to the end of CRT. And results were graded according to the Radiation Therapy Oncology Group acute radiation toxicity scoring criteria [29], with HT2+ or gastrointestinal toxicity (GT) of grade ≥ 2 (GT2+) noted as an event.

Regarding the the sample size calculation and methods for power calculation, We set α and β to be 0.05 and 0.1, respectively. The incidence of HT2+ in control group was estimated to be 0.72 according to the pre-test in our center. The incidence of HT2+ in PBMS group was estimated to be 0.46 according to multiple retrospective studies. The sample size of two groups was estimated to be 74 by two-sided test.

Statistical analysis was performed using SPSS 24.0 software (IBM SPSS Statistics for Windows, Armonk, NY, USA). Body mass index (BMI, calculated as body weight (kg)/height (m)²) and dosimetric parameters were coded as continuous variables. Categorical variables included age (dichotomized by 60 years), pathology, comorbidity, ECOG score and clinical stage (dichotomized by IIIb). The enumeration data following a normal distribution are represented by mean ± standard deviation (\( \overline{x} \) ± s). Categorical data are represented by the number of cases (n) and percentage (%). According to different data, Student’s t-test, the chi-square test or the Mann-Whitney U rank test were used to compare the differences in means and proportions. Univariate logistic regression was used to test the correlation between clinical and dosimetric parameters with HT2+. Multivariate logistic regression models controlling for clinical stage, ECOG score, BMI, age, pathology and comorbidity were then used to examine the effect of significant dosimetric parameters on HT2+. Due to large inter-variable correlations within the dosimetric variables, only one of the dosimetric variables and the clinical factors were included in pairs for each multivariable analysis. Receiver operating characteristic (ROC) curves were used to evaluate the value of dosimetric parameters for predicting hematologic toxicity. Area under curve (AUC) was used to evaluate the goodness of the curve model. The ROC curve was plotted using Medcal 15.0.

Between March 2018 and March 2019, 164 patients consented to the study, of which two patients in the PBMS group were discontinued for financial reasons. Data of all patients who initiated protocol therapy were analyzed according to intention-to-treat. Our study showed no difference in the clinical parameters and the first full blood count between the studied groups (Table 1).

The dosimetric parameters were significantly different between the PBMS group and the control group (P < 0.01). Descriptive statistics of radiation dose volume parameters are presented in Table 2.

In the PBMS group, 40 patients completed all six cycles, 33 completed five cycles, and seven completed four cycles, one completed two cycles and one completed one cycle. In the control group, 37 patients completed all six cycles, 31 completed five cycles, and 14 completed four cycles Cumulatively. Some patients have withheld or postponed CRT for some reasons, as detailed in Table 1s. There was no significant difference in the total number of chemotherapy cycles between the two groups (P = 0.47). 17 patients received granulocyte-monocyte colony stimulating factors and 2 patients received platelet transfusions during treatment. No patients received red blood cell transfusions or recombinant human erythropoietin injection.

All patients enrolled had full blood count tests weekly and none of the patient missed a blood sample. The incidence of HT2 + in the PBMS group was significantly lower than that in the control group (50% vs 69.5%) (Table 3). The median WBC, ANC, hemoglobin (HGB), and PLT count nadirs in the PBMS group were 3.2 × 10^{^9}/L (range, 1.63–5.33), 2.02 × 10^{^9}/L (range, 0.86–4.20), 116.10 g/L (range, 82.00–138.00), 142.5.00 × 10^{^9}/L (range, 111.00–232.00). Meanwhile the median WBC, ANC, HGB, and PLT count nadirs in the control group were 2.53 × 10^{^9}/L (range, 1.07–4.97), 1.68 × 10^{^9}/L (range, 0.23–3.97), 114 g/L (range, 78.00–133.00), and 142 × 10^{^9}/L (range, 21.00–322.00), with significant differences (P < 0.05) (Fig. 2). The GT2+ events did not differ significantly between the two groups (P > 0.05).

For PB, LSS V10, LSS mean, PB V40 were significantly associated with HT2+. For PBM, V10, V20, V40, mean of LSS marrow and PB marrow V40 were significantly associated with HT2+, correspondingly (P < 0.05). No dosimetric parameters of the hip bone were associated with HT2+ in this analysis (Table 4).

To identify optimal thresholds for dosimetric planning, we analyzed the ROC curves for HT2+ versus LSS V10, LSS mean and PB V40. The optimal LSS V10 cutoff indicated by the ROC curve was 87%. The sensitivity and specificity for this threshold were 71.4 and 56.1. The patients with LSS V10 ≥ 87% were more likely to experience HT2+ (OR = 3.13, 95%CI = 1.62–6.05, P = 0.001). The optimal cutoff of LSS mean was 39.05 Gy. The sensitivity and specificity for this threshold were 36.7 and 89.4. The patients with LSS mean ≥ 39 Gy were more likely to experience HT2+ (OR = 3.03, 95%CI = 1.49–6.17, P = 0.002). The optimal cutoff of PB V40 was 28%. The sensitivity and specificity for this threshold were 36.7 and 89.4. The patients with PB V40 ≥ 28% were more likely to experience HT2+ (OR = 2.85, 95%CI = 1.35–6.01, P = 0.006) (Fig. 3).