Open Access * Accepted Blockinfor Blockinpublication: Blockin*

Objective: To identify the association between radiation dose volume and acute hematological toxicity (HT) in postoperative gynecological cancer patients receiving whole pelvic radiotherapy (RT) or intensity-modulated RT (IMRT), a principal component regression model was used to calculate HT. Methods: Women (n=100) receiving with or without chemotherapy RT were retrospectively analyzed, 52 of whom received chemotherapy (paclitaxel and nedaplatin). The pelvis and lumbar vertebrae, defined as the prolong-pelvic bone marrow, were divided into the (1) combined ilium, ischium and pubis and the (2) lumbar vertebrae and the sacrum. The V5-V40 of subsides was calculated. The complete blood counts were recorded weekly. The principal component analysis was performed on volumes which generated the principal components (PCs), followed by using a logistic regression model. Results: Forty-seven patients presented with grade 2-3 HT during RT. Chemotherapy increased the incidence of HT compared with RT alone (70.21% vs. 29.79%; p=0.001). Fifty-three patients with persistent HT developed more serious HT at an earlier stage of RT. The chemotherapy cycles and three PCs associated with grade 2-3 HT was identified to form the resulting principal logistic regression model. Conclusion: A new method to calculate the NTCP was achieved by PCs logistic regression.


.79%;p=0.001)
Fifty-three patients with persistent HT developed more serious HT at an earlier stage of RT.The chemotherapy cycles and three PCs associated with grade 2-3 HT was identified to form the resulting principal logistic regression model.Conclusion:A new method to calculate the NTCP was achieved by PCs logistic regression.

INTRODUCTION

Hematological toxicity (HT) is a common sid effect during the course of cancer treatments, reaching up to 60%. 1,2Chemoradiotherapy is the standard treatment for most locally advanced gynecological cancer patients, 3,4 but this has been observed to increase the risk of HT, which may prolong hospitalization and increase the

equirements for t
ansfusions and growth factors.

Approximately 60% of the body's total bone marrow (BM) is located in the pelvis and vertebrae, and it is usually included within the pelvic radiotherapy (RT) fields.The alleviation of HT theoretically depends on reducing the BM irradiation areas.BM stem cells are radiosensitive and predisposed to apoptosis damage in RT, resulting in myelosuppression. 5ntensity-modulated RT (IMRT) is designed to conform the rigid dose coverage to the planning target volume (PTV) by sparing the surrounding normal tissues. 6,7IMRT can reduce the volume of irradiated BM, especially with lower doses. 8,9owever, it is more difficult to constrain the large xtent of the irradiated BM because the pelvic bone and vertebrae are close to the PTV during IMRT, and these bones are contained in the fields during conventional whole pelvic RT.

Normal tissue complication probabilities (NTCP) models derived from dose volume histograms (DVHs) optimize the treatment planning according to biological cost functions, e.g., n, m, and D 50 for the Lyman-Kutcher-Burman model [10][11][12] or s, γ and D 50 for the seriality model. 13However, the collinearity among the dose volume parameters confounds the complication probabilities, resulting in inaccurate statistical results.We analyzed the collinearity and summarized the principal components (

s) of th
dose array to a logistic regression.We used a PCs logistic regression model rather than the original dose volume model to formulate a new method of calculating NTCP.


METHODS

One hundred patients with gynecologic malignancies who received conventional whole pelvic RT (AP/PA fields) or IMRT after transabdominal surgery between January 2012 and June 2013 were analyzed.All of the patients initially underwent pelvic computed tomography (CT).Complete blood counts were obtained on a weekly basis.Patients were excluded if they had received either extended-field (paraaortic) RT or four-field box technique treatments.The study design was approved by the appropriate ethics review boards, and written consent has been obtained from all patients.Chemotherapy administration: The concurrent and previous history of chemotherapy was recorded.Some patients (6 in Group one and two in Group II) had both concurrent and prior chemotherapy; thus, these patients were repeatedly calculated in the chemotherapy subgroups.The concurrent chemotherapy was delivered every three weeks (180 mg of paclitaxel on day one, and 100 mg or 30 mg of nedaplatin on day one or on days one, two and three).The patients who received chemotherapy before RT had a higher dosage of paclitaxel and nedaplatin (180-240 mg of paclitaxel and 100-120 mg of nedaplatin).The patients received one (21%), two (17%), three (4%), four (5%), five (1%) or more than six (4%) cycles of chemotherapy.Radiation planning: Group I consisted of the patients receiving conventional whole pelvic RT, and Group II consisted of the patients receiving IMRT.The radiation dose was 40-50 Gy given through five treatments per week and 1.8-2.0Gy per daily fraction.The patients' characteristics are listed in Table-I.

Seventy-seven of the patients in Group I received a conventional whole pelvic RT, extending from the bottom of the obturator foramina to the L4-5 interspace or above L4.The lateral border was 1.5 cm lateral to the widest diameter of the pelvic inlet.Blocks were used in 55 of the patients' treatment plans.

Twenty-three patients received IMRT within co-planar seven-or nine-field plans using 6-MV photons that were generated by commercial inverse treatment planning (Varian Eclipse platform 10.0).The clinical target volume (CTV) was contoured following the previous consensus on the recommendation for all patients, including the upper segment of the vagina, parametrial tissues and lymph ode regions (common, external, internal iliac and presacral lymph node regions), 14 and a 7-mm margin around the vessels was maintained, excluding bone or muscle.

The pelvis and lumbar vertebrae were deemed to be prolong-pelvic bone marrow (P-PBM), which was divided into two, subsides The combined ilium, ischium and pubis (IL extending from the iliac crests to the inferior border of the ischial tuberosities, including the acetabula and pubes)

The lumbar vertebrae and sacrum (LA) extending from the superior border of L1 to the entire sacrum.Dose-volume histograms (DVHs) were calculated for each patient from 5 Gy to 40 Gy at a 5-Gy dose-level interval (20 p tients did not receive CT examinations in our center; thus only the absolute volume parameters were recorded).These dosimetric parameters were noted as follows:

From IL-V5 to IL-V40 From LA-V5 to LA-V40 Hematological toxicity: HT was graded according to the Radiation Therapy Oncology Group's acute radiation toxicity scoring criteria. 15The endpoints of interest were the white blood count (WBC), absolute neutrophil count (ANC), hemoglobin (Hgb), and platelet count nadirs, and the highest grade for each of these endpoints was recorded 120 days after the start of the RT.An HT grade of 2-3 was noted as an event.None underwent grade 4 HT.Tran fusion was recorded in twenty-five patients in Group I during surgeries, while no patient in Group II had transfusion or none of the patients did during RT.

A proportion of the patients had HT before RT due to prior chemotherapy, and these patients received treatments including granulocyte-monocyte colony stimulating factor (CSF), erythropoietin (EPO) and interleukin-11 (IL-11) before RT to ensure normal blood cell levels.We recorded the first appearance of HT and the start of RT; thus, the interval time in these patients was negative.We also observed that some of the patients had persistent HT during RT; therefore, we defined persistent HT as weekly complete blood counts below the normal value ≥3 times.Statistical analysis: All of he DVHs original parameters were transformed to standardized values by the Z-score formula.The standardized values were used to construct a new dataset.

A principal component analysis (PCA) was performed on this new data array (N×N, where N is the number of DVHs parameters) by a correlation matrix.λ 1 λ 2 ≥…≥λ i denote the eigenvalues of PC

d e i denotes the eigenvector associat
d with eigenvalue λ i which is calculated as follows:

Dose volume and hematological toxicity A stepwise logistic regression approach was adopted to identify all PCs with higher grade HT for further analyses.The patients' factors (age, weight, chemotherapy, dose fraction, block, blood transfusion) were also analy ed using regression models.The principal component score function is expressed as follows:

F i = e 1i zx 1 + e 2i x 2 + e 3i zx 3 +...+ e ni zx n (zx n = standardized values of original data; n=1, 2, 3…N)

The univariate analysis of HT was calculated by the independent t test or chi-square/Fisher exact test.The collinearity was assessed by the tolerance and the variance inflation factor (VIF).A tolerance value of ≤1 or a VIF value of ≥10 demonstrated that the collinearity correlation existed.The statistics was obtained using SPSS 11.0.A p value of <0.05 was consid

tatistically significant.
he P-values were obtained from a two-sided test.


RESULTS


Patient characteristics:

Seventy-seven patients were treated with conventional RT, and 23 patients were treated with IMRT.In all, 47 patients had received prior chemotherapy, and 15 patients received concurrent chemotherapy.Table-II summarizes the dose volumes of the BM in the two groups.Significant differences were achieved from V5 to V15 and from V25 to V40 of the IL (p<0.01) and from V20 to V40 of the LA (p<0.05).Hematological toxicity: Overall, 47 patients (30 from Group I and 17 from Group II) experienced grade 2-3 HT during RT.Four patients ended the treatment due to the unendurable toxicity symptom of pain due to hemorrhoids.An increased risk of a higher grade acute HT was observed in the patients with total chemotherapy compared with RT alone (70.21% vs. 29.79%;p=0.001).Similarly, patients treated with prior chemotherapy experienced more grade 2-3 HT (63.83% vs. 36.17%;p=0.01,Table-III).However, patients in the concurrent chemotherapy subgroup had less of a risk of suffering HT in both of the groups (23.40% vs. 76.60%;p=0.027).Collectively, chemotherapy deliver d before RT had a more marked impact on HT than concurrent chemotherapy.

Medical treatments (CSF, EPO or IL-11) were provided to the patients with myelosuppression.However, a proportion of these patients still had abnormal blood counts in their next weekly examinations.A total of 53 patients encountered persistent HT (36 patients in Group I) and required more attention during their daily medical treatments.As shown in Table-IV, the patients with persistent HT developed more serious BM complications and were particularly prone to blood count reductions in the early stage of RT (for

ll: 69.81% vs. 21.27%;2.68±10.13days vs.14.85±12.5 days; p=0
000).


Principal component analysis (PCA) of the dosevolume array:

The volumes irradiated from 5 to 40 Gy are associated with myelosuppression and are therefore correlated with each ot er.We analyzed the collinearity among the different volume levels.The mathematical results showed that collinearity existed in all data levels (from V5 IL to V40 IL and The patients received chemotherapy before and/or during RT. from V5 LA to V40 LA ), and the minimum VIF value was 29 .019 of V20 IL , whereas all tolerance values decreased to values less than 0.1 (the maximum value was 0.03).

We aligned all DVHs parameters into a volume array for each patient.PCs were calculated from the standardized DVH parameters using PCA.Sixteen eigenvectors corresponding to the nonzero eigenvalues of the sample correlation matrix were generated by PCA

ig. 1 shows a screen plot displaying t
e percentage of the variation in the data array carried by each PC.


Principal component regression (PCR):

PCR used all PCs of the predictor variables as regressors.Three PCs (1st, 2nd and 16th) and chemotherapy ycles were significantly associated with grad 2-3 HT, as identified in the logistic regression (Table-V).

To predict a new patient's toxicity outcome,


DISCUSSION

Hematological toxicity (HT) rarely occurs in patients undergoing RT alone.Hematopoiesis injury is rapidly repaired because of the marrow's regeneration or the migration of unirradiated BM hem

opoietic st
m cells. 5In our center, most of patients receive whole pelvic RT (AP/PA fields) instead of four-field or box technique due to the prior clinical observation.The patients who receive four-field or box RT have a high risk of osteonecrosis of the femoral head and low-control of paraaortic lymph node tumor metastasis.

PCR, as one of the important multivariate calibration methods, can fully utilize all the dosimetric data and repeatedly perform factor analysis to identify the number of components in a mixed sample and pred ct the components in unknown samples with higher precision than univariate calibration. 16CR offers several analytical advantages and can determine PCs that describe relevant information.Most of the data from PCs contain noise.PCA is a technique used to reduce the dimensionality of a dataset by orders of magnitude to retain the maximum relevant information and to generate a few PCs to replace an entire dataset.This method is different from prior NTCP calculation which can better usage of volume data.The metrics of the dose volume effect associated with toxicity is based simply on the differences among the parameter values, which are therefore correlated to one another; thus this correlation should be of concern.The complications of multicollinearity are widely acknowledged but are not an area of focus in this study.More focused research on the strength of this correlation is warranted because it presents a problem for the current methodology.Further studies should be designed with consideration of collinearity.This analytical technique can be applicable to other diseases, such as rectal and anal cancer.

Our report notes that increased numbers of chemotherapy cycles increase the risk of a higher grade of HT.At the same time, we found that the patients with persistent HT were more likely to suffer abnormal blood counts at an early stage of RT and develop severe BM toxicity.Clinicians may therefore wish to be more attentive to patients who ha e several abnormal blood examination results because these patients have a higher likelihood of break in treatment and missed chemotherapy cycles.All of the patients with HT received drug therapy; however, patients with persistent HT do not have robust blood cells levels and require individually customized treatment strategies.


Source of funding: None.

Declaration of interest: None.



we applied the coefficient estimated from the regression model to the volume vector such that the principal logistic regression resulting model was expressed as logit P=-1.268+1.082×chemotherapycycles-0.255×F1+0.457×F2+158.632×F16PCswere returned to the original data by the Z-score f