Int. J. Radiation Oncology Biol. Phys., Vol. 70, No. 5, pp. 1431–1437, 2008Copyright � 2008 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/08/$–see front matter

doi:10.1016/j.ijrobp.2007.08.074

CLINICAL INVESTIGATION Anus

ASSOCIATION BETWEEN BONE MARROW DOSIMETRIC PARAMETERS AND ACUTEHEMATOLOGIC TOXICITY IN ANAL CANCER PATIENTS TREATED WITH

CONCURRENT CHEMOTHERAPY AND INTENSITY-MODULATED RADIOTHERAPY

LOREN K. MELL, M.D.,* DAVID A. SCHOMAS, M.D.,y JOSEPH K. SALAMA, M.D.,*z

KIRAN DEVISETTY, M.D.,* BULENT AYDOGAN, PH.D.,* ROBERT C. MILLER, M.D.,yASHESH B. JANI, M.D.,x

HEDY L. KINDLER, M.D.,k ARNO J. MUNDT, M.D.,{ JOHN C. ROESKE, PH.D.,*

AND STEVEN J. CHMURA, M.D., PH.D.*

*Department of Radiation and Cellular Oncology, University of Chicago and University of Illinois at Chicago, Chicago, IL;yDepartment of Radiation Oncology, Mayo Clinic, Rochester, MN; zThe Cancer Research Center, and kSection of

Hematology/Oncology, University of Chicago, Chicago, IL; xDepartment of Radiation Oncology, Emory University,Atlanta, GA; and {Department of Radiation Oncology, University of California San Diego, La Jolla, CA

Purpose: To test the hypothesis that the volume of pelvic bone marrow (PBM) receiving 10 and 20 Gy or more(PBM-V10 and PBM-V20) is associated with acute hematologic toxicity (HT) in anal cancer patients treated withconcurrent chemoradiotherapy.Methods and Materials: We analyzed 48 consecutive anal cancer patients treated with concurrent chemotherapyand intensity-modulated radiation therapy. The median radiation dose to gross tumor and regional lymph nodeswas 50.4 and 45 Gy, respectively. Pelvic bone marrow was defined as the region extending from the iliac crests tothe ischial tuberosities, including the os coxae, lumbosacral spine, and proximal femora. Endpoints included thewhite blood cell count (WBC), absolute neutrophil count (ANC), hemoglobin, and platelet count nadirs. Regressionmodels with multiple independent predictors were used to test associations between dosimetric parameters and HT.Results: Twenty patients (42%) had Stage T3–4 disease; 15 patients (31%) were node positive. Overall, 27 (56%),24 (50%), 4 (8%), and 13 (27%) experienced acute Grade 3–4 leukopenia, neutropenia, anemia, and thrombocy-topenia, respectively. On multiple regression analysis, increased PBM-V5, V10, V15, and V20 were significantly as-sociated with decreased WBC and ANC nadirs, as were female gender, decreased body mass index, and increasedlumbosacral bone marrow V10, V15, and V20 (p < 0.05 for each association). Lymph node positivity was significantlyassociated with a decreased WBC nadir on multiple regression analysis (p < 0.05).Conclusion: This analysis supports the hypothesis that increased low-dose radiation to PBM is associated withacute HT during chemoradiotherapy for anal cancer. Techniques to limit bone marrow irradiation may reduceHT in anal cancer patients. � 2008 Elsevier Inc.

IMRT, Bone marrow, Anal cancer, V20, Hematologic toxicity.

INTRODUCTION

Concurrent chemoradiotherapy (CRT) is a standard treatment

approach for patients with locoregionally confined carcinoma

of the anus (1). Two randomized trials have established that

concurrent 5-fluorouracil (5-FU), mitomycin C (MMC) and

radiotherapy (RT) improves locoregional control and colos-

tomy-free survival compared with RT alone (2, 3). 5-Fluoro-

uracil, MMC, and RT remains the standard CRT regimen

after two randomized trials that did not find superior out-

comes with alternative chemotherapy approaches (1, 4, 5).

Note—An online CME test for this article can be taken at http://asro.astro.org under Continuing Education.

Reprint requests to: Loren K. Mell, M.D., Department of Radia-tion and Cellular Oncology, University of Chicago Hospitals, 5758S. Maryland Ave./MC 9006, Chicago, IL 60637. Tel: (773) 702-6870; Fax: (773) 834-7340; E-mail: lmell@radonc.uchicago.edu

14

Despite excellent therapeutic results, acute hematologic

toxicity (HT) is common with this regimen, particularly

when 2 cycles of MMC are given. For example, Grade 4–5

and 3–4 HT in two multi-institutional studies were 18%

and 60%, respectively (4, 5). The large volume of bone mar-

row (BM) irradiated during conventional pelvic–inguinal RT

is likely a significant contributor to HT, because up to 50% of

a patient’s total hematopoietically active BM is within the

conventional treatment fields (6). Bone marrow stem cells

are highly radiosensitive, and the destruction of these cells

is a principal cause of acute myelosuppression (7).

Presented in part at the 48th Annual Meeting of the AmericanSociety of Therapeutic Radiology and Oncology (ASTRO),November 5–9, 2006, Philadelphia, PA.

Conflict of interest: none.Received July 9, 2007, and in revised form Aug 16, 2007.

Accepted for publication Aug 29, 2007.

31

1432 I. J. Radiation Oncology d Biology d Physics Volume 70, Number 5, 2008

One strategy to limit HT is to use intensity-modulated RT

(IMRT) to reduce the volume of BM irradiated during CRT

while maintaining adequate target volume coverage. Inten-

sity-modulated RT has been shown to reduce radiation

dose to the bowel, bladder, and genitalia compared with con-

ventional RT plans in the treatment of anal cancer (8, 9). Pre-

liminary results indicate that anal cancer patients treated with

IMRT have favorable outcomes and toxicity, with infrequent

treatment breaks, while delivery of 45 Gy to regional (pelvic

and inguinal) lymph nodes is enabled (9, 10). Intensity-mod-

ulated RT plans can be optimized to reduce BM irradiation as

well (BM-sparing IMRT, or BMS-IMRT) (11), but it is not

known whether this technique will reduce HT.

Identifying dosimetric parameters associated with HT is an

important step in designing an optimal BMS-IMRT approach.

In a previous study of cervical cancer patients treated with

IMRT and concurrent cisplatin, we found an association be-

tween acute leukopenia and the volume of pelvic BM

receiving 10 and 20 Gy or more (PBM-V10 and PBM-V20, re-

spectively) (12). The present study primarily aimed to test the

hypotheses that PBM-V10 and PBM-V20 were associated with

acute leukopenia in anal cancer patients treated with CRT.

METHODS AND MATERIALS

PatientsWe retrospectively analyzed 48 consecutive patients with carci-

noma of the anal canal treated with concurrent chemotherapy and

IMRT either at the University of Chicago and affiliated institutions

(n = 38) or the Mayo Clinic (n = 10) between October 2000 and Feb-

ruary 2006. This study was approved by the institutional review

board at each institution.

ChemotherapyThirty-five patients were treated according to a protocol using 2

cycles of 5-FU (1000 mg/m2/d on Days 1–4 and 29–32) with

MMC (10 mg/m2, maximum dose 20 mg, on Days 1 and 29).

Twenty-nine received full-dose chemotherapy, whereas 6 received

reduced doses of MMC or had the second MMC cycle omitted. Eight

patients received 2 cycles of 5-FU as above and 1 cycle of MMC (15

mg/m2 on Day 1) according to a different protocol. One patient re-

ceived 2 cycles of 5-FU as above and 2 cycles of cisplatin (50 mg/

m2 on Days 1 and 29). One patient received 2 cycles of 5-FU (250

mg/m2/d on Days 1–4 and 29–32) and 2 cycles of MMC (10 mg/

m2/d on Day 1 and 8 mg/m2/d on Day 29). One patient received 2

cycles of 5-FU (750 mg/m2/d on Days 1–4 and 29–32) and 1 cycle

of MMC (11.25 mg/m2/d on Day 1). Two patients received 2 cycles

of 5-FU (400–1000 mg/m2/d on Days 1–4 and 29–32) alone.

Transfusions and growth factorsGranulocyte-monocyte colony stimulating factor was given to 12

patients for severe neutropenia. Eight patients received erythropoi-

etin during CRT. Seven patients received transfusions of packed

red blood cells during treatment. No patients received platelet trans-

fusions.

Radiotherapy simulation, planning, and deliveryAll patients underwent computed tomography (CT) simulation in

the supine or frog-leg position with custom immobilization. The

gross tumor volume (GTV) consisted of radiographically visible dis-

ease, including the primary tumor and grossly involved lymph no-

des. The clinical target volume (CTV) consisted of the GTV and

pelvic and inguinal lymph nodes. In 3 patients, fields were limited

to the primary tumor, obturator, and internal and external iliac

lymph nodes. A margin between 1.0 and 1.5 cm was added to ac-

count for setup uncertainty and organ motion to generate the plan-

ning treatment volume (PTV).

The IMRT planning systems used were Helios or Eclipse (Varian

Medical Systems, Palo Alto, CA) or CORVUS versions 3.0 and 4.0

(NOMOS Corporation, Sewickley, PA). In general, seven to nine

equally spaced coplanar fields were used. The photon energy was

6 MV. The bowel and bladder, and for some patients the genitalia,

were delineated on the planning CT and used as avoidance struc-

tures for IMRT planning. In 15 patients the iliac BM was also

used as an avoidance structure for IMRT planning. Target goals

specified that at least 95% of the PTV receive the prescription

dose, no more than 20% of the PTV receive >110% of the prescrip-

tion dose, and no more than 1% of the PTV receive <93% of the pre-

scription dose. Normal tissue dose was kept as low as possible. Input

parameters for IMRT optimization were variable but in general

closely followed guidelines published for gynecologic cancer (13).

At the University of Chicago and affiliated hospitals, patients re-

ceived 45 Gy to the initial PTV and a subsequent 9–14.4 Gy boost to

the PTVboost (GTV plus margin), in 1.8-Gy daily fractions over 6

weeks. At the Mayo Clinic, patients were treated using a simulta-

neous integrated boost technique, typically delivering 25 fractions

(range, 25–28 fractions) of 2 Gy (range, 1.8–2.1 Gy), 1.8 Gy (range,

1.36–1.85 Gy), and 1.65 Gy (range, 1.36–1.85 Gy) to the PTVboost,

PTV1, and PTV2, respectively, where CTV2 consisted of the GTV

and regional lymph nodes, CTV1 consisted of the portion of the

CTV below the sacroiliac joints, and PTV2 and PTV1 were obtained

by expanding CTV2 and CTV1 with a 1–1.5 cm margin. In 1 case,

electron fields were added to increase dose to the inguinal lymph no-

des; the small contribution to PBM dose from these fields was dis-

regarded.

Bone marrow delineationThe external contour of the PBM was delineated on the planning

CT using bone windows. Pelvic BM was divided into three subsites:

(1) iliac BM (IBM), extending from the iliac crests to the superior

border of the femoral head, (2) lower pelvis (LP), consisting of

the pubes, ischia, acetabula, and proximal femora, extending

from the superior border of the femoral heads to the inferior border

of the ischial tuberosities, and (3) lumbosacral spine (LS), extending

from the superior border of the L5 vertebral body to the coccyx but

not extending below the superior border of the femoral head.

Cumulative dose–volume histograms (DVHs) corresponding to

the delivered IMRT plan were generated for each contoured BM

region. The volume of each region receiving 10 Gy or more was

quantified and designated PBM-V10, IBM-V10, LPBM-V10, and

LSBM-V10, for the pelvic, iliac, lower pelvic, and lumbosacral

BM, respectively. Dosimetric parameters for other dose levels

(5, 15, 20, 30, and 40 Gy) were designated similarly.

Nonhematologic toxicityPatients were monitored weekly during RT for acute skin, gastro-

intestinal, and genitourinary toxicity. Acute nonhematologic toxic-

ity was scored according to the Common Terminology Criteria for

Adverse Events, version 3.0.

Bone marrow parameters and acute hematologic toxicity d L. K. MELL et al. 1433

Hematologic toxicityAll patients had weekly complete blood counts with differential

during treatment. Hematologic toxicity was graded according to

the Radiation Therapy Oncology Group acute radiation morbidity

scoring criteria (14). Endpoints of interest were the white blood

cell count (WBC), absolute neutrophil count (ANC), hemoglobin

(Hgb), and platelet count nadirs and highest grade of each toxicity

occurring within 60 days of initiation of CRT.

Statistical analysisMultiple linear and logistic regression models (i.e., models with

multiple independent predictors) were used to test correlations be-

tween dosimetric parameters and HT endpoints. Demographic

(age, race, gender, body mass index [BMI]) and tumor (tumor

size, nodal status) characteristics were included in the model as co-

variates. Body mass index was calculated as weight (in kilograms)/

height (in meters) squared. Number of cycles and dose of MMC (in

milligrams) given in the first cycle were also included in the model

to test for confounding. Fisher’s exact and Student’s t tests were

used to test differences between dichotomous and continuous vari-

ables, respectively, on univariate analyses. Pearson’s product–mo-

ment correlation coefficient was used to quantify correlations

between explanatory variables. R2 and adjusted R2 values were

used to evaluate goodness of fit of linear models. Tests for

normality were performed using the Shapiro-Wilk statistic. The

Box-Cox transformation, i.e.,

tðx; lÞ ¼ ðxl � 1Þ=l if ls0lnðxÞ if l ¼ 0

where l is a fitted parameter, was used to eliminate skew in non-nor-

mally distributed explanatory variables. The Mann-Whitney U sta-

tistic was used to compare areas under DVH curves. Statistical

analyses were performed with SAS (SAS Institute, Cary, NC) and

STATA (StataCorp, College Station, TX) software.

RESULTS

Patient and tumor characteristicsPatient demographic characteristics and stage are listed in

Table 1. The mean and median age were 58 and 55 years, re-

spectively (range, 32–88 years). The mean and median BMI

were both 27 kg/m2 (range, 18–59 kg/m2). Twenty patients

(42%) had Stage T3–4 disease, and 15 patients (31%) had

node-positive disease.

Radiation delivery and dosimetric parametersThe median total dose was 50.4 Gy (range, 32.0–59.4 Gy)

to the primary. The median duration of CRT was 43 days

(range, 21–58 days). Nineteen patients (40%) required a treat-

ment break, with a median total duration of CRT of 48 days.

Reasons for treatment breaks were HT in 10 patients (21%),

skin toxicity in 8 patients (17%), and gastrointestinal toxicity

in 1 patient (2%).

Descriptive statistics of dose–volume parameters are listed

in Table 2. The lower pelvis constituted the largest subsite of

PBM (43% of the total PBM volume on average). The mean

PBM-V10 and PBM-V20 were 85% and 75%, respectively.

The median PBM-V10 and PBM-V20 were 88% and 78%,

respectively. The PBM-V10 and PBM-V20 were closely cor-

related (Pearson’s coefficient 0.87, p < 0.0001).

Nonhematologic toxicityDetailed results of nonhematologic toxicity for patients in-

cluded in this study have been reported previously (10). Of

the 44 patients with available nonhematologic toxicity

data, 59% experienced acute Grade 2 skin desquamation,

and 36% experienced Grade 3 skin desquamation. Acute

Grade 2 and 3 gastrointestinal toxicity was seen in 59%

and 11% of patients, respectively. Acute Grade 2 and 3

genitourinary toxicity was seen in 11% of patients and 0 pa-

tients, respectively. No patients had Grade 4 nonhematologic

toxicity.

Hematologic toxicityThe mean baseline WBC, ANC, Hgb, and platelet counts

were 7.0 � 109/L (range, 2.8–17.9 � 109/L), 4.5 � 109/L

(range, 1.5–13.1 � 109/L), 12.9 g/dL (range, 9.6–16.2 g/

dL), and 262� 109/L (range, 83–443� 109/L), respectively.

The median follow-up period for collecting complete blood

count data was 45 days from the start of treatment (range,

30–58 days). Overall, 27 (56%), 24 (50%), 4 (8%), and 13

(27%) experienced acute Grade 3–4 leukopenia, neutropenia,

anemia, and thrombocytopenia, respectively, during treat-

ment (Table 3). The median nadir WBC, ANC, Hgb, and

platelet count during treatment were 1.9 k/mL (range,

0.2–4.9 k/mL), 1.2 k/mL (range, 0.0–4.2 k/mL), 10.0 g/dL

(range, 3.1–15.2 g/dL), and 94 k/mL (range, 10–222 k/mL),

respectively.

On univariate analysis, increased PBM-V10, PBM-V15,

LSBM-V10, LSBM-V15, and LSBM-V20, were significantly

associated with a decreased WBC nadir during CRT (Table

4). The mean WBC nadirs for patients with LSBM-V20

#65% vs. >65% were 2.5 k/mL vs. 1.6 k/mL, respectively

Table 1. Demographic and tumor characteristics

Characteristic n %

Total 48 100Female 28 58African American 15 31BMI >27 kg/m2 (median) 24 50HIV positive 7 15Age >55 y (median) 24 50TNM Stage

T1 N0 3 6T1 N1 1 2T2 N0 19 40T2 N1 4 8T2 N2 1 2T3 N0 9 19T3 N1 4 8T3 N2 2 4T3 N3 3 6T4 N0 2 4

Abbreviations: BMI = body mass index; HIV = human immuno-deficiency virus; TNM = tumor, node, metastasis.

1434 I. J. Radiation Oncology d Biology d Physics Volume 70, Number 5, 2008

(p = 0.02). Comparisons of WBC nadirs for other dosimetric

parameters, using mean values as a cutoff, were not statisti-

cally significant.

Covariates significantly associated with a lower WBC na-

dir were female gender, decreasing BMI, and lymph node–

positive status (Table 4). Female patients, for example, had

an average WBC nadir 1.18 k/mL lower than male patients

(p = 0.001), despite having equivalent baseline WBC counts

(6.9 vs. 7.0, p = 0.95). In fact, of the 14 patients with Grade 4

leukopenia, 12 (86%) were female, compared with 16 of 34

(47%) with Grade #3 leukopenia (p = 0.023). Mitomycin

C dose and number of cycles were not significantly associ-

ated with WBC nadir.

Table 2. Descriptive statistics of dose–volume parameters

Parameter Mean SD

PBMVolume (cm3) 1204 227V5 (%) 90 13V10 (%) 85 15V15 (%) 81 16V20 (%) 75 17V30 (%) 56 19V40 (%) 32 17

IBMVolume (cm3) 366 79V5 (%) 85 20V10 (%) 75 21V15 (%) 66 21V20 (%) 55 20V30 (%) 34 18V40 (%) 17 13

LSBMVolume (cm3) 324 66V5 (%) 82 21V10 (%) 75 22V15 (%) 71 23V20 (%) 66 23V30 (%) 49 26V40 (%) 30 23

LPBMVolume (cm3) 514 104V5 (%) 99 6V10 (%) 98 12V15 (%) 96 14V20 (%) 92 16V30 (%) 74 21V40 (%) 43 21

Abbreviations: SD = standard deviation; PBM = pelvic bone mar-row (BM); IBM = iliac BM; LSBM = lumbosacral BM; LPBM =lower pelvic BM; Vx = volume receiving $ x Gy.

Table 3. Acute hematologic toxicity

Grade

Toxicity 0 1 2 3 4

Leukopenia 4 (8) 6 (13) 11 (23) 13 (27) 14 (29)Neutropenia 11 (23) 2 (4) 11 (23) 9 (19) 15 (31)Anemia 17 (35) 11 (23) 16 (33) 3 (6) 1 (2)Thrombocytopenia 22 (46) 6 (13) 7 (15) 8 (17) 5 (10)

Values are number (percentage).

Increasing PBM-V5 and LSBM-V5 were associated with

a decreased WBC nadir on univariate analysis, but the

associations were not statistically significant (p = 0.07 in

each case). Similarly, increasing V5, V10, and V20 of IBM

and LPBM were associated with a decreased WBC nadir

on univariate analysis, but the associations were not statisti-

cally significant (p > 0.10 in each case). V30 and V40 of PBM

and each subsite were not significantly correlated with the

WBC nadir (p > 0.46 in each case). No significant correlation

between WBC or ANC nadirs and either the use of IBM spar-

ing or the length of follow-up period was observed.

On multiple regression analysis, PBM-V5, PBM-V10,

PBM-V15, PBM-V20, LSBM-V5, LSBM-V10, LSBM-V15,

and LSBM-V20 were significantly associated with WBC

and ANC nadirs (Table 5). Female gender, decreasing BMI,

and lymph node–positive status were also significantly asso-

ciated with decreasing WBC nadir on multiple regression

analysis, whereas female gender and decreasing BMI were

significantly associated with decreasing ANC (Table 5). R2

and adjusted R2 values ranged from 0.41 to 0.54 and from

0.35 to 0.45, respectively, indicating that the models ex-

plained a moderate degree of the variation in WBC and

ANC nadirs. No significant correlation between dosimetric

parameters and Hgb or platelet count nadirs was observed.

The mean dose to PBM and LSBM were not significantly as-

sociated with WBC nadir (p = 0.07 and p = 0.10, respec-

tively), although it is possible that a significant relationship

might be detected with a larger sample.

With linear modeling of the WBC nadir, the best fits were

obtained with PBM-V15 (adjusted R2 = 0.45) and LSBM-V15

(adjusted R2 = 0.43). Plots of the WBC nadir with respect to

PBM-V15 and LSBM-V15 are shown in Fig. 1. Although the

probability of HT is predicted to be a continuous function of

BM dose or dosimetric parameters, these models may be use-

ful for identifying cutoff values to be used as dose–volume

Table 4. Univariate analysis of factors associated with whiteblood cell count nadir

Variable b* SE p

PBM-V5 (%) �0.025 0.014 0.073PBM-V10 (%) �0.025 0.012 0.034PBM-V15 (%) �0.024 0.011 0.037PBM-V20 (%) �0.020 0.010 0.058LSBM-V5 (%) �0.016 0.008 0.068LSBM-V10 (%) �0.019 0.008 0.019LSBM-V15 (%) �0.020 0.008 0.010LSBM-V20 (%) �0.018 0.008 0.020Age 0.023 0.013 0.086African American race �0.207 0.398 0.605Female gender �1.184 0.332 0.001BMI 0.098 0.027 0.001HIV positive 0.133 0.524 0.801T3 or T4 �0.249 0.376 0.512Lymph node positive �0.808 0.381 0.039

Abbreviations: SE = standard error; PBM = pelvic bone marrow(BM); LSBM = lumbosacral BM; BMI = body mass index; HIV =human immunodeficiency virus.

* Change in white blood cell count nadir (k/mL) per unit changein independent variable.

Bone marrow parameters and acute hematologic toxicity d L. K. MELL et al. 1435

constraints for IMRT planning. Our analysis indicates that

women and patients with lower BMI or lymph node–positive

disease are more likely to experience leukopenia, so the pre-

dicted constraints would depend on these factors. The best

predicted dose–volume constraints for PBM-V15 and

LSBM-V15, for patients of average BMI (27 kg/m2) according

to gender and lymph node–positive status are given in Table 6.

Multiple logistic regression analysis did not identify sig-

nificant associations between dosimetric parameters and

Grade 3–4 leukopenia or neutropenia. Lumbosacral spine

BM-V10 was, however, associated with a non–statistically

significant increased odds of Grade 4 leukopenia on univar-

iate analysis (odds ratio [OR], 1.06; 95% confidence interval

[CI], 1.00–1.12; p = 0.051) and multiple regression analysis

(OR, 1.06; 95% CI, 0.99–1.12; p = 0.092). Analysis of

LSBM DVHs illustrates the disparity in LSBM irradiation

for patients with and without Grade 4 leukopenia (Fig. 2).

Pelvic BM-V5, PBM-V10, PBM-V15, PBM-V20, LSBM-V5,

LSBM-V10, LSBM-V15, and LSBM-V20 were not normally

distributed (Shapiro-Wilk p < 0.05 in each case), owing to

the inclusion of 3 patients with low volumes of BM irradiated.

As illustrated in Fig. 1, 3 patients had PBM-V15 < 60%. The

Table 5. Multiple linear regression (adjusted) coefficientestimates for dosimetric parameters associated with acute

hematologic toxicity

Variable b* SE p

WBC nadirPBM-V5 (%) �0.025 0.011 0.028PBM-V10 (%) �0.023 0.009 0.019PBM-V15 (%) �0.022 0.009 0.019PBM-V20 (%) �0.019 0.009 0.031LSBM-V5 (%) �0.016 0.007 0.029LSBM-V10 (%) �0.016 0.006 0.017LSBM-V15 (%) �0.015 0.007 0.026LSBM-V20 (%) �0.015 0.006 0.024Female gendery �1.062 0.350 0.004BMI (kg/m2)y 0.084 0.028 0.004LN positivey �0.653 0.320 0.048

ANC nadirPBM-V5 (%) �0.020 0.010 0.041PBM-V10 (%) �0.018 0.008 0.028PBM-V15 (%) �0.018 0.008 0.026PBM-V20 (%) �0.015 0.007 0.045LSBM-V5 (%) �0.012 0.006 0.056LSBM-V10 (%) �0.012 0.006 0.045LSBM-V15 (%) �0.013 0.005 0.026LSBM-V20 (%) �0.011 0.006 0.043Female gendery �0.679 0.300 0.029BMIy 0.058 0.024 0.018LN positivey �0.465 0.274 0.098

Abbreviations: SE = standard error; WBC = white blood cellcount; PBM = pelvic bone marrow (BM); LSBM = lumbosacralBM; BMI = body mass index; LN = lymph node; Vx = volume re-ceiving $ x Gy; ANC = absolute neutrophil count.

Models included one dosimetric parameter and each of the fol-lowing covariates: gender, BMI (continuous), LN positive status,age (continuous), African American race, HIV positive status, andT3 or T4 stage.

* Change in WBC or ANC nadir (k/mL).y Results presented for model including PBM-V10.

associations between BM dosimetric parameters and WBC

and ANC nadirs became statistically nonsignificant when these

patients were excluded from the analysis (PBM-V10:

b =�0.035, p = 0.17; PBM-V15: b =�0.017, p = 0.46; PBM-

V20: b =�0.005, p = 0.78; LSBM-V10: b =�0.018, p = 0.15;

LSBM-V15: b = �0.018, p = 0.15; LSBM-V20: b = �0.012,

p = 0.29). Results from models using Box-Cox transformations

yielded borderline significant coefficient estimates for PBM-

V10 (p = 0.08), LSBM-V10 (p = 0.05), LSBM-V15 (p = 0.06),

and LSBM-V20 (p = 0.05), and nonsignificant coefficient esti-

mates for PBM-V5 (p = 0.11), PBM-V15 (p = 0.17), PBM-V20

(p = 0.34), and LSBM-V5 (p = 0.14).

DISCUSSION

The purpose of this study was to validate findings from our

previous study in which the V10 and V20 of PBM and LSBM

were found to correlate with HT in cervical cancer patients

receiving concurrent pelvic IMRT and weekly cisplatin

(12). In general, the results from the present study support

the hypothesis that these parameters are also associated

Fig. 1. (a) Plot of white blood cell count (WBC) nadir vs. volume ofpelvic bone marrow receiving 15 Gy or more (PBM-V15). Regres-sion coefficient (b) = �0.024 k/mL/%, p = 0.037. (b) Plot of WBCnadir vs. volume of lumbosacral bone marrow receiving 15 Gy ormore (LSBM-V15). b = �0.020 k/mL/%, p = 0.010.

1436 I. J. Radiation Oncology d Biology d Physics Volume 70, Number 5, 2008

with acute HT in patients undergoing CRT for anal cancer.

These two studies yielded remarkably similar results, despite

the differences in patients studied, total radiation dose, treat-

ment volume, and chemotherapy delivered. Both studies

identified significant associations between WBC and ANC

nadirs and the V10 and V20 of PBM and LSBM, suggesting

that these associations might be true generally for patients re-

ceiving concurrent chemotherapy and radiation to the pelvic

region. The lack of correlation between Hgb or platelet nadirs

and BM dosimetric parameters, and between HT and BM V30

and V40, was also consistent in both studies.

The hypothesized correlation between HT and PBM-V10

and V20 is based on numerous studies demonstrating the

high radiosensitivity of BM stem cells (7, 15–18). Destruc-

tion of BM stem cells is a principal cause of acute myelosup-

pression (7), although effects on peripheral blood stem cells

and stromal tissue also seem to be important (7). Radiation

doses below 30 Gy cause acute myelosuppression and both

radiographic (19) and pathologic (18, 20) changes to BM.

Chronic myelosuppression and irreversible morphologic

BM changes can occur at doses above 30 Gy (20).

In patients with gynecologic cancer, Brixey et al. (21)

showed that IMRT was associated with both lower HT and

lower BM dose compared with patients treated with conven-

tional 4-field box techniques, indicating that IMRT could be

used to reduce HT. Subsequent investigations have shown

that IMRT plans can be further optimized to reduce BM (i.e.,

BMS-IMRT) in comparison with conventional techniques in

gynecologic (22–25) and anal (11) cancer. Recently described

approaches to BMS-IMRT have shown reductions in LSBM ir-

radiation and PBM-V20 compared with anteroposterior–poster-

oanterior techniques in gynecologic and anal cancer (11, 16);

however, PBM-V10 seems to be slightly increased with this

technique. It is therefore unclear what effect current BMS-

IMRT techniques, which use conventional doses, fractionation,

and planning margins, might have on HT.

A strength of this study was the use of a cohort of consec-

utively treated patients to test a specific hypothesis using

methods determined a priori. However, care in interpreting

Table 6. Predicted dose–volume constraints to maintaina WBC nadir $2.0 k/mL, according to gender and nodal

status

Gender/node status PBM-V15 LSBM-V15

Female/LN negative 68 61Female/LN positive 44 24Male/LN negative 100* 100*Male/LN positive 80 70

Abbreviations: WBC = white blood cell count; PBM = pelvic bonemarrow (BM); LSBM = lumbosacral spine BM; LN: lymph node.

Values are percentages.Models (variables in parentheses take on a value of 1 if the con-

dition is true, 0 otherwise):WBC nadir (k/mL) = 4.5 – 0.87*(female) � 0.57*(node positive)� 0.024* PBM-V15.

WBC nadir (k/mL) = 3.7 – 0.73*(female) � 0.58*(node positive)� 0.016* LSBM-V15.

* Values greater than 100% shown as 100%.

these results is warranted, because the conclusions were sen-

sitive to the exclusion of data from a few patients with very

low volumes of BM irradiated. It is thus unclear the extent

to which the relationship between HT and BM dosimetric pa-

rameters is valid over the entire range of parameter values,

and there is still unexplained variation with these models,

owing to the presence of some high WBC nadirs in patients

with high PBM-V10 and PBM-V20. These models are there-

fore presently unlikely to provide robust predictions of nor-

mal tissue complication probability, and more sophisticated

models may be required to characterize accurately the depen-

dence of HT on dosimetric and clinical factors.

Further research into strategies to reduce low-dose BM ir-

radiation is therefore needed to validate the observed associ-

ations and evaluate the potential benefits of BM sparing. Use

of image-guided RT to improve patient setup reproducibility

and reduce required planning margins is one such strategy

under investigation. Imaging techniques such as single pho-

ton emission computed tomography (SPECT) (26) and mag-

netic resonance imaging (MRI) (27) have also been proposed

to optimize BMS-IMRT plans, by identifying hematopoieti-

cally active regions of BM that could be spared preferentially.

These studies have identified a high concentration of active

BM in the LS spine, medial ilium, and iliac crest (26, 27). Re-

sults of this and our previous study (12) indicate a stronger

association between HT and irradiation of LSBM than other

subsites of PBM, suggesting that ‘‘partial’’ BM-sparing ap-

proaches might be successful, in addition to being simpler

than PBM sparing. Using SPECT or MRI to correlate acute

HT with BM changes during CRT would also be informative

in the study of adaptive RT. To our knowledge, however,

such studies have not yet been performed. Finally, future in-

vestigation could examine the ability of simultaneous inte-

grated boost planning to reduce BM irradiation.

A practical aspect of this study was to help identify target

constraints for BMS-IMRT planning, for both the optimiza-

tion process (i.e., penalty or input parameters) and plan eval-

uation. The significant dependence of HT on gender, nodal

status, and BMI suggests that appropriate constraints may

need to be tailored to the patient on the basis of these

Dose (Gy)0 5 10 15 20 25 30 35 40 45 50 55

0102030405060708090

100

No Grade 4 LeukopeniaGrade 4 Leukopenia

Lum

bosa

cral

Vol

ume

(%)

Fig. 2. Lumbosacral bone marrow dose-volume histograms for pa-tients with (n = 14) and without (n = 34) Grade 4 leukopenia. p =0.62 for comparison of area under the curve, p < 0.05 for comparisonof area under the curve from 0 to 20 Gy (Mann-Whitney U test).

Bone marrow parameters and acute hematologic toxicity d L. K. MELL et al. 1437

demographic characteristics. For example, women with low

BMI and node-positive disease would be predicted to derive

the greatest benefit from BMS-IMRT, whereas men with a nor-

mal to high BMI and node-negative disease would be predicted

to derive less benefit. However, we do not presently recom-

mend application of BMS-IMRT outside of a study setting.

Several randomized controlled trials have demonstrated

improved therapeutic outcomes with CRT, compared with

RT alone, in cervical, anal, and rectal cancer. Chemoradiother-

apy is also standard treatment for many patients with vulvar and

vaginal cancer and is under investigation in the treatment of

advanced endometrial cancer. Hematologic toxicity remains

a significant problem that limits chemotherapy delivery, neces-

sitates transfusions and growth factors, and occasionally results

in life-threatening infections. To the extent that BM sparing can

reduce HT and improve the therapeutic ratio of CRT for can-

cers of the pelvic region, it has broad potential utility.

In conclusion, this analysis supports the hypothesis that the

V10 and V20 of PBM and LSBM are associated with acute

leukopenia and neutropenia during CRT for anal cancer.

Techniques to limit BM irradiation, such as BMS-IMRT,

may reduce HT in these patients, but further efforts to reduce

PBM-V10 may be required, and prospective analyses of BM-

sparing techniques are needed.

REFERENCES

1. Cummings BJ. Current management of anal canal cancer. SeminOncol 2005;32:S123–S128.

2. Bartelink H, Roelofsen F, Eschwege F, et al. Concomitant ra-diotherapy and chemotherapy is superior to radiotherapy alonein the treatment of locally advanced anal cancer: Results ofa phase III randomized trial of the European Organization forResearch and Treatment of Cancer Radiotherapy and Gastroin-testinal Cooperative Groups. J Clin Oncol 1997;15:2040–2049.

3. Epidermoid anal cancer: Results from the UKCCCR rando-mised trial of radiotherapy alone versus radiotherapy, 5-fluoro-uracil, and mitomycin. UKCCCR Anal Cancer Trial WorkingParty. UK Co-ordinating Committee on Cancer Research.Lancet 1996;348:1049–1054.

4. Flam M, John M, Pajak TF, et al. Role of mitomycin in combi-nation with fluorouracil and radiotherapy, and of salvage chemo-radiation in the definitive nonsurgical treatment of epidermoidcarcinoma of the anal canal: results of a phase III randomized in-tergroup study. J Clin Oncol 1996;14:2527–2539.

5. Gunderson LL, Winter KA, Ajani JA, et al. Intergroup RTOG9811 phase III comparison of chemoradiation with 5-FU andmitomycin vs 5-FU and cisplatin for anal canal carcinoma: Im-pact on disease-free, overall and colostomy-free survival[Abstract]. Int J Radiat Oncol Biol Phys 2006;66:S24.

6. Ellis RE. The distribution of active bone marrow in the adult.Phys Med Biol 1961;5:255–258.

7. Mauch P, Constine L, Greenberger J, et al. Hematopoietic stemcell compartment: Acute and late effects of radiation therapy andchemotherapy. Int J Radiat Oncol Biol Phys 1995;31:1319–1339.

8. Milano MT, Jani AB, Farrey KJ, et al. Intensity-modulated ra-diation therapy (IMRT) in the treatment of anal cancer: Toxicityand clinical outcome. Int J Radiat Oncol Biol Phys 2005;63:354–361.

9. Chen YJ, Liu A, Tsai PT, et al. Organ sparing by conformalavoidance intensity-modulated radiation therapy for anal can-cer: Dosimetric evaluation of coverage of pelvis and inguinal/femoral nodes. Int J Radiat Oncol Biol Phys 2005;63:274–281.

10. Salama JK, Mell LK, Schomas DA, et al. Concurrent chemo-therapy and intensity-modulated radiation therapy for anal canalcancer patients: A multicenter experience. J Clin Oncol 2007;25:4581–4586.

11. Mell LK, Aydogan B, Salama J, et al. Bone marrow-sparing in-tensity modulated radiotherapy versus conventional techniquesfor pelvic-inguinal irradiation in anal cancer [Abstract]. Pre-sented at the Annual Meeting of the American Radium Society,May 5–9, 2007, Amsterdam, The Netherlands.

12. Mell LK, Kochanski JD, Roeske JC, et al. Dosimetric predictorsof acute hematologic toxicity in cervical cancer patients treatedwith concurrent cisplatin and intensity-modulated pelvic radio-therapy. Int J Radiat Oncol Biol Phys 2006;66:1356–1365.

13. Mell LK, Roeske JC, Mehta N, et al. Gynecologic cancer over-view. In: Mundt AJ, Roeske JC, editors. Intensity modulated

radiation therapy: A clinical perspective. Toronto: BC Decker;2005. p. 497.

14. Radiation Therapy Oncology Group. Acute radiation morbidityscoring criteria. Available at: http://www.rtog.org/members/toxicity/acute.html. Accessed May 14, 2007.

15. Sacks EL, Goris ML, Glatstein E, et al. Bone marrow regener-ation following large field radiation: Influence of volume, age,dose, and time. Cancer 1978;42:1057–1065.

16. Rubin P, Landman S, Mayer E, et al. Bone marrow regenerationand extension after extended field irradiation in Hodgkin’s dis-ease. Cancer 1973;32:699–711.

17. Sykes MP, Chu FC, Savel H, et al. The effects of varying dos-ages of irradiation upon sternal-marrow regeneration. Radiol-ogy 1964;83:1084–1088.

18. Hall EJ, Giaccia AJ. Clinical response of normal tissues. In:Hall EJ, Giaccia AJ, editors. Radiobiology for the radiologist.6th ed. Philadelphia: Lippincott Williams & Wilkins; 2006. p.333–337.

19. Blomlie V, Rofstad EK, Skjonsberg A, et al. Female pelvicbone marrow: Serial MR imaging before, during, and after radi-ation therapy. Radiology 1995;194:537–543.

20. Fajardo LF, Berthrong M, Anderson RE. Hematopoietic tissue.In: Fajardo LF, Berthrong M, Anderson RE, editors. Radiationpathology. Oxford: Oxford University Press; 2001. p. 379–388.

21. Brixey CJ, Roeske JC, Lujan AE, et al. Impact of intensity-modulated radiotherapy on acute hematologic toxicity inwomen with gynecologic malignancies. Int J Radiat OncolBiol Phys 2002;54:1388–1396.

22. Lujan AE, Mundt AJ, Yamada SD, et al. Intensity-modulatedradiotherapy as a means of reducing dose to bone marrow in gy-necologic patients receiving whole pelvic radiotherapy. Int JRadiat Oncol Biol Phys 2003;57:516–521.

23. Aydogan B, Tiryaki H, Mell LK. Bone marrow-sparing inten-sity modulated radiotherapy versus conventional techniquesfor cervical cancer [Abstract]. Int J Radiat Oncol Biol Phys2007;69(suppl.):S401.

24. Wong E, D’Souza DP, Chen JZ, et al. Intensity-modulated arctherapy for treatment of high-risk endometrial malignancies. IntJ Radiat Oncol Biol Phys 2005;61:830–841.

25. Ahmed RS, Kim RY, Duan J, et al. IMRT dose escalation for pos-itive para-aortic lymph nodes in patients with locally advancedcervical cancer while reducing dose to bone marrow and other or-gans at risk. Int J Radiat Oncol Biol Phys 2004;60:505–512.

26. Roeske JC, Lujan A, Reba RC, et al. Incorporation of SPECTbone marrow imaging into intensity modulated whole-pelvic ra-diation therapy treatment planning for gynecologic malignan-cies. Radiother Oncol 2005;77:11–17.

27. Roeske JC, Mundt AJ. Incorporation of magnetic resonance im-aging into intensity modulated whole-pelvic radiation therapytreatment planning to reduce the volume of pelvic bone marrowirradiated. Int Congress Series 2004;1268:307–312.