C

Nlrsttiinp

RC5

S

Cj

Int. J. Radiation Oncology Biol. Phys., Vol. 68, No. 4, pp. 1030–1035, 2007Copyright © 2007 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2007.01.031

LINICAL INVESTIGATION Lung

RADIATION PNEUMONITIS: LOCAL DOSE VERSUS[18F]-FLUORODEOXYGLUCOSE UPTAKE RESPONSE IN IRRADIATED LUNG

THOMAS GUERRERO, M.D., PH.D.,* VALEN JOHNSON, PH.D.,† JUSTIN HART, M.D., PH.D.,*TINSU PAN, PH.D.,‡ MEENA KHAN, B.S.,* DERSHAN LUO, PH.D.,* ZHONGXING LIAO, M.D.,*

JAFFER AJANI, M.D.,§ CRAIG STEVENS, M.D., PH.D.,� AND RITSUKO KOMAKI, M.D.*

*Division of Radiation Oncology, and Departments of †Biostatistics and Applied Mathematics, ‡Imaging Physics, and§Gastrointestinal Medical Oncology, University of Texas M.D. Anderson Cancer Center, Houston, TX; �Division of Radiation

Oncology, H. Lee Moffitt Cancer Center, University of South Florida, Tampa, FL

Purpose: To quantify the relationship between the local radiation dose received and the posttreatment positronemission tomography/computed tomography (PET/CT) [18F]2-fluoro-2-deoxyglucose (FDG) uptake in the lung.Methods and Materials: The data from 36 patients treated for esophageal cancer with thoracic radiotherapy whounderwent restaging PET/CT imaging between 4 and 12 weeks after radiotherapy completion were evaluated.Their treatment planning CT was registered with the restaging PET/CT. Using histogram analysis, the voxelaverage FDG-PET uptake vs. radiation dose was obtained for each case. Hierarchical linear regression modelsfor each patient were applied to study the variation in the linear trends between cases. Deviation of thedose–response curve from a linear model was tested.Results: The median time between radiotherapy completion and FDG-PET imaging was 40 days (range, 26–70days). The median of the mean standard uptake value in the lung that received 0–5 Gy was 0.63 (range,0.36–1.27), 5–10 Gy was 0.77 (range, 0.40–1.35), 10–20 Gy was 0.80 (range, 0.40–1.72), and >20 Gy was 1.08(range, 0.44–2.63). A hierarchical linear regression model of the radiation dose and normalized FDG uptake percase found an adequate fit with the linear model, and the addition of quadratic and logarithmic functions did notimprove the fit. The 36 cases had a posterior mean of slopes range of 0.0048–0.069.Conclusion: The regional dose vs. radiation pneumonitis response was evaluated with FDG-PET/CT imaging.Statistical modeling found a linear relationship. The slope of this relationship varied over an order of magnitude,reflecting the range of the underlying biological response to radiation among the study population. © 2007 ElsevierInc.

Radiation pneumonitis, Pulmonary injury, Computed tomography, Positron emission tomography.

mr

i[pdpait(

TAvCFf

INTRODUCTION

early all patients undergoing radiotherapy for thoracic ma-ignancies develop some degree of lung injury detectable byadiographic studies, and a subset of patients with the greatestensitivity have the most severe complications. In clinicalrials, these subsets of patients experience the dose-limitingoxicities, typically radiation pneumonitis, and define the max-mal tolerable dose. A recent clinical series of hemithoracicntensity-modulated radiotherapy after extrapleural pneumo-ectomy for malignant mesothelioma reported a 47% fatalneumonitis rate (1). A better understanding of radiation pneu-

Reprint requests to: Thomas Guerrero, M.D., Ph.D., Department ofadiation Oncology, Unit 97, University of Texas M.D. Andersonancer Center, 1515 Holcombe Blvd., Houston, TX 77030. Tel: (713)63-2300; Fax: (713) 563-2366; E-mail: tguerrero@mdanderson.org

Presented at the 89th Annual Meeting of the American Radiumociety, May 5–9, 2007, Amsterdam, The Netherlands.Supported by funding from the University of Texas M.D. Anderson

ancer Center Multi-Disciplinary Research Project titled “Lung In-

ury Following Cancer Treatment” and through the University of A

1030

onitis, including the individual biologic variation and theelation between the local dose and response, is needed.

Radiation pneumonitis is an inflammatory reaction withinrradiated lung tissue in response to radiation injury (2, 3).18F]-2-fluoro-2-deoxy-D-glucose positron emission tomogra-hy (FDG-PET) can visualize and quantitate endotoxin-in-uced pneumonitis in normal healthy volunteers (4) and inatients with cystic fibrosis (5). Radiation pneumonitis is char-cterized by the migration of leukocytes from the blood torradiated lung tissue; thus, on FDG-PET imaging, more in-ense inflammatory responses will result in greater FDG uptake4). Previous studies with FDG-PET imaging in the postradio-

exas M.D. Anderson Cancer Center’s Physician Scientist Program.cknowledgments—We extend our gratitude to Dr. Elizabeth Tra-is who organized the University of Texas M.D. Anderson Cancerenter Multi-Disciplinary Research Project titled “Lung Injuryollowing Cancer Treatment,” which provided the motivation andunds for this project.

Conflict of interest: none.Received Nov 21, 2006, and in revised form Jan 6, 2007.

ccepted for publication Jan 7, 2007.

tpr

ranetise

lt

P

ogspawtwwaifC3Cs(tFrPa(whs

I

iitgrmsiuop

Pamr

ebpvatsitt

C

f

S

Tctdemw3

fseusrotnf

N

Aris

S

rd

1031Radiation pneumonitis FDG uptake ● T. GUERRERO et al.

herapy setting have focused on tumor response (6, 7) orerformed only qualitative evaluation of the lung radiationesponse (8).

At our institution, esophageal cancer patients undergoestaging FDG-PET/CT imaging approximately 6 weeksfter chemoradiotherapy (9). These patients have no pulmo-ary malignant involvement and to adequately treat theirsophageal cancer they received incidental irradiation ofheir lungs. Although variations in the lung’s response torradiation is expected with differing diseases and treatmentettings, this group comprises a good set of patients for thestimation of the normal lung response.

In this study, we performed quantitative evaluation of theocal dose—FDG-PET response in normal lung tissue afterhoracic radiotherapy for esophageal cancer.

METHODS AND MATERIALS

atient dose and PET dataThirty-six patients treated at the Department of Radiation Oncol-

gy, University of Texas M.D. Anderson Cancer Center for esopha-eal cancer between November 1, 2004 and October 31, 2005 wereelected for this study. Patients who had undergone CT treatmentlanning and follow-up PET-CT imaging between 24 and 84 daysfter radiotherapy completion at the M.D. Anderson Cancer Centerere selected. Patients whose imaging studies were obtained with

heir arms at the side or whose PET/CT contained image artifactsere excluded. The patient identifiers were removed in accordanceith an M.D. Anderson Cancer Center institutional review board-

pproved retrospective study protocol and Health Insurance Portabil-ty and Accountability Act (HIPAP) regulations were strictly en-orced. Each patient underwent a treatment planning session in whichT images of the entire thorax and upper abdomen were obtained at-mm slice spacing on a Phillips MX8000 IDT treatment-planningT scanner (Philips Medical Systems, Cleveland, OH) or at 2.5-mm

lice spacing on a General Electric Discovery ST PET/CT scannerGE Medical Systems, Waukesha, WI) with a 70-cm bore. Thereatment planning CT scans were obtained during free breathing. AnDG-PET/CT scan was obtained between 24 and 84 days afteradiotherapy completion on the General Electric Discovery STET/CT scanner (GE Medical Systems). The use of average CT forttenuation correction was introduced during the period of this data set10). Approximately one-half of the PET/CT images were acquiredith mid-inspiratory breath-hold CT for the CT component, the otheralf used an average CT obtained from four-dimensional CT acqui-ition.

mage registration and lung segmentationThe treatment plan for each patient was restored and imported

nto the Pinnacle-3, version 7.6c, treatment planning system (Phil-ps Medical Systems, Andover, MA). The radiation dose distribu-ions were recalculated using the collapsed-cone convolution al-orithm (11), which includes lung heterogeneity correction. Theestaging FDG-PET/CT scan was imported into the Pinnacle treat-ent planning system and spatially registered to the planning CT

can using the Syntegra image fusion software, version 1.2b (Phil-ps Medical Systems). Manual segmentation of the vertebral col-mn was performed on each CT volume. A CT-to-CT six-degree-f-freedom rigid body registration of the vertebral column was

erformed between the treatment planning CT scan and the f

ET/CT scan using a mutual information algorithm (12). A radi-tion oncologist visually verified each registration and performedinor adjustments of �1 cm along each axis to improve the

egistration of the lung regions.The images, registration parameters, and dose distribution were

xported for analysis. A pulmonary segmentation algorithm (13),ased on an eight-point connectivity scheme and a set of three seedoints, was applied to each treatment planning CT image set. CToxels between �920 and �250 Hounsfield units (HU) were selecteds lung parenchyma if they were positive for connectivity to one of thewo lung seed points and not connected to the tracheal-bronchial treeeed point. The resulting lung parenchyma region of interest was usedn the subsequent analysis. The lung segmentation was also applied tohe CT scan from the restaging PET/CT scan and used in analysis ofhe pulmonary standard uptake values (SUVs).

alculation of 18F-FDG uptakeThe SUVs were calculated from the PET count rate using the

ollowing equation (14):

tandard Uptake Value

�

18F-FDG count rate per mL � body weight (gm)

decay corrected 18F-FDG injected dose (Bq)(1)

he mean SUV within the pulmonary tissue was obtained for eachase. Using the registration between the treatment planning CT andhe PET/CT scans, the mean SUVs within the lung tissue for theose ranges of 0–5, 5–10, 10–20, and �20 Gy were obtained forach case. The median of the mean SUVs and the range of theean values for the 36 cases were determined. The maximal SUVithin the pulmonary tissue irradiated to �5 Gy was found for all6 cases.Using the registered dose and PET images, histograms were

ormed of the FDG-PET count rate vs. the radiation dose usingingle-voxel values, resulting in an average count rate per voxel atach dose. The histogram bin size was 1 Gy (range, 1–50). Thepper limit of 50 Gy was chosen because most patients in thistudy were treated to a prescribed dose of 50.4 Gy; few lung voxelseceived �50 Gy. The average value for the 1–5 Gy range wasbtained as an estimate of the FDG uptake in nonirradiated lungissue. The normalized FDG uptake was calculated using theonirradiated lung tissue estimate as an internal control and theollowing equation:

ormalized 18F-FDG Uptake

�count rate irradiated lung per mL

count rate non-irradiated lung per mL(2)

hierarchical linear regression model was applied to estimate theegression line on the normalized FDG uptake vs. the dose for eachndividual case and to estimate the variation in these lines acrossubjects.

tatistical analysisLetting yij denote the normalized FDG uptake for subject i at

adiation dose j, and letting Np��, �� denote a p-variate normalistribution with mean � and covariance matrix �, a model of the

orm

y

wtwsTrdt

P

tarpG

v

mHm7(t(�v3

t(M(Tc1SatS

N

Pumw0wb�mS2fpF3ts

irorcr0sfa0

A

G

S

L

P

P

I

C

v

1032 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 4, 2007

ij � �0i � �1i xi � �ij, �ij � N1(0, (xi � 1)�2),

��0i

�1i� � N2(�, �) (3)

as applied to the data. Jeffrey’s prior densities were assumed forhe hyperparameters �, �2, and �. The adequacy of the model fitas assessed by constructing a Bayesian chi-square statistic from

tandardized, sampled residuals at each dose level, across patients.his process produced a test statistic distributed as a chi-square

andom variable on 150 degrees of freedom. Additions of qua-ratic and logarithm functions of dose into the regression equa-ions were tested.

RESULTS

atient and image characteristicsThe patient characteristics of the 36 cases evaluated in

his study are summarized in Table 1. All the cases evalu-ted were mid-thoracic or lower esophageal cancer prima-ies. The median prescription dose was 50.4 Gy, all theatients received that dose, except for one who received 45y and one who received 59.4 Gy.On the pretreatment planning CT scan, the median lung

olume was 3250 cm3 (range, 1681–7188 cm3) and the

Table 1. Patient characteristics

Characteristic Value

ge (y)Median 68Range 48–85

ender (n)Male 26Female 8

tage (n)I 1 (2.8)II 10 (27.8)III 19 (52.8)IVA 4 (11.1)IVB 2 (5.6)

ocationMiddle 12 (33.3)Middle to GE junction 2 (5.6)Lower 7 (19.4)GE junction 15 (41.7)

TV (mL)Median 893Range 437–1,406

rescription dose (Gy)Median 50.4Range 45–59.4

nterval between RT completion and PET (d)Median 40Range 26–70

hemotherapyInduction before RT 21 (58.3)Concurrent with RT 32 (88.9)

Abbreviations: GE � gastroesophageal; PTV � planning targetolume; RT � radiotherapy.

sData in parentheses are percentages.

edian lung CT value was �715 HU (range, �815 to �561U). On the posttreatment PET/CT scan, the CT-deter-ined median lung volume was 3601 cm3 (range, 2140–

116 cm3) and the median lung CT value was �715 HUrange, �796 to �554 HU). From the radiotherapy plans,he median lung volume receiving �5 Gy was 61.8%range, 10.9–87.0%). The median lung volume receiving10 Gy was 41.2% (range, 8.1–59.8%). The median lung

olume receiving �20 Gy (V20) was 23.2% (range, 4.7–7.7%).The median time between radiotherapy completion and

he restaging FDG-PET/CT imaging session was 40 daysrange, 26–70 days). The median injected dose was 633.3

Bq (range, 447.5–740.0 MBq), and the uptake periodbefore imaging) was 85.8 min (range, 57.3–144.0 min).he mean SUV in the lung was calculated for each of the 36ases; the median lung mean SUV was 0.81 (range, 0.40 to.53). No correlation was found between the mean lungUV value and the time between radiotherapy completionnd the PET/CT imaging session (restaging PET delayime). No correlation was found between the mean lungUV value and either the injected dose or the uptake period.

ormalized FDG uptakeThe registration of the treatment planning CT with the

ET/CT scan was performed, and correlative values derivedsing the radiation dose and FDG-PET were calculated. Theedian of the mean SUV from lung that received 0–5 Gyas 0.63 (range, 0.36–1.27), 5–10 Gy was 0.77 (range,.40–1.35), 10–20 Gy was 0.80 (0.40–1.72), and �20 Gyas 1.08 (range, 0.44–2.63). No correlation was foundetween these values and the median lung volume receiving5 Gy, median lung volume receiving �10 Gy, V20, orean lung dose. The median for the 36 cases of the maximalUV within the lung irradiated to �5 Gy was 5.1 (range,.3–18.2). The isodose distribution and FDG-PET imagesrom 2 cases are shown in Fig. 1 to illustrate the range ofulmonary response, with high (Case 1) and low (Case 2)DG uptake response shown in the irradiated lung. For all6 cases, the normalized FDG uptake per dose was ob-ained. Fig. 2 shows the normalized FDG uptake for theame 2 cases from Fig. 1.

The correlation between the radiation dose and normal-zed FDG uptake found that all 36 cases had a positiveesponse slope. No correlation was found between the slopef the normalized FDG uptake vs. dose and either theestaging PET delay time or the median lung volume re-eiving �5 Gy (Fig. 3). All 36 cases had a positive FDGesponse slope, with median of 0.016 (range, 0.0048 to.069); a histogram distribution of the posterior mean of thelopes found is given in Fig. 4. Also, no difference wasound in the FDG response slope with the presence orbsence of induction chemotherapy, 0.023 (SD, 0.018) vs..018 (SD, 0.012). The response slope was varied in this

tudy population over an order of magnitude.

A

f

�

p1dar

rldTwppri

evsffba

uptake

FerCfC6

1033Radiation pneumonitis FDG uptake ● T. GUERRERO et al.

nalysis of uptake vs. dose relationThe posterior mean values of the population parameters

rom the model given by Eq. 3 were �� � (0.947, 0.020)T,

Fig. 1. Radiation dose and [18F]2-fluoro-2-deoxygluc(a) Isodose distribution delivered to this esophageal cancfrom treatment planning computed tomography (CT) scanresponse), after image registration, shown overlaying trhigh FDG uptake indicated by horizontal arrow. (c) Istreatment planning CT scan for Case 2 (low response). (overlaying treatment planning CT scan. Pulmonary regiodiagonal arrow. These 2 cases represent range of FDG

ig. 2. Normalized [18F]2-fluoro-2-deoxyglucose (FDG) positronmission tomography (PET) response. Normalized FDG uptakeesponse per voxel vs. radiotherapy dose (in Gray) shown forases 1 and 2 from Fig. 1. Linear regression analysis performed

or each case, and resulting lines shown; correlation coefficient forases 1 and 2 was 0.96 and 0.97, respectively. Case 1 had a slope

t.3 times greater than that for Case 2.

�2 � 0.00263, and �� � � .00155 .000330

.000330 .000212 . The

osterior mean of the chi-square statistics was 173, with only1% of values greater than the 0.95 quantile of their nominalistribution, demonstrating an adequate model fit (15). Theddition of quadratic and logarithm functions of dose into theegression equations did not improve the model fit.

DISCUSSION

This report is the first to quantify the local pneumonitisesponse to thoracic radiotherapy in a patient population. Ainear relationship was found between the local radiationose and the voxel-averaged posttreatment FDG uptake.he slope of that relationship varied across the patients andas independent of the interval between radiotherapy com-letion and PET imaging (within 1–3 months) and of theercentage of volume of lung irradiated. We propose thisesponse slope is a measure of the underlying individualntensity of radiation pneumonitis.

Clinical studies of radiation pneumonitis have assumedqual susceptibility among patients and that the dose andolume of the lung irradiated determine the risk of theubsequent pneumonitis. In a review of 99 patients treatedor non–small-cell lung cancer using three-dimensional con-ormal radiotherapy, Graham et al. (16) found a correlationetween the total lung volume irradiated to V20 (p � 0.001)nd the development of worse than Grade 2 pneumonitis. In

DG) positron emission tomography (PET) response.ent (Case 1) shown overlaying single transaxial sectionorresponding restaging FDG-PET scan for Case 1 (high

al treatment planning CT scan. Pulmonary region withdistribution shown overlaying transaxial section fromresponding restaging FDG-PET scan shown for Case 2igh dose and its corresponding PET region indicated byresponse found in all 36 cases evaluated.

ose (Fer pati. (b) C

ansaxiodosed) Corn of h

hat study, the fatal pneumonitis rate was 4%. They noted

omed

lcbutsrowedpir

ttrtc

terua

F(h0e

1034 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 4, 2007

utliers in their data, for example 1 patient with fatal pneu-onitis had a V20 of only 22%, which is clinically consid-

red a low risk value. Similarly, the V20 of the patients whoeveloped worse than Grade 3 (severe) pneumonitis, over-

ig. 4. Histogram distribution of [18F]2-fluoro-2-deoxyglucoseFDG) uptake response. Distribution of response slopes shown inistogram format for 36 cases evaluated. Slopes ranged from.0048 to 0.069/Gy, with more than an order of magnitude differ-

Fig. 3. [18F]2-fluoro-2-deoxyglucose (FDG) response vseach of 36 cases plotted vs. time between radiotherapy cwas found within this range (24–78 days). (b) Slope of llung volume that received �5 Gy. No correlation wasvolume receiving �5 Gy.

bnce from smallest to largest.

apped the V20 of patients who did not experience anyomplications. Kwa et al. (17) evaluated 540 patients withreast or lung cancer who received thoracic radiotherapysing dose–volume analysis. Of these 540 patients, 73 pa-ients developed Grade 2 pneumonitis or worse, with aignificant difference (p � 0.02) between the pneumonitisates in the lung cancer and breast cancer groups. In a studyf children with refractory or relapsed Hodgkin’s diseaseho received high-dose therapy and autologous hematopoi-

tic cell transplantation, atopy was found to predict foriffuse lung injury syndrome (18). In that study, 44% of theatients had a history of allergic rhinitis and/or asthma. Themportance of the variation in individual susceptibility foradiation pneumonitis can be seen in these studies.

In this study, we have demonstrated a method to measurehe underlying individual intensity of radiation pneumonitis,he slope of the radiation dose posttreatment FDG uptakeesponse. Additional development of this method, includinghe evaluation of a larger cohort and the correlation withlinical toxicity observations, is in progress.

CONCLUSION

In this study, the local dose effect after irradiation of lungissue was evaluated using restaging FDG-PET imaging insophageal cancer patients. The data supported a linearelation between the radiation dose and the normalized FDGptake in the lung. The slope of this relationship varied overn order of magnitude, reflecting the range of the underlying

and volume. (a) Slope of linear regression analysis fortion and positron emission tomography. No correlationegression for each of 36 cases plotted vs. percentage ofbetween FDG response slope and percentage of lung

. timeompleinear rfound

iologic response to radiation among the study population.

1035Radiation pneumonitis FDG uptake ● T. GUERRERO et al.

REFERENCES

1

1

1

1

1

1

1

1

1

1. Allen AM, Czerminska M, Janne PA, et al. Fatal pneumo-nitis associated with intensity-modulated radiation therapyfor mesothelioma. Int J Radiat Oncol Biol Phys 2006;65:640 – 645.

2. Fajardo LF, Berthrong M, Anderson RE. Radiation pathol-ogy. New York: Oxford University Press; 2001; p. 200 –206.

3. Roberts CM, Foulcher E, Zaunders JJ, et al. Radiation pneu-monitis: A possible lymphocyte-mediated hypersensitivity re-action. Ann Intern Med 1993;118:696–700.

4. Chen DL, Rosenbluth DB, Mintun MA, et al. FDG-PETimaging of pulmonary inflammation in healthy volunteersafter airway instillation of endotoxin. J Appl Physiol 2006;100:1602–1609.

5. Chen DL, Ferkol TW, Mintun MA, et al. Quantifying pulmo-nary inflammation in cystic fibrosis with positron emissiontomography. Am J Respir Crit Care Med 2006;173:1363–1369.

6. MacManus MP, Hicks RJ, Matthews JP, et al. Positron emis-sion tomography is superior to computed tomography scan-ning for response-assessment after radical radiotherapy orchemoradiotherapy in patients with non-small-cell lung can-cer. J Clin Oncol 2003;21:1285–1292.

7. Hicks RJ, Kalff V, MacManus MP, et al. 18F-FDG PETprovides high-impact and powerful prognostic stratification instaging newly diagnosed non-small cell lung cancer. J NuclMed 2001;42:1596–1604.

8. Nestle U, Hellwig D, Fleckenstein J, et al. Comparison ofearly pulmonary changes in 18FDG-PET and CT after com-bined radiochemotherapy for advanced non-small-cell lungcancer: A study in 15 patients. Front Radiat Ther Oncol

2002;37:26–33.

9. Ajani JA, Komaki R, Putnam JB, et al. A three-step strategyof induction chemotherapy then chemoradiation followed bysurgery in patients with potentially resectable carcinoma of theesophagus or gastroesophageal junction. Cancer 2001;92:279–286.

0. Pan T, Mawlawi O, Luo D, et al. Attenuation correction ofPET cardiac data with low-dose average CT in PET/CT. MedPhys 2006;33:3931–3938.

1. Ahnesjo A. Collapsed cone convolution of radiant energy forphoton dose calculation in heterogeneous media. Med Phys1989;16:577–592.

2. Studholme C, Hawkes DJ, Hill DLG. A normalised entropymeasure for multi-modality image alignment. Proc SPIE MedImaging 1998;3338:132–142.

3. Hu S, Hoffman EA, Reinhardt JM. Automatic lung segmen-tation for accurate quantitation of volumetric X-ray CT im-ages. IEEE Trans Med Imaging 2001;20:490–498.

4. Strauss LG, Conti PS. The applications of PET in clinicaloncology. J Nucl Med 1991;32:623–648.

5. Johnson VE. Bayesian chi-squared test for goodness-of-fit.Annals of Statistics 2004;32:2361–2384.

6. Graham MV, Purdy JA, Emami B, et al. Clinical dose–volumehistogram analysis for pneumonitis after 3D treatment fornon–small-cell lung cancer (NSCLC). Int J Radiat Oncol BiolPhys 1999;45:323–329.

7. Kwa SL, Lebesque JV, Theuws JC, et al. Radiation pneumonitisas a function of mean lung dose: An analysis of pooled data of540 patients. Int J Radiat Oncol Biol Phys 1998;42:1–9.

8. Frankovich J, Donaldson SS, Lee Y, et al. High-dose therapyand autologous hematopoietic cell transplantation in childrenwith primary refractory and relapsed Hodgkin’s disease:Atopy predicts idiopathic diffuse lung injury syndromes. Biol

Blood Marrow Transplant 2001;7:49–57.