DOI 10.1378/chest.106.6_Supplement.377S 1994;106;377S-381SChest

 Slamon, Adi Gazdar and John MinnaPiantadosi, Valerie Rusch, Jan A. Nowak, Donald McIntire, Dennis David P. Carbone, Tetsuya Mitsudomi, Itsuo Chiba, Steven Cancer : A Preliminary Report of LCSG 871

LungMutations in Resected Non-small Cell Imperfectly Correlated With GeneAssociated With Reduced Survival and Is p53 Immunostaining Positivity Is

  http://chestjournal.chestpubs.org/content/106/6_Supplement/377S

services can be found online on the World Wide Web at: The online version of this article, along with updated information and 

ISSN:0012-3692)http://chestjournal.chestpubs.org/site/misc/reprints.xhtml(

prior written permission of the copyright holder.this article or PDF may be reproduced or distributed without theDundee Road, Northbrook, IL 60062. All rights reserved. No part of Copyright1994by the American College of Chest Physicians, 3300Physicians. It has been published monthly since 1935.

is the official journal of the American College of ChestChest

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

p53 Immunostaining Positivity IsAssociated With ReducedSurvival and Is ImperfectlyCorrelated With Gene Mutationsin Resected Non-small CellLung Cancer*

A Preliminary Report of LCSG 871

David P. Carbone, MD, PhD; Tetsuya Mitsudomi, MD;Itsuo Chiba, MD; Steven Piantadosi, PhD;Valerie Rusch, MD, FCCP; Jan A. Nowak, MD;Donald McIltire, PhD; Dennis Slamon, MD;Adi Gazdar, MD; and John Minna, MD

We investigated the correlation of p53 abnormalitieswith survival in 85 patients with non-small cell lungcancer (NSCLC) who had undergone resection withcurative intent as part of Lung Cancer Study Group(LCSG) 871. Our previous studies showed that only a

subset of p53 mutations in lung cancers result inoverexpression. In addition, protein overexpression hasbeen described in the absence of mutation. Therefore,we determined both p53 protein overexpression (by im-munostaining) and p53 and ras gene mutations (by sin-gle-strand conformation polymorphism and DNA se-

quencing) in this set of resected tumor specimens. Clin-ical follow-up data were available for 75 cases. Of thestudied patients, 64% showed p53 overexpression and51% had mutant p53 sequences; however, the concor-

dance rate was only 67%. There was a negative survivalcorrelation with positive p53 immunostaining (p=0.05),but not with the presence of gene mutations (p=0.62) inthis group of patients. Overexpression of p53 proteindetermined by immunostaining may contribute to ad-verse outcome due to the ability of p53 to act as a dom-inant oncogene, or alternatively, overexpression may

reflect ongoing DNA damage in the tumor as a markerfor a more aggressive behavior. When adjusted for stage,age, and gender by multivariate analysis, however, therewas no independent impact of p53 overexpression on

survival.

Lung cancer is the leading cause of cancer death in theUnited States and in much of the world, and p53 is

the gene most commonly found to be mutant in lung can-

cer and many other common cancers.1 Overall 5-year

*From the Simmons Cancer Center, University of Texas South-western Medical Center, Dallas, (Drs. Carbone, McIntire, Gaz-dar, and Minna); Department of Surgery II, Faculty of Medi-cine, Kyushu UniversitK, Fukuoka, Japan (Dr. Mitsudomi);Hokkaido University School of Dentistry, Department of FirstOral Surgery, Sapporo, Japan (Dr. Chiba); Johns Hopkins Uni-versity School of Medicine, Baltimore (Dr. Piantadosi); Memo-rial Sloan Kettering Cancer Center, New York (Dr. Rusch);Department of Pathology, Illinois Masonic Medical Center,Chicago (Dr. Nowak); and University of California at Los An-geles (Dr. Slamon).

Reprint requests: Dr. Carbone, UT Southwestern Med Ctr, 5323Harry Hines NB8.206, Dallas, TX 75235-8593

survival for lung cancer remains low at less than 15%. Inthe subgroup of patients with resectable non-small cell lungcancer (NSCLC stages I to IIIA), long-term survival ispossible, but even in this group of patients, overall survivalis only 25 to 30%. Many clinical studies are ongoing to testthe ability of chemotherapeutic and radiotherapeuticregimens to improve the survival of such "curable"patients, and molecular tests that help to distinguishhigh-risk from low-risk patients would be of clinical util-ity.

p53 is a complex, multifunctional protein. Early studiesshowed p53 to act in a dominant fashion, cooperating withan activated ras gene to transform primary rat embryo fi-broblast cells.2 It was subsequently discovered that the p53clones used for transfection were, in fact, mutant, withsingle amino acid residue substitutions, and that the wild-type p53 gene is not only inactive, but can suppress trans-formation by other oncogenes.3 Thus, normal p53 came tobe viewed as a "tumor suppressor gene."3

p53 mutations in lung cancer can be of many types, in-cluding missense, nonsense (truncations), splicing abnor-malities, or deletions. We have previously shown that p53protein overexpression was associated with only the mis-sense mutant subset of these mutants.4 Other studies haveshown the half-life of such mutant proteins is increasedfrom 20 to 30 min for wild type to several hours for themutant.5 In addition, there are reports of p53 proteinoverexpression occurring in the absence of a detectablemutation, especially after DNA damage due to a variety ofsources, including ionizing radiation.6 In this case, a path-way is activated that includes products of the ataxia-telangectasia gene locus and results in posttranscriptionalincrease of wild-type p53 protein occurring within 30 minof the DNA damaging event.

While the importance of p53 mutations in the patho-genesis of human lung cancer is clear, it is not clearwhether the presence or absence of p53 mutations oroverexpression of p53 protein adversely affects an indi-vidual patient's chances for survival. Some authors reportadverse prognosis associated with p53 mutation,7 somestate that p53 overexpression detected by positive immu-nostaining adversely affects outcome,7-9 and some say thatthese changes have no effect10 or even improve prognosis.11In none of these studies was mutation analysis withsequencing as well as immunohistochemistry performedon the same set of samples with significant follow-up. Thus,the available data do not allow us to determine whether itis mutation of p53 with loss of suppressor function, muta-tion with gain of dominant transforming ability, or whetherit is simply overexpression of wild-type p53 protein as areflection of underlying and ongoing DNA damage (or allor none of these) that play a role in an individual patient'sprognosis.

This study was designed to address this question by se-lecting a cohort of carefully staged and evaluated patientswith NSCLC who had undergone resection for cure as partof Lung Cancer Study Group (LCSG) study 871, andevaluating both p53 gene mutations via single-strand con-formation polymorphism analysis and DNA sequencing,and p53 overexpression by immunohistochemistry. Muta-tions in ras genes were determined in this same population

CHEST / 106 / 6 / DECEMBER, 1994 / Supplement 377S

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

for comparative analysis. This is a preliminary report of apartial data set.

METHODSPatient Population

We studied tumor material from 85 patients with NSCLC whohad undergone resection with curative intent and entered LCSGstudy 871 between June 26, 1987 and February 29, 1988. The ageat resection ranged from 44 to 83 years (median, 61 years); therewere 21 women and 63 men, and all patients were of good per-formance status at the time of surgery (Karnofsky, 80 to 100). Themedian follow-up was 752 days, and survival data were availableon 75 of the 85 patients. Thirty-seven of the 75 were alive at thetime of last follow-up. No clinical data were available on one pa-tient.

Tumor CharacteristicsOf the 85 resected tumors, there were 43 adenocarcinomas, 34

squamous cell carcinomas, 7 large cell carcinomas, and 1 carci-noid tumor. Thirty tumors were TI, 47 were T2, and 7 were T3.Pathologic staging revealed the following distribution of nodalinvolvement: 58 NO, 10 Ni, and 16 N2.

Molecular AnalysisTumors were flash frozen at the time of surgery, and DNA and

RNA were prepared by standard techniques. 12 All 85 tumors wereanalyzed for mutations in p53 exons 4 through 8 by single-strandconformation polymorphism analysis. In our hands, this tech-nique is about 90% sensitive, has no false positives, and can de-tect mutations in the presence of 90% contaminating normal tis-sue.13 In addition, most of these samples were analyzed by a sec-ond technique as well, including sequencing of the full openreading frame of 6 cDNA clones in both directions, RNase pro-tection, or chemical cleavage techniques.' All 85 samples werealso analyzed for mutations in ras using designed restrictionfragment length polymorphism and polymerase chain reaction-single-strand conformation polymorphism analysis.14

Immunostaining AnalysisParaffin sections containing tumor with good preservation were

available in 39 of these cases. Four- to 5-Am-thick sections weredeparaffinized and immunostained with the mouse monoclonalantibody BP53-12 (BioGenex Laboratories, San Ramon, Calif35)using standard immunohistochemical methods. Sections of tumorwere incubated with antibody (1:500 in phosphate-buffered salinesolution, 2 h at room temperature) and subsequently detectedusing an immunostaining reagent kit (Vectastain Elite ABC,Vector Laboritories, Burlingame, Calif) with diaminobenzidineas the chromagen. Slides were counterstained with hematoxylin,dehydrated, and coverslipped.

Sections of tumor were examined for evidence of nuclearstaining, with a semiquantitative assessment of signal intensity

Table I-p53 and ras Mutation Correlationin the Subset of 50 Adenocarcinoma or Large

Cell Carcinoma Cases*

ras Mutant ras Normal Total(%) (%) (%)

p53 mutant 3 (6)1 18 (36) 21 (42)p53 normal 15 (30) 14 (28) 29 (58)Total 18 (36) 32 (64) 50 (100)

*The presence of p53 and ras mutations were determined in the subsetof adenocarcinoma tumors, and their correlation compared.tThe presence of tumors doubly mutant for p53 and ras was signif-icantly less than predicted by chance (p=0.006).

Table 2-Concordance of p53 Mutation Analysis andImmunohistochemistry (IHC)*

IHC + IHC-

p53 mutant 16 4p53 wild type 9 10

*The presence of p53 mutations determined by genetic analysis andimmunohistochemical positivity were correlated in the subset ofsamples where both pieces of information were available. The con-cordance rate is 67%.

and proportion of cells staining. In most positive cases, positivecells were present diffusely throughout the section of tumor; inother cases, the positive staining was focal. In most of the positivecases, the density of positive nuclei ranged from 50 to 100%. Inthree cases, only about 30% of the tumor cells had stained nuclei.One positive case had fewer than 10% of the tumor cell nucleistaining for p53 antigen. They were read without prior knowledgeof the p53 mutation status.

Statistical MethodsComparisons of the patient groups for survival were based on

the use of time-to-event methods, including Kaplan-Meier esti-mation,16 stratified log-rank tests,17 and proportional hazards re-gression models.18 A stepwise variable selection approach wasused in a proportional hazards model to select statistically signif-icant classification variables. Statistical significance was set at asignificance level of 0.05 and all tests were two-tailed as were thereported p values.

RESULTSp53 and ras Mutation AnalysisForty-five of the 85 samples (53%) had DNA abnormal-

ities in p53. These were found to be distributed in thevarious exons of p53 in a manner similar to those previouslyreported.19 Of the 3.3 squamous cell tumors, none had rasmutations, also consistent with the literature.20 In the sub-set of tumors, including pure adenocarcinoma, adenobron-choalveolar, adenosquamous, pure bronchoalveolar, orcarcinoid, 16 of 43 (or 36%) were mutant for ras (14 atK-ras codon 12, 1 at K-ras codon 13, and 1 at K-ras codon61), also consistent with previously reported frequencies.When the coincidence of p53 and ras mutations was com-pared in this subset of tumors, there were statistically fewertumors with both mutations than would be expected if the

Table 3-Distribution of p53 Mutations andImmunostain Positivity by Exon*

Exon No. Mutant IHC +

4 2 0/15 19 6/76 1 1/17 9 3/48 13 6/6Splice 1 0/1Totals 45 16/20

*The locations of the 45 p53 mutations are listed in the first columnby exon, and the numbers in each exon that are immunostain-pos-itive are listed in the second column over the total number evaluated.It can be seen that most of the mutations were located in exons 5,7, and 8, and that most of these mutations resulted in protein over-expression. IHC =immunohistochemistry.

p53 Immunostaining & Reduced Survival (Carbone et al)378S

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

0 250 500 750 1000Days post-op

FIGURE 1. Survival of patients who have undergoneresection for NSCLC by the status of the p53 genein their tumors. Mutations in p53 were determinedusing single-strand conformation polymorphismanalysis and DNA sequencing (see text). No differ-ence in survival was seen between patients with and

1250 1500 1750 without detected p53 genetic abnormalities in theirtumors.

._

cnCD)CD

EU

FIGURE 2. Survival of patients who have undergoneresection for NSCLC, stage III disease, by the sta-tus of the p53 gene in their tumors. No differencein survival was seen between patients with andwithout detected p53 genetic abnormalities in theirtumors.

100

90

80

70

60

50

40

30

20

100

0 200 400 600 800 1000Days Post-op

'a

E

100

90

80

70

60

50

40

30

20

10

0

0 250 500 750 1000Days Post-op

1250 150C

two occurred independently (p=0.006, Table 1).

p53 Immunohistochemical AnalysisTwenty-five of the 39 tumor paraffin blocks evaluable

for p53 expression (64%) were positive by immunostainingin 10% or more of the tumor cells. When these were cor-

related with p53 mutations, there was a 67% concordance(Table 2). The p53 mutations that resulted in positive im-munostaining were distributed over exons 5 to 8 (Table 3).

FIGURE 3. Survival of patients who have undergoneresection for NSCLC, stage I and II disease, by thestatus of the p53 gene in their tumors. No differencein survival was seen between patients with and

1750 without detected p53 genetic abnormalities in theirtumors.

Survival AnalysisWhen survival was analyzed by log-rank analysis, no

prognostic significance was identified either in the groupas a whole or in patients with early- or late-stage diseasecomparing those with and without detected p53 genemutations (Fig 1 through 3). When the subgroup withpositive immunostaining was evaluated for survival, the

group with positive staining had a worse outcome withborderline statistical significance (Fig 4).

CHEST / 106/ 6 / DECEMBER, 1994/ Supplement 379S

100

90

80

70

60

50

40

30

20

10

0

U)

.E

1200 1400 1600

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

FIGURE 4. Survival of patients who have undergoneresection for NSCLC by the p53 overexpression intheir tumors. Paraffin sections were immunostainedfor p53 as described in the text. Only a marginallysignificant difference in survival was seen betweenpatients with and without detected p53 overexpres-sion in their tumors.

-a

cna)

EQ

10090

80706050403020100

0 250 500 750 1000Days Post-op

Multivariate AnalysisStepwise multivariate analysis was performed using sex,

age, histologic feature, TNM stage, performance status, rasmutations, p53 mutations, and immunohistochemistry asvariables. The only independently significant adverse pa-rameters were sex (male), advanced age (>60), T stage >1,and N stage >1 (Table 4). Neither p53 mutations nor im-munohistochemical positivity were independently predic-tive.

DISCUSSIONCancer is thought to arise after accumulation of a

threshold number of lesions in key regulatory molecules inpreviously normal cells, and these lesions are being in-creasingly well characterized. As we uncover the molecu-lar origins of cancer, the ultimate hope is that new thera-pies can be rationally designed based on this knowledge,but until that becomes reality, these molecular markers canstill be useful in helping to predict outcome and guide theuse of aggressive standard therapies. p53 is the most com-mon somatically acquired molecular lesion yet discoveredin lung cancer, and in fact is so common in small cell lungcancer (about 90% of cases2l) that it is unlikely to be ofprognostic utility. In NSCLC, about 50% of tumors containmutant p53, and neoadjuvant and adjuvant therapies areavailable that might be more rationally utilized in the tu-mors with the highest risk of recurrence. There is signif-icant controversy over the prognostic importance of ab-normalities in the p53 gene in resected NSCLC,2223and few authors have evaluated both molecular abnor-malities and protein overexpression in a cohort of patients

Table 4-Independent Prognostic Factors*

Variable p Value Risk Ratio

Male 0.0035 5.06Age 0.0483 1.04T stage 0.0004 2.97

*The independently prognostic variables derived from a stepwisemultivariate analysis are shown along with p value and the risk ra-tio associated with each. p53 mutations, p53 immunostaining posi-tivity, and the presence of ras mutations all were not independentlypredictive in this dataset.

with adequate staging and follow-up. This study reportsour preliminary data on the usefulness of genetic analysisand immunostaining of the p53 gene in a subset of patientsenrolled in LCSG 871.We find in this dataset that (1) molecularly detected

mutations in the p53 gene are not adverse prognostic fac-tors for survival, (2) that positive p53 immunostaining isweakly associated with poor survival in a single variableanalysis, but is not independently predictive, (3) that thereis only a 66% concordance between these two tests, and that(4) double mutants for ras and p53 are infrequent.The lack of strong independent prognostic significance

to abnormalities in p53 may be due to the universaloccurrence of abnormalities in a "p53 pathway," some-

times in p53, sometimes in other genes. This may under-score the importance of this pathway in the developmentof cancer, and a similar situation has been proposed for theretinoblastoma protein and a cell cycle kinase p16.24 Themodest prognostic impact of overexpression as detected byimmunostaining may reflect the action of the dominanttransforming function of the mutant p53 protein, but sev-

eral of the tumors showing positive immunostaining had noabnormalities detected in the open reading frame. Whileit is possible that these are false negatives due to screening

errors or mixtures of tumor and normal tissue, other fac-tors such as ionizing radiation25 are known to upregulatethe expression of normal p53. It is therefore possible thatin some of the cases with overexpression of p53, there is a

high level of ongoing DNA damage due to other cellularderangements that are reflected as an increased expressionof wild-type p53 protein.

Conversely, absence of p53 immunostaining does notimply that the p53 protein is normal either, as several in-stances were found of mutations in the absence of overex-

pression. Immunostaining positivity should not, therefore,be taken as equivalent to molecular abnormality in the p53protein.

These data do not imply that p53 is clinically irrelevantfor directing standard therapies for lung cancer, as clini-cal response to various chemotherapy agents or ionizing

radiation may indeed strongly depend on the functionalstatus of the p53 gene, and this awaits testing in futurestudies.

p53 Immunostaining & Reduced Survival (Carbone et al)

1250 1500 1750

380S

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

REFERENCES1 Chiba I, Takahashi T, Nau MM, et al. Mutations in the p53 gene

are frequent in primary, resected non-small cell lung cancer.Oncogene 1990; 5:1603-10

2 Parada LF, Land H, Weinberg RA, et al. Cooperation betweengene encoding p53 tumour antigen and ras in cellular trans-formation. Nature 1984; 312:649-51

3 Finlay C, Hinds P, Levine AJ. The p53 protooncogene can actas a suppressor of transformation. Cell 1989; 57:1083-93

4 Bodner SM, Minna J, Jensen SM, et al. Expression of mutant p53proteins in lung cancer correlates with the class of p53 genemutation. Oncogene 1992; 7:743-49

5 Harvey DM, Levine AJ. p53 alteration is a common event in thespontaneous immortalization of primary BALB/c murine em-bryo fibroblasts. Gene Dev 1991; 5:2375-85

6 Kastan MB, Zhan Q, El DW, et al. A mammalian cell cyclecheckpoint pathway utilizing p53 and GADD45 is defective inataxia-telangiectasia. Cell 1992; 71:587-97

7 Horio Y, Takahashi T, Kuroishi T, et al. Prognostic significanceof p53 mutations and 3p deletions in primary resected non-small cell lung cancer. Cancer Res 1993; 53:1-4

8 Quinlan DC, Davidson AG, Summers CL, et al. Accumulationof p53 protein correlates with a poor prognosis in human lungcancer. Cancer Res 1992; 52:4828-31

9 Ebina M, Steinberg SM, Mulshine JL, et al. Relationship of p53overexpression and up-regulation of proliferating cell nuclearantigen with the clinical course of non-small cell lung cancer.Cancer Res 1994; 54:2496-2503

10 McLaren R, Kuzu I, Dunnill M, et al. The relationship of p53immunostaining to survival in carcinoma of the lung. Br JCancer 1992; 66:735-38

11 Morkve 0, Halvorsen OJ, Skjaerven R, et al. Prognostic signif-icance of p53 protein expression and DNA ploidy in surgicallytreated non-small cell lung carcinomas. Anticancer Res 1993;13:571-78

12 Reissmann PT, Koga H, Takahashi R, et al. Inactivation of theretinoblastoma susceptibility gene in non-small-cell lung can-

cer: the Lung Cancer Study Group. Oncogene 1993; 8:1913-1913 Mitsudomi T, Steinberg SM, Nau MM, et al. p53 gene mutations

in non-small-cell lung cancer cell lines and their correlationwith the presence of ras mutations and clinical features.Oncogene 1992; 7:171-80

14 Mitsudomi T, Steinberg SM, Oie HK, et al. ras gene mutationsin non-small cell lung cancers are associated with shortenedsurvival irrespective of treatment intent. Cancer Res 1991;51:4999-5002

15 Bartek J, Iggo R, Lane D. Genetic and immunochemical anal-ysis of mutant p53 in human breast cancer cell lines. Oncogene1990; 5:893-99

16 Kaplan EL, Meier P. Nonparametric estimation from incom-plete observations. J Am Stat Assoc 1958; 53:457-81

17 Lee ET. Nonparametric estimation from incomplete observa-tions. Belmont, Calif: Lifetime Learning Publications, 1980

18 Cox DR. Regression models and life-tables. J R Stat Soc. 1972;34:187-220

19 Caron de Fromental C, Soussi T. TP53 tumor suppressor gene:a model for investigating human mutagenesis. Genes Chromo-somes Cancer 1992; 4:1-15

20 Rodenhuis S, Slebos RJ. Clinical significance of ras oncogeneactivation in human lung cancer. Cancer Res 1992; 52:2665-69

21 D'Amico D, Carbone D, Mitsudomi T, et al. High frequencyof somatically acquired p53 mutations in small cell lung can-cer cell lines and tumors. Oncogene 1992; 7:339-46

22 Passlick B, Izbicki JR, Riethmuller G, et al. p53 in non-small celllung cancer. J Natl Cancer Inst 1994; 86:801-03

23 Mitsudomi T, Oyama T, Kusano T, et al. Mutations of the p53gene as a predictor of poor prognosis in patients with non-small-cell lung cancer. J Natl Cancer Inst 1993; 85:2018-23

24 Kamb A, Gruis NA, Weaver-Feldhaus J, et al. A cell cycle reg-ulator potentially involved in genesis of many tumor types.Science 1994; 264:436-40

25 Kuerbitz SJ, Plunkett BS, Walsh WV, et al. Wild-type p53 is acell cycle checkpoint determinant following irradiation. ProcNatl Acad Sci USA 1992; 89:7491-95

CHEST / 106/6 / DECEMBER, 1994 / Supplement 381S

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from

DOI 10.1378/chest.106.6_Supplement.377S 1994;106; 377S-381SChest

Gazdar and John MinnaValerie Rusch, Jan A. Nowak, Donald McIntire, Dennis Slamon, Adi

David P. Carbone, Tetsuya Mitsudomi, Itsuo Chiba, Steven Piantadosi,Non-small Cell Lung Cancer : A Preliminary Report of LCSG 871and Is Imperfectly Correlated With Gene Mutations in Resected

p53 Immunostaining Positivity Is Associated With Reduced Survival

 April 17, 2012This information is current as of

 

http://chestjournal.chestpubs.org/content/106/6_Supplement/377SUpdated Information and services can be found at:

Updated Information & Services

http://chestjournal.chestpubs.org/content/106/6_Supplement/377S#related-urlsThis article has been cited by 15 HighWire-hosted articles:

Cited Bys

http://www.chestpubs.org/site/misc/reprints.xhtmlfound online at: Information about reproducing this article in parts (figures, tables) or in its entirety can bePermissions & Licensing

http://www.chestpubs.org/site/misc/reprints.xhtmlInformation about ordering reprints can be found online:

Reprints

"Services" link to the right of the online article.Receive free e-mail alerts when new articles cite this article. To sign up, select the

Citation Alerts

PowerPoint slide format. See any online figure for directions. articles can be downloaded for teaching purposes inCHESTFigures that appear in Images in PowerPoint format

 1994, by the American College of Chest Physicians by guest on April 17, 2012chestjournal.chestpubs.orgDownloaded from