bridging the gap between biologic, individual, and ... · biologic level (e.g., cellular biomarkers...

Editorial

Cancer Epidemiology in the 21st Century

Bridging the Gap between Biologic, Individual, andMacroenvironmental Factors in Cancer: A Multilevel Approach

Shannon M. Lynch and Timothy R. Rebbeck

AbstractTo address the complex nature of cancer occurrence and outcomes, approaches have been developed to

simultaneously assess the role of two or more etiologic agents within hierarchical levels including the:

(i) macroenvironment level (e.g., health care policy, neighborhood, or family structure); (ii) individual level

(e.g., behaviors, carcinogenic exposures, socioeconomic factors, and psychologic responses); and (iii)

biologic level (e.g., cellular biomarkers and inherited susceptibility variants). Prior multilevel approaches

tend to focus on social and environmental hypotheses, and are thus limited in their ability to integrate

biologic factors into a multilevel framework. This limited integration may be related to the limited

translation of research findings into the clinic. We propose a "Multi-level Biologic and Social Integrative

Construct" (MBASIC) to integrate macroenvironment and individual factors with biology. The goal of this

framework is to help researchers identify relationships among factors that may be involved in the

multifactorial, complex nature of cancer etiology, to aid in appropriate study design, to guide the

development of statistical or mechanistic models to study these relationships, and to position the results

of these studies for improved intervention, translation, and implementation. MBASIC allows researchers

from diverse fields to develop hypotheses of interest under a common conceptual framework, to guide

transdisciplinary collaborations, and to optimize the value of multilevel studies for clinical and public

health activities. Cancer Epidemiol Biomarkers Prev; 22(4); 485–95. �2013 AACR.

MotivationCancer is etiologically complex and its causes are mul-

tifactorial. Risk factors associated with cancer develop-ment have been identified that represent a variety oflevels of influence on health and disease (Table 1). Macro-environment factors including health system, neighbor-hood, or community characteristics, have increasinglybeen linked to cancer incidence and mortality (1, 2). Inaddition, social determinants and processes (1, 3, 4) havebeen identified as cancer risk factors, including socioeco-nomic status or self-reported race (5–8). Environmentalexposures at the level of the individual (5) includingcigarette smoking (9), radon (10), asbestos (11), diet

(12), and physical activity (13) are causally associatedwith some cancers. Applied and fundamental investiga-tions have identified a wide array of biologic factorsmechanistically involved in carcinogenesis includingthose of the tumor microenvironment, metabolome, pro-teome, transcriptome, and genome. For example, hun-dreds of novel genetic susceptibility loci have been iden-tified through candidate and genome-wide associationstudies (GWAS; ref. 14).

Studies of factors at a single level have provided a greatdeal of insight into the etiology of disease. Despite suc-cesses in identifying cancer risk factors, these approachesare limited and at some point the information obtainedfrom these single-level studies reach a saturation point,and have provided as much information as they can. It isclear that the factors reported to date do not fully explaincancer incidence in the general population. For example,while smoking is strongly associated with lung cancer(15),most smokerswill not bediagnosedwith lung cancer,whereas some nonsmokers will (16). While BRCA1 orBRCA2mutation carriers have a greatly increased lifetimerisk of developing breast cancer (17), some BRCA1/2mutation carriers are never diagnosed with breast orovarian cancer, even at an advanced age. GWAS haveidentified a wealth of susceptibility genes, but the iden-tification of novel genes using this approach is unlikely tocontinue ad infinitum. Therefore, risk factors studied in

Authors' Affiliation: Center for Clinical Epidemiology and Biostatistics,Department of Biostatistics and Epidemiology and Abramson CancerCenter, University of Pennsylvania, Philadelphia, Pennsylvania

Note from theEditor-in-Chief: This is one in a series of commentaries thathave arisen from an initiative of the National Cancer Institute to advanceepidemiological science in the 21st century.

Corresponding Author: Shannon M. Lynch, Department of Biostatisticsand Epidemiology, University of Pennsylvania School of Medicine, 243Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104. Phone: 215-898-5300; Fax: 215-573-1050; E-mail: [email protected]

doi: 10.1158/1055-9965.EPI-13-0010

�2013 American Association for Cancer Research.

CancerEpidemiology,

Biomarkers& Prevention

www.aacrjournals.org 485

on March 28, 2021. © 2013 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

Published OnlineFirst March 5, 2013; DOI: 10.1158/1055-9965.EPI-13-0010

http://cebp.aacrjournals.org/

isolation and identified by standard approaches areunable to fully explain the complex, multifactorial causesof cancer. For this reason, cancer research has evolvedfrom focusing on single factors to studies of complexrelationships between social, behavioral, molecular, andenvironmental factors.

Overview of Current Multilevel ApproachesTo address the complex nature of cancer etiology,

multilevel approaches have been developed to simulta-neously assess the role of 2 ormore etiologic agentswithina hierarchical or nested structure (18). A number of con-ceptual frameworks have been proposed that integrateinformation across levels of disease etiology, includingthe "web of disease" of MacMahon and Pugh (19), the"wheel" of Mausner and colleagues (20), "systems epide-miology" (21), and more recent models of multifactorialetiology (22–27). Multilevel approaches are generallycharacterized by 3 main levels: (i) macroenvironment,referred to elsewhere as "eco-level" (23, 24); (ii) individual;and (iii) biology (Table 1). Each of these levels is furthercharacterized by sublevels (Table 1) that define domainsof variables involved in cancer etiology or outcomes.Multilevel conceptual frameworks are based on the prem-ise that factors affecting disease act within and acrosslevels to collectively affect disease. These approachesgenerally hypothesized that cancer outcomes can resultfrom the complex relationship of factors at multiple levelsin at least 2 ways (Table 1). First, factors at the macro-environment and individual levels can directly affect thebiologic events and result in cancer. Second, factors mayconfer risk in a hierarchal fashion, such that biologic-leveleffects are affected by behaviors or exposures of theindividual, and individual level effects are affected bythe macroenvironment (28).

The relationships described earlier in the context of amultilevel model refer to both statistical and biologicinteractions. Here, we use the term "interaction" generi-cally to refer to any nonadditive statistical structure that

canbe constructedbetween2ormore factors. This conceptincludes that of effectmodification,mediation (29), aswellas biologic structures that may be defined between 2 ormore factors (e.g., epistasis among genetic loci). The goalof the multilevel framework wewill present here is not todefine a specific form for interaction. A variety of statis-tical approaches have been developed to guide imple-mentation of hierarchal, longitudinal, or multilevel mod-els (18, 30–32). Instead, we hope to provide a frameworkaround which a researcher can generate hypothesesabout the relationship among etiologic agents in a con-sistentmanner. When results of these hypothesis tests areknown, investigators using our proposed frameworkmaybe better able to compare and combine their results toform coherent multilevel inferences.

Most multilevel approaches lack a detailed focus onmechanisms that can be used to frame the relationshipsbetweenmacroenvironment or individual-level factors. Inpart, the limited incorporation of mechanistic hypothesesstems from the early multilevel frameworks havingevolved from research focused on social factors. Thus,multilevel conceptual approaches have tended to take a"top-down" approach that is focused on the role of socialdeterminants at the macroenvironment level (Table 2).Only more recently has a detailed consideration of thebiologic level been included in multilevel studies. For in-stance, themodel ofWarnecke and colleagues (22) centersonhealthdisparities as the outcomeof interest anddefinesmacroenvironment level factors by policies, institutions,and social or physical factors. They also include a singlelevel including biologic factors. Similarly, the models ofTaplin and colleagues (23) and Gorin and colleagues (24)focus on improved cancer care, subdividing the macro-environment level by national and state health policy,local community environment, organization or practicesetting,health careprovider teams,and family/social sup-port. Across proposed multilevel frameworks, the tradi-tional individual level risk factors for cancer (e.g., smok-ing, race, diet, etc.) are also considered, whereas biologic

Table 1. Hierarchical level definitions

Level Sublevel Factors at this level can serve as:

Macroenvironment * Health policy (national, state, local)* Community, neighborhood* Social and built environment* Practice setting and health care providers* Family and social support

* Exposures that affect individual risk factors* Exposures that affect biologic processes* Contextual variables (87)

Individual * Behaviors* Exposures* Psychologic determinants* Socioeconomic factors

* Exposures leading to disease* Intermediates between the macroenvironmentand disease

Biologic * Tissue* Cell* Somatic genome* Inherited genome

* Processes leading to disease* Intermediates and biomarkers reflectingthe relationship between macroenvironmentaland individual factors

Lynch and Rebbeck

Cancer Epidemiol Biomarkers Prev; 22(4) April 2013 Cancer Epidemiology, Biomarkers & Prevention486




Tab

le2.

Multilev

elfram

eworkex

amples

Leve

lHiattan

dBreen

(3)

Warne

ckean

dco

lleag

ues(22)

Tap

linan

dco

lleag

ues(23)

Gorinan

dco

lleag

ues(24)

Morrisse

yan

dco

lleag

ues(26)

Mac

roen

vironm

ent

Defi

nedbyfactors:

social

determinan

tsan

dhe

alth

care

system

s

Defi

nedby

subleve

ls(from

larges

tto

smallest):so

cialco

ndition

s(e.

g.,d

iscrim

ination),ins

titutions

(e.

g.,fam

ilies

),ne

ighb

orho

od,

social

relatio

nships

Defi

nedbysu

bleve

ls:n

ationa

l/statepolicy,

loca

lcom

mun

ity,

orga

niza

tionor

practicese

tting,

health

care

providers,

family/

social

support

Defi

nedbysu

bleve

lsfrom

Taplin

andco

lleag

ues(23)Gorin

and

colleag

ues(24)

Individua

lDefi

nedbyfactors:

social

determinan

ts,b

ehav

ioral/

psy

cholog

icfactors

Defi

nedby

factors:

age,

socioe

cono

mic

status

,ed

ucation,ob

esity

,tob

acco

use,

accu

lturatio

n,diet,rac

e

Defi

nedbyfactors:

biologic

factors,

sociod

emog

raphics

,insu

ranc

eco

verage

,riskstatus

,co

morbidities

,kno

wledge

,attitud

es,b

eliefs,d

ecision-

mak

ingpreferenc

es,

psych

olog

icreac

tion/co

ping

Sim

ilarto

Taplin

andco

lleag

ues

(23)

Gorin

andco

lleag

ues(24)

whe

reindividua

lisdes

cribed

asthepa

tient.

Biologic

Defi

nedbyfactors:

gene

san

dbiomarke

rsDefi

nedby

factors:

allostatic

load

(e.g.,co

mbinationof

stress

marke

rsor

othe

rbiom

arke

rs),

metab

olic

proce

sses

,gen

etic

mec

hanism

s

Defi

nedbysu

bleve

ls(larges

tto

smallest

leve

l):orga

n,tis

sue,

cell,

gene

,molec

ule,

atom

Prim

aryou

tcom

eof

interest

Can

cerco

ntrolc

ontin

uum

(predisea

se,p

reclinical,

inciden

ce,m

orbidity

/su

rvival,m

ortality)

interven

tions

(preve

ntion,

early

detec

tion,

diagn

osis/

trea

tmen

t,qua

lityof

life)

Can

cerhe

alth

disparities

Can

cerca

reco

ntinuu

m(risk

asse

ssmen

t,prim

arypreve

ntion,

detec

tion,

diag

nosis,

trea

tmen

t,su

rvivorsh

ip,e

ndof

life)

Can

cerca

reco

ntinuu

m

Multilevel Approach in Cancer

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 22(4) April 2013 487




factors in these constructs remain broadly defined bygenes, proteins, enzymes, and other somatic changes inthe cellular environment (23, 24, 33). In these models, allbiologic processes are treated in amanner similar to thoseof other levels without accounting for the extensiveknowledge of biologic processes, pathways, and etiologicrelationships that are involved in carcinogenesis.

Approaches that focus on macroenvironment have hadan impact on the conceptual advancement of our under-standing of disease etiology. While the NIH has increas-ingly recognized and encouraged the use of multilevelapproaches to go beyond investigating individual levelfactors to include macroenvironment level exposures (34,35), many of the current approaches (34, 35), come fromthe perspective of social and environmental research, andthe full integration of biologic level factors has yet to berealized. A search of PubMed for the term "cancer" and"multilevel analysis" or "multilevel model" resulted in 55articles published between 2002 and 2012, although themajority of these (26 of 55, 47%)were published since 2010.Most of these studies focused on individual-level andmacroenvironmental factors, and few incorporated biolog-ic factors.Thus,work isneeded to improve theunderstand-ing ofwhich factors at each level are relevant to thedisease,the hierarchical nature of the relationship of those factors,and the effective application of integrative multilevelapproaches to achieve meaningful etiologic inferences.

Multilevel Biologic and Social IntegrativeConstruct (MBASIC)

To the degree that a researcher has knowledge ofbiologic mechanisms of human cancer, multilevel models

could be used to harness this information and to generatehypotheses that link macroenvironment or individuallevel factors with mechanisms of carcinogenesis. Currentmultilevel conceptual approaches, while created to pro-mote multidisciplinary research, often lack detaileddescriptions of the biologic level that could be used tounite traditionally distinct fields (e.g., molecular biologyand social epidemiology).

MBASIC defines the multilevel framework (construct)to include 3 main hierarchal levels that contribute tocancer etiology and levels of carcinogenesis (i.e., macro-environment, individual, and biologic factors; Fig. 1),where the biologic level is more specifically defined. Thismultilevel etiologic model is then placed in the contextof interventions, and translation/implementation (i.e., T0-T4; refs. 36, 37; Fig. 1). This framework allows researchersfrom the fields of public health, health policy, prevention,behavioral sciences, sociology, epidemiology, biology,clinical medicine, and others to test hypotheses of interestunder a common conceptual framework, to address thedynamic nature of carcinogenesis, to facilitate translationof multilevel studies to clinical and public health strate-gies, and to support multidisciplinary collaborations.

The primary goal of MBASIC is to consistently andsystematically frame complex hypotheses about canceretiology. As may be expected with any comprehensiveconceptual framework, the full range of MBASIC compo-nents is not meant to be implemented in any one study.Instead, MBASIC is meant to aid the researcher in statinghypotheses for individual studies that address a part ofthe complete framework. Thus, MBASIC provides theframework for hypotheses that allow comparison andcompilation of individual study results using formal

Macroenvironment

Individual

Biologic

IG

SG

Progression

Ini�a�on

Metastasis

Promo�on

Diagnosis

Risk assessment

Treatment

Detec�on

Surv

ivor

ship

Health care systems change

Popula�on level changes in health status

Evidenced-based policy

Levels of carcinogenesis

Levels of interven�on

Levels of implementa�on/evalua�on

T0/T1

T2

T3

T4

EBE

Levels of e�ology

BED ASF

1o , 2o ,

3oPr

even

�on

Figure 1. Multilevel Biologic andSocial Integrative Construct. Thisframework includes levels ofetiology, carcinogenesis,intervention, and implementation/evaluation, as well as previouslydefined phases of translation (i.e.,T0-T4; refs. 36, 37). Biologic levelsinclude inherited genome (IG),somatic genome (SG), as well asrelated biomarkers of biologicallyeffective dose (BED), biomarkersof early biological effect (EBE), andbiomarkers of altered structureand function (ASF) (41).

Lynch and Rebbeck





(e.g., meta-analytic) or ad hoc means. Individual studiesbuilt around theMBASIC framework could also motivatemultidisciplinary collaborations and could rationalizesingle, large-scale multilevel studies in the future.

Predictive andmechanistic links between and amonghierarchical levels of etiologyAprimary goal of theMBASIC is to guide researchers to

consistently and systematically incorporate biologicmechanisms into a multilevel framework. Despite thesubstantial limitations in our ability to generate meaning-ful statistical or epidemiologic models of mechanism andbiologic events (38, 39), knowledge of existing biologicpathways emerging from animal, tumor, and other in vivostudies can be used to improve generation of hypothesesabout how each of the 3 hierarchal levels relates with theothers to frame questions about the complexity of canceretiology (40). The well-known molecular epidemiologyparadigm (41–45) provides a useful structure into whichbiology can be incorporated into a multilevel framework.As shown in Fig. 2 and defined later, the effect of expo-sures can be measured by biomarkers of biologicallyeffective doses (BED), early biologic effects (EBE), andaltered structure and function (ASF) that are predictive ofdisease (42–45). The formation of these biomarkers can beinfluenced by inherited genotypes. These factors can giverise to somatic genomic changes involved in carcinogen-esis. Note that while prior constructs include markers ofinternal dose, which have great value as biomarkers forresearch, clinical, or screening purposes,we exclude thesein the present framework to emphasize biologic andmechanistic effects in the multilevel etiology of cancer.While spontaneousmutation may give rise to the biomar-kers of disease and effect shown in Fig. 2, the multilevelconstruct assumes that each of the biomarkers occur in

response to an initial macroenvironment or individuallevel exposure, even though that exposure may not beknown or measurable.

We adapt the traditional molecular epidemiologyapproach (42–45) in 2 ways: by considering the nestedhierarchical nature of themultilevelmodel (Fig. 2); and byexpanding the definition of "exposure" to include bothmacroenvironment level and individual level exposures.As noted in Table 1, relevant etiologic factors can bemeasured by biomarkers (i.e., BED, EBE, and ASF) ofexposure ordisease at the biologic level. These biomarkersreflect somatic changes and are often measured at thetissue or cellular level. For example, biomarkers of expo-sure to cigarette smoking at the individual level can bemeasured by exposure biomarkers such as DNA adducts(42–45) in blood; prostate-specific antigen (PSA) levels orchromosomal instability (45)measured in blood can serveas markers of disease. Thus, these factors may be framedas both processes leading to disease and as intermediatesreflecting the relationship between macroenvironmentaland individual factors, separately, and disease (Table 1).For instance, macroenvironment level variables caninduce a psychologic response, which can be directlymeasured at the biologic level. Witnessing a crime in aneighborhood environment can lead to flight or fightcellular responses that cause increases in cortisol levels.Thus, cortisol is a biomarker of a macroenvironmentexposure.An example of a linkage between the individuallevel and the biologic level is that of the human exposome(46). The exposome is defined by environmental expo-sures (including lifestyle factors) that represent combinedexposures from all sources, from the prenatal periodonward (46). The exposome can be measured by biomar-kers at the cellular level via bodily fluids or tissue that canserve as surrogates for exogenous or endogenous envi-ronmental exposures. For instance, exposure to organo-phosphate pesticides can be measured by certain meta-bolites, and dietary factors, such as vitamin intake, can bemeasured by antioxidant metabolites. Like the GWASapproach, epidemiology has used environment-wideassociation studies (EWAS; refs. 46, 47) that use an agnos-tic approach to identifying environmental factorsinvolved in disease. Future EWAS studies in cancer arewarranted to provide practical evidence for a linkbetween individual level exposures and the biologic level.While EWAS and GWAS share some conceptual similar-ities, there are numerous methodologic differencesbetween the 2 approaches (48). However, the results ofeach can provide information that may promote thedevelopment of multilevel hypotheses in cancer etiology.

While the examples earlier show how macroenviron-ment and the individual level factors can each separatelyaffect the biologic level as an exposure, we can also showthe hierarchal effect among exposures at multiple levelson the biologic level. For example, exposure to a group offriendswho smoke cigarettes could prompt an individualto change her behavior and also start to smoke cigarettes.This change in behavior at the individual level influences

SG

IG

EBE

BED

ASF

DNA adducts, point muta�on

Biomarker:

Examples: Transcriptome changes

Chromosomal damage

Exposure (Macroenvironment or individual level)

Tissue and cell (microenvironment, metabolome, proteome)

e ent

Figure 2. Incorporating molecular epidemiology and biomarkers in themultilevel framework. Biologic levels include inherited genome (IG),somatic genome (SG), as well as related biomarkers of biologicallyeffective dose (BED), biomarkers of early biological effect (EBE), andbiomarkers of altered structure and function (ASF) (41).






molecular carcinogenesis at the biologic level (i.e., DNAadducts; BED) and chromosomal damage (ASF). Despitesymptoms of decreased lung function over the course of15 to 20 years, the individual is genetically predisposed tonicotine dependence, is unable to quit smoking, andultimately ends up developing lung cancer. Here, thebehavior change served as an intermediate between themacroenvironment and biologic events involved in car-cinogenesis. Thus, this example shows the biologic plau-sibility of howamacroenvironmental factor can impact anindividual, affecting her biologic environment, ultimatelyresulting in disease. When the macroenvironment, indi-vidual, and biologic factors are collectively considered topredict or explain a cancer outcome, statistical methodswill be needed to determine which levels or which riskfactors within each level are most relevant to the canceroutcome under study. Thus, it is possible for intermedi-ates to serve as surrogates of exposure anddisease, but theimportance of each level and each factor within each levelwill need to be determined statistically based on availablemethods.

Biology in a Multilevel FrameworkStarting with the levels of etiology (Fig. 1), the biologic

level can be subdivided into sublevels with a hierarchicalorder based on our knowledge of biology and carcino-genesis: tissues are composed of cells, which containgenes. Somatic mutations and cellular events (e.g., DNAreplication) may be involved in carcinogenesis. In thefollowing sections,webuild the framework aroundwhichthe biologic level can be optimally incorporated intomultilevel analysis (Fig. 2).

TissuesTissues warrant consideration as a unique biologic sub-

level in amultilevel framework for 2 reasons. First, cellularmarkers and processes that aremeasured in normal tissue,preneoplasia, ormalignant tumors could serve as potentialmarkers of exposure, disease, orprognosis. Second, tumorsoccur at the tissue level. Most cancers are diagnosed andstaged using tissue samples or by imaging techniques thatmay identify lesions in a particular organ. A growing areaof research is focused on the tumor microenvironment,definedbynormal cells, signalingmolecules,matrices, andblood vessels that surround and feed a tumor cell (49). Atumor can alter its microenvironment (as defined by cel-lular and genomic sublevel factors), and the microenvi-ronment can affect how a tumor grows and spreads. Dataabout the role of the tumor microenvironment are rapidlybecoming available via initiatives such as The CancerGenome Atlas (TCGA; http://cancergenome.nih.gov).

CellsThe cellular sublevel is characterized by proteins,

enzymes, and other biomarkers that can be detected inbodily fluids and tissues. In the context of our model, thecellular level includes the transcriptome, proteome, and

metabolome, where biomarkers of exposure and diseasecan be measured (Fig. 2). The transcriptome includes thevarious forms of RNA in the cell that affect gene expres-sion and cellular function (50–52). The proteome includesthe total set of proteins expressed in a given cell at a giventime (51). Examples of factors measured in the proteomeinclude PSA and CA-125 (45, 50, 53). Complex proteininteractions are referred to as the metabolome (51, 54).Therefore, even within the cellular sublevel, there is anemerging hierarchy (51). Many approaches for diseasebiomarker discovery focus on a single biomarker at thecellular level, despite an emerging expectation that panelsof biomarker analytes will be needed to provide sufficientsensitivity and specificity for cancer screening, diagnosis,or prognosis (50, 54). Therefore, there is a shifting focus tothe role of the pathway-based and statistical interactionsamong cellular factors, but progress in this area is limitedby available, high-throughput technologies that candetect and organize themillions of proteins obtained froma given biologic sample.

Somatic genomicsThe somatic genome sublevel (Figs. 1 and 2) is defined

by acquired somatic genomic changes over the course of aperson’s lifetime. The somatic genome level is defined byfactors that can be bothmarkers of disease andmarkers ofexposure. Somatic genome examples include mutations,copy number variants, and epigenetic changes occurringin DNA (50, 55). Early studies of somatic genome usedmethods that identify potential susceptibly loci a priori,but this approachused a small number of geneticmarkers,rarely identified robust associations between candidategenes and cancer, andmost findingswere not replicable inother studies (14).

Inherited genomicsThe inherited genome sublevel (Figs. 1 and 2) is com-

posed of inherited susceptibility loci that serve asmarkersof disease risk and outcome. Inherited genome includeshereditary cancer syndromes (56), which confer a highrisk of cancer development. Inherited genome researchmay use family-based linkagemethods to identify impor-tant inherited, high-penetrance genes, such asBRCA1 andBRCA2. However, themutations in these genes are rare inthe general population (4, 14, 17), and only explain a smallfraction of familial aggregation and cancer risk. GWAShave identified many dozens of cancer susceptibility loci(43, 57), most of which were not previously hypothesizedto be involved in cancer susceptibility (14). Despite thissuccess, genetic risk variants identified fromGWAS, aloneand in combination, explain a relativelyminor proportionof disease risk, and have had limited translational value tothe clinic. This has led to a focus on the identification ofrare variants that may account for larger proportions ofcancer genetic risk (58).

Given the limited clinical use of somatic genome andinherited genome findings focused on single disease lociand statistical interactions thereof, there has been a

Lynch and Rebbeck





renewed interest in studyingepistasis, definedasgenes at 2or more loci that produce phenotype effects that are dif-ferent than theexpectedeffectsof the individual loci (59).Atboth the somatic genome and inherited genome sublevels,gene–gene interaction studies arebeingconducted toascer-tain the independent and joint effects of risk loci on canceroutcomes (60). These studies may use multiple cancer risksusceptibility loci based on pathway or shared biologicfunction,orbecombinedusingstatisticalpredictivemodelsindependent of biologic knowledge.

Nonhierarchical effects within and across levels ofetiologyMechanisms and example methodologies have been

proposed to build on the definition of the biologic leveland to illustrate how interactions between and amongfactors at each level relates to one another, assuming ahierarchal structure for levels of etiology (Fig. 1). Hypoth-eses that consider the hierarchical framework of MBASICare readily constructed from the discussion providedearlier.However, the effects of factorswithin each of theselevels need not follow a strict hierarchy. In the context ofpredictive (as opposed to mechanistic) models, each levelcan dynamically affect another. Thus, statistical (causal)inferences need not be constrained in a linear hierarchicalfashion (61). Concepts in social science and genetics sup-port this assertion. According to the social ecologic per-spective (62, 63), human health results from the complexinteraction of personal factors (e.g., behaviors, biology,psychology, etc.) as well as physical and social environ-ments (e.g., geography, built environment, culture, eco-nomics, politics, and social relationships; ref. 62). Forinstance, a combination of geography, psychology, andbehavior without a clear hierarchal or biologic link couldinteract (statistically) and affect disease outcomes. Inaddition, changes in eating habits at the individual levelmay affect social relationships at the macroenvironmentlevel as a person who is more conscious of their eatinghabits may prefer to be around other healthy eaters; theeffect of each level on the other is not necessarily linear,top-down, or bottom-up. In the field of genetics, pene-trance (64) is defined by the probability of a phenotypegiven genotype. Even though a person is born with adisease genotype, lack of exposure to harmful environ-mental factors or carcinogens may prevent the diseasefromoccurring.While it is likely that the disease genotypeand exposure have some biologic link, in the absenceof this knowledge, specific methodologies aimed at ana-lyzing gene–environment interactions, more recently,gene–environment interaction-wide association studies(GEIWAS; ref. 65), can be developed to help elucidatestatistical interactions across levels. Because it is clear thatbiological, social, and environmental factors interact insome way in cancer etiology, a multilevel framework isneeded to both organize and guide traditionally separatefields of cancer research; however, these frameworksshould also account for the dynamic nature of the disease.

Expanding MBASIC: Levels of Intervention,Implementation, and Evaluation

MBASIC expands the use of themultilevel approach byincluding levels of etiology and carcinogenesiswith levelsof intervention and implementation/evaluation, all ofwhich can influence one another in a nonlinear manner.Levels of intervention are characterized by primary, sec-ondary, and tertiary prevention strategies and survivor-ship that range from risk assessment to detection todiagnosis and treatment (Fig. 1). Previous multilevelstudies have focused on assessing factorswithin the levelsof intervention (3), particularly cancer care outcomes suchas detection or screening at the individual level or practicesetting sublevel (23, 24).

The implementation/evaluation level is characterizedby changes made through the application and transla-tion of relevant interventions (66). Implementation/evaluation may occur through national, state, or localpolicy or health care systems changes, and the impact ofinterventions and implementation will ultimately beseen in changes to the health status of a population.The levels of implementation/evaluation are based onthe translational model of Khoury and colleagues (36,37), which describes 5 translational phases (Fig. 1): T0/T1 (determination of mechanisms, etiology and devel-opment of interventional strategies); T2 (developmentof evidence-based policy and practice); T3 (implement-ing evidence-based guidelines to elicit health care sys-tem changes); and T4 (surveillance and monitoring theeffect of changes on health outcomes in populations).Appropriate consideration of the dissemination, imple-mentation, and evaluation of research findings intohealth systems is critical if the potential of multilevelmodels is to be realized. For instance, knowledge of therole of macroenvironmental factors (e.g., residentiallocation, social environment) in individuals with spe-cific biologic characteristics and risk factor profile couldprovide a resource-efficient approach to early detectionor screening for cancer.

Simultaneous consideration of multiple levels in theMBASIC framework may impact a number of canceroutcomes. The levels or sublevels of inference (etiology,carcinogenesis, intervention, or implementation/evaluation; Fig. 1) could serve as the outcome or exposureof interest. For instance, health care system changes (e.g.,insurance coverage) can affect individual level behavior(e.g., participation in smoking cessation programs),whichcan affect the cellular environment (e.g., carcinogen levelsand formation of DNA adducts). Therefore, interactionswithin and across levels can be modeled in a variety ofways, and extent to which the 4 levels of inference impactthe trait of interest will vary depending on the etiologicsetting. For instance, during cancer initiation, living in acommunity that promotes cancer screening and havingaccess to primary caremayplay aprominent role in cancerearly detection. After cancer is diagnosed, the oncologyprovider and social support may become a predominant






influence on clinical and psychosocial outcomes. Both ofthese scenarios may be imposed on a common biologiccontext (e.g., a cancer having a specificmechanistic cause),but the relevant individual and macrolevel factors maydiffer substantially.

MBASIC Example: Prostate CancerIn the United States, prostate cancer is the second

leading cause of cancer-related death inmen (67). Prostatecancer is of public health concern because it dispropor-tionately affectsdifferent races.AfricanAmericanmenaremore likely to be diagnosed with and die from prostatecancer thananyother racial group, and thisdisparity is thelargest observed for any cancer (68). Despite the burden ofprostate cancer, particularly for African American men,little is known about the etiology and predictors of poorprognosis for the disease. At present, the only widelyagreed-upon risk factors for prostate cancer are at theindividual level: race, age, and family history of prostatecancer (69). Tumor and patient characteristics used toidentify men with a poor prognosis include tumor stage,Gleason score or grade, and PSA level at diagnosis. How-ever, these clinical characteristics are imperfect in theirability to determine long-term prognosis and appropriatetreatment options. Thus, prostate cancer is a good exam-ple of the potential value of the MBASIC framework.

PSA screening: from the cellular level to T4implementation

In the 1980s, studies on the cellular level showed thatPSA levels could serve as markers for prostate cancerrecurrence (70). The use of PSA screening for patientsundergoing treatment was approved in 1986 (70–72).Despite studies in the late 1980s suggesting that PSAmight not be an ideal biomarker for screening and earlydetection of prostate cancer (70–72), the U.S. Food andDrug Administration (FDA) also approved PSA as anearly detection screening test in 1994, and PSA becameone of the first FDA-approved early-detection biomarkersfor cancer (70). The FDA based its approval on a largeclinic study whose results suggested that men with PSAvalues above 4.0mg/mLcould be biopsied for cancer (73).As a result of this bench to beside clinical translation (T1phase), screening guidelineswith often conflicting recom-mendations from different organizations such as theUnited States Preventive Services Task Force (74), theAmerican Cancer Society (75), and the National Compre-hensive Cancer Network (76), started to emerge. Theseguidelines affected clinical practice (T2 phase), resultingin more men being screened for and diagnosed withprostate cancer (70). The health care system was alsoaffected by these guidelines: insurance companies, par-ticularly the VeteransAssociation andMedicare, incurredlarge costs covering routine PSA screening (T3 phase;ref. 77). Continued research at both the population level(T4 phase) and levels of causation (cellular and individuallevels) in more recent years have shown that PSA screen-ing may not improve prostate cancer mortality rates, that

early detection of prostate cancer can often lead to unnec-essary treatment for some and insufficient treatment inothers (70, 78, 79). As a result, researchers continue todevelop enhanced PSA screening tests that are moresensitive and specific (70). Guidelines for routine PSAscreening are continually being revised in the context ofindividual level factors. These include questioning theuse of screening for men under the age of 75 years (74),focusing on screening high-risk groups (5), and recom-mending baseline PSA measures in men under age 50years (58). Despite its limitations and controversies,PSA screening illustrates how cellular and individuallevels of causation, resulting biomarker interventions,and health care implementation (Fig. 1) can inform oneanother to optimize the early detection of prostatecancer.

The PSA scenario also suggests that a comprehensiveevaluation of PSA in early detection of prostate cancermay benefit from the use ofMBASIC to frame the hypoth-eses and approaches needed to improve screening andtreatment of prostate cancer. While macrolevel factorshave yet to bewidely used in the context of PSA screening,it is not hard to imagine that screening strategies may beoptimized by having a better understanding of those menwho are most likely to have unfavorable prostate canceroutcomes based on their socioeconomic situation, accessto health care, or other macroenvironmental factors. Therole of macroenvironmental factors in prostate cancer riskand mortality are beginning to emerge from the healthdisparity and PSA screening literature. Screening beha-viors can be affected by economic, physical, and socialcharacteristics of residential neighborhoods (80). Neigh-borhoods considered to be disadvantaged or low-incomehave been correlated with higher levels of pollutants,overcrowding, violence, less social cohesion, and lessaccess to services (81). Screening practices can affectprostate cancer incidence, and low-income neighbor-hoods often have fewer medical facilities that are over-burdenedwith indigent care to provide optimal screening(80). This can lead to differential screening practices byneighborhood (82) and differences in both the diagnosisand treatment of prostate cancer, particularly amongCaucasian versus African American men (83, 84). There-fore, neighborhood measures could serve as a surrogatefor access to care in prostate cancer and seem to be arelevant macroenvironment level measure to investigatefor this cancer outcome. In the setting of an MBASICapproach, men with known biologic risk profiles maytherefore benefit from targeted intervention if they alsoreside in defined disadvantaged neighborhoods.

This concept is further illustrated by a multilevel anal-ysis that investigated the role of individual level charac-teristics and census-tract neighborhood variables andstage of prostate cancer. Consistent with data showingan association between race, stage, and socioeconomiccircumstances such as living in a low income area, usinggeographic information systems technology, Xiao andcolleagues (85) went beyond identifying factors

Lynch and Rebbeck





associated with prostate cancer stage and suggest com-munity education and outreach in areas with unfavorableneighborhood characteristics. In the context of MBASIC,discovery and early translation can be leveraged in asingle study and can provide additional insights thatwould not be as readily apparent in studies focusing ona single etiologic level.

Prostate cancer disparities: piecing together studieson biology and neighborhoodBecause of the complex etiology of prostate cancer, an

understanding of prostate cancer disparities may benefitfrom a multilevel approach. A growing body of literaturesupports this hypothesis. Rundle and colleagues (86)reported that neighborhood socioeconomic status (basedon median income level of a census tract) modifies theassociation between individual smoking status andPAH–DNA adduct levels in prostate tissue (BED). We reportedan interaction between prostate cancer genetic suscepti-bility loci identified in GWAS and census-tract levelneighborhood variables on time to PSA failure in menwho had undergone radical prostatectomy (87). We iden-tified no main effects of the genetic variants or neighbor-hood factors on PSA failure by themselves but foundstatistically significant interactions between neighbor-hood variables and the susceptibility loci. Specifically,genotypes at MSMB and HNF1B/TCF2 predicted time toPSA failure in men from disadvantaged neighborhoods.This suggests that context-specific effects of genotypeshould be explored and may improve the ability to iden-tify groups that may experience poor prostate canceroutcomes. It is important to note that these studies rep-resent predictive models that may have implications forimplementation or translation but themselves do notprovide direct mechanistic conclusions. In general, thesestudies may motivate a continued focus on multilevelapproaches and provide rationale for the use ofmultilevelmodels in cancers such as prostate cancer, where typicalsingle disciplinary approaches provide limited insightinto disease etiology.

Charge to the Scientific CommunityWehave proposed a unifying conceptual framework that

allows researchers from public health, policy, oncology,health services research, behavioral science, epidemiology,and the biomedical sciences to test hypotheses of interestunder a common framework. The MBASIC frameworkallows researchers to generate common inferences fromotherwise disparate individual research findings by usinga common conceptual model. As illustrated by the prostatecancer example, taking a multilevel approach can help toexpedite translation of etiologic findings into translationalefforts, more than would occur in studies focused on singlelevels of etiology alone. By providing a stronger basis forinclusion of biologic factors in amultilevel hierarchy,MBA-SIC bridges the gap between social science and biology tofoster multidisciplinary collaboration and streamline inter-vention, implementation, and translation efforts. Emergingbiomedical technologies enable population-based studies toinclude biomarker data such that the landscape of cancerresearch is changing and the lines between disciplines areincreasingly blurring. MBASIC can serve as a roadmap forhypothesis generation and the development of emergingmultidisciplinary teams. The MBASIC framework allowsindividual studies to more effectively piece together indi-vidual research findings under a common conceptualmod-el. Knowledge gained from this integration can be used torationalize the costsof future, large-scale,multilevel studies.Finally, MBASIC represents a framework around whichtransdisciplinary research (i.e., research that generates newfields of inquiry) can be built.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Grant SupportThis research was supported by grants from the Public Health Service

(P50-CA105641, P60-NM006900, and R01-CA85074 to T.R. Rebbeck andF31-AG039986 to S.M. Lynch).

Received January 3, 2013; revised February 20, 2013; accepted February22, 2013; published OnlineFirst March 5, 2013.

References1. Krieger N. Defining and investigating social disparities in cancer:

critical issues. Cancer Causes Control 2005;16:5–14.2. Krieger N, Quesenberry C, Peng T, Horn-Ross P, Stewart S, Brown S,

et al. Social class, race/ethnicity, and incidence of breast, cervix,colon, lung, and prostate cancer among Asian, black, Hispanic, andwhite residents of the San Francisco Bay Area, 1988–92 (UnitedStates). Cancer Causes Control 1999;10:525–37.

3. Hiatt RA, Breen N. The social determinants of cancer: a challenge fortransdisciplinary science. Am J Prev Med 2008;35:S141–50.

4. Holmes JH, LehmanA, Hade E, Ferketich AK, Gehlert S, Rauscher GH,et al. Challenges for multilevel health disparities research in a trans-disciplinary environment. Am J Prev Med 2008;35:S182–92.

5. National Cancer Institute and National Institutes of EnvironmentalHealth Sciences. Cancer and the Environment. National Institutes ofHealth. Bethesda, MD; 2003.

6. Aarts MJ, Lemmens VEPP, Louwman MWJ, Kunst AE, CoeberghJWW. Socioeconomic status and changing inequalities in colorectal

cancer? A reviewof the associationswith risk, treatment and outcome.Eur J Cancer 2010;46:2681–95.

7. Bradley CJ, Given CW, Roberts C. Race, socioeconomic status,and breast cancer treatment and survival. J Natl Cancer Inst 2002;94:490–6.

8. Dayal HH, Polissar L, Dahlberg S. Race, socioeconomic status, andother prognostic factors for survival from prostate cancer. J NatlCancer Inst 1985;74:1001–6.

9. The United States Department of Health and Human Services. Thehealth consequences of smoking: A Report of the Surgeon General.Washington, D.C., 2004.

10. Lubin JH. Environmental Factors inCancer: Radon.RevEnvironHealth2010;25:33–38.

11. Orenstein MR, Schenker MB. Environmental asbestos exposure andmesothelioma. Curr Opin Pulm Med 2000;6:371–7.

12. Gibson TM, Ferrucci LM, Tangrea JA, Schatzkin A. Epidemiologicaland clinical studies of nutrition. Semin Oncol 2010;37:282–96.






13. Na H-K, Oliynyk S. Effects of physical activity on cancer prevention.Ann N Y Acad Sci 2011;1229:176–83.

14. Varghese JS, Easton DF. Genome-wide association studies in com-mon cancers—what have we learnt? Curr Opin Genet Dev 2010;20:201–9.

15. HechtSS. Tobacco smokecarcinogensand lung cancer. JNatlCancerInst 1999;91:1194–210.

16. Thun MJ, Henley SJ, Burns D, Jemal A, Shanks TG, Calle EE. Lungcancer death rates in lifelong nonsmokers. J Natl Cancer Inst 2006;98:691–9.

17. Rebbeck TR, Kauff ND, Domchek SM. Meta-analysis of risk reductionestimates associated with risk-reducing salpingo-oophorectomy inBRCA1 or BRCA2mutation carriers. JNatl Cancer Inst 2009;101:80–7.

18. Plewis I. Statistics in education. London: Edward Arnold; 1997.19. MacMahon B, Pugh TF, editors. Epidemiology: principles and

methods. Boston, MA: Little, Brown and Company, 1970.20. Mausner J, Kramer S, Bahn. Epidemiology: An Introductory Texts,

Saunders, Philadelphia, PA; 1985.21. Galea S, Riddle M, Kaplan GA. Causal thinking and complex system

approaches in epidemiology. Int J Epidemiol 2010;39:97–106.22. Warnecke RB, Oh A, Breen N, Gehlert S, Paskett E, Tucker KL, et al.

Approaching health disparities from a population perspective: theNational Institutes of Health Centers for Population Health and HealthDisparities. Am J Public Health 2008;98:1608–15.

23. Taplin SH, Anhang Price R, Edwards HM, Foster MK, Breslau ES,Chollette V, et al. Introduction: understanding and influencing multi-level factors across the cancer care continuum. J Natl Cancer InstMonogr 2012;2012:2–10.

24. Gorin SS, Badr H, Krebs P, Das IP. Multilevel interventions and racial/ethnic health disparities. J Natl Cancer Inst Monogr 2012;2012:100–11.

25. Glass TA, McAtee MJ. Behavioral science at the crossroads in publichealth: extending horizons, envisioning the future. Soc Sci Med2006;62:1650–71.

26. Caporaso NE. Integrative study designs—next step in the evolution ofmolecular epidemiology? Cancer Epidemiol Biomarkers Prev 2007;16:365–6.

27. SpitzMR,CaporasoNE, Sellers TA. Integrative cancer epidemiology—the next generation. Cancer Discov 2012;2:1087–90.

28. Rappaport SM. Implications of the exposome for exposure science.J Expos Sci Environ Epidemiol 2011;21:5–9.

29. Gelfand LA, Mensinger JL, Tenhave T. Mediation analysis: a retro-spective snapshot of practice and more recent directions. J GenPsychol 2009;136:153–76.

30. Hox JJ. Applied multilevel analysis. Amsterdam, the Netherlands: TT-Publikaties; 1994.

31. Kreft I, De Leeuw J. Introducing multilevel modelling. London: Sage;1998.

32. Raudenbush SW, Willms JDE. Schools, classrooms and pupils. SanDiego, CA: Academic Press; 1991.

33. Morrissey JP, Lich KH, Price RA,Mandelblatt J. Computational model-ing and multilevel cancer control interventions. J Natl Cancer InstMonogr 2012;2012:56–66.

34. Office of Behavioral and Social Sciences Research, National Institutesof Health (NIH). Toward higher levels of analysis: progress and promisein research on social and cultural dimensions of health. NIH: Bethesda,MD; 2000.

35. Lam TK, Spitz M, Schully SD, Khoury MJ. Drivers of translationalcancer epidemiology in the 21st century: needs and opportunities.Cancer Epidemiol Biomarkers Prev 2013;22:181–8.

36. Khoury MJ, Gwinn M, Yoon PW, Dowling N, Moore CA, Bradley L. Thecontinuum of translation research in genomic medicine: how can weaccelerate the appropriate integration of human genome discoveriesinto health care and disease prevention? Genet Med 2007;9:665–74.

37. Khoury MJ, Gwinn M, Bowen MS, Dotson WD. Beyond base pairs tobedside: a population perspective on how genomics can improvehealth. Am J Public Health 2011;102:34–7.

38. Siemiatycki J, Thomas DC. Biological models and statistical interac-tions: an example from multistage carcinogenesis. Int J Epidemiol1981;10:383–7.

39. Thompson WD. Effect modification and the limits of biological infer-ence from epidemiologic data. J Clin Epidemiol 1991;44:221–32.

40. Cortessis V, ThomasDC. Toxicokinetic genetics: an approach togene-environment and gene-gene interactions in complex metabolic path-ways. IARC Sci Publ 2004:127–50.

41. Perera FP, Schulte, PA. Chapter 1: A Conceptual and HistoricalFramework for Molecular Epidemiology. Molecular epidemiology:principles and practices. In: Schulte PA, Perera FP, editors. SanDiego,CA: Academic Press; 1993.

42. Perera F, Weinstein IB. Molecular epidemiology and carcinogen-DNAadduct detection: new approaches to studies of human cancer cau-sation. J Chronic Dis 1982;35:581–600.

43. Guy M, Kote-Jarai Z, Giles GG, Al Olama AA, Jugurnauth SK, Mulhol-land S, et al. Identification of new genetic risk variants for prostatecancer. Asian J Androl 2009;11:49–55.

44. Perera F. Molecular cancer epidemiology: a new tool in cancer pre-vention. J Natl Cancer Inst 1987;78:887–98.

45. Schulte PA, Rothman N, Schottenfeld D. Chapter 6: Design Consid-erations in Molecular Epidemiology. Molecular epidemiology: princi-ples and practices. In: Schulte PA, Perera FP, editors. San Diego, CA:Academic Press; 1993.

46. Patel CJ, Bhattacharya J, Butte AJ. An environment-wide associationstudy (EWAS) on type 2 diabetes mellitus. PLoS ONE 2010;5:e10746.

47. ThomasDC, Siemiatycki J, DewarR, Robins J,GoldbergM,ArmstrongBG. The problem of multiple inference in studies designed to generatehypotheses. Am J Epidemiol 1985;122:1080–95.

48. Ioannidis JP, Loy EY, Poulton R, Chia KS. Researching genetic versusnongenetic determinants of disease: a comparison and proposedunification. Sci Transl Med 2009;1:7ps8.

49. Swartz MA, Iida N, Roberts EW, Sangaletti S, Wong MH, Yull FE, et al.Tumor microenvironment complexity: emerging roles in cancer ther-apy. Cancer Res 2012;72:2473–80.

50. Tainsky MA. Genomic and proteomic biomarkers for cancer: a mul-titude of opportunities. Biochim Biophys Acta 2009;1796:176–93.

51. Dove A. Proteomics: translating genomics into products. Nat Biotech-nol 1999;17:233–6.

52. Meltzer PS. Cancer genomics: small RNAs with big impacts. Nature2005;435:745–6.

53. Andriole GL, Crawford ED, Grubb RL, Buys SS, Chia D, Church TR,et al. Prostate cancer screening in the randomized prostate, lung,colorectal, and ovarian cancer screening trial: mortality results after 13years of follow-up. J Natl Cancer Inst 2012;104:125–32.

54. Spratlin JL, Serkova NJ, Eckhardt SG. Clinical applications of meta-bolomics in oncology: a review. Clin Cancer Res 2009;15:431–40.

55. McDermott U, Downing JR, Stratton MR. Genomics and the contin-uum of cancer care. N Engl J Med 2011;364:340–50.

56. FoulkesWD. Inherited susceptibility to commoncancers. NEngl JMed2008;359:2143–53.

57. Manolio TA. Genome-wide association studies and disease riskassessment. N Engl J Med 2010;363:166–76.

58. ChenG,Wei P, DeStefanoAL. Incorporating biological information intoassociation studies of sequencing data. Genet Epidemiol 2011;35:S29–S34.

59. Cordell HJ. Epistasis: what it means, what it doesn't mean, andstatistical methods to detect it in humans. Hum Mol Genet 2002;11:2463–8.

60. Fehringer G, Liu G, Briollais L, Brennan P, Amos CI, Spitz MR, et al.Comparisonof pathwayanalysis approaches using lung cancerGWASdata sets. PLoS ONE 2012;7:e31816.

61. Weiner BJ, Lewis MA, Clauser SB, Stitzenberg KB. In search ofsynergy: strategies for combining interventions at multiple levels.J Natl Cancer Inst Monogr 2012;2012:34–41.

62. Stokols D. Establishing andmaintaining healthy environments: towarda social ecology of health promotion. Am Psychol 1992;47:6–22.

63. Stokols D. Translating social ecological theory into guidelines forcommunity health promotion. Am J Health Promot 1996;10:282–98.

64. Lobo I. Same genetic mutation, different genetic disease phenotype.Nat Educ 2008;1(1). Available from: http://www.nature.com/scitable/topicpage/same-genetic-mutation-different-genetic-disease-phe-notype-938 [accessed November 30, 2012].

Lynch and Rebbeck





65. Cornelis MC, Tchetgen EJ, Liang L, Qi L, Chatterjee N, Hu FB, et al.Gene–environment interactions in genome-wideassociation studies: acomparative study of tests applied to empirical studies of type 2diabetes. Am J Epidemiol 2012;175:191–202.

66. Yano EM, Green LW, Glanz K, Ayanian JZ, Mittman BS, Chollette V,et al. Implementation and spread of interventions into the multilevelcontext of routine practice and policy: implications for the cancer carecontinuum. J Natl Cancer Inst Monogr 2012;2012:86–99.

67. American Cancer Society. Cancer facts and figures 2010 tables andfigures; 2010. http://www.cancer.org/research/cancerfactsfigures/cancerfactsfigures/most-requested-tables-figures-2010. [cited 2011;retrieved 2011 Apr 19].

68. Lansdorp PM. Stress, social rank, and leukocyte telomere length.Aging Cell 2006;5:583–4.

69. CrawfordED. Epidemiology of prostate cancer. Urology 2003;62:3–12.70. De Angelis G, Rittenhouse HG, Mikolajczyk SD, Blair Shamel L,

Semjonow A. Twenty years of PSA: from prostate antigen to tumormarker. Rev Urol 2007;9:113–23.

71. Stamey TA, Yang N, Hay AR, McNeal JE, Freiha FS, Redwine E.Prostate-specific antigen as a serum marker for adenocarcinoma ofthe prostate. N Engl J Med 1987;317:909–16.

72. Catalona WJ, Smith DS, Ratliff TL, Dodds KM, Coplen DE, Yuan JJ,et al. Measurement of prostate-specific antigen in serum as a screen-ing test for prostate cancer. N Engl J Med 1991;324:1156–61.

73. Catalona W, Hudson MA, Scardino PT, Richie JP, Ahmann FR, Fla-nigan RC, et al. Selection of optimal prostate specific antigen cutoffsfor early dection of prostate cancer: receiver operating characteristiccurves. J Urol 1994;152:2037–42.

74. United States Preventive Task Force. Screening for prostate cancer;2012. Available from: http://www.uspreventiveservicestaskforce.org/prostatecancerscreening.htm. [accessed December 1, 2012]

75. American Cancer Society. Guidelines for the early detection ofcancer (prostate cancer). In American Cancer Society. Atlanta,Georgia, 2012.

76. National Comprehensive Cancer Network (NCCN). Prostate cancerearly detection guidelines, version 2. NCCN Clinical Practice Guide-lines in Oncology; Fort Washington, PA. 2007.

77. Vickers AJ, Roobol MJ, Lilja H. Screening for prostate cancer: earlydetection or overdetection? Annu Rev Med 2012;63:161–70.

78. Schroder FH, Hugosson J, RoobolMJ, Tammela TL, Ciatto S, Nelen V,et al. Screening and prostate-cancer mortality in a randomized Euro-pean study. N Engl J Med 2009;360:1320–8.

79. Andriole GL, Grubb RL III, Buys SS, Chia D, Church TR, Fouad MN,et al. Mortality results from a randomized prostate-cancer screeningtrial. N Engl J Med 2009;360:1310–9.

80. Zeigler-Johnson C, Tierney A, Rebbeck TR, Rundle A. Prostate cancerseverity associations with neighborhood deprivation. Prostate Cancer2011;2011:PMID 22111000.

81. Subramanian SV, Kubzansky L, Berkman L, Fay M, Kawachi I. Neigh-borhood effects on the self-rated health of elders: uncovering therelative importance of structural and service-related neighborhoodenvironments. J Gerontol B 2006;61:S153–60.

82. Fowke JH, Schlundt D, Signorello LB, Ukoli FA, Blot WJ. Prostatecancer screening between low-income African-American and Cauca-sian Men. Urol Oncol 2005;23:333–40.

83. Berglund A, Garmo H, Robinson D, Tishelman C, Holmberg L, Bratt O,et al. Differences according to socioeconomic status in the manage-ment andmortality in men with high risk prostate cancer. Eur J Cancer2012;48:75–84.

84. Carpenter W, Howard D, Taylor Y, Ross L, Wobker S, Godley P. Racialdifferences in PSA screening interval and stage at diagnosis. CancerCauses Control 2010;21:1071–80.

85. Xiao H, Gwede CK, Kiros G, Milla K. Analysis of prostate cancerincidence using geographic information system and multilevel model-ing. J Natl Med Assoc 2007;99:218–25.

86. Rundle A, Richards C, Neslund-Dudas C, Tang D, Rybicki BA. Neigh-borhood socioeconomic status modifies the association betweenindividual smoking status and PAH–DNA adduct levels in prostatetissue. Environ Mol Mutagen 2012;53:384–91.

87. Rebbeck TR, Weber AL, Walker AH, Stefflova K, Tran TV, Spangler E,et al. Context-dependent effects of genome-wide association studygenotypes and macroenvironment on time to biochemical (prostatespecific antigen) failure after prostatectomy. Cancer Epidemiol Bio-markers Prev 2010;19:2115–23.






2013;22:485-495. Published OnlineFirst March 5, 2013.Cancer Epidemiol Biomarkers Prev Shannon M. Lynch and Timothy R. Rebbeck Macroenvironmental Factors in Cancer: A Multilevel ApproachBridging the Gap between Biologic, Individual, and

Updated version

10.1158/1055-9965.EPI-13-0010doi:

Access the most recent version of this article at:

Cited articles

http://cebp.aacrjournals.org/content/22/4/485.full#ref-list-1

This article cites 69 articles, 7 of which you can access for free at:

Citing articles

http://cebp.aacrjournals.org/content/22/4/485.full#related-urls

This article has been cited by 6 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/22/4/485To request permission to re-use all or part of this article, use this link



http://cebp.aacrjournals.org/lookup/doi/10.1158/1055-9965.EPI-13-0010

http://cebp.aacrjournals.org/content/22/4/485.full#ref-list-1

http://cebp.aacrjournals.org/content/22/4/485.full#related-urls

http://cebp.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cebp.aacrjournals.org/content/22/4/485


bridging the gap between biologic, individual, and ... · biologic level (e.g., cellular biomarkers...

Documents