• No se han encontrado resultados

3.7 Resultados

3.7.1 Conocimientos

Abroad,butstilllimitedrangeofsocialandfamilyfactorswereconsideredforinclusionin this dissertation. The selection of characteristics was driven by the a priori defined hypotheses(seeChapter5.1)butwasrestrictedbytheavailabilityofcertaincharacteristics inthedatasourcesofthedissertation.Race,SEP,placeofresidence,birthorder,numberof siblings, parental age and cohabitation status of the parents were studied within the differentarticles.Asaproxyforchildren’sSEPseveralmakersofparentalSEPwereused. 

From the South African Cancer Registry data racial group was the only available social characteristic. Since racial group is highly correlated with socioeconomic circumstances includingaccesstoprivatehealthcareservicesinSouthAfrica[30],raceintheseanalysesis primarilyunderstoodasasocialcharacteristicratherthanageneticfactor(articleI&II). FortheGermansurvivalstudiesinformationonsocioeconomicfactors(articleIII)andfamily factors (article IV) were derived from the former German caseͲcontrol study collected via standardised telephone interview and were therefore predefined. The socioeconomic factors disposable family income, parental education and parental occupational training were available and the family characteristics birth order, number of siblings, parental age anddegreeofurbanizationoftheplaceofresidence.

TheDanishregistriesenabledtheinvestigationofabroadvarietyofsocialandfamilyfactors includingmaternalincome,parentaleducation,numberoffullandhalfsiblings,cohabitation status of the parents, parental age, birth order and degree of urbanization of place of residence(articleV&VI).However,astheanalysesforarticleVIwereconductedatIARCin France and the social registries of Denmark do not allow analyses of their data outside of Denmarksocioeconomicfactorscouldnotbeincludedinthatstudybutwerecoveredbythe otherDanisharticle(articleV).



More detailed information on the definition and data collection of the social and family factors used for the various investigations can be found in the respective articles (s. Appendix).

5.4Studypopulations

An overview of the different study populations analysed in thecontext of this dissertation and their special features is given below. The term childhood cancer was defined in two waysdependingonthedatasource.Childhoodcancerwasdefinedasanycancerdiagnosed beforetheageof15forthearticlesI,IIIandIV(GermanandSouthAfricandata).ArticlesV,

VI and VII, based on the Danish registry data, used the alternative definition of childhood

cancer:anycanceroccurringinpatientsbelowtheageof20. 

Article I Ͳ Childhood cancer incidence patterns by race, sex and age for 2000 – 2006: A reportfromtheSouthAfricanNationalCancerRegistry:

4,601newlydiagnosedchildhoodcancercaseswerereportedtotheSouthAfricanNational CancerRegistryduringtheperiod2000to2006;withagreaternumberofcasesamongboys thangirls(55%vs.45%).67.9%ofthereportedchildhoodcancercaseswereBlackAfricans, 8.7%hada mixedancestry,3.2% wereofIndian/Asianancestry,15%wereWhitesandfor 5.2% information on racial group was missing. The percentage of cases reported by the public laboratories of the NHLS ranged from 92% among Black children to 66.5% among Whites.Themostcommonlyreportedchildhoodcancerwasleukaemia(19%).



ArticleIIͲHaematologicalmalignanciesinSouthAfrica:2000Ͳ2006:

Between2000and2006,therewereatotalof14,662haematologicalmalignanciesinadults (defined as over the age of 14 years at diagnosis) reported to the South African National CancerRegistry.Almost50%ofthecaseswerereportedamongtheBlackpopulation,oneͲ third among Whites, 10% among individuals of mixed ancestry, 4% among Indians/Asians and for just below 5% information on racial group was missing. The distribution of racial groups differed considerably on whether cases where notified by public or private laboratories. 84% of the cases reported among Blacks were reported by the public laboratoriescomparedto50%ofthereportedcasesamongWhites.Regardlessofgenderor race, NHL was the most commonly reported haematological malignancy, accounting for at least 50% of cases in most gender and racial groups. In all groups, this was followed by leukaemia,contributing15Ͳ25%ofcasesinthevarioussubgroups.



Article III Ͳ Survival from childhood acute lymphoblastic leukaemia in Germany – Does socioͲdemographicbackgroundmatter?:

ThestudypopulationconsistedofallchildhoodALLcasesdiagnosedbetweenOctober1992 and September 1994 in West Germany and identified by the GCCR in the context of the former German caseͲcontrol study. Children were followed for 10 years after diagnosis, usingsurvivaldatafromtheGCCR.Ofthe788casesidentified,58.4%wereboysandmore than60%were1Ͳ5yearsofageatdiagnosis.Informationonsocioeconomiccharacteristics wasavailablefor647cases.Informationwasmissingforafurther6.5%onincome,5%on maternaleducationand10%onpaternalschooleducation.Mostcommonly,familiesranked in the second lowest categories – a monthly family income between 2,000 and 4,000 DM (53%) and education with “low degree” (“Hauptschulabschluss”) (paternal: 43% and maternal: 38%). Over the followͲup period of 10 years 137 children had died. The 10Ͳyear overallsurvivalofallcaseswas82.5%.



Article IV Ͳ Family circumstances and survival from childhood acute lymphoblastic leukaemiainWestGermany:

The article studied the same study population as analysed in article III but solely included cases for which family characteristics were available. Of the 647 cases, 334 (52%) were firstbornsand159(25%)weretheonlychild;almosthalfofthefamilieshadtwochildren. With respect to place of residence, most families were living in urban areas, and most parentswereagedч30yearsatdiagnosis.Numbersofmissingvalueswereverylowforthe key variables, ranging between 0.5% for maternal age and 1.6% for paternal age. Children werefollowedfor10yearswith10Ͳyearoverallsurvivalbeing84.7%,basedon98deaths. 

ArticleVͲEffectofsocioeconomicpositiononsurvivalafterchildhoodcancerinDenmark: The nationwide survival study was conducted by linking data from the Danish public administrative registries with the Cancer Registry. 3,797 children diagnosed with cancer belowtheageof20between1January1990and31December2009wereidentified.Most of the children were diagnosed in the age range of 0Ͳ4 years (29%) or 15Ͳ19 years (33%), with slightly more boys diagnosed with cancer than girls. The most common cancers were CNStumours(26%)followedbyleukaemia(24%).Abouthalfoftheparentshadavocational education,andabout80%werecohabiting.62%ofthechildhoodcancerpatientshadoneor

more full siblings. FollowͲup time was not censored for this study. Median followͲup time was9.0years(range0Ͳ22years)with841deathsandanoverallsurvivalof78%.



ArticleVIͲFamilycharacteristicsandsurvivalfromchildhoodhaematologicalmalignancies inDenmark,1973Ͳ2006:

Similar to article V this survival study analysed data from the Danish administrative registries. The study population comprised all children born and diagnosed with any haematologicalmalignancyinDenmarkbetween1January1973and31December2006.Of the 1,819 children diagnosed with haematological malignancies in this time period, 59% wereboysand40%ofthecasesoccurredinchildrenages1to4.Morethanhalfofthecases werediagnosedwithALL(56%),followedbyHL(compromising13%ofallcases),AML(12%) andNHL(9%).Among allcohortmembers,834(46%)werefirstborns and664(37%)were secondͲborn.285(16%)weretheonlychildatdateofdiagnosis.Mostfamilieswerelivingin provincialcities(53%).Mothersandfathersweremostfrequentlyaged31Ͳ35yearsattheir child’s diagnosis. Children were followed for 10 years. Overall 10Ͳyears survival was 72.1% basedon504deaths.



ArticleVIIͲBirthorderandtheriskofchildhoodcancerintheDanishbirthcohortof1973Ͳ 2010:

This cohort study was also based on data from the Danish registries. The birth cohort comprised2,461,283childrenbornbetween1973and2010inclusively,ofwhich1,262,979 (51.3%) were boys. Among the total cohort, 1,099,058 children were firstborn (44.7%), 906,852 (36.8%) secondͲborn, 336,017 (13.7%) thirdͲborn, and 119,356 (4.8%) had a birth order of four or higher. 227,913 (9.3% of total and 20.7% of firstborn) remained only children.Inthestudypopulationaccruingatotalof38.6millionpersonͲyearsoffollowͲup; 5,699 childhood cancers were observed, with leukaemias and CNS tumours each representing approximately one quarter of cases. 45.6% of cancer cases were firstborn, similartotheproportionintheoverallcohort.



Moredetailedinformationonthestudypopulationscanbefoundintherespectivearticles (s.Appendix).

5.5Statisticalmethods

Thestatisticalmethodsusedforthisdissertationdependedontherespectivearticle.Articles

IandIIanalysedincidencedatafromtheSouthAfricanCancerRegistryapplyingdescriptive

epidemiologicalmethods.ArticleIIItoVIanalysedsurvivaldatafromGermanyandDenmark using KaplanͲMeier curves and Cox proportional hazard models. Article VII analysed the childhood cancer risk in a birth cohort using Poisson regression models. A more detailed descriptionisgivenbelowandintherespectivearticles(s.Appendix). Generally,childhoodcancercasesweregroupedaccordingtotheInternationalClassification ofChildhoodCancer,dependingonthetimeperiodofthedataeitherICCCͲ1orICCCͲ3(see Chapter1.2).  Analysesofchildhoodcancerandadulthaematologicalmalignanciesincidencedatafrom theSouthAfricanNationalCancerRegistry(articlesIandII)

Descriptive epidemiological methods, in particular calculation of (relative) frequencies, sex ratios, ageͲspecific and ageͲstandardised incidence rates stratified by racial group and/or certain diagnostic subgroups were used to analyse the reported incidence of childhood cancerandadulthaematologicalmalignanciesinSouthAfrica(articleIandII). Thedirectly ageͲstandardisedincidencerates(ASR)werecalculatedusingtheweightsoftheSegiworld standardpopulation[197].Consistentwiththeapproachusedintheannualreportsofthe South African National Cancer Registry, the Alternative South African midͲyear population estimatesfromtheCentreforActuarialResearch,UniversityofCapeTownwereusedasthe denominator[181,198]forcalculatingincidencerates.ThesemidͲyearpopulationestimates aresimilarinmagnitudetotheofficialmidͲyearestimatesbutmaintainanagedistribution that is consistent with that of the most recent Census in 2011 [181]. Therefore they were considered as the more appropriate population estimates for the purposes of studying differencesbyagegroup.



For article I, the incidence rate proportions in South Africa compared to the reference incidenceratesofGermanywerecalculatedtoinvestigatedifferencesforspecificsubgroups ofchildhoodcancer.Forthispurpose,theASRsinGermanyintherespectivesubͲcategories (bycancertypeandbysex),aswellasageͲspecificrates,weresetto100%andthesewere comparedtothereportedincidenceratesamongWhiteandBlackSouthAfricansseparately.

The German childhood cancer incidence rates were provided by the GCCR according to specifications defined by me to exactly match the eligibility criteria of the South African CancerRegistryintermsofincludeddiagnosesanddiagnostictimeperiod.



ForarticleII,Incidencerateratios(IRRs)and95%confidenceintervals(CIs)wereestimated using Poisson regression models with the number of cases for a given category of haematologicalmalignancyastheoutcome,thepopulationsizeasthelogoffsetandalog linkfunction,overallandstratifiedbyracialgroup,comparingratesofadulthaematological malignancies among females to that of males. Similar models, stratified by gender, were usedtocompareincidenceratesacrosstheracialgroups,usingtheBlackpopulationasthe referencegroup.  Survivalanalyses(articlesIII–VI) TwoprimaryoutcomesweredefinedforthetwoGermansurvivalstudies(articlesIII&IV): overallsurvival,withdeathfromanycauseastheendpoint,andeventͲfreesurvival,withthe first (if any) relapse (defined as >5% lymphoblasts in bone marrow), second malignant neoplasm or death as events. Since information on relapse is not recorded in the Danish CancerRegistryonlyoverallsurvivalwasdefinedasoutcomefortheDanishstudies(articleV

&VI).Childrenwereobservedfromthedateofdiagnosisuntilthedateofevent(previous

definedevent,e.g.deathfromanycause,relapseorsecondarymalignancy),emigrationor lost to follow up, or date of 10 years of followͲup, whichever came first. FollowͲup period wascensoredat10yearsasveryfewdiseaseͲrelatedeventsoccurafterwards,whereasthe incidence of competing risks rises. An exception is the Danish survival study on SEP and survivalfromchildhoodcancer(articleV)forwhichthefollowͲupperiodwasnotcensored. 

Forgraphicalillustration(unadjusted)survivalprobabilitiesstratifiedbyselectedsocialand family characteristics were calculated, using KaplanͲMeier curves. Statistical significance (definedaspч0.05)ofdifferencesinsurvivalprobabilitieswasassessedbythelogͲranktest [199]. Cox proportional hazards models were used to assess the impact of selected social andfamilycharacteristicsonsurvivalwithtimesincediagnosisastheunderlyingtimescale. Itisthemostcommonlyusedmultipleregressionmodelforanalysingsurvivaldatainhealth researchandillustratestherelationshipbetweentheeventincidenceandasetofmutually

adjusted covariates using hazard ratios [200, 201]. Models were adjusted for the well– establishedprognosticfactorschild’sageatdiagnosis[31Ͳ33,120]andsex[120,132]aswell as for the possible mediating effect of other social variables (adjustment varied between

articles and models). Results were expressed as adjusted hazard ratios (HRs) with

corresponding95%confidenceintervals. 

AnalysesofDanishbirthcohortdataandchildhoodcancerrisk(articleVII)

Allchildrenwerefollowedupfromdateofbirthuntil20yearsofage,dateofdeath,dateof firstcancerdiagnosis,orendofstudyperiod(October,31,2013),whicheveroccurredfirst. Poisson regression models were used to evaluate associations between birth order and differenttypesofchildhoodcancer,estimatingtherateratios(RRs)andcorresponding95% confidenceintervals,withandwithoutcontrollingformaternalage,paternalage,andbirth weight.Thefirstbornchildrenservedasthereferencegroupforthecomparisons.