epidemiologic mapping of florida childhood cancer clusters

Pediatr Blood Cancer 2010;54:511518

Epidemiologic Mapping of Florida Childhood Cancer Clusters

Raid Amin, PhD,1,2* Alexander Bohnert,1 Laurens Holmes, PhD, DrPH,2,3 Ayyappan Rajasekaran, PhD,2

and Chatchawin Assanasen, MD2,4**

INTRODUCTION

Cancer remains the leading cause of disease-related deathamong children in theUnitedStates despite progress in clinical trialsand significant improvements in survival rates [1]. Over the past20 years in theUnited States, increases in the incidence of childhoodcancer have also been observed from 11.5 cases per 100,000children in 1975 to 14.8 per 100,000 children in 2004 [2]. In 2009,approximately 10,730 children under the age of 15will be diagnosedwith cancer and about 1,480 are projected to die from the disease [3].

Despite the burden of childhood cancer and the many years ofepidemiologic investigations, its causes remain largely unknownbut have been linked in small percentages to certain geneticpredispositions and exposures to chemotherapy agents and ionizingradiation [47]. A number of studies continue to examine thecomplexities of other possible risk factors for childhood cancers [811]. These include early-life exposures to infectious agents;parental, fetal, or childhood exposures to environmental toxins;parental occupational exposures to radiation or chemicals; parentalmedical conditions during pregnancy or before conception;maternal diet during pregnancy; early postnatal feeding patternsand diet; and maternal reproductive history [1224]. Environ-mental factors may play an important etiologic role in childhoodmalignancies and can be evidenced by excessive numbers of cases ina defined geographic area relative to other areas, termed clusters.

A cancer cluster can be defined as the occurrence of a greaterthan expected number of cases of a malignancy within a group ofpeople, a geographic area, or a period of time. There exist variousdefinitions of the terms cluster and clustering in the context ofspatial epidemiology and cancer research, respectively [25,26].Identification of spacetime variations in incidence rate patternscan provide important clues for further in-depth studies into theetiology and control of cancer [27]. Spatial clustering is defined as ageneral irregular spatial distribution of cases that is not confined toone particular small area. Spacetime cancer clustering is observedwhen an excess number of cases occur within a geographicallocation over very limited periods of time and cannot be explainedin terms of general excesses in these locations or time frames.Regional, national, and international registries have been utilized to

investigate possible spatial and spacetime clustering and anyassociated risks of cancer predisposition [2124,2831].

In Florida, overall cancer statistics are similar to the rest of theUnited States. From 1981 to 2000, 10,238 new cases of cancer werediagnosed among Florida children and adolescents, representing0.7% of all cancer cases diagnosed in the state [28]. The FloridaAssociation of Pediatric Tumor Programs (FAPTP) was founded in1970 as a statewide network of childrens cancer programs under theauspices of the Florida Regional Medical Program (FRMP). TheFlorida legislature established a pediatric hematology/oncologyprogramwithinChildrensMedical Services (CMS) andFAPTP thatwas given the responsibility and authority to monitor and evaluatepediatric cancer care statewide. This reporting system provides thestate and the public with data on cancer incidence, clinical trialparticipation, and survivorship. The Statewide Patient InformationRecording System (SPIRS) registers patients from the 16 pediatrichematology/oncology centers statewide. In addition, the FloridaCancer Data System (FCDS) captures the data from patients treatedoutside the FAPTP system and can be linked with SPIRS data tostudy the larger patient data base.

Background. Childhood cancer remains the leading cause ofdisease-relatedmortality for children.Whereas, improvement in carehas dramatically increased survival, the risk factors remain to be fullyunderstood. The increasing incidence of childhood cancer in Floridamay be associated with possible cancer clusters. We aimed, in thisstudy, to identify and confirm possible childhood cancer clusters andtheir subtypes in the state of Florida.Methods.We conducted purelyspatial and spacetime analyzes to assess any evidence of childhoodmalignancy clusters in the state of Florida using SaTScanTM. Datafrom the Florida Association of Pediatric Tumor Programs (FAPTP) forthe period 20002007 were used in this analysis. Results. In thepurely spatial analysis, the relative risks (RR) of overall childhoodcancer persisted after controlling for confounding factors in southFlorida (SF) (RR1.36, P0.001) and northeastern Florida (NEF)

(RR1.30, P 0.01). Likewise, in the spacetime analysis, therewas a statistically significant increase in cancer rates in SF(RR1.52, P0.001) between 2006 and 2007. The purely spatialanalysis of the cancer subtypes indicated a statistically significantincrease in the rate of leukemia and brain/CNS cancers in both SFand NEF, P< 0.05. The spacetime analysis indicated a statisticallysignificant sizable increase in brain/CNS tumors (RR2.25,P 0.02) for 20062007. Conclusions. There is evidence of spatialand spacetime childhood cancer clustering in SF and NEF. Thisevidence is suggestive of the presence of possible predisposingfactors in these cluster regions. Therefore, further study is needed toinvestigate these potential risk factors. Pediatr Blood Cancer 2010;54:511518. ! 2010 Wiley-Liss, Inc.

Key words: cancer cluster; childhood neoplasm; cluster analysis; epidemiology; florida

! 2010 Wiley-Liss, Inc.DOI 10.1002/pbc.22403Published online 6 January 2010 in Wiley InterScience(www.interscience.wiley.com)

1Department of Mathematics and Statistics, University of West Florida,Pensacola, Florida; 2Nemours Center for Childhood Cancer Research,Wilmington, Delaware; 3Department of Orthopedics, Alfred I duPontHospital for Children, Wilmington, Delaware; 4Nemours ChildrensClinic, Pensacola, Florida

Conflict of interest: Nothing to report.

Grant sponsor: Nemours Childrens Clinic, Pensacola; Grant sponsor:Nemours Foundation; Grant sponsor: Caitlin Robb Foundation.

*Correspondence to: Raid Amin, Department of Mathematics andStatistics, University of West Florida, 11000 University Parkway,Pensacola, FL 32514. E-mail: [email protected]

**Correspondence to: Chatchawin Assanasen, Nemours Center forChildhood Cancer Research, Nemours Childrens ClinicPensacola,5153 North 9th Avenue, Pensacola, FL 32504.E-mail: [email protected]

Received 8 September 2009; Accepted 17 November 2009

The recently founded Nemours Center for Childhood CancerResearch (NCCCR) has three of its oncology clinics in Jacksonville,Orlando, and Pensacola, Florida as well as one in Wilmington,Delaware. One of the initial goals of this center was to evaluatepediatric cancer epidemiology data in the states of Florida andDelaware. In 2008, the Delaware childhood cancer rates wereevaluated byNCCCR in collaborationwithDelawareDepartment ofHealth and Social Services for possible childhood cancer clusters.This assessment failed to confirm clusters probably due to smallnumber of cases aswell as absence of clusters. The current studywasinitiated about 2 years ago in collaboration with the University ofWest Florida. We sought to identify and confirm overall childhoodcancer clusters as well as to determinewhether or not clusters couldbe confirmed by cancer subtypes. We utilized the data from FAPTPand modeled our analysis using SaTScanTM to test the followingnull hypotheses: (1) The pediatric cancer rate of all cancer typesis randomly distributed over space in Florida from 2000 to 2007,(2) The pediatric cancer rate of all cancer types is randomlydistributed over time and space in Florida from 2000 to 2007,(3) The rates for specific pediatric cancer types are randomlydistributed over space in Florida from 2000 to 2007, and (4) Therates for specific pediatric cancer types are randomly distributedover time and space in Florida from 2000 to 2007.

MATERIALS AND METHODS

We conducted purely spatial and spacetime analyzes to assessthe evidence of childhood cancer clusters in the state of Floridausing SaTScanTM.

Study Area and Population

We identified 67 counties and 972 zip code areas in Florida inthe year 2000. While the clustering evaluations could have beenbased on Florida counties, we decided to obtain more detailedinformation by using the zip code areas. The statistical analysis usedin this study requires that the geographic information for each zipcode area be represented by some form of a centroid. To obtain thegeographical centroid of each zip code area and to create maps withinformation on the cancer clusters, the geographical informationsystem ArcGIS was utilized. We used consistent geographical datafor zip code areas, Zip Code Tabulation Area (ZCTA), from the year2000 from the Florida Geographic Data Library (FGDL). For zipcode areas that were created after 2000 with an identified cancercase, we manually assigned in the software package ArcGIS a zipcode based on its position in the 2000 geographical data set. Amarginal number of cases for which we could not determine theirposition relative to the 2000 zip code area file were discarded. Thestudy population included the entire population of children 019 years of age in the state of Florida during the time period 20002007. These included children with and without the diagnosis of achildhood cancer. During this time, there were 4,591 cases ofpediatric cancer diagnosed, of which 1,254 (27%) had leukemia,839 (18%) had brain/central nervous system (CNS) cancer, and 252(5.5%) had lymphoma.

Data Sources

The data for this study were available from FAPTP, an existingde-identified dataset, that is, publicly available. FAPTP has been

shown to be a valid and reliable source for pediatric cancer incidencedata in Florida [28,32,33]. The dataset included information oncancer cases such as the diagnosis code for the study period 20002007 designated by the International Classification for ChildhoodCancer (ICCC) [34], incorporating the new codes introduced inICD-O-2 and the updated ICD-O-3. Demographic information wasalso included, such as date of birth, age at cancer diagnosis, sex, andzip code of residence. This study involved age-adjusted data. Weobtained Florida demographic population data such as age and race/ethnicity from the 2000 census. For each ZCTA, we obtained thetotal population at risk, stratified by age, sex, and race/ethnicity.

Data Analyzes

Clusters have been analyzed previously using several statisticaland epidemiologic approaches [35]. In this study, we usedSaTScanTM. The software package SaTScanTM [26] uses spatialscan statistics to identify and test for the significance of cancerclusters. The incidence counts in each zip code area are used eitherin two dimensions for a purely spatial analysis or in a three-dimensional setting for a spacetime analysis with the additionaldimension representing time. We assumed that the incidence ofcancer in each zip code area is distributed according to a Poissonmodel [36,37]. This method tests the null hypothesis that the age-adjusted risk of cancer incidence is the same for all zip code areas.With the covariates included in the model, we tested the nullhypothesis that within any age group, the risk of cancer incidence isthe same for the entire area covered in this study [37]. To include theeffect of urbanicity in our analysis, we used the population densityinformation for postal code level [37] that is available through theFlorida Geographic Data Library (FGDL). Possible associations tosocioeconomic status (SES) were investigated by using theeconomics wealth index by Woods & Poole Economics Inc., whichwe obtained from the HAAS Business Center at the University ofWest Florida [38]. Since neither of these two covariates resulted inany changes in the SaTScanTM computations, results on populationdensity or the socioeconomic status are not presented.

The spatial scan statistics in SaTScanTM identifies clusters byimposing a window that moves over a map, including different setsof neighboring zip code areas represented by their correspondingcentroids [29]. If the window includes the centroid of a specific zipcode area, then this zip code area is included in the window. Assuggested by Kulldorff et al. [29], the center of the window ispositioned only at the 972 zip code centroids. For each window, thespatial scan statistic tests the null hypothesis of equal risk ofchildhood cancer incidence for all zip code areas against thealternative hypothesis that there exists an elevated risk of childhoodcancer incidencewithin the scanwindowwhen comparedwith areasoutside the window. The likelihood function for the Poisson modelcan be shown to be proportional to

n

E

! "n N" nN" E

# $N"nI n > E

where n is the number of cancer incidences within the scan window,N is the total number of incidences in Florida, and E is the expectednumber of cancer incidences under the null hypothesis [29,37].Sincewe are using a one-tailed test that rejects the null hypothesis ifthere exists elevated cancer risk, an indicator function I is used suchthat I 1 when the scan window has a larger number of cancerincidences than expected if the null hypothesis were true, and zero

Pediatr Blood Cancer DOI 10.1002/pbc

512 Amin et al.

otherwise. It can be shown that for a given N and E, the likelihoodincreases as the number of incidences, n, increase in the scanwindow. How the spatial scan statistic within SaTScanTM actuallyidentifies cancer clusters is described elsewhere in detail [37]. By aMonte Carlo simulation, we generated 999 random replications ofthe data set to obtain the statistical stability for the identified cancerclusters in the program SaTScanTM. The Monte Carlos test alsoallows for the simultaneous controlling of multiple confounderssuch as age, sex, race, income level, etc. The identified cancerclusters are listed by SaTScanTM in order of significance such thatthe P-value for each cluster is compared with a pre-set significancelevel of 0.05.

There exist different types of the spatial scan statistics. Circularor elliptical windows can be used to identify circular clusters andelliptical shaped clusters, respectively. Both approaches were used,and we arrived at virtually identical cluster results. In this study, wepresent only the cancer clusters identified by circular windows.While the spatial scan statistic requires specification of theunderlying distribution of the data used in SaTScanTM, making ita parametric statistical method, a non-parametric smoothingmethod was also used to check whether similar or identical clusterresults would be obtained. In particular, we used a weighted HeadBanging algorithm based on median smoothing which removes thebackground noise of random variability so that the underlyingspatial pattern becomes more clear [3941]. Both parametric andnon-parametric methods were used for the purpose of resultsvalidation. In this study, Head Bang was used to statistically doublecheck the results from SaTScanTM by removing local variations incancer incidence age-adjusted rates for the 972 zip code areas. Thisparticular smoother retains the important features, such as edges, butsmoothes out unreliable data points and spikes for low populationareas based on the chosen weights in the algorithm. To ensureadequate statistical power, all cancer cases for the period 20002007 were used to perform a purely spatial analysis. For the spacetime analysis, which is a temporal extension of the spatial analysis,the algorithm searches within 20002007 for time periods in whichclusters appear.

RESULTS

The SaTScanTM purely spatial analysis of the FAPTP datarevealed two significant clusters in the state, one in southern Floridaand the other in northeastern Florida (NEF). The south Florida (SF)cluster encompasses the southwest, south central and southeastregions. The NEF cluster incorporates areas of the northeast andnorth central regions. After adjusting for age, sex, and race ascovariates, a total of 4,181 cases were identified with a correspond-ing incidence rate of 14.4 average annual cases per 100,000. In SF,there were 465 observed cases and 352 expected cases, with arelative risk of 1.36, implying that comparedwith the state, there is astatistically significant 36% increased risk of childhood cancer(P 0.001). In the NEF cluster, there were 466 and 375 observedand expected cases, respectively. This region appears to be smallerin size, although it may represent a more densely populated area. Asimilar increase in the rates of childhood cancerwas identifiedwith aRR 1.30 (P 0.01). In addition, a third overall childhood cancercluster was identified in a small area of central Florida in which theobserved number of cases was 31 as compared to 11 expected cases.The rates were statistically significantly higher in this area relativeto the state with a RR 2.82 (P 0.008), which implies that

compared with the state of Florida, those in this area are almostthree times as likely to be diagnosed with childhood cancer (Fig. 1).

Since a purely spatial analysis for the period 20002007does notindicate when the cluster appeared, a spacetime analysis wasperformed, assessing these clusters using the Poisson model withinSaTScanTM.We observed that the spatial dimensions of the clusterspersisted during these periods. SF emerged as the most likelytemporal cluster with elevated risk during 20062007 (Fig. 2).Whereas the observed cases were 403, the expected were 274,RR 1.52, P 0.001, implying a significant 52% increase inchildhood cancer rate in SF compared with the state of Florida.Similarly, the NEF emerged as a secondary temporal cluster for20012004, with the observed and expected cases as 136 and 87respectively, RR 1.59,P 0.06. This suggested a 59% increase inthe rate of overall childhood cancer in NEF relative to the state, butthe increase was not statistically significant.

To confirm the clusters, we compared cancer rates within SF tothe state. The cancer rates of the state for this time period was 14.1per 100,000 in 2005 and increased slightly to 16.4 per 100,000 and15.7 per 100,000 for 2006 and 2007, respectively. By contrast, from2000 to 2007, the SF cancer rates have been consistently higher thanthe corresponding Florida rates. In particular, the rates computed for2006 and 2007 increased significantly from 13.8 per 100,000 in2005 to 23.9 and 21.1 per 100,000, in 2006 and 2007, respectively.


Fig. 1. Purely spatial analysis of FAPTP database for all cancer types20002007. Clustering representation of SaTScanTM purely spatialanalysis is illustrated utilizing zip code data with age, sex, and race ascovariates. Clusters are represented in colors. The red area represents theSouth Florida cluster. SaTScanTM computed results include: Coor-dinates/radius (26.3N, 81.3W)/101.6 km, Population 294,119,Observed cases 465, Expected Cases 352. The orange arearepresents the North Central Florida cluster. SaTScanTM computedresults include: Coordinates/radius (29.9N, 82.4W)/95.8 km, Popu-lation 375,761, Observed cases 530, Expected cases 420. Theyellow area represents the Central Florida cluster. SaTScanTM

computed results include: Coordinates/radius (28.2N, 81.5W)/13.4km, Population 9,213, Observed cases 31, Expected cases 11.

Florida Childhood Cancer Clusters 513

However, when we excluded the SF cases from the overall Floridacancer cases, the rates in Florida significantly decreased (Table I,Fig. 3).

Purely spatial analysis of leukemia rates identified two regions ofFlorida (during the period of 20002007) similar to the clusterareas identified when all cancer types were combined. A total of1,254 leukemia cases in the state were identified and utilized inthis analysis. There was a statistically significant cluster in SF(RR 1.53, P 0.001) (Fig. 4). A second cluster was identified inthe north central region of the state, shifting somewhat from theNEFcluster andwas statistically significant aswell, RR 1.45,P 0.03.Likewise, in the spacetime analysis of leukemia cases, there was astatistically significant cluster in SF (RR 1.74, P 0.05) (Fig. 5).The time period identified for the peak rate of the cluster was 20002002. During this time period, the number of observed cancer caseswas 105 while the expected number of cases was 63. While thespacetime analysis points to 20002002 as the time of the peak inleukemia rates, the purely spatial analysis indicated that leukemiarates in the SF cluster area remained elevated throughout the entireperiod (20002007), when compared to the state.

A purely spatial analysis of brain/CNS cancer identified one areain southern Florida. Of the 839 cases identified in the state, there

were 60 observed and 33 expected cases in this region. The relativerisk comparing Florida to SF was not statistically significant,RR 1.86, P 0.07 (Fig. 6). A spacetime analysis (52 observedcases and 24 expected cases) for the brain/CNS cancer identified acluster corresponding to the SF cluster, with a statisticallysignificant increased incidence rate RR 2.25, P 0.02, implyingthat children in SF were two times as likely to develop brain/CNScancerwhen comparedwith children in the state of Florida. The timeperiod identified for this clusterwas 20062007 (Fig. 7). In contrast,lymphoma rates were not statistically significant probably due tosmall numbers.


TABLE I. Childhood Age and Sex Adjusted Cancer IncidenceRates for Florida, SF Cluster, and Florida Without SF Cluster

Area Year Rate 95% CI Rate ratio

FL 2006 16.4 15.1, 17.6 1.02007 15.7 14.5, 16.9 1.0

Aggregate 16.05 14.8, 17.3 1.0FL w/o SF 2006 14.8 13.5, 16.1 0.9024

2007 14.5 13.2, 15.8 0.9236Aggregate 14.65 13.2, 15.7 0.9128

SF 2006 23.9 20.3, 27.5 1.45732007 21.1 17.7, 24.4 1.3439

Aggregate 22.5 19.2, 26.1 1.4019

Incidence countswere utilized directly to compute incidence rates usingthe FAPTP Dataset for 20002007 and Florida population statistics for2000. Confidence intervals are provided, as uncertainty still existswithin ideal registry datasets and computed cancer statistics (UnitedStates Cancer Statistics: 1999 incidence). Aggregate refers to the ratesfor 2006 and 2007 combined. The state of Florida is the reference group,hence the ratio is 1.0 for FL, Florida. SF is the southern Florida cluster.The time frame for the SF cluster was noted to be January 1, 2006 toDecember 31, 2007. CI, Confidence intervals.

Fig. 2. Spacetime analysis of FAPTP database for all cancer types20002007. Clustering representation of SaTScanTM spacetimeanalysis is illustrated utilizing zip code data with age, sex, and race ascovariates. Clusters are represented in colors. Spatial representationswere not affected significantly however time frame results for theSouthern Florida (SF) cluster (20062007) are noted to be representa-tive of a recent surge in incidence rates. The red area represents theSouth Florida cluster. SaTScanTM computed results include: Coor-dinates/radius (26.0N, 81.4W)/121.1 km, Time frame January 1,2006 to December 31, 2007, Population 963,643, Observed cases403, Expected cases 274. The orange area represents the NorthCentral Florida cluster. SaTScanTM computed results include: Coor-dinates/radius (29.5N, 82.0W)/65.9 km, Time frame January 1,2001 to December 31, 2004, Population 155,681, Observed cases136, Expected cases 87.

Fig. 3. Age-adjusted pediatric cancer incidence rates 20002007.Incidence countswere utilized directly to compute incidence rates usingFAPTP Dataset for 20002007 and Florida population statistics for2000. Southern Florida cluster (SF) is shown in comparison to rates forthe entire state of Florida and to rates for the state of Florida excludingthe influence of the SF. Differences between these rates during 2006 and2007 suggest that the rise in Florida rates during this period wasinfluenced by the surge in incidence rates in the SF cluster.

514 Amin et al.

DISCUSSION

These purely spatial and spacetime clustering studies ofchildhood cancer in Floridawere conducted using data from FAPTPand the Census data of 2000. The accuracy of case ascertainment ishigh with FAPTP and has been described and validated elsewhere[28,32,33]. This epidemiologic mapping study of Florida revealsthreemajor findings. First, childhood cancer clusters were identifiedin SF and NEF. Second, the childhood cancer clusters persistedafter controlling for age, sex, and race/ethnicity. Third, whereassignificant increase in cancer rates was observed in leukemia andbrain/CNS cancer, there was no significant increase in thelymphoma rate among children in SF and NEF.

There are several methodologic issues in identification andconfirmation of childhood cancer clusters, especially leukemia [42].In general, these studies are limited by low statistical power [43].Therefore, the identification of cancer clustersmay be driven by biassuch as the practice of defining geographical boundaries of thecluster and improved case ascertainment in the areas suspected ofhaving clusters, as well as error, namely, random variation [44].Cancer cluster studies utilizing multiple comparisons over a smallperiod of time or differentmethods have shown false positive results[45]. Further, population density, age, migration, sex, and race/ethnicity are potential confounding elements affecting childhoodcancer cluster confirmation [46,47].

This study utilized statistical software (SaTScanTM) that isreliable in the assessment of cancer clusters, as well as other disease

clusters, in the human population [2931,35,36]. By utilizing thedata fromFAPTP,we ensured the accuracy and reliability of the dataused. FAPTP routinely reviews the cancer data for discrepanciesincluding duplications and provides the most comprehensiveincidence data of childhood cancer in Florida. Thus, FAPTPfacilitates assessment of patterns of cancer rates and geographicaltrends within the state of Florida.Whereas the limitations addressedin previous studies on clusters could not be avoided completely,our chances of repeating similar methodologic issues weresubstantially minimized as described below.

The large sample of cases with overall childhood cancer as wellas significant cases in cancer subtypes should ensure a sufficientlyhigh statistical power. It has been shown in a simple power study[36] for the likelihood ratio test used in SaTScanTM that a relativerisk of 1.35 can result in an estimated power (1" b> 0.80) to detectthe differences in cancer cases between the clusters and non-clusterareas (in the state of Florida), if one does exist. For example, from2006 to 2007, the observed cases were 403 in SF, which is a largesample for comparison between areas with and without clusters(Fig. 2). Because we used cancer data from a highly reliable source(FAPTP), both selection and misclassification biases weredramatically minimized in our study. The observed clusters in SFand NEF are not driven by improved case ascertainment followingthe increased childhood cancers in certain geographic areas inFlorida. In addition, because this study started 2 years ago, it ishighly unlikely that our findings are influenced by other recentstudies on Florida clusters.


Fig. 4. Purely spatial analysis of leukemia cases 20002007.Clustering representation of SaTScanTM purely spatial analysis isillustrated utilizing FAPTP zip code data with age and sex covariates.Clusters are represented in colors. The SF cluster remains durable andsignificantwith respect to the specific leukemia cases in Florida. The redarea represents the South Florida cluster. SaTScanTM computed resultsinclude: Coordinates/radius (26.2N, 81.7W)/141.6 km, Population417,327, Observed cases 190, Expected cases 131. The orange arearepresents the North Central Florida cluster. SaTScanTM computedresults include: Coordinates/radius (29.1N, 82.7W)/120.1 km, Pop-ulation 435,669, Observed cases 190, Expected cases 138.

Fig. 5. Spacetime analysis of leukemia cases 20002007. Cluster-ing representation of SaTScanTM Spacetime analysis is illustratedutilizing FAPTP zip code data with age and sex as covariates. Spacetime clusters were statistically significant. Time frame results for theSouthern Florida (SF) cluster were noted to be 20002002. The yellowarea represents the South Florida cluster. SaTScanTM computed resultsinclude: Coordinates/radius (26.0N, 81.6W)/128.5 km, Time frameJanuary 1, 2000 to December 31, 2002, Population 553,592, Observedcases 105, Expected cases 63.


To better understand the increased cancer rates in SF, it isimportant to consider changes in the population for that region aswell. Otherwise, the possible environmental factors affecting cancerrates could be confounded with population migrations andincreases. While estimates for the pediatric population counts forall ages in each of the Florida zip code areas were not available forthe period 20012007, we utilized population estimates for thepediatric population by county for 2001 and 2007 from the FloridaLegislature [48] and the estimates of the pediatric population for a 3-year-period 20052007 from the American Community Survey ofthe Census Bureau [49]. Considering the relative annual populationincrease, defined by the ratio r as follows:

r pediatric pop 2007" pediatric pop 2000pediatric pop 2000

where the change in the pediatric population count in 2007 isobtained relative to the pediatric population count in 2000, wecompared the average values for the ratio r for the SF area with thecorresponding annual relative population increase for the rest ofFlorida. Similarly, we also obtained a ratio based on the 20052007estimates. Our results indicated that relative population increases inthe SF cluster area are not significantly different from the rest of thestate. It is also possible that zip code population shifts over timecould have altered the results between 2000 and 2007. Such shiftscould result in an apparently elevated cancer ratewhen using 2000 asthe population standard. Using population estimates for larger areassuch as counties would limit the effects of such small-areamigrations on the cancer rates. Florida county population estimates

between 2005 and 2007 were available for 53 counties in Floridawith populations greater than 20,000. We analyzed purely spatialand spacetime SaTScanTM results for these 53 counties from 2000to 2007 and found that the brain tumor cluster persisted. Analysis ofleukemia clusters persisted during the spacetime analysis but notfor the purely spatial analysis. While our initial analysis was basedon zip codes, limited analysis based on counties indirectly suggeststhat population shifts did not play a significant role in altering thecancer clusters. Thus, it is highly unlikely that our findings ofchildhood cancer clusters are driven primarily by migration sincepopulation changes in these geographic areas were non-differential,thus minimizing any misclassification bias and confounding fromthe observed clusters.

In this study, we have shown that there is a relative increase inchildhood cancer crude incidence rate in SF and NEF duringthe years 20002007. Since this findingmight have been influencedby potential confounders of childhood cancer [44], we adjusted forage at diagnosis, sex,and race/ethnicity and still observed astatistically significant relative increase in SF and NEF comparedwith the state of Florida. Therefore, given these adjustments, it ispossible to suspect geographic variation as the potential riskvariable for the clusters. Although the cluster areas identified arequite large geographically, it is possible that localized environ-mental factors or person-to-person spread of viral or bacterialpathogen [12,13,2124], may be involved in these suspected


Fig. 6. Purely spatial analysis of brain/CNSTumor cases 20002007.Clustering representation of SaTScanTM purely spatial analysis isillustrated utilizing FAPTP zip code data with age and sex covariates.The SF cluster significant although size of area is altered with respectto prior cluster maps identified in Florida. The red area representsthe South Florida cluster. SaTScanTM computed results include:Coordinates/radius (26.0N, 80.4W)/15.6 km, Population 157,361,Observed cases 60, Expected cases 33.

Fig. 7. Spacetime analysis of brain/CNS Tumor Cases 20002007.Clustering representation of SaTScanTM Spacetime analysis isillustrated utilizing FAPTP zip code data with age and sex as covariates.Clusters are represented in colors. The red area represents a Northeast-ern Florida cluster. SaTScanTM computed results include: Coordinates/radius (30.1N, 81.8W)/20 km, Time frame January 1, 2005 toDecember 31, 2007, Population 111,133, Observed cases 29,Expected cases 9. The orange area represents the South Floridacluster. SaTScanTM computed results include: Coordinates/radi-us (26.3N, 81.3W)/105.2 km, Time frame January 1, 2006 toDecember 31, 2007, Population 455,519, Observed cases 52,Expected cases 24.

516 Amin et al.

geographic areas. Finally, despite these adjustments, we cannot ruleout unmeasured confounding elements as a possible explanation ofthe observed clusters. Furthermore, residual confounding elementsmay influence this confirmation especially by race/ethnicity, sincethis information may have suffered from misclassification bias.Therefore, statistical modeling cannot completely remove the effectof confounding [27,50].

Our study found the crude incidence rate of childhood leukemiaand brain/CNS cancers to be significantly higher in the SF and NEFclusters when compared with the state of Florida. As describedearlier, these findings are unlikely to be driven by non-factualattributes of cancer clusters but are suggestive of environmentalfactors or common risk factors in the areas. Consequently, thesefindings could be etiologically driven, indicating the need for furtherinvestigation to identify the potential risk factors in the observedleukemia and brain/CNS cancer clusters in these areas. We did notfind spatial or spacetime clustering with lymphoma in the adjustedmodels. The negative finding with lymphoma may be due to thesmall number of cases in this subset, which limits the statisticalpower to detect significant clusters with these data [47] or due to thelack of a lymphoma cluster.

Despite the strengths of this article, there are also somelimitations. First, we used a preexisting dataset that may beassociated with information and selection bias, thus influencing thevalidity of our findings. However, since the FAPTP data are highlyreliable, it is unlikely that our confirmation of cancer clusters in SF isdriven solely by information or selection bias. Second, confoundingelements such as race/ethnicity and agemay very well influence ourresults. But this is unlikely since we focused on childhoodmalignancy with no reference to adult tumors. Finally, as with allepidemiologic studies, unmeasured and residual confoundingelements may also partly influence the findings reported.

In summary, we found evidence of spatial and spacetimechildhood cancer clustering in SF and NEF. Statistically significantcancer subtype clustering was found for leukemia and brain/CNScancer but not for lymphomas, which may be due to low statisticalpower of our study to detect smaller clusters. This evidence issuggestive of the presence of some environmental and possiblysocial conditions that may act individually or collectively topredispose children in these cluster regions to increased risk ofchildhood cancer. Further study is needed to investigate the possiblepredisposing factors in the elevated childhood cancer rates in SF andNEF.

ACKNOWLEDGMENT

This study was supported in part by Nemours Childrens Clinic,Pensacola, the Nemours Foundation and the Caitlin RobbFoundation. We also thank Gulf Coast Wings of Hope for theirsupport. We thank Brian Calkins andWendyMcLeod of FAPTP fortheir assistance in obtaining data and Dr. Elliot Daniel, ChristianaCare, DE, for the critical reading of the manuscript.

REFERENCES

1. Heron MP, Hoyert DL, Murphy SL, et al. Deaths: Final data for2006. National vital statistics reports. Hyattsville, MD: NationalCenter for Health Statistics; 2009. http://www.cdc.gov/nchs/fastats/children.htm. 57(14) Available from URL.

2. American Cancer Society. Cancer Facts and Figures2007. Available from http://www.cancer.org/downloads/STT/CAFF2007PWSecured.pdf.

3. American Cancer Society. Cancer in Children. http://www.cancer.org/docroot/CRI/content/CRI_2_4_1X_Introduction_7.asp?rnavcri (accessed 9/1/2009).

4. Strahm B, Malkin D. Hereditary cancer predisposition in children:Genetic basis and clinical implications. Int J Cancer 2006;119:20012006.

5. Pakakasama S, Tomlinson GE. Genetic predispositions andscreening in pediatric cancer. Pediatr Clin North Am 2002;49:13931413.

6. Davies SM. Subsequent malignant neoplasms in survivors ofchildhood cancer: Childhood cancer survivor study (CCSS)studies. Pediatr Blood Cancer 2007;48:727730.

7. Shuryak I, Hahnfelt P, Hlatky L, et al. A new view of radiationinduced cancer: Integrating short- and long-term processes. Part II:Second cancer risk estimation. Radiat Environ Biophys 2009;48:275286.

8. Scott D. Chromosomal radiosensitivity and low penetrance pre-disposition to cancer. Cytogenet Genome Res 2004;104:365370.

9. Taylor M, Hussain A, Urayama K, et al. The human majorhistocompatibility complex and childhood leukemia: An etiolog-ical hypothesis based on molecular mimicry. Blood Cells Mol Dis2009;42:129135.

10. Pereira TV, Rudnicki M, Pereira AC, et al. 5,10-Methylenetetrahy-drofolate reductase polymorphisms and acute lymphoblasticleukemia risk: A meta-analysis. Cancer Epidemiol BiomarkersPrev 2006;15:19561963.

11. Ogilvy-Stuart AL, Gleeson H. Cancer risk following growthhormone use in childhood: Implications for current practice. DrugSaf 2004;27:369382.

12. McNally RJ, Eden TO. An infectious aetiology for childhood acuteleukemia: A review of the evidence. Br J Haematol 2004;127:243263.

13. Buffler PA, Kwan ML, Reynolds P, et al. Environmental andgenetic risk factors for childhood leukemia: Appraising theevidence. Cancer Invest 2005;23:6075.

14. Kim AS, Eastmond DA, Preston RJ. Childhood acute lymphocyticleukemia and perspectives on risk assessment of early-life stageexposures. Mutat Res 2006;613:138160.

15. Chang JS. Parental smoking and childhood leukemia.MethodsMolBiol 2009;472:103137.

16. Dal Maso L, Bosetti C, LaVecchia C, et al. Risk for thyroid cancer:An epidemiological review focused on nutritional factors. CancerCauses Control 2009;20:7586.

17. Infante-Rivard C,Weichenthal S. Pesticides and childhood cancer:An update of Zahm and Wards 1998 review. J Toxicol EnvironHealth B Crit Rev 2007;10:8199.

18. Howard SC, Metzger ML, Wiliams JA, et al. Childhood cancerepidemiology in low-income countries. Cancer 2008;112:461472.

19. Wigle DT, Arbuckle TE, TurnerMC, et al. Epidemiologic evidenceof relationships between reproductive and child health outcomesand environmental chemical contaminants. J Toxicol EnvironHealth B Crit Rev 2008;11:373517.

20. Schieve LA, Rasmussen SA, Buck GM, et al. Are children bornafter assisted reproductive technology at increased risk for adversehealth outcomes? Obstet Gynecol 2004;103:11541163.

21. Bellec S, Hemon D, Rudant J, et al. Spatial and spacetimeclustering of childhood acute leukaemia in France from 1990 to2000: A nationwide study. Br J Cancer 2006;94:763770.

22. McNally RJ, Alexander FE, Bithell JF. Spacetime clustering ofchildhood cancer inGreat Britain: a national study, 19691993. IntJ Cancer 2006;118:28402846.



23. McNally RJ, Bithell JF, Vincent TJ, et al. Spacetime clustering ofchildhood cancer around the residence at birth. Int J Cancer2009;124:449455.

24. McNally RJ, Alexander FE, Vincent TJ, et al. Spatial clustering ofchildhood cancer inGreat Britain during the period 19691993. IntJ Cancer 2009;124:932936.

25. Lawson AB. Statistical Methods in Spatial Epidemiology. 2ndedition. New York: John Wiley & Sons; 2006.

26. Kulldorff M. InformationManagement Services, Inc. SaTScanTMv8.0: Software for the Spatial and SpaceTime Scan Statistics,2009. http://www.SaTScan.org/.

27. ThunMJ, Sinks T.Understanding cancer clusters. CACancer JClin2004;54:273280.

28. Pinheiro PS, Button JH, Fleming LE, et al. Pediatric cancer inFlorida, 19812000. Miami, FL: Florida Cancer Data System;Available from URL: http://fcds.med.miami.edu/inc/statistics.shtml.

29. KulldorffM, FeuerEJ,Miller BA, et al. Breast cancer clusters in thenortheast United States: A geographic analysis. Am J Epidemiol1997;146:161170.

30. Kulldorff M, Athas WF, Feurer EJ, et al. Evaluating clusteralarms: A spacetime scan statistic and brain cancer in LosAlamos. New Mexico: Am J Public Health; 1988. pp. 13771380.

31. Kearney G. A procedure for detecting childhood cancer clustersnear hazardous waste sites in Florida. J Environ Health 2008;70:2934.

32. Roush SW, Krischer JP, Cox MW, et al. Progress in ChildhoodCancer Care in Florida 19701992. J Fla Med Association1993;80:747751.

33. Krischer JP, Roush SW, Cox MW, et al. Using a population-basedregistry to identify patterns of care in childhood cancer in Florida.Cancer 1993;71:33313336.

34. Steliarova-Foucher E, Stiller C, Lacour B, et al. InternationalClassification of Childhood Cancer, third edition. Cancer 2005;103:14571467.

35. Knox E. Detection of clusters. In: Elliott P, editor. Methodology ofenquiries into disease clustering. Small area health statistics unit.1989.

36. KulldorffM, Nagarwalla N. Spatial disease clusters: Detection andinference. Stat Med 1995;14:799810.

37. Kulldorff M. A spatial scan statistic. Comm Stat Theory Methods1997;26:14811496.

38. Woods& Poole Economic, Inc., 1794; Columbia Road, N.W. Suite4 Washington, DC 20009-2808 . http://woodsandpoole.com/.

39. Hansen KM. Head-banging: Robust smoothing in the plane. EETrans Geosci Remote Sens 1991;29:369378.

40. Hansen KM. and Statistical Research and Applications Branch,NCI (2003). Headbang Software v3.0. http://srab.cancer.gov/headbang/.

41. Mungiole M, Pickle LW, Simonson KH. Application of a weightedhead-banging algorithm to mortality data maps. Stat Med 1999;18:32013209.

42. Hyams J. Leukemia and the clusters enigma: The towns united tofear. Australian Women Weekly 1988;Oct: 44-6.

43. Linet MS. The Leukemia: Epidemiologic aspects. New York:Oxford University Press; 1985.

44. Fraumeni JF, Jr., Miller RW. Epidemiology of human leukemia:Recent observations. J Natl Cancer Inst 1967;38:593605.

45. Wartenbeberg D, Greenberg MM. Detecting disease clusters: Theimportance of statistical power. Am J Epidemiol 1990;132:S156S166.

46. WartenbebergD,GreenbergMM. Solving the cluster puzzle: Cluesto follow and pitfalls to avoid. Statist Med 1993;12:17631770.

47. Alexander FE, Boyle P, In: editors. Methods for inves-tigating localized clustering of disease. IARC scientific publicationNo 135. Lyon: International Agency for Research onCancer; 1996.pp. 1247.

48. The Office of Economic and Demographic Research (EDR),Florida Legislature, http://edr.state.fl.us/conferences/population/demographic.htm.

49. U.S. Census Bureau, http://factfinder.census.gov/servlet/Dataset-MainPageServlet?_programACS&_submenuId&_langen&_ts.

50. Holmes L, Chan W, Jiang Z, et al. Effectiveness of androgendeprivation therapy in prolonging survival of men treated for loco-regional prostate cancer. Prostate cancer and Prostatic Dis 2007;10:388395.


518 Amin et al.

epidemiologic mapping of florida childhood cancer clusters

Documents