A New Model for Annual Cancer Incidence Repor5ng Bruce L. Riddle, Ph.D. 1,2, Alexander D. Fuld, M.D., M.S. 1 , Marc S. Ernstoff, M.D. 1 , Angeline Andrew, Ph.D., 1 Maria O. Celaya, M.P.H., C.T.R. 1,2 , GM Monawar Hosain, M.B., B.S., Ph.D. 2,3 , Judy R. Rees, B.M., B.Ch., Ph.D. 1,2 1 Geisel School of Medicine at Dartmouth College; 2 New Hampshire State Cancer Registry; 3 New Hampshire Division of Public Health Services Funding for this work was provided in part by a contract with the New Hampshire Department of Health and Human Services and funding from the CDC CooperaPve Grant UA/DP00393002. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and PrevenPon or the New Hampshire Department of Health and Human Services. Conclusions We believe that adding a wide and robust collecPon of molecular markers is an essenPal step for cancer registries to conPnue their role in tracking cancer incidence, measuring success of treatment, and developing care strategies for the populaPon based on the inherent heterogeneity of molecular markers pertaining to ePology and biological behavior.. Cancer registries need a way to collect very complex data on markers in the immediate future, and to adapt the data dicPonary rapidly as new markers are introduced into clinical pracPce. The complexiPes of molecular markers will soon present us with a vast challenge: where the reports do not provide a simple binary yes/no or the scale 1 to 4, but rather provide a set of data elements, sequences, and text summaries relaPng to the possible benefits of specific therapies. Some cancers will require mulPple markers for diagnosis and the selecPon of treatment. We also need a way to collect biomarkers at AND a\er diagnosis, for example at recurrence, because in some cases, tumors take on different molecular characterisPcs than the primary from which they originated. The conPnued presentaPon of cancer incidence and survival rates by anatomic site distorts the understanding of disease by stakeholders and the public. Cancer registries need to consider how the rise of molecular markers is about to affect our collecPon of relevant data, and our response to that challenge will determine the usefulness of cancer registry data for research, oncology, and making improvements in public health. In Table 1, examples are shown of a small number of molecular markers and the ways in which they may be used in cancer diagnosis and management. For example, predicPve markers are a criPcal step in the pathological diagnosis of disease and are directly related to the response of the targeted therapy. As we begin to understand more about the roles these tumor markers play within the cancer cell, and as new treatments are developed, the classificaPons shown above may change. Table 2 shows only a few examples of tumor markers for which targeted therapies are already in development, clinical trials, or in common use. Background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able 1. Examples of selected molecular markers used for diagnosis, predicPon (treatment), prognosis, and disease monitoring. (Adapted from: Gonzalez de Castro, et al. Personalized Cancer Medicine: Molecular DiagnosJcs, PredicJve Biomarkers, and Drug Resistance. Clinical Pharmacology & TherapeuJcs, 93(3): 252295. Used with permission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able 2. Examples of tumor biomarkers with potenPally targeted therapies Table 3. A new cancer incidence report format for 2015 It is clear that genePc data on tumors is becoming increasingly complex, and that soon registries will need to decide whether to invest in the technology and staffing to capture highly detailed tumor data not only at diagnosis but also potenPally at tumor recurrence. The heterogeneity of tumors has been recognized for decades (e.g. Howell et al, 1987. Endocrine Therapy for Advanced Carcinoma of the Breast: Effect of Tumor Heterogeneity and Site of Biopsy on the PredicJve Value of Progesterone Receptor EsJmaJons . Cancer Res January 1987 47; 296; Allred, DC et al. Ductal Carcinoma In situ and the Emergence of Diversity during Breast Cancer EvoluJon Clin Cancer Res 2008;14(2); 370). For example, the tumor markers present at diagnosis of breast cancer and in a subsequent recurrence are discordant in up to 40% of cases (reviewed in Foukakis T 2012. When to order a biopsy to characterize a metastaJc relapse in breast cancer. Ann Oncology 23 Supplement 10: x349– x353). When a primary cancer and a recurrence may have different tumor markers, a strong argument can be made for the collecPon of data on both events rather than only on the primary. An event driven data model is one possible way to collect molecular markers in a robust fashion. In this model, the diagnosis and treatment of cancer are recorded as a series of events, each with its own structure and definiPon, and common elements that allow connecPons to a master record. Table 6. Structure of an EventDriven Cancer Registry Dataset !"#$"%&' )'*+#$,-$./ 0.-'* !"#$"#% '()*)#+, +- ./,+)0 1)2-03 45),) /0) )()*)#+, +0/#,*"++)3 6"+5 )/25 )7)#+ 8-0 ("#$/%) 9:09-,), 1"*- 0"2' 3$#*- 0"2' 4$55&' 0"2' 6'7 8$#-9 )"-' :"*9'5 6.+$"& 6'+;#$-< 0;2%'# =9' 9"*9$/> .? 660 *''2* $/'@$-"%&'A B'>$*-#< )"-' .? =;2.# )$">/.*$* ;)#3"#% 1)%",+0< '()*)#+, C"-$'/- D) 0;2%'# 0EEFFB D-'2 GHI 6'/5$/> B'>$*-#< D) 0;2%'# '7)#+ '()*)#+, J@'/- D) J"+9 *,'+$?$+ '@'/- $* 5'?$/'5 K #'>$*-'#'5 L$-9 " +'/-#"& ";-9.#$-<A J@'/- !'#*$./ 0;2%'# E&& 5'?$/'5 '@'/-* 9"@' @'#*$./ /;2%'#* "* 5'?$/$-$./* +9"/>' J@'/- )"-' .-()2:(/0 ="-*/0$)0 '()*)#+, F.22./ 2"#M'# /"2' J> :JBH *-"-;* ='*- /"2' J> :JBH ,.*$-$@$-< F.2,./'/-* D22;/.9$*-.+9'2$*-#< *+.#' ND:FO PQ R'/' "2,&$?$+"-$./ N:JBHSFJCTUO 3D6: #"-$. HAT VW Introduc5on For at least 40 years, the tradiPonal cancer report has been based primarily on anatomic site; a few more detailed reports may include reference to histology. Rapid and significant changes in our understanding of the molecular biology of tumors have rendered these historical classificaPon schemes outdated. It is clear that there is heterogeneity within the cancers of a specific anatomic site, and recent data demonstrate that the clinically tractable molecular features may actually be similar across tumors of different organ sites. Molecular markers allow cancers to be categorized into subtypes of the disease that may reflect different molecular underpinnings and ePologies. These molecular markers are being used to assign tailored treatment strategies and define management opPons. Molecular markers are being invesPgated as a tool to beder define prognosis too. Clinicians now rouPnely test cancers for an increasing number of relevant markers, well beyond the bederknown molecular markers such as ER, PR and HER2 for breast cancer. TradiPonal cancer incidence reports do not typically incorporate these rapidly changing yet important biomarker data. To remain relevant, cancer reporPng by the cancer surveillance community needs to align itself with the clinical science by collecPng and reporPng data on all the relevant factors defining the disease ePology, management, and prognosis. IncorporaPon of this informaPon into registries will provide more complete and robust data that subsequently will be highly useful to the medical community and research funding agencies. SubclassificaPon of tumors by molecular markers is especially important as the surveillance community begins to report on prognosis and survival. Because the tradiPonal cancer incidence report has changed relaPvely lidle over several decades, it remains a useful tool to show some trends over Pme. However, there may be advantages to supplemenPng this format with tumor subclassificaPons that reflect the evolving clinical paradigms in cancer care. We propose a new model for an annual incidence report, which will supplement anatomic site with tumor markers. Because cancerrelated data are changing extremely rapidly, so must the incidence report and registry required variables must keep pace with these changes. We recognize that a new model (as well as the underlying data collected by Cancer Registries) will be dynamic as new observaPons will need to be incorporated in “real Pme” manner. • +(,)&-'. *')" ,(/ 0'*)&1&2# ,3" 4".&-'(2 , 1"** /&-'(,() .&-5&("() 6&3 .1,**'6'.,)'&(* &6 '(.'/"() .,(."37 • 80" '-5&3),(." &6 )9-&3 4'&-,3:"3* 6&3 /',2(&*'*; )3",)-"(); 53&2(&*'*; ,(/ /'*",*" -&(')&3'(2 '* '(.3",*'(2 3,5'/1# • <=">? )9-&3 4'&-,3:"3* ,3" 4"'(2 '/"()'6'"/ ,11 )0" )'-"@ )0" 6'"1/ '* .0,(2'(2 A"3# B9'.:1# • C,(# )9-&3 4'&-,3:"3* ,3" .&--&( )& .,(."3* '( /'66"3"() *')"* "727 DEFG@ 43",*) H 2,*)3'. .,(."3 • I) >'11 4" '-5&3),() )& .0,3,.)"3'J" )0" /'*)3'49)'&(* &6 )9-&3 4'&-,3:"3* '( 5&591,)'&(* • +* 3".933"() )9-&3* <"A&1A"? 63&- )0" 53'-,3#; 53'-,3# ,(/ 3".933"() )9-&3* -,# 0,A" /'*.&3/,() )9-&3 4'&-,3:"3*@ ,( ,329-"() .,( 4" -,/" 6&3 .&11".)'(2 /,), &( 4&)0 )#5"* &6 )9-&3 "A"( '( ,( '(.'/"(." .,(."3 3"2'*)3# • 80" .,(."3 3"2'*)3# .&--9(')# *0&91/ 3"*5&(/ 3,5'/1# )& .0,(2"* '( )9-&3 .1,**'6'.,)'&(; 4&)0 '( )"3-* &6 )0" /,), >" .&11".); ,(/ )0" /,), >" 3"5&3) Key Points The Accelera5ng Pace & Promise of Personalized Medicine In Table 3, we take some currently known molecular biomarkers used in diagnosis and treatment to breakdown anatomic sites into sub classificaPons. Not all anatomic sites have associated molecular markers while other molecular markers appears across sites (See Table 2, e.g. HER2, BRAF, KRAS, NRAS) and sPll other relevant factors relate to viral infecPons such as Human Papilloma Virus (HPV), HepaPPs C Virus (HCV) and Epstein Barr virus (EBV). So while this table is based on anatomic site, we can envision a new table(s) with increased emphasis on specific molecular biomarkers. We acknowledge that this report is only possible if the underlying data are rouPnely collected by cancer registries. In Table 4, we break down cancers of the breast by ER, PR, and HER2 using New Hampshire data 20102012. This table brings into focus several issues. (1) ER, PR and HER2 represent only a Pny fracPon of the tumor markers that are currently being invesPgated in breast cancer research, many of which may rapidly be brought into rouPne clinical pracPce. (2) MulPple combinaPons of tumor markers may be needed to break down an anatomic site. (3) The tumor markers or combinaPons of markers are unlikely to be evenly distributed. Small counts of some markers may make reporPng difficult, yet sPll be potenPally important in determining treatment opPons. (4) Changes in tumor marker tesPng are taking place at different speeds in different healthcare seings. To keep the cuing edge in sight, cancer registry variables must be updated more quickly, and decisions must be taken on how to collect very complex genePc data. In the future it seems likely that personalized medicine will involve the genePc characterizaPon not only of tumor markers, but also of the paPents themselves, in an effort to define those who will respond best to specific treatments. Table 4. CollaboraPve Stage Factor 16 201012 NH State Cancer Registry November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	 :2%+2 :%)) ),#$ 5. &' $,+%$, :2,&2," &' 0#..%;,)4 ,8-#($ ",3%.&"4 $#&#/#.,. 7(+",#.%(3 #5&'0#&%'( '< 0'),+5)#" +2#"#+&,"%=#&%'( '< /'&2 &50'" 1 -#&%,(&9 :%&2 ",.5)&%(3 .%0-)%<%+#&%'( '< +#(+," ",+'"$. >8#0-), ",-'"&? >8#0-), ",-'"&? >8#0-), ",-'"&? >8#0-), ",-'"&? @')'(? A$,('+#"+%('0# B&2," C,+&50? A$,('+#"+%('0# B&2," DCAE FGH ICAJ FGH KE7 FGH B&2," !50'" /%'0#"6,".? E,L5,(+, M? .+'", '" FGH E,L5,(+, N? .+'", '" FGH E,L5,(+, O? .+'", '" FGH ,&+ P%"#) <#+&'".? P%"#) M? FGH P%"#) N? FGH ,&+ !50'" -"'<%), <'" &",#&0,(& ",.-'(., A3,(& M? A3,(& N? ,&+ Q#&%,(& -"'<%), <'" &",#&0,(& ",.-'(., Q"'<%), M? Q"'<%), N? ,&+ Q#&%,(& -"'<%), <'" &",#&0,(& &'),"#(+, Q"'<%), M? Q"'<%), N? ,&+ Q#&%,(& -"'<%), <'" -"'3('.%. Q"'<%), M? Q"'<%), N? ,&+ B;,"#)) &",#&0,(& .5%&#/%)%&4 .+'",? A3,(& M? .+'", '" FGH A3,(& N? .+'", '" FGH >&+ !50'"H-#&%,(&H&",#&0,(& .500#"4 -"'<%),? Q"'<%), M Q"'<%), N Q"'<%), ORS Q"'<%), TUVO Q"'<%), TUVW ,&+ In Table 5, we speculate about the direcPon in which cancer classificaPon will be recorded in cancer registries of the future (Table 5). The first column (2005) represents our tradiPonal cancer incidence report based on site and histology. The second (2015) reflects recent changes in our understanding of tumor markers, including those shown in Table 3. The third column (2025) reflects our hypotheses that (1) tumors will be fully sequenced and acPonable sequences will be reported; (2) more extensive infecPous disease tesPng may become rouPne for some cancers e.g. Human Papilloma Virus, Epstein Barr Virus; (3) paPent characterisPcs will be assessed to idenPfy innate characterisPcs that predict response to and tolerance to treatment agents; and that (4) treatment agents will be ranked in a measurable way for individual paPents. Within the next decade, we anPcipate seeing a huge increase in the volume of data that cancer registries could collect. Further, we need internal data from molecular markers reports rather than just binary summaries because definiPons and ranges change over Pme. In the fourth column (2035), we hypothesize that the extensive informaPon collected in the preceding decade will be assimilated using genePc and treatment algorithms, and provided back to the oncologist along with simple summary paPent and tumor profiles. In other words, things will get beder a\er they get worse. However, to reach the simplified model, the registry community needs to contribute to the understanding of tumor markers in the context of cancer management, and in doing so, we must decide how much complex genePc data are needed in the cancer registry database. It is difficult to envision the complexity and size of a populaPonbased registry that collected the key genePc sequences within incident tumors. Table 5. Progression of possible cancer registry classificaPons for cancer incidence reporPng, by decade into the future: example, colorectal cancer Model Image courtesy of the NaPonal Human Genome Research InsPtute's Talking Glossary (hdp://www.genome.gov/glossary/). hdp://geneed.nlm.nih.gov/topic_subtopic.php?Pd=15&sid=16 The Double Helix Trastuzumab (Breast CA: Her2/neu) Depicts date of FDA approval for targeted cancer drugs, the relevant disease and acPonable biomarker & select milestone dates. Ima5nib (Chronic Myelogenous Leukemia (CML): BcrAbl) Human Genome Sequenced Erlo5nib (Lung CA: EGFR) Dasa5nib (CML BcrAbl) Nilo5nib (CML BcrAbl) Mammaprint 70 gene assay approved for Breast CA Risk recurrence Lapa5nib (Breast CA: Her2/neu) Cetuximab (Colorectal/Head & neck CA: EGFR, KRAS wild type) Crizo5nib (Lung CA: ALK) Vemurafenib (Melanoma: BRAF) Ima5nib (GI Stromal Tumors: ckit) Pertuzumab (Breast CA: Her2/neu) Ceri5nib (Lung CA: ALK) 1998 2000 2002 2004 2006 2008 2010 2012 2014 Dabrafenib (Melanoma: BRAF) Tramatenib (Melanoma: BRAF) Guanine Thymine Base Pair Sugar Phosphate Backbone Adenine Cytosine