Nomograms provide predictions of outcomes for prostate cancer patients based on known treatment outcomes of similar patients. However, nomograms have several limitations including bias from the development cohort, lack of external validation, and lack of updates using contemporary patient populations. Additionally, nomograms often use surrogate endpoints rather than clinically meaningful endpoints and predictive accuracy is not 100%. While nomograms can help guide clinical decision making, good clinical judgement is still needed and nomograms may not accurately capture all risk factors or change clinical decisions for individual patients.

1. Do I need a NOMOGRAM to make a good decision ? Hein Van Poppel, MD,PhD Steven Joniau, MDLeuven, BelgiumSecondInterdisciplinary ConferenceMSKCC – ESO New York, 9-4-2011

2. In 2008 about 33.000 man died of PCa in N.Am., and 90.000 in EuropeHowcan we do better?

3. Do I need a NOMOGRAM to make a good decision ? “The answer is no” Hein Van Poppel“The answer is yes” Michael Kattan

5. Situation in the pastFor decades, staging and prognosticationinvolvedwell-knownprognostic factors:cTNMGleason scorePSAHowever, these parameters provide ratherlimitedinformationwhenconsideredseparately

6. Whydecision aids?Physicianjudgement is the basis for risk estimation, patient counseling, and decisionmakingThisjudgementmaybebiasedbecause of subjective and objectiveconfoundersDevelopment of predictive and prognostic toolsRecent explosion in the field of PCadecision aidsShariat SF et al. Cancer 2008;113:3075-99

7. NomogramsCurrently the most accurate tool to predictoutcomes in PCapatientsBasedonknowntreatmentoutcomesfor a groupwithsufficientsimilarities to the patient, and having been treated in the samefashion, a number of yearsago

8. Whatcan we usenomogramsfor?1. Pre-biopsyprostate cancer risk calculation2. LocalPCastagingpost-biopsyEstimation of pTNM3. Pre-treatmentestimation of outcomeBiochemicalrelapseClinicalrelapseDeathfrom prostate cancer

9. Whatcan we usenomogramsfor?4. Post-treatmentestimation of outcomeBiochemicalrelapseClinicalrelapseDeathfrom prostate cancer5. Predictionof local / systemicfailurewhenbiochemicalrelapse6. Predictionof death in HRPC

10. Predictiveaccuracy of existingnomograms Chun F et al. World J Urol 2007;25:131-42

11. Limitations of nomogramsBias due to development cohortOftenbasedonsingle-centre series and/or data fromtertiary care centresRetrospectivestatisticalapproachDespiteprospective data collectionSpecific model selection criteriaModel selection criteria excludecertainsubgroups, e.g. patientswho had neoadjuvant HT are excluded in most models Lack of externalvalidationChun F et al. World J Urol 2007;25:131-42

12. Limitations of nomogramsImply a concept of stability:No change, noevolution in surgerytechniques and methods,…. impossible to consider recent improvements in technique, knowledge in tumourbiology, and disease characteristics1Lack of periodic updates in contemporarycohortsDevelopment in non-contemporarysituations = inaccurate predictions in contemporary patients2Stage migration/ screen detectedpopulationsChange in diagnostic and therapeuticstandardsFor example: sextant biopsies vs. 10-12 core biopsiesType and dose of Radiotherapy, Surgical techniquesGuillonneau B. EurUrol2007 2.Chun F et al. World J Urol2007

13. Limitations of nomogramsSurrogateendpoints in most nomogramsPathologic stage predictionBiochemicalrecurrenceLack of “hard endpoint” nomogramsRequiredendpoints:Local and distantrecurrenceDisease- specific and overall survivalCorrect and long-term follow-up and competingcomorbidityanalysisChun F et al. World J Urol 2007;25:131-42

14. Summary: limitations of nomogramsBias due to development cohortHigh volume, tertiary care centersRetrospectivestatisticalmethodologySpecific model selection criteriaLack of externalvalidationLack of periodic updates in contemporarycohortsLack of novel more specific markersSurrogate versus clinicallymeaningfulendpointsPredictiveaccuracynot 100%

15. Howmany more nomograms do we need?Confusingvariety of available toolsHow do we decidewhich (ifany) nomogram to use in clinicalpractice?PCa metagram (Nuygen and Kattan 2009)frameworkonwhichexistingPCaprediction tools are organised and stratifiedbyaccuracy, quality and usefulness in a clinical settingadditionalnomograms are needed …….NguyenCT and Kattan MW. Cancer 2009

16. Thismay all beveryinteresting, BUT…Do these predictionsreallychangeyourclinicaljudgement?Will theyinfluenceyourdecision-makingprocess?Are these predictionsreallyhelpfulforyourindividualpatient?

17. Questions to Michael Kattanabout a givenpatientShould weomit a biopsy…?Should weadvocate Active Surveillance…?Shall we not go for a RPr…?Can we safelyomit a LND…?Do we give give adjuvant therapy to all…?Will our patient appreciate an estimation of the chance of his time to failure, or to death…?

19. No proofonadvantage of use of nomogramsUse of nomograms has notyet been implementedsufficientlyinto routine urologicalpracticeStudies providingevidence-basedproofon the advantage of usingnomograms over clinicaljudgement are virtually ABSENTNo nomogramwill ever take the place of goodclinicaljudgement and information to the patient

20. How to improve PAPredictiveAccuracy (PA) is suboptimal - Novelbiomarkersassociatedwithbiologicbehaviour8% increase in PA withinclusion of IL-6 and TGF-β11Plasminogen activator inhibitor type 1, humanglandular kallicrein-2, gene expressionsignatures, plasma endoglin, …- Larger datasets and systematic and clean data collectionMore sophisticated modeling procedures2Kattan MW et al. J ClinOncol 2003;21:3573-9Shariat SF et al. Cancer 2008; 113:3075-99

21. Whatis the future?Ultimately, improvedimagingstudies and high-throughputgenomicswillreplace the use of nomograms, as theywill provide a realpatient-specificstaging and prognostication, and allowpatient-tailoredtreatmentdecisionsGenomicSignaturesforPersonalisedTherapy Mammaprint, Coloprint, “Prostaprint “?

22. Do patientsneednomograms ? No, ifthey are whatthey are today, onlyprovidingestimations of a chance, whileforthemit is rather a black or white scenarioPatientsneed to befullyinformed at all decision- making stepsTheyneed to find centers where experts are willing to provide thisinformation

23. The experts canorcannotusenomograms to this endI do notneed the actuallyavailablenomograms …to make a gooddecisionApril 2011

26. BACK - UP26

29. 29Prostate cancer: Decision aidsUnivariate and multivariableanalysisRisk groupingsProbability tablesArtificialneuralnetworks (ANN)Classification and regression tree (CART) analysisNomogramsShariat SF et al. Cancer 2008;113:3075-99

30. 30Prostate cancer: Evaluatingpredictive toolsPredictiveaccuracy of the modelInternal and externalvalidation to ensuregeneralizabilityModel calibrationLevel of complexicityClinicalimplicationHead-to-headcomparisonsShariat SF et al. Cancer 2008;113:3075-99Capitanio U et al. The Prostate 2010;70:1371-78

31. Definition of nomogramStatisticaldefinitionGraphicalrepresentation of a mathematicalformulaoralgorithmIncorporatingseveralpredictors modeled as continuous variables To predict a particularendpointUsing traditional statisticalmethods - Multivariablelogisticregression - Cox proportional hazard analysis

32. A smallexercise…Patient XXPSA 6.4cT2aBiopsyGleason score 6 in 2 cores

33. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion?5% 10% 15% 20%pN+?2% 4% 6% 8%

34. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion?5% 10% 15% 20%pN+?2% 4% 6% 8%

35. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion? 5% 10% 15% 20%pN+?2% 4% 6% 8%

36. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion?5% 10% 15% 20%pN+?2% 4% 6% 8%

37. What is the risk of biochemicalrelapseaftersurgery… 25% 50% 75% 100%

38. What is the risk of biochemicalrelapseaftersurgery…25% 50% 75% 100%

39. Anothersmallexercise…Patient YYPSA 8.6cT2cBiopsyGleason score 7 on 2 biopsies

40. What is the risk of…ECE?20% 40% 60% 80%SVI invasion? 5% 10% 15% 20%pN+?2% 4% 6% 8%

41. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion?5% 10% 15% 20%pN+? 2% 4% 6% 8%

42. What is the risk of…ECE? 20% 40% 60% 80%SVI invasion?5% 10% 15% 20%pN+? 2% 4% 6% 8%

43. What is the risk of…ECE?20% 40% 60% 80%SVI invasion? 5% 10% 15% 20%pN+?2% 4% 6% 8%

44. What is the risk of biochemicalrelapseaftersurgery…25% 50% 75% 100%

45. What is the risk of biochemicalrelapseaftersurgery… 25% 50% 75% 100%

46. The American World

47. The Real World

48. BRUSSSELSBELGIUMBelgiëBelgiqueBelgienLEUVEN

49. Nomogramtopredictlowvolumeinsignificantprostatecancer (n=258)58 years30 gramsPSA 3,04mm tumorLow-volume/low-grade cancer was defined as pathologic organ-confined disease and a tumor volume < 0.5 cc with no Gleason grade 4 or 5 cancer. Nakanishi et al., Cancer 2007

50. Prediction of biopsyoutcomeKarakiewicz PI and Hutterer GC.Nat Clin Pract Urol 2008;5:82–92Karakiewicz PI and Hutterer GC.Nat ClinPractUrol 2008;5: 82–92

51. Prediction of pathological features clinically localized PCa(before treatment)Karakiewicz PI and Hutterer GC.Nat ClinPractUrol 2008;5: 82–92

52. Prediction of pathological features clinically localized PCa(before treatment)Karakiewicz PI and Hutterer GC.Nat ClinPractUrol 2008;5: 82–92

53. Prediction of biochemical recurrence with preoperative variablesKarakiewicz PI and Hutterer GC.Nat ClinPractUrol 2008;5: 82–92

54. PCa metagramPCa metagram is constructed of 16 different treatmentoptions10 outcomesrelated to cancercontrol, survival and morbidity160 treatment/outcomecombinationsOnly 31 cells are populatedwithavailable toolsAreas of deficiency in the currentcatalog of prediction toolsNguyen CT and Kattan MW. Cancer 2009;115(Suppl 13):3160-2

55. PCa metagramData willbeincorporatedinto a software programPhysicianwill enter patient-specific variableswillgenerategraphical and tabularpresentation of predictions of treatmentendpoints, with all availablealternativestailored to the individualpatientLimitations of the metagramNot all cells are populated  more tools are neededLack of prediction tools for LRP, cryoablation and HIFUSurvival and morbidityoutcomes are poorlyrepresentedAdditionalprediction tools assessing risk of metastasis and cancer-specificmortality are neededNguyen CT and Kattan MW. Cancer 2009;115(Suppl 13):3160-2

56. PCa metagram

57. What to do to implement the use of nomograms?Update nomograms to contemporarypatientpopulationsNovelbiomarkers to improvepredictionsHead-toheadcomparisonsbetweennomogramsto select the best-suited model in selectedfields of PCaoutcomesWe neednomogramsthatprovide accurate predictions of hard clinicalendpoints (clinicalfailure, deathfrom the disease)accuratelypredictdeathfromcomorbiddisease in men withlocalizeddiseaseselectedforradicaltreatmentpredicttreatment-relatedtoxicityEfforts are needed to improve the accuracy, accessibility and flexibility of nomograms and to provide more evidence to justifytheir routine use in clinical practice1Lughezzani G et al. EurUrol 2010;58:687-700

58. Whymightnomogramsbeimplementedintoclinicalpractice?Appropriatepatient counseling / decision-making ?Betterdiseaseprognostication ?Follow-up monitoring ?Selection of patientsforclinical trials !

59. Limitations of nomogramsSuboptimalpredictiveaccuracyNomogramprediction is not 100% accurateLack of consideration of all predictive risk factorsInability to assemble all knownprognostic factors optimally1To improvepredictiveaccuracy (PA) we needNovelbiomarkersassociatedwith the biologicbehaviour of PCa8% increase in PA withinclusion of IL-6 and TGF-β12Plasminogen activator inhibitor type 1, humanglandular kallicrein-2, gene expressionsignatures, plasma endoglin, …Larger datasets and systematic and clean data collectionMore sophisticated modeling procedures3Chun F et al. World J Urol 2007;25:131-42Kattan MW et al. J ClinOncol 2003;21:3573-9Shariat SF et al. Cancer 2008; 113:3075-99

60. EAU PCaguidelinesUse of nomograms is onlyincluded 3 timesPreoperativestaging(Kattannomogram, Partin tables (= look-up tables)Indication of extendedlymph node dissection(Briganti nomogram)Indication of nerve-sparingsurgery(Partin tables)