1. Studying the Risk of Cancer with Topical Calcineurin Inhibitor Use in Children: Design Issues Lois La Grenade, MD, MPH Center for Drug Evaluation & Research Office of Drug Safety Lois La Grenade, MD, MPH Center for Drug Evaluation & Research Office of Drug Safety

2. 2 OutlineOutline Observational studies - general methods Methods for risk of cancer with long term use of calcineurin inhibitors (CI) in children with atopic dermatitis (AD) Observational studies - general methods Methods for risk of cancer with long term use of calcineurin inhibitors (CI) in children with atopic dermatitis (AD)

3. 3 Epidemiologic Methods Observational Studies –Case Control –Cohort –Registries Observational Studies –Case Control –Cohort –Registries

4. 4 Case Control Studies Retrospective –Start with disease of interest (cases), –Compare with people without disease (controls) –Compare frequency of the exposure of interest between cases & controls Retrospective –Start with disease of interest (cases), –Compare with people without disease (controls) –Compare frequency of the exposure of interest between cases & controls

5. 5 Case Control Studies - Advantages Less expensive Relatively quick Generally useful for studying most rare events, with common exposures Less expensive Relatively quick Generally useful for studying most rare events, with common exposures

6. 6 Case Control Studies - Disadvantages Subject to several biases –Recall –Selection Unsuitable for studying diseases with very long latency periods, e.g. cancer Difficult for rare exposures Subject to several biases –Recall –Selection Unsuitable for studying diseases with very long latency periods, e.g. cancer Difficult for rare exposures

7. 7 Cohort Studies Compare exposed vs. non-exposed persons Start with a defined group of people (defined by exposure, disease, place of residence etc.) Followed through time for occurrence of disease(s) of interest Prospective Compare exposed vs. non-exposed persons Start with a defined group of people (defined by exposure, disease, place of residence etc.) Followed through time for occurrence of disease(s) of interest Prospective

8. 8 Cohort Studies - Advantages Cases & exposure determined prospectively, so recall bias minimized All cases can potentially be captured, so selection bias reduced Can study several outcomes or diseases Most closely resemble experimental design – toxin administered, then follow for outcome Consequent high acceptance by scientific community Cases & exposure determined prospectively, so recall bias minimized All cases can potentially be captured, so selection bias reduced Can study several outcomes or diseases Most closely resemble experimental design – toxin administered, then follow for outcome Consequent high acceptance by scientific community

9. 9 Cohort Studies - Disadvantages More expensive Large sample sizes for rare diseases Long Problems from losses to follow-up More expensive Large sample sizes for rare diseases Long Problems from losses to follow-up

10. 10 Retrospective Cohort Studies Can be used to overcome disadvantages of prospective cohorts viz. length Use a preexisting cohort, e.g. occupational cohort, cohort exposed to a drug of interest Instead of comparison controls, can compare disease with population incidence rates (Standardized Incidence Ratio - SIR) Can be used to overcome disadvantages of prospective cohorts viz. length Use a preexisting cohort, e.g. occupational cohort, cohort exposed to a drug of interest Instead of comparison controls, can compare disease with population incidence rates (Standardized Incidence Ratio - SIR)

11. 11RegistriesRegistries Exposure based – occupational, drug exposures Disease based – State & National Cancer Registries Complete (usually mandatory) – all subjects with exposure/disease captured Incomplete (usually voluntary) – only some cases/exposed persons captured Exposure based – occupational, drug exposures Disease based – State & National Cancer Registries Complete (usually mandatory) – all subjects with exposure/disease captured Incomplete (usually voluntary) – only some cases/exposed persons captured

12. 12 Uses of Registries in Epidemiology Exposure registries – cases ascertained in exposure cohort e.g. cohort of drug exposed people Case based registries – as source of cases for case control study Complete registries generally far more useful –Can be used to determine incidence rates when based in defined population –Incidence of rare events with rare exposures Exposure registries – cases ascertained in exposure cohort e.g. cohort of drug exposed people Case based registries – as source of cases for case control study Complete registries generally far more useful –Can be used to determine incidence rates when based in defined population –Incidence of rare events with rare exposures

13. 13 Topical Calcineurins & atopic dermatitis in children

14. 14 Cancer Risk with Calcineurin Inhibitor (CI) Use in Children with AD Special problems with cancer studies –Rare events, particularly in the young –Very long latency, i.e. many years between exposure & clinically apparent cancer Prospective method (cohort or registry) ideal Special problems with cancer studies –Rare events, particularly in the young –Very long latency, i.e. many years between exposure & clinically apparent cancer Prospective method (cohort or registry) ideal

15. 15 Cohort Design Prospective –Exposure assessment, accurate, standardized Dose, duration of topical CI can be assessed –Case ascertainment, complete, accurate –Data on confounding/risk factors Prospective –Exposure assessment, accurate, standardized Dose, duration of topical CI can be assessed –Case ascertainment, complete, accurate –Data on confounding/risk factors

16. 16 Cohort Study - Indication There is good evidence of an association between a disease & an exposure –Clinical studies –Case control studies –Other studies e.g. animal studies* Introduction of a new agent that requires monitoring for its possible association with several diseases** There is good evidence of an association between a disease & an exposure –Clinical studies –Case control studies –Other studies e.g. animal studies* Introduction of a new agent that requires monitoring for its possible association with several diseases** *Gordis, Epidemiology, 1996 **Kelsey et al, Methods in Observational Epidemiology,1996

17. 17 “International Monitoring of Adverse Drug Reactions of Long Latency” Fletcher AP, Griffin JP Adverse Drug React. Toxicol. Rev. 1991, 10 (4) 209 – 230 “for adverse reactions of long latency to be detected methods have to be used that permit observation of the patient to be followed for many months or years…..An essential requirement is the identification of a cohort of patients exposed to a particular drug ….who may be accessed at specified intervals of time.” Fletcher AP, Griffin JP Adverse Drug React. Toxicol. Rev. 1991, 10 (4) 209 – 230 “for adverse reactions of long latency to be detected methods have to be used that permit observation of the patient to be followed for many months or years…..An essential requirement is the identification of a cohort of patients exposed to a particular drug ….who may be accessed at specified intervals of time.”

18. 18 Cohort Study Design - Cancer Risk with Calcineurin Inhibitor (CI) Use in Children with AD Background Objectives Population Sample Size/Power Exposure definition & assessment Endpoint (cancers) definition & ascertainment Measures to reduce bias – losses to follow-up Analysis Plan Background Objectives Population Sample Size/Power Exposure definition & assessment Endpoint (cancers) definition & ascertainment Measures to reduce bias – losses to follow-up Analysis Plan

19. 19BackgroundBackground Enhanced photocarcinogenicity with both topical CI vehicles Carcinogenicity studies – signal for lymphoma & other systemic malignancies Oral CI use associated with ↑ risk of lymphoma, cutaneous & other malignancies Enhanced photocarcinogenicity with both topical CI vehicles Carcinogenicity studies – signal for lymphoma & other systemic malignancies Oral CI use associated with ↑ risk of lymphoma, cutaneous & other malignancies

20. 20 ObjectiveObjective To investigate the risk of developing cutaneous & systemic malignancies in children with atopic dermatitis (AD) who have long term intermittent treatment with topical CIs

21. 21 Outcomes of Interest Cutaneous malignancies –Melanoma –Non-melanoma (basal & squamous) Systemic –Lymphoma (Hodgkin’s, NHL) –Others Additional cutaneous end points, e.g. actinic keratoses? Cutaneous malignancies –Melanoma –Non-melanoma (basal & squamous) Systemic –Lymphoma (Hodgkin’s, NHL) –Others Additional cutaneous end points, e.g. actinic keratoses?

22. 22PopulationPopulation Traditional study –Cohort of children (aged 2 - 16 years) with AD –Follow for 10 – 15 years –Document treatment type, response, confounding factors e.g. sunlight exposure, skin type, disease severity –Document occurrence of malignancies as they occur At end of follow up, compare incidence in CI treated vs. non-CI treated group Traditional study –Cohort of children (aged 2 - 16 years) with AD –Follow for 10 – 15 years –Document treatment type, response, confounding factors e.g. sunlight exposure, skin type, disease severity –Document occurrence of malignancies as they occur At end of follow up, compare incidence in CI treated vs. non-CI treated group

23. 23 Cohort - Population Traditional study - difficulties –Very long –Very large sample size –May find that most patients have used both CI and non-CI treatment regimens (or vice versa) and comparison may be difficult Traditional study - difficulties –Very long –Very large sample size –May find that most patients have used both CI and non-CI treatment regimens (or vice versa) and comparison may be difficult

24. 24 Cohort – Population (Alternate method) Cohort of CI users (aged 2 – 16 years) with AD –Follow subjects for minimum 10 year period –Use age specific population incidence rates for cancer as comparator –(Standardized Incidence Ratio, SIR) –Similar to occupational epidemiology methods Cohort of CI users (aged 2 – 16 years) with AD –Follow subjects for minimum 10 year period –Use age specific population incidence rates for cancer as comparator –(Standardized Incidence Ratio, SIR) –Similar to occupational epidemiology methods

25. 25 SIR Approach - Difficulties No US national incidence rates for most cutaneous malignancies May have to extrapolate from data from other countries, e.g. Finland, or from US local data e.g. Southeastern Arizona No US national incidence rates for most cutaneous malignancies May have to extrapolate from data from other countries, e.g. Finland, or from US local data e.g. Southeastern Arizona

26. 26 *US Age Specific Incidence of all Malignancies per 100,000, by gender *Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) SEER*Stat Database: Incidence - SEER 9 Regs Public-Use, Nov 2002 Sub (1973-2000), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2003, based on the November 2002 submission.

27. 27 Age Specific Incidence of SCC per 100,000, by gender, Southeastern Arizona* *Harris RB et al, 2001, JAAD (43):4, 528-536 Males Females

28. 28 Age Specific Incidence of BCC per 100,000, by gender, Southeastern Arizona* *Harris RB et al, 2001, JAAD (43):4, 528-536 Males Females

29. 29 Power/Sample Size Calculations* Background Rate/10,000 pop RR# in each group Total 6310,30020,600 64600012000 6540008000 *Background rate 6/10,000 represents annual incidence rate for all cancers in 25 - 29 age group, SEER Cancer Statistics Review 1973-1999 Power 80%,  0.05

30. 30 Power & Sample Size Calculations Background Rate* RR# in Group 2/10,000262,000 414,000 68,000 86,000 *Annual incidence rate for all cancers in 15 - 19 age group, SEER Cancer Statistics Review 1975-2000 Power 80%,  0.05

31. 31 Power & Sample Size Calculations Background Rate* 24.1/million RR# in Group 2514,000 4115,000 662,000 842,000 1032,000 *Annual incidence rate for lymphoma in 0 - 19 age group, SEER Cancer Statistics Review 1973-1999 Power 80%,  0.05

32. 32 Probability of finding no event with defined sample sizes – Lymphoma* N = 10,000N = 20,000 RRProbability (%)RRProbability (%) 261238 4 415 62466 81582 109 1 *Incidence 24.1/million, Age 0-19 years, *Incidence 24.1/million, Age 0-19 years, SEER Cancer Statistics Review 1973-1999

33. 33 Probability of finding no event with defined sample sizes N=10,000 N=20,000 *Lymphoma incidence 24.1/million, Age 0-19 years

34. 34 Sample Size/Power Multicenter, multinational cohort would boost sample size and power considerably Use of additional end points, e.g. AK, might also increase power Multicenter, multinational cohort would boost sample size and power considerably Use of additional end points, e.g. AK, might also increase power

35. 35 AK as additional End Point Actinic Keratoses (AK) – precancerous skin lesions, proliferation of abnormal keratinocytes within epidermis –SCC in situ –may become SCC (60% of all SCC may start as AK*) –Very rare in young people Actinic Keratoses (AK) – precancerous skin lesions, proliferation of abnormal keratinocytes within epidermis –SCC in situ –may become SCC (60% of all SCC may start as AK*) –Very rare in young people * Fu W & Cockerell C, 2003 Arch Dermatol,139:66-70

36. 36 Exposure Definition Need to define “long term intermittent exposure” to CIs –At enrollment Suggestions –6 weeks/3 months continuous or intermittent use –30 GM over 6 weeks Need to define “long term intermittent exposure” to CIs –At enrollment Suggestions –6 weeks/3 months continuous or intermittent use –30 GM over 6 weeks

37. 37 Exposure Assessment (Enrollment & During Study) Possibilities –Prescription + Self report by care giver –Return unused portions of tubes –Weigh unused tubes/portions Consider additional burden on participants (losses to follow-up) vs. more accurate information Possibilities –Prescription + Self report by care giver –Return unused portions of tubes –Weigh unused tubes/portions Consider additional burden on participants (losses to follow-up) vs. more accurate information

38. 38 Ascertainment of Malignancies Pathologic/histologic definitions ICD codes Systemic malignancies could be ascertained by linkage with state/national cancer registries, once a unique identifier was obtained for each subject at baseline But not true for most cutaneous malignancies Self reporting of cutaneous malignancies unreliable Pathologic/histologic definitions ICD codes Systemic malignancies could be ascertained by linkage with state/national cancer registries, once a unique identifier was obtained for each subject at baseline But not true for most cutaneous malignancies Self reporting of cutaneous malignancies unreliable

39. 39 Ascertainment of Cutaneous Malignancies Most state & national cancer registries collect limited data on skin malignancies (usually data on invasive melanomas only) Non-melanoma skin cancers (basal & squamous cell cancers) often treated in office procedure Basal cell cancers (BCC) & AK may be treated with cryotherapy or electrocautery, without even histology Most state & national cancer registries collect limited data on skin malignancies (usually data on invasive melanomas only) Non-melanoma skin cancers (basal & squamous cell cancers) often treated in office procedure Basal cell cancers (BCC) & AK may be treated with cryotherapy or electrocautery, without even histology

40. 40 Ascertainment of Cutaneous Malignancies Best done by periodic (annual?) physical examination (PE) of the skin by physician, preferably a dermatologist PE particularly important for good capture of all cutaneous outcomes & hence early, accurate assessment of the risk Best done by periodic (annual?) physical examination (PE) of the skin by physician, preferably a dermatologist PE particularly important for good capture of all cutaneous outcomes & hence early, accurate assessment of the risk

41. 41 Follow-upFollow-up Duration –Minimum 10 years for each subject, in keeping with long latent period for cancers Duration –Minimum 10 years for each subject, in keeping with long latent period for cancers

42. 42 Minimizing Losses to Follow-up Losses to follow-up –Important source of bias in cohort studies of long duration So vigorous methods will have to be used to reduce losses to follow-up Statistical methods to handle losses to follow-up Losses to follow-up –Important source of bias in cohort studies of long duration So vigorous methods will have to be used to reduce losses to follow-up Statistical methods to handle losses to follow-up

43. 43 Statistical Analysis Plan Crude & adjusted incidence rate, SIR Methods for handling losses to follow-up Crude & adjusted incidence rate, SIR Methods for handling losses to follow-up

44. 44 Registry - Cancer Risk with Calcineurin Inhibitor (CI) Use in Children with AD Ideal: Exposure registry Registration of all users of CIs (probably mandatory), prospective collection of unique identifier (e.g. SSN) Link to state/national cancer registries for detecting systemic malignancies Would still need periodic examination by physician to ascertain skin cancers Ideal: Exposure registry Registration of all users of CIs (probably mandatory), prospective collection of unique identifier (e.g. SSN) Link to state/national cancer registries for detecting systemic malignancies Would still need periodic examination by physician to ascertain skin cancers

45. 45 Mandatory Registry - Advantages All users registered Minimal survivor bias, minimal losses to follow-up If ascertainment of malignancies were also complete, registry would provide fastest incidence results All users registered Minimal survivor bias, minimal losses to follow-up If ascertainment of malignancies were also complete, registry would provide fastest incidence results

46. 46 Mandatory Registry - Disadvantages Poor acceptance by physicians & patients Expensive No accurate exposure assessment No data on confounding factors Not possible to ascertain most skin cancers Poor acceptance by physicians & patients Expensive No accurate exposure assessment No data on confounding factors Not possible to ascertain most skin cancers

47. 47 SummarySummary CohortRegistry (mandatory) Case Control Exposure Assess. GoodFairPoor Outcome Assess. GoodIncomplete DurationLong Shorter Expense++++++ Sample sizeLarge Smaller Risk factorsYesIncompletePoor IncidenceYesIncompleteNo RRYesNoNo (OR) SIRYesNot for skin cancers No

48. 48 Practical Issues Duration of follow-up Sample Size/Power Endpoint (cancers) ascertainment –How often, by whom Measures to reduce bias –losses to follow-up What level of uncertainty? Duration of follow-up Sample Size/Power Endpoint (cancers) ascertainment –How often, by whom Measures to reduce bias –losses to follow-up What level of uncertainty?

49. 49 AcknowledgementsAcknowledgements Yi Tsong, PhD. Actg. Deputy Director, QMR, OB David Graham, MD, MPH, Associate Director for Science, ODS Yi Tsong, PhD. Actg. Deputy Director, QMR, OB David Graham, MD, MPH, Associate Director for Science, ODS