1. Draft Quantitative Risk Assessment of vCJD Risk Potentially Associated with the Use of Human Plasma-Derived Factor VIII Manufactured Under United States (US) License From Plasma Collected in the US Steven Anderson, PhD, MPP Office of Biostatistics & Epidemiology Center for Biologics Evaluation and Research U.S. Food and Drug Administration December 15, 2006 TSEAC Meeting

2. Background: Draft Risk Assessment for U.S. Factor VIII and vCJD l Emergence of vCJD transmission via red cell transfusion in December 2003 l Concerns vCJD may potentially be transmitted through plasma- derived products including clotting factors l Clotting factors such as plasma-derived FVIII (pdFVIII) made from human plasma are used in large quantities by many US patients l Transmission of vCJD via FVIII is a potential hazard but magnitude of potential risk was unknown l Fall 2004 FDA began developing a risk assessment of potential vCJD risk via pdFVIII manufactured from plasma collected in US

3. Background: Draft Risk Assessment for U.S. Factor VIII and vCJD l FDA presented conceptual vCJD-FVIII risk assessment model at February 8, 2005 TSEAC l FDA sought Committee discussion and advice on several risk assessment inputs at October 31, 2005 TSEAC l Peer review of FDA Risk Assessment model and document by 3 external experts in Summer 2006 l Presentation of “Draft Quantitative Risk Assessment of vCJD Risk Potentially Associated with the Use of Human Plasma- Derived Factor VIII Manufactured Under United States (US) License From Plasma Collected in the US” at December 15, 2006 TSEAC l FDA risk communication activities

4. Advice from TSEAC on vCJD-pdFVIII risk assessment data inputs TSEAC Oct 31, 2005 UK vCJD prevalence estimate 1. Epidemiological Model Estimate (mean = ~1.8 in 1 million) 2. Tissue Surveillance method – (mean =1 in 4,225) Efficiency of vCJD deferral 85% - 99% Quantity of infectivity present in blood 5th Percentile: 2 Minimum: 0.1 Most likely: 10 Maximum: 1,000 95th Percentile: 30 i.c ID 50 /ml Adjustment for efficiency of transmission i.v. vs i.c. 0.1 - 1 Infectivity presence in blood Last half of incubation period Clearance of vCJD agent during manufacturing 3 Categories stratified by clearance: 7-9 log 10, 4-6 log 10, 2-3 log 10 FVIII usage 1. CDC Sponsored 6 state surveillance study 1993-1998 2. CDC UDC project used to estimate patient population sizes

5. Risk Assessment Framework used by FDA (NAS, 1983) 1. Hazard identification Establishes causality between hazard and adverse effects 2. Dose response (Hazard characterization) Probability of response – infection or illness 3. Exposure assessment Frequency and level of exposure Estimates potential DOSE vCJD ID 50 4. Risk characterization Probability of occurrence, severity of adverse effects Uncertainty Sensitivity analysis

6. FDA Risk Assessment: Scope and Type of risk assessment Risk Assessment Problem Statement: What is the potential vCJD risk for recipients of pdFVIII manufactured from plasma collected in US? Scope of FDA pdFVIII Risk Assessment:  Estimates potential vCJD risk for US pdFVIII recipients  Severe Hemophilia A  Severe von Willebrand disease (vWD) (Type 3)  Potential vCJD risk estimated for one-year treatment period in 2002  Model results generally applicable to current year General Analytic Approach:  Quantitative risk assessment (QRA)  Input data incorporated into model as statistical distributions  Probabilistic computer-based model  Monte Carlo methods

7. Factor VIII (FVIII): a plasma protein necessary for blood clotting Two types of bleeding disorders associated with deficiency of FVIII: Hemophilia A (HA):  Deficiency of FVIII,  Severe HA patients have <1% FVIII activity von Willebrand disease (vWD):  Deficiency of von Willebrand factor (glycoprotein carrier of FVIII)  Severe vWD patients have reduced levels of FVIII Clinical usage of pdFVIII:  Human pdFVIII used by ~25% HA patients  Recombinant FVIII available since early 1990s  vWD patients use pdFVIII: no recombinant vWF, some pdFVIII contains vWF FDA Risk Assessment Background pdFVIII product usage in clinical setting

8. Hazard Identification: Potential vCJD risk via pdFVIII  3 vCJD infections probably acquired through red-cell transfusions in UK since 2003  Potential presence of vCJD agent in human plasma (and plasma- derived products including pdFVIII??) may be a hazard to human health  Hemophilia A and vWD patients who used pdFVIII over long treatment period may have been exposed to vCJD agent  To date – no cases vCJD identified in recipients of plasma-derived clotting factors

9.  Plasma pooled from thousands of donors – Increased chance plasma pools contain vCJD donations compared to single donors  HA & vWD patients may use large amounts FVIII over lifetime  TSE clearance during manufacturing  Clearance likely decreases potential vCJD infection risk by decreasing exposure to vCJD agent  FVIII products vary in level of reported clearance  Challenge to evaluate and study clearance Hazard Identification: vCJD Risk issues associated with pdFVIII

10. Dose-response for Human vCJD  Challenges for determining dose-response  Human data absent  Quantity agent in human blood? Plasma?  Present throughout incubation period?  Genetics & susceptibility of humans  Threshold? Or not?  Accumulation of agent in humans?  Use of animal data: Approximate human vCJD well? FDA risk assessment model assumes:  vCJD agent present in last half incubation period  Linear dose-response (below 2 ID 50 ), no threshold  Accumulation of infectious agent in the body

11.  Donor travel history UK, FR, other Europe  Adjustments for duration & year traveled, donor age  Screening questionnaire  Plasma pool size  Quantity vCJD agent in pool  Reduction of vCJD agent during manufacture  Annual dose Factor VIII  Severity of disease & treatment regimen Module 1 vCJD Prevalence UK Module 2 vCJD Prevalence US Donors Module 3 FVIII Processing Module 4 Utilization FVIII INPUTMODULEOUTPUT  Number US vCJD donors  Number vCJD donors post- screening  Total number vCJD donations  Percentage plasma pools / vials with vCJD agent  Quantity vCJD agent per FVIII vial  Annual exposure (DOSE) severe HA and vWD recipients to vCJD agent  LOWER estimate of UK vCJD case prevalence  HIGHER estimate of UK vCJD infection prevalence  Epidemiological modeling based on UK vCJD cases  UK surveillance data Exposure Assessment Overview: Potential vCJD risk via pdFVIII

12. Monte Carlo Method used in FDA Risk Assessment model  Tool for combining input data as distributions rather than using and propagating data as summary statistics  Without Monte Carlo methods process of combining more than 2 distributions of possible values of variables would be challenging  Method:  Draws randomly from defined distributions  Performs the programmed mathematical functions and stores number each time  Process repeated thousands of times or “iterations”  Results displayed as a new, aggregate distribution

13. Monte Carlo Analysis – Example calculation i.v. ID 50 per ml plasma Mean= 5.2 2 10 30 0.1 1.0 58% Adjustment i.c. to i.v. Percent ID 50 in plasma i.c. ID 50 per ml Blood* x x Perform Multiplication 10,000 times Distribution i.v. ID 50 per ml plasma Result of 10,000 iterations (5 th ) (95 th ) * Actual distribution used in FDA risk assessment included a range of 0.1 to 1,000 i.c.ID 50 /ml

14. Exposure Assessment: Module 1 Prevalence of vCJD in United Kingdom A. Epidemiological modeling vCJD cases Clarke and Ghani 2005  Predicted 70 (10 – 190 95% CI) vCJD cases, 2002-2080  Prevalence vCJD ~1.8 per million UK population B. Tonsil/appendix tissue surveillance in UK patients Hilton, et al. 2004  3 prion positive samples in 12,674 samples tested  Mean of 1 positive in 4,225 individuals, or 237 per million Mostly in 20 – 30 yr old patients  UK prevalence is Critical model input – used to estimate vCJD prevalence for: France, other Europe, Military bases in Europe, Plasma donors in United States

15. Exposure Assessment: Module 1 vCJD Prevalence in United Kingdom (cont.) Uncertainties of Estimating UK vCJD Prevalence  Epidemiological modeling based on vCJD cases in UK  All vCJD cases-MM codon 129 PrP individuals (40% population)  Assumptions – incubation period, time of infection, effectiveness feed ban, etc  Tonsil/appendix surveillance data  Prion protein in appendix – but may not have agent in blood or develop disease  Sample size relatively small for rare disease  Underestimate vCJD prevalence - in one vCJD infection no agent found in appendix  Tissue surveillance lacks controls (survey of non-BSE exposed population), patient outcomes unknown

16. Exposure Assessment: Module 2 vCJD Prevalence in US plasma donors Modeling Approach for Module 2 Estimation of US vCJD donor risk  Estimate size of US plasma donor population travel history to UK, France or other countries Europe since 1980  Model donor/traveler risk using survey data blood donors  Adjust travel risk by several factors (duration of stay, year, etc.)  Estimate probability of infection in individual donors  Add up potential number infections for US plasma donor groups  Apply effectiveness of donor deferral policy  Calculate total number potential vCJD infected donors & donations

17. Relative Risk metric was used to estimate probability vCJD infection (prevalence) in US donors Estimates vCJD prevalence in donors who traveled to France and other countries of Europe relative to UK vCJD prevalence  Relative Risk estimated in year 2001 using country-specific info on BSE cases, amt imported UK beef, number vCJD cases, etc. Adjustment for Relative Risk Prevalence UK1P vCJD-UK France0.05P vCJD-France = 0.05 х P vCJD-UK Other Countries in Europe0.015P vCJD-EU = 0.015 х P vCJD-UK Military Bases in Europe0.35P vCJD-DOD = 0.35 х P vCJD-UK Exposure Assessment: Module 2 vCJD Prevalence in US plasma donors (cont.)

18. FDA Model estimated relative risk for US plasma donors with travel history to UK, France or other Europe since 1980 Model proportionally adjusted relative risk for vCJD for US donors over 23 year period (1980 – 2002) based on:  Duration of Travel  Relative risk is adjusted on a per month or per day basis  Specific year(s) of travel  Accounts for variation in BSE epidemic / exposure  Age of donor  To apply age specific rates for vCJD in UK (median age 28yrs)  Age specific donation rates  Type of Donor : Source and Recovered

19. Model estimates two sources residual risk for US donors: (1) Donors with deferrable criteria but not deferred because limitations in screening (2) Not deferrable: short-term travel Approximately 90% vCJD risk eliminated for US donors by current donor deferral policy for travel: UK: 1980-1996, > 3 mos France: since 1980, > 5 years Other Europe: since 1980, > 5 years (recovered plasma donors only) Military Bases in Europe: 1980-1996 Efficiency of deferral: 85-99% Exposure Assessment: Module 2 vCJD Prevalence in US plasma donors (cont.)

20. Module outputs for Module 2  Potential vCJD infected US plasma donors  Potential donations with vCJD agent Uncertainties in the data  No travel data for source plasma donors-source plasma donors may travel less than whole blood donors (group surveyed)  Estimation deferral effectiveness a challenge –self-deferred population unknown  MV and VV at codon 129 PrP individuals: Potential disease attack rate and incubation period unknown  Percentage infections that become symptomatic disease

21. Exposure Assessment: Module 3 Factor VIII Processing Modeling Approach for Module 3  Estimate probability plasma pool contains vCJD donation  Estimate quantity vCJD ID 50 per ml plasma and per pool  Efficiency of exposure through i.v. vs i.c. route  Log 10 reduction in quantity i.v. ID 50

22. Infectivity clearance in product plasma pools:  Each product has different purification steps and clearance  Product specific data: not available for all products and process steps  Data in published literature: available only for some purification steps, variation among studies  FDA model stratified by 3 clearance levels: 7-9 log 10, 4- 6 log 10, 2-3 log 10  FDA believes that most pdFVIII products have at least 4 log 10 of clearance Exposure Assessment: Module 3 Factor VIII Processing (cont.)

23. Model output from Module 3 predicts:  Percentage plasma pools with vCJD agent  Percentage vials with vCJD agent  Quantity vCJD agent per vial

24. Uncertainties in the data Quantity of infectivity in plasma and plasma pools:  Detection of low level infectivity difficult  Infectivity in animal blood may not be representative of infectivity in human blood Infectivity clearance  No standard method for clearance study spiking materials  No standard method for animal study selection of donor and recipient animals  Reduction in laboratory scale with high concentration spiked infectivity may not reflect reduction in real processing systems  Reduction in different purification steps additive? Exposure Assessment: Module 3 Factor VIII Processing (cont.)

25. Exposure Assessment: Module 4 Utilization of Factor VIII (cont.) Modeling Approach for Module 4 Inputs:  Percentage vials with vCJD agent  Quantity vCJD agent per vial  Annual utilization / dose Factor VIII per patient Model output to predicts:  Annual potential dose vCJD ID 50 per patient  Prediction of risk of vCJD infection based on animal dose-response information

26. Exposure Assessment: Module 4 Utilization of FVIII (cont.) Factors considered for utilization  Type of disease:  Severe Hemophilia A,  Severe von Willebrand disease (vWD)  Treatment regimens: Prophylaxis and Episodic  Inhibitor and immune tolerance Data Sources: CDC – UDC data to estimate size of HA & vWD populations CDC sponsored 6 states hemophilia surveillance study, 1993-1998 (total records: 17,848) Information on: age, sex, factor of deficiency, baseline level of factor, product and amount used, inhibitor and immune tolerance

27. Risk Characterization Integration of information from Exposure Assessment and Dose-Response Exposure assessment of FDA model provides estimate of potential vCJD ID 50 DOSE + FDA assumed a LINEAR DOSE-RESPONSE relationship for vCJD (TSEAC recommendation: vCJD ID 50 is a linear dose-response) Estimated RISK = Probability vCJD infection DoseProbability of infection 1 ID 50 50 % 0.1 ID 50 5 % etc.

28. Risk Characterization (cont’d) Uncertainty- 2 major sources: (1) Arises from lack of information –Inadequate quantity / quality of data Express outcomes uncertainty with confidence intervals –95% CI, 5 th and 95 th percentiles, etc. Outcomes expressed with measures central tendency – mean, median, etc. (2) Inaccuracies of models –Peer-review to identify errors or omissions, etc.

29. LOWER PREVALENCE Epidemiological model prevalence ~1.8 in 1million HIGHER PREVALENCE Surveillance prevalence 1 in 4,225 Scenario Source Mean (5 th - 95 th ) Recovered Mean (5 th - 95 th ) Source Mean (5 th - 95 th ) Recovered Mean (5 th - 95 th ) Percent pools with vCJD agent 0.01% (0 – 0%) 0.10% (0 – 0%) 0.96% (0 – 5.88%) 9.12% (0 – 40.17%) Average percent pools containing vCJD agent 0.027 % (0 – 0 %) 2.41 % (0 – 10 %) Risk Characterization (cont’d) Factor VIII Risk Assessment Table 4.5 Annual Percentage plasma pools containing vCJD donation Results from the model as shown but considerable UNCERTAINTY is associated with estimates !

30. LOWER vCJD Case Prevalence ~1.8 in 1 million HIGHER vCJD Prevalence 1 in 4,225 Treatment regimes Inhibitor status Mean annual vCJD risk (5 th -95 th ) Mean annual vCJD risk (5 th -95 th ) Prophylaxis No Inhibitor 1 in 4.0 million (0-0) 1 in 54,000 (0 - 1 in 12,000) With Inhibitor – No Immune Tolerance 1 in 4.8 million (0-0) 1 in 41,000 (0 - 1 in 9,000) With Inhibitor – With Immune Tolerance 1 in 1.3 million (0-0) 1 in 15,000 (0 - 1 in 3,700 ) Episodic No Inhibitor 1 in 9.4 million (0-0) 1 in 105,000 (0 - 1 in 24,000 ) With Inhibitor 1 in 8.0 million (0-0) 1 in 48,000 (0 - 1 in 12,000 ) Risk Characterization (cont’d) Table 5.2A Individual vCJD risk- severe Hemophilia A FVIII Manufacture reduction of 4-6 log 10 clearance

31. LOWER PREVALENCE Epidemiological model prevalence ~1.8 in 1million HIGHER PREVALENCE Surveillance prevalence 1 in 4,225 Total number patients Mean Total quantity FVIII used by all patients per yr Mean annual vCJD risk (5 th -95 th ) Mean annual vCJD risk (5 th -95 th ) 1,800243 million IU 1 vCJD infection in 3,077 years (0 - 0) 1 vCJD infection in 35 years (0 - 1 in 8) Risk Characterization (cont’d) Table 5.2B Population vCJD risk for severe Hemophilia A Manufacture reduction of 4-6 log 10 clearance Results from the model as shown but considerable UNCERTAINTY is associated with estimates !

32. Risk Characterization (cont’d) Table 5.3A Range of vCJD risk-severe Hemophilia A 7 - 9 Log 10 Reduction 4 – 6 Log 10 Reduction 2 – 3 Log 10 Reduction LOWER vCJD Prevalence ~1.8 in 1 mill HIGHER vCJD Prevalence 1 in 4,225 LOWER vCJD Case ~1.8 in 1 mill HIGHER vCJD Prevalence 1 in 4,225 LOWER vCJD Prevalence ~1.8 in 1 mill HIGHER vCJD Prevalence 1 in 4,225 Treatment Regimen Inhibitor Status Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Prophylaxis No Inhibitor 1 in 4.1 billion (0-0) 1 in 50 mill (0 - 1 in 11 mill) 1 in 4 mill (0-0) 1 in 54,000 (0- 1 in 12,000) 1 in 15,000 (0-0) 1 in 82 (0 - 1 in 17) With Inhibitor – No Immune Tolerance 1 in 3.5 bill (0-0) 1 in 40 mill (0 - 1 in 8.8 mil) 1 in 4.8 mill (0-0) 1 in 41,000 (0- 1 in 9,000) 1 in 12,000 (0-0) 1 in 65 (0 - 1 in 13 ) With Inhibitor – With Immune Tolerance 1 in 551 mill (0-0) 1 in 15 mill (0 - 1 in 3.4 mill) 1 in 1.3 mill (0-0) 1 in 15,000 (0- 1 in3,700) 1 in 2,700 (0-0) 1 in 24 (0 - 1 in 3 ) Episodic No Inhibitor 1 in 3.2 bill (0-0) 1 in 100 mill (0 - 1 in 24 mill) 1 in 9.4 mill (0-0) 1 in 105,000 (0- 1 in 24,000) 1 in 21,500 (0-0) 1 in 159 (0 - 1 in 34 ) With Inhibitor 1 in 4 bill (0-0) 1 in 50 mill (0 - 1 in 11 mill) 1 in 8 mill (0 - 0) 1 in 23,000 (0- 1 in 12,000) 1 in 23,000 (0-0) 1 in 73 (0 - 1 in 16)

33. Risk Characterization (cont’d) Table 5.3A Model results of Range of vCJD risk-severe Hemophilia A Excerpt: 7 - 9 Log 10 Reduction 4 – 6 Log 10 Reduction LOWER vCJD Prevalence ~1.8 in 1 mill HIGHER vCJD Prevalence 1 in 4,225 LOWER vCJD Case ~1.8 in 1 mill HIGHER vCJD Prevalence 1 in 4,225 Treatment Regimen Inhibitor Status Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Mean potential vCJD risk per person per year (5 th - 95 th perc) Prophylaxis No Inhibitor 1 in 4.1 billion (0-0) 1 in 50 mill (0 - 1 in 11 mill) 1 in 4 mill (0-0) 1 in 54,000 (0- 1 in 12,000) With Inhibitor – No Immune Tolerance 1 in 3.5 bill (0-0) 1 in 40 mill (0 - 1 in 8.8 mil) 1 in 4.8 mill (0-0) 1 in 41,000 (0- 1 in 9,000) With Inhibitor – With Immune Tolerance 1 in 551 million (0-0) 1 in 15 mill (0 - 1 in 3.4 mill) 1 in 1.3 mill (0-0) 1 in 15,000 (0- 1 in 3,700)

34. Risk Characterization (cont’d) Table 5.2A Range of vCJD risk severe vWD Manufacture reduction of 4-6 log 10 clearance Excerpt: YOUNG vWD (< 15 yrs of age) LOWER vCJD Prevalence ~1.8 in 1,000,000 HIGHER vCJD Prevalence 1 in 4,225 Mean potential vCJD risk per person per year b (5 th - 95 th perc) c Mean potential vCJD risk per person per year b (5 th - 95 th perc) c Prophylaxis 1 in 4.7 million (0 - 0) 1 in 52,000 (0 - 1 in 13,000) Episodic 1 in 48 million (0 - 0) 1 in 971,000 (0 - 1 in 293,000) ADULT vWD ( > 15 yrs of age) Prophylaxis 1 in 4.1 million (0 - 0) 1 in 46,300 (0 - 1 in 11,000) Episodic 1 in 10 million (0 - 0) 1 in 1 million (0 - 1 in 24,000)

35. LOWER PREVALENCE Epidemiological model prevalence ~1.8 in 1million HIGHER PREVALENCE Surveillance prevalence 1 in 4,225 Total number patients Mean annual vCJD risk (5 th -95 th ) Mean annual vCJD risk (5 th -95 th ) 250 1 vCJD infection in 28,450 years (0 - 0) 1 vCJD infection in 405 years (0 - 1 in 76) Risk Characterization (cont’d) Table 5.2B Population vCJD risk for severe vWD Manufacture reduction of 4-6 log 10 clearance Results from the model as shown but considerable UNCERTAINTY is associated with estimates !

36. Risk Characterization (cont’d) Sensitivity and Importance Analysis Analyses identify key inputs or “drivers” of vCJD risk l (1) Conducted by varying value of input and observing impacts on risk estimate l (2) Importance analysis –Ranks inputs according to level of influence on final risk estimate l (3) Most important model inputs –Clearance of vCJD agent during manufacture –Quantity of pdFVIII used by patients –Prevalence of vCJD in UK –Efficiency i.c. vs i.v. route

37. Risk Characterization (cont’d) Importance Analysis Fig 2.A. Importance Analysis ranking influential factors for predicted vCJD exposure (yr) using prevalence estimates from 0.7 to 700 cases per million + _

38. Uncertainties and Data Gaps More data are needed on: Clearance of vCJD agent during manufacturing steps Prevalence of vCJD in UK, USA, etc. Utilization: How do patients use pdFVIII? By lot? Several lots per year? Amount vCJD agent present in human blood & plasma Progression of vCJD and variability of levels of infectivity in blood & plasma Dose response relationship in animal models Many other parameters

39. Risk Characterization  It is not possible to precisely estimate potential vCJD risk for pdVIII recipients because of uncertainties in data and knowledge of vCJD  vCJD risk from use of pd FVIII may not be zero, but most likely extremely small  FDA risk assessment results consistent with absence of observed vCJD cases in clotting factor recipients  Current donor deferral greatly reduced the risk by deferring individuals with history of extended travel to UK, France and other countries in Europe since 1980  Risk assessment shows that manufacturing processes shown to reduce the Infectivity is the most influential factor in model for reducing vCJD risk  Risk assessment highlights data gaps in level of clearance, product usage, vCJD prevalence and dose-response

40. Acknowledgements OBE Hong Yang OBRR David M. Asher Jay Epstein Jonathan Goldsmith Dorothy Scott Mark Weinstein Alan Williams OD Jesse Goodman Karen Midthun Diane Maloney Peer Reviewers David Gaylor (Gaylor & Assoc.) Mark Powell (USDA) Sonja Sandberg (Univ. Framingham)