1. Containing Pandemic Influenza at the Source Ira M. Longini, Jr. Dept. Biostatistics U. Washington Hutchinson Rsh Ctr

2. Collaborators M. Elizabeth Halloran Azhar Nizam Shufu Xu Depts. Biostatistics, U Wash and Emory U Derek Cummings Johns Hopkins U. Kumnuan Ungchusak Wanna Hanshaoworakul Thai Ministry of Health Timothy C. Germann Kai Kadau Catherine A. Macken Los Alamos National Laboratory

4. How Bad Could it Get? Current Avian A(H5N1) Influenza is SE Asia –165 cases, 88 deaths, 53% case fatality ratio Global pandemic, first wave about 6 - 9 months, 2 billion cases –1918 scenario: 10 - 50 million deaths –Other scenarios: 2 – 7 million deaths –Contrast 20 million AIDS deaths over 25 years 811 SARS deaths over 8 months

5. Containing Pandemic Influenza at the Source It is optimal to stop a potential pandemic influenza strain at the source –Longini, et al. Science 309, 1083-7 (2005). Longini and Halloran. Science 310, 1117 ‑ 18 (2005). –Ferguson, et al. Nature 437, 209-14 (2005) –Targeted antiviral prophylaxis with mobile stockpile (WHO) ~ 5 million courses –Quarantine, social distancing, school closing, travel restrictions –Pre or rapid vaccination with a possibly poorly matched vaccine

6. Pandemic Influenza in the US or Other Countries Once Spread is Global Hard to contain as it comes in Once widespread, slow transmission until well-match vaccine is available –Targeted antiviral prophylaxis –Quarantine, social distancing, school closing, travel restrictions –Rapid vaccination with a possibly poorly match vaccine Germann, T.C., Kadau, K., Longini I.M. and Macken C.A.: Mitigation strategies for pandemic influenza in the United States. (accepted – Feb or March, 2006) Halloran, M.E. and Longini, I.M.: Community studies for vaccinating School children against influenza. Science 311 (Feb. 3, 2006).

7. CONTACTS Household Household cluster Preschool/daycare School Workplace TAP: Targeted antiviral prophylaxis using neuraminidase inhibitors (oseltamivir/zanamivir) 80% ascertainment 100% household + HH cluster 80% preschool 80% school 60% workplace

8. Antiviral efficacies used in the model: Oseltamivir Antiviral efficacy of reducing susceptibility to infection: AVE S = 0.48, [0.17, 0.67] 95% CI * Antiviral efficacy of reducing illness given infection: AVE D = 0.56, [0.10, 0.73] 95% CI * Antiviral efficacy of reducing illness with infection: AVE SD = 0.80, [0.35, 0.94] 95% CI * –Mult.: AVE SD = 1 – (1- AVE S ) (1- AVE D ) = 0.77 Antiviral efficacy of reducing infectiousness to others: AVE I = 0.80, [0.45, 0.93] 95% CI * * data from Welliver, et al. JAMA (2001); Hayden, et al. JID (2004); analysis by Yang, Longini, Halloran, Appl Stat (in print); Halloran, et al. (in prep).

9. Prevaccination Prevaccination with low efficacy vaccine –Low efficacy vaccine: VE S = 0.30, VE I = 0.5 –50% and 70% prevaccination of the population and evaluate above interventions

10. Preliminaries

11. Basic Reproductive Number: R 0 R 0 > 1 for sustained transmission For pandemic influenza: 1 < R 0 ≤ 2.4 –A(H3N2) 1968-69, R 0 ≈ 1.7 –A(H1N1) 1918, second wave, R 0 ≈ 2.0 –New variant, early spread: 1 < R 0 ≤ 1.6

12. Reed-Frost Model Stochastic process: discrete state space and time t 0, t 1, t 2 …. Infectious agent natural history –Infectious for one time unit Social contact structure –Random mixing – p = 1 – q, probability two people make contact sufficient to transmit R 0 = (n-1)p

13. Reed-Frost Model See chain binomial chapter in the Encyclopedia Biostat., Vol 1, 593-7

14. Reed-Frost Model Threshold theorem: When R 0  1, then no epidemic, When R 0 >1, then epidemic with probability

15. Simulated Reed-Frost Model * n Start with (S 0,I 0 ≥ 1) n For each S 0, generate random number x  [0,1] n If x ≥ q Io, then person becomes infected n Repeat for next generation and update states n Stop when S 0 = 0 or I 0 = 0 *First done by Elveback and Varma (1965)

16. * Source: Elveback and Varma (1965) *

17. Containment in SE Asia

18. Rural population of 500,000 in Thailand Population matched to non- municipal area household-size and age distributions. * *Population and Housing Census 2000 data used where available (www.nso.go.th); other National Statistical Office reports and tables used as necessary.

19. Population Characteristics 36 localities each of size ~ 14,000 Total area: 75 km X 75 km = 5,625 km 2 Population density ~89/km 2 12.5km

20. Locality Characteristics ~ 28 villages, each of size ~ 138 households, ~ 500 people Villages are clustered Within village clusters: Household are clustered Small & large playgroups Elementary, lower-secondary and upper-secondary school mixing groups Social groups Work groups

21. Social network incorporated from the Nang Rong District study * n 310 villages under study n Village size average  100 households n Main mixing groups under study u Households u Villages u Hiring tractors u Temples u Elementary schools u Secondary schools u Workplaces * Faust, et al., Soc Net (1999)

22. Distribution of travel distance to work, school and social groups

23. Secondary school, work and social group assignment n Localities are linked by secondary schools, work groups and social groups n For residents of each locality, secondary school, work group and social group locality is selected according to distance distribution shown below (using most Southwesterly locality as an example) Zone% 1 90 2 7 3 2 4-6 1 Zone1 Zone2 Zone3 Zone4 Zone5 Zone6

24. Distribution of travel distance to work, school and social groups* For residents of most Southwesterly locality: Zone% 1 90 2 7 3 2 4-6 1 Zone1 Zone2 Zone3 Zone4 Zone5 Zone6 90% stay in zone 1 7% go to zone 2 2% go to zone 3 1% go beyond zone 3

25. Model calibration Illness Attack Rate Modeled Asian A(H2N2) PandemicHK-Like 1957-58Strain ’68-69 Young Children 35% 32% 34% Older Children 62% 46% 35% Adults 24% 29% 33% Overall 33% 33% 34%

26. Social Connectivity

27. Transmission n c daily adequate contact probability u c(n-1) average mixing group degree n x transmission probability given adequate contact n y relative susceptibility n p = cxy overall transmission probability

28. Bipartite Graph 1 2 n 1 2 m PeoplePlaces

30. Weighted Person-to-Person Graph 1 4 n 2 3 c 12 c 2n c 1n c 2j c 3r c 4s

34. Mean degree 4.6 Clustering coefficient 0.2 Mean shortest path 10.6 Small World Network

35. Infection Transmission Process

36. Latency Incubation Asymptomatic (33%) 0 Probability of infecting others days Exposure and infection 1.2d 1.7d 3.5d Natural History Used for Influenza Possibly symptomatic Symptomatic (67%) Case serial interval = 3.2 days

37. Intervention

38. Interventions considered n All interventions carried out in the localities as triggered n 80% targeted antiviral prophylaxis (TAP) n 90% geographically targeted antiviral prophylaxis (GTAP) n Localized household and household cluster quarantine. Lifted when there are no more local cases.

39. Interventions considered n TAP + pre-vaccination n TAP + localized household quarantine n TAP + localized household quarantine + pre- vaccination n Localized interventions begin 7, 14 and 21 days after outbreak is recognized, one day after as cases appear locally

40. Results

41. Frequency Number of secondary infections Average R 0 =1.4 R 0 * : Number of people infected by a single initial infective * Based on 1000 simulations

42. No. of cases 25 Day 11 25 Day 11 18 No Intervention R 0 = 1.4 166,408 total cases Day # Cases 11 6 18 47 25153 No. of cases Day

43. Illness attack rate by age-group and R 0 (No Intervention) R 0 =1.1 R 0 =1.4 R 0 =1.7 R 0 =2.1 R 0 =2.4

44. No. of cases Day 18 Contained GTAP 14 days after the first detected case(~ day 18) R 0 = 1.4 44 total cases No. of cases Day

45. No. of cases Day 18 Not contained GTAP 14 days after the first detected case(~ day 18) R 0 = 1.4 1925 total cases No. of cases Day

47. No Intervention, R 0 = 1.4

48. 80% TAP, 14 day delay, R 0 = 1.4

49. Cases per 1000 EscapesCourses Containment Ratio Intervention R 0 = 1.4 R 0 = 1.7 R 0 = 1.4 R 0 = 1.7 R 0 = 1.4 R 0 = 1.7 R 0 = 1.4 R 0 = 1.7 No Intervention 2293617421184---- 80% Targeted Antiviral Prophylaxis (TAP) 0.111400.355007713552591.000.35 90% Geographically Targeted Antiviral Prophylaxis (GTAP) 0.06540.16183269463024881.000.60 80% TAP + 50% Pre-vaccination 0.010.490.0626631111.000.85 80% TAP + 70% Pre-vaccination 0.010.030.030.13462171.001.00 70% Quarantine 0.170.770.602--0.980.70 80% TAP + 70% Quarantine 0.050.140.2137911491.001.00 80% TAP + 70% Quarantine + 50% Pre- vaccination 0.020.040.060.191322851.001.00 Simulated mean cases, escapes, courses and containment proportion for various interventions

50. R0 sensitivity - cases

51. 90% GTAP Threshold

52. R0 sensitivity - courses

53. Delay in interventions sensitivity - cases

54. Conclusions n 80% TAP and 90% GTAP would be effective in containing pandemic influenza at the source if R 0 ≤ 1.4 u This is true even up to a 21 day delay after the first detected case, about 25 days after first infection u At least 70% TAP or GTAP would be needed u Fewer than 120,000 courses would generally be needed n Neither 80% TAP nor 90% GTAP would be effective in containing pandemic influenza at the source if R 0 ≥ 1.7 u 350,000 courses would generally be needed

55. n Prevaccination of the population with a low efficacy vaccine makes a big difference, even at the 50% coverage level u With 50% of the population vaccinated, 80% TAP would be effective in containing pandemic influenza at the source if R 0 ≤ 1.7, even up to 56 days delay. Prevaccination lowers the effective R before the epidemic. u With 70% of the population vaccinated, 80% TAP would be effective in containing pandemic influenza at the source if R 0 ≤ 2, up to 56 day delay

56. n Household and neighborhood quarantine is effective for R 0 ≤ 1.7 up to 35 days delay, becomes ineffective for R 0 ≥ 2.1 n A combination of 80% TAP + quarantine is effective even for an R 0 as high as 2.4, while adding prevaccination makes this combination even more effective even up to 56 days delay.

57. Policy Implications n A mobile stockpile of oseltamivir has been created by WHO. It will be deployed quickly after the initial infection cluster is detected. n The outbreak is containable with targeted antiviral prophylaxis if transmissibility is reasonably low (R 0 ≤ 1.4) and intervention occurs with 21 days of first detected case. n Localized quarantine and other social distancing measures would be important for containment for viruses with higher transmissibility (R 0 ≥ 1.7).

58. Policy Implications Continued n The development and deployment of vaccine for potential pandemic strains for at-risk populations should move forward as quickly as possible. n Surveillance and detection of early pandemic influenza transmission is extremely important in all potentially at-risk regions of the world.

59. Oseltamivir Stockpiles n WHO: 120,000 courses, soon 5 million courses n US: 4 - 5 million courses, increase to 75 million courses? H5N1 vaccine stockpile? n Other Countries (e.g., U K, France, Finland, Norway, Switzerland, New Zealand) u 20 - 40% population (one course) ordered

60. The End