US NAE

1
US - NAE IOM meeting - 10th April 2006

• Scientific methods underpinning policy
  formulation for an influenza pandemic

Roy Anderson Department of Infectious Disease
Epidemiology Faculty of Medicine Imperial College
London University
2
Epidemics of Infectious Diseases
Plague London 1665
Human fascination with epidemics of infectious
diseases and the associated patterns of mortality
has a long history. Some examples include the
lists of the epidemics compiled by the Chinese
scholar, Ssu Kwong, who lived during the Sing
Dynasty (AD 960-1279), the epidemics of the Greek
scholar Hippocrates (458-377 BC), the rudimentary
statistics of John Grant (1620-74) and William
Petty (1623-87), who studied the London Bills of
Mortality in the seventeenth century, and the
detailed description of epidemics by Richard Mead
(Discourse on Plagues 1673-1754) who so nearly
interested Isaac Newton in the shape of
epidemics. Well summarised in Plagues and
People William H. McNeill (1976) and more
recently by Christopher Willis in Plagues their
origin, history and future (1996).
SARS Hong Kong 2003
3
The changing world
4
World population projections
5
Record of increasing travel over four
male generations of the same family.(A)
Great-grandfather. (B) Grandfather. (C) Father.
(D) Son. Each map shows in a simplified manner
the individuals life-time tracks in a widening
spatial context, with the linear scale increasing
by a factor of 10 between each generation
(Bradley, 1994 Geog. Ann. 7691-104).
Great-grandfather
Grandfather
Father
Son
6
Global aviation network. Civil aviation traffic
among the 500 largest international airports in
100 different countries. Each line represents a
direct connection between airports. The colour
encodes the number of passengers per day (see
colour code at the bottom) travelling between two
airports. The network accounts for 95 of the
international civil aviation traffic. (Hufnagel
et al, 2004 PNAS).
7
Worldwide increase in Megacities
Hong Kong
Re-assortment of bird and human influenza viruses
H5N1?
Less Developed Regions
1970 1994 2000
2015 Africa
0 2 2
3 Asia 2
10 12 19 Latin
America 3 3
4 5 More Developed
Regions Europe
2 2 2
2 Japan 2
2 2 2 North
America 2 2
2 2
8
SARS
9
The emergence of a new disease urgent tasks

• Indication unusual clusters of
  morbidity/mortality in space and time (e.g.
  Quangzhou China November 2002).
• Identify aetiological agent.
• Develop diagnostic tests.
• Determine route(s) of transmission.
• Put in place, or activate, data capture and
  communication systems.
• Identify clinical algorithms for care to reduce
  morbidity and mortality.
• Identify and implement key public health
  measures.
• Keep public informed at all stages.

10
The Hong Kong epidemicA mix of person-to-person,
within hospital and point source spread
characterized by super spreading events
11
Basic Reproductive number, Ro
Chains of transmission between hosts
R0 Basic reproduction number average no. of
secondary cases generated by 1 primary case in a
susceptible population Rt Effective
reproduction number number of infections caused
by each new case occurring at time, t
Generation
2
1
4
3
5
6
7
8

• The key determinant of incidence and prevalence
  of infection is the basic reproductive number Ro.
• Many factors determine its magnitude, including
  those that influence the typical course of
  infection in the patient and those that determine
  transmission between people.

Number Infected
1
7
1
2
6
3
2
4
R
1.5
1.16
1
1.33
1.5
12
Basic principles in Infectious Disease
Epidemiology
13
Data capture in real time essential for
analysis and public health policy formulation
plus intervention evaluation
For each case

• SARS status (inc virology / serology)
• Date of onset
• Date of admission
• Date of discharge
• Date of death
• Still in hospital (true / false)
• Hospital, community or household exposure
• If hospital exposure, which hospital?
• Symptoms (over time)

• Unique identifier
• Age at admission
• Gender
• Home district
• Work district
• Occupation
• Hospital attended
• Identifier for contacts
• Address (Tel No)

For each contact
For virological / serological sample
For each hospital

• Identifier of case / contact
• Result
• Dates of sample

• Unique identifier
• District
• Dates of quarantine
• Treatment received

• As for cases
• Dates of quarantine
• Dates of exposure

Will enable the estimation of key distributions,
the characterization of transmission to define
treatment success
14
SARS - distribution of R
Super spreading Events (person, place setting)
15
Changes over time in the onset to admission
distribution Hong Kong
Probability distribution from onset of clinical
symptoms to admission to hospital from Hong Kong
early in the epidemic (Donnelly et al, 2003)
Two outliers influence this value
16
Incubation period of SARS
Hong Kong, China 113 cases. Mean 4.2
days. Fitted gamma distribution
Quangdong, China 70 cases. Mean 4.5 days
Average 4.5 days
17
SARS case fatality by age Hong Kong (modified
Kaplan-Meier) Leung et al (2004)
Right hand tail censoring
Case
Time
tm
18
RT-PCR sensitivity Peiris et al (2003)
SARS CoV
Influenza A
Viral load
Symptom score
Hayden et al. (1998) J. Clin Invest 101
p643 Experimental human influenza A/Texas/36/91
(H1N1) intranasal inoculation 105 dose
19
(No Transcript)
20
Hong Kong 2003 SARS outbreak model predictions
and R0 estimates(Riley et al, 2003 Science)
No. of secondary cases generated by each primary
case
Pattern of the epidemic stochastic
mathematical model - fit to observations
21
A new parameter to denote ease of control by
quarantine isolation

• Define ? as the fraction of transmission
  (infectiousness) that occurs before the onset of
  clinical symptoms. Isolation of a proportion e of
  symptomatic individuals can control an outbreak
  if the following expression is satisfied

Contact tracing introduces considerable
complications. Best to use simulation approaches
for evaluation
22
The relationship between clinical symptoms and
infectiousness
SARS HIV Smallpox Influenza A
2-7 days
Symptoms
Infectiousness
23
Effect of patient isolation (within 2 days of
onset of clinical symptoms) (Fraser et al, 2004)
Control can be augmented by other measures such
as - contact tracing plus isolation of
contacts travel restrictions.
90 success in isolation
24
Conclusions what worked

• Simple public health measures worked well.
• Isolation post onset of illness even with vague
  clinical case definition.
• Contact tracing and quarantine (voluntary or
  compulsory).
• Movement restrictions.
• Travel advisory notices (WHO).
• International cooperation

25
Influenza A pandemics
26
Four separate - but linked issues
1) Bird Influenza 2) Bird Influenza H5N1
high pathogenicity. 3) Bird Flu (H5N1)
transmission from birds to humans 4) Human
to human transmission of an H5N1 variant
with high pathogenicity? (175 cases, 90
deaths)
27
Influenza A viruses
2006-2007 H5N1?
Two sets of protein spikes Haemagluttinin or
HA (15 subtypes), and Neuraminidase or NA (9
subtypes). All 15 HA subtypes and 9 NA subtypes
have been detected in free-flying birds and they
form a huge and mobile pool of genetic diversity
Mortality in USA 1918-19 epidemic
28
Evolution of Influenza A (1)(Ferguson Anderson
(2002) Nature Medicine 8 562-3)

• Antigenic drift the evolutionary replacement of
  existing strains by new lineages characterizes
  human influenza evolution.
• Given the high transmissibility and mutation rate
  of the influenza virus, the constancy of genetic
  change is surprising.
• Antigenic shift is the introduction of a new
  avian influenza A subtype into the human
  population via reassortment events

29
Types of Influenza circulating in the USA - end
of 2005
H3N2
H1N1
H1N2
30
Control options for Influenza pandemic

• Any control requires very early detection and
  a well-planned, rapidly executed response.
• Control arguably only feasible if the nascent
  pandemic is small and geographically contained.
• Containment made more feasible if evolution of
  transmissibility is incremental.
• Options
• Prophylactic vaccination (if H5N1 trial
  vaccine available).
• Antiviral (NAI) prophylaxis/treatment.
• Increasing social distance (school/workplace
  closure, movement restrictions, isolation).
• Restricting entry to and exit from UK
• Key questions is any combination of the above
  capable of controlling a pandemic? What resources
  are required? What is the best combination?

31
Model design for individual based stochastic
simulations

• High performance, object-oriented code. Intended
  to be scaleable to allow 000s of model
  simulations of 60 million population to be
  performed.
• Computationally intensive (gt5GB memory use for 60
  million).
• Three levels of population structure
• Flexible modelling of disease biology (arbitrary
  distributions, number of disease stages), and
  interventions (ring vaccination, mass
  vaccination, quarantine, anti-viral treatment,
  movement constraints).

Probability
Distance
Network (disease specific)
Spatial
Household
32
Mobile phone data location of call for sample
data set (May 2004 - 1000 phones over 3 months)
Mobile data (scatter plot of 6700 locations
listed in dataset)
Population density data UK 2002
33
Spatial kernels mobile phone data frequency
versus distanced moved per defined time
unit(Anderson, Ferguson Donnelly, 2006)
Number of people
Distance metres
34
Example output Influenza A UK outbreak (Ferguson
et al, 2005)
Influenza A parameters, 5 seed infections in
London, realistic movement patterns, no past
immunity (i.e. human H5N1), six month epidemic.
35
Example output effect of school closures
(Ferguson et al, 2006)
36
Mean time between exports with differing degrees
of travel restrictions(Hollingsworth, Ferguson
Anderson, Nature Medicine 2006)
Stochastic effects Incidence as a function of
time for four realizations of a stochastic
simulation of an influenza epidemic. Incidence is
highly variable during the early stages of any
epidemic, here ranging between 10 and 120 cases
per day by day 20. Incidence increases during the
course of an epidemic and the interval between
exports will decrease rapidly.
effectiveness of travel restrictions
37
Conclusions influenza A

• Simulation studies suggest that antiviral
  prophylaxis may be able to contain the earliest
  stage of an influenza pandemic - providing
• R0 is low (1-2).
• The original cluster is identified rapidly.
• Treatment can be delivered rapidly to a large
  proportion of the population of a rural
  area.
• Enough courses of drug are available (100k-1m).
• Case detection is good once first cluster is
  identified.
• Restricting air travel via departure or arrival
  screening unlikely to be of great value.
• Restricting social/work interactions can have a
  significant effect (economic cost?). Probably
  happened in previous influenza pandemics via
  changed behaviour once epidemic established.
  Certainly happened during the SARS outbreaks in
  China and other countries.
• Build/enhance vaccine/drug production
  capabilities - worldwide.

38
End