Neil Ferguson

1
Evidence supporting the use of non-pharmaceutical
interventions in a pandemic
Neil Ferguson MRC Centre for Outbreak Analysis
and ModellingImperial College London
2
Timescale of spread

• 2-4 months to peak at source, 1-3 months to
  spread to West (in absence of seasonality).
• 1/3 of population might become ill, 1 million
  new sick people per day at peak.
• 15 absenteeism at peak.
• 1st wave over 3 months after 1st UK case.

2
3
What can we expect from NPIs?

• Developed world reduce attack rates until
  vaccine available.
• Developing world reduce attack rates
  difficult, since measures dont give
  permanent immunity .

The most NPIs can do is eliminate the overshoot
inherent in an unmitigated epidemic.
4
Recent reviews

• WHO working group (EID 2006) on National
  measures
• recommended hand hygiene, and that other measures
  considered based on circumstances.
• but highlighted the very limited evidence base
  for the community impact of most non-pharm.
  measures (e.g. avoiding crowding, school closure,
  hand hygiene, masks, travel restrictions).
• IOM report on reusability of face masks concluded
  more research critically needed to evaluate the
  effectiveness of face mask use.
• IOM report on community mitigation
• The evidence base is scant for use of case
  isolation Neither modeling nor historical
  analyses provide support for these
  interventions.
• The evidence suggests a role for community
  restrictions ... but does not allow for
  differentiating ... specific types of community
  restrictions.
• ... any discussion of using these interventions
  should consider not only their potential health
  benefits, but also their likely ethical, social,
  economic, and logistical costs.

5
Closing the evidence gap

• CDC NPIs studies on seasonal flu 8 projects,
  now in 2nd year looking at masks, hand hygiene
  etc.
• Analysis of historical pandemic and seasonal flu
  data.
• Surveys of public attitudes to NPIs and what
  measures people may take spontaneously (key issue
  did people modify behaviour in past
  pandemics?).
• Other studies.

6
Learning from the past 1918

• Very different epidemic patterns seen in
  different US cities in fall 1918 (much more
  variation than UK).
• Timing and nature of public health interventions
  varied between US cities.
• Can public health interventions provide a
  plausible quantitative explanation of the
  variation between US cities?
• What if? measures hadnt been imposed, or were
  imposed earlier
• 3 papers recently Bootsma Ferguson PNAS,
  Hatchett et al. PNAS, Markel et al., JAMA.

7
Trends in mortality

• Both peak and total mortality weakly correlated
  with timing of epidemic and previous years
  mortality.
• Peak mortality strongly correlated with early
  interventions.
• Peak mortality strongly correlated with presence
  of 2 autumn peaks, total mortality weakly so.

8
Results of 1918 analysis

• Public health measures explain 1918 pattern well.
• San Francisco, St Louis, Milwaukee and Kansas
  City had most effective policies (gt30 drop in
  transmission).
• But measures often started too late, always
  lifted too early.
• Also evidence of spontaneous behaviour change.

9
A study of the household use of masks
C Raina MacIntyre (1), Simon Cauchemez (2),
Dominic Dwyer (3), Holly Seale (1), Mary Iskander
(1), Pamela Cheung (1), Gary Browne (5), Michael
Fasher (6), Robert Booy (1), Zhanhai Gao (1),
Noemie Ovdin (1), Neil Ferguson(2).
Affiliation's 1. Discipline of Pediatrics
and Child Health, faculty of Medicine, University
of Sydney and National Centre for Immunisation
Research and Surveillance of Vaccine Preventable
Diseases, Childrens Hospital at Westmead, 2.
Imperial College, London, UK 3. Centre for
Infectious Diseases and Microbiology, Westmead
Hospital, Westmead, 5. Emergency Department,
Children's Hospital at Westmead, Westmead, NSW,
6. Wenwest Division of General Practice
10
Study design

• Recruitment
• Sydney, winter/spring of 2006 and 2007
• Families of children presenting to the emergency
  department and general practice with ILI a temp
  of gt37.8 and at least one respiratory symptom.
• At least 2 well adults in household.
• Intervention
• Random allocation of the 2 adults to one of three
  groups surgical mask, P2 mask, and control
  groups.
• 1-week follow up
• Incidence of ILI (phone call)
• Adherence to mask use (phone call)
• Nose throat viral swab obtained for PCR testing
  for influenza and other respiratory viruses (at
  recruitment for index case visit for secondary
  cases).

11
Outcomes

• Primary outcome presence of ILI or respiratory
  virus infection within 1 week of enrollment
• Intention to treat analysis.
• Secondary outcome time lag between recruitment
  and infection
• Important to assess the impact of time-dependent
  variables such as adherence
• Important to demonstrate a temporal association
  between mask use and reduction in the risk of
  infection
• Multivariate Cox proportional-hazard survival
  analysis, with random effects for household
  clustering Viboud et al, BMJ, 2004.

12
Intention to treat analysis
Control All mask All mask All mask
N() Total100 N() Total186 RR P
ILI 16 (16) 33 (18) 1.11 (0.64-1.91) 0.75
Laboratory confirmed 3 (3) 14 (8) 2.51 (0.74-8.52) 0.19
No significant reduction in incidence due to mask
use
13
Compliance
Low - 30 no significant difference between
mask groups.
compliant
Day of mask wearing
14
Survival analysis time lag from recruitment to
infection
Hazard Ratio of infection (95 CI) P value
Daily adherence with surgical or P2 mask 0.26 (0.09-0.77) 0.015
Nb adults 1.07 (0.66-1.71) 0.80
Nb siblings 0.86 (0.55-1.35) 0.52
Index lt5 years old 0.88 (0.41-1.89) 0.75
Frailty 0.005

Daily adherence with surgical or P2 mask 0.32 (0.11-0.98) 0.046
Nb adults 1.13 (0.71-1.81) 0.60
Nb siblings 0.80 (0.51-1.27) 0.34
Index lt5 years old 1.02 (0.46-2.24) 0.96
Frailty 0.004
Assumption on incubation period
1-day incubation period
2-day incubation period
15
Mask study summary

• No significant impact of mask use on
  transmission of respiratory virus during winter
  seasonal outbreaks. Low compliance.
• Compliant mask use was associated with a
  reduction in the hazard ratio of infection.
• If compliance is higher in during a severe
  pandemic or an emerging disease outbreak, mask
  use might reduce the risk of transmission.
• Study limitations
• Relatively small population size - underpowered
  to detect reductions in incidence smaller than
  75 inconclusive comparison of surgical and P2
  masks.
• Effect needs to be confirmed for the household
  context and investigated for other settings. More
  trials needed.

16
Whats next
Angus Nicoll, ECDC class dismissal/school
closure Ben Schwartz, CDC community mitigation
in the US
17
Back up slides
18
Recruitment
19
Intention to treat
Control Surgical Surgical Surgical P2 P2 P2
N() Tot.100 N() Tot.94 RR P N() Tot.92 RR P
ILI 16 (16) 19 (20) 1.29 0.69-2.31 0.46 14 (15) 0.95 0.49-1.84 gt0.99
Laboratory confirmed 3 (3) 6 (6) 2.13 0.55-8.26 0.32 8 (9) 2.90 0.79-10.60 0.12