Lessons from FMD by mikesanye


									   Lessons from FMD 2001

           Nick Honhold
   Veterinary Epidemiology Unit
(TVI, Epidemiology Section, Carlisle)
FMD UK 2001: Location
of infected premises in
Great Britain

Total of 2,026 IPs

Widespread across GB,
particularly on west side

Dispersion largely followed
the seasonal sheep trade

Biggest cluster was Cumbria
and Dumfries/Galloway,
1,069 IPs (>50%),
related to Longtown Market

Biggest sheep market in Europe
       Where did the disease come from?

• Source unknown, but untreated swill certainly being
  fed at Burnside Farm

• Farmer failed to report disease which he must have
  known to be present

• “Bone in” hams illegally imported from China found in
  nearby warehouse in April

• Airborne spread to sheep on nearby farm
  (but NOT contiguous)

• Other “conspiracy” theories have no credibility
    Why did 15 sheep lead to all that bother?
• See paper by Mansley et al. for forensic tracing of the
  movements of these sheep
• Moved in vehicle with around 100 other sheep
• Major problem is that they went to Longtown Mart,
  the biggest sheep market in Europe at a busy time of
• “Rolling” market, never empty so constant infection
• Mixing of sheep in market and outside
• Staff moving between pens
• “Mouthing” of sheep to check teeth
• Dealers from all over UK spread infected sheep over
  long distances rapidly
      Infectivity of infected animals and farms

The incubation period for individual animals in FMD can range
from 2 to 14 days for contact spread. Usually 5-8 days.

For farm-to-farm spread by airborne virus the doses are likely to
be much less and so a 4 to 14 day incubation period can be
      Infectivity of infected animals and farms

Sheep excrete the maximum amount of airborne virus just
before the peak of viraemia which generally occurs at around 1
or 2 days before the clinical phase of the disease.

In pigs or cattle, airborne virus excretion occurs maximally
during the early clinical phase of disease but that is generally
after the peak of viraemia.
      Infectivity of infected animals and farms

Air borne spread was not significant in FMD UK 2001.

In the absence of significant airborne spread, the time when
vesicles are rupturing i.e. around 2 to 3 days into the clinical
phase, is when the greatest amounts of virus are liberated into
the environment. This is the time when a farm will be most
heavily contaminated and infectious.
                           FMD in sheep

• Hidden, occult, silent, sub-clinical, inapparent??? NOT in my

• Undetected? YES because clinically looks like other diseases

• If it is there, you know what to look for, and you look at enough
  animals, you will find it if the disease is still active

• If there isn’t enough to find clinically, it is unlikely the disease will

• May move very slowly through extensive flocks or even die out

• Goats may be hard to spot but still have lesions at some point
                Methods of spread of FMD
• Well studied/documented methods of spread
  – Infected shedding animals directly to other animals
  – Vehicles which have carried infected shedding animals
  – Air borne over short and substantial distances

• Poorly studied/documented methods of spread
   – Non-livestock vehicles moving between farms
   – Personnel
   – How does virus transfer from vehicles/personnel to livestock
   – Biosecurity measures

• BUT: in both 67/68 and 2001, 70-80% of all cases were
  attributed to “local spread” i.e. unidentified
              How did FMD spread in 2001?

• Air- borne : Long distances possible, plume.
   – Rare in 2001

• “Over the fence”: nose to nose contact between susceptible
  stock or short distance aerosol spread (<100m)

• Through the gate:
   – Livestock early on
   – Fomite spread: vehicles, personnel
   – Can be from next door farm, may be long distance
   – Commonest route
                    Distance categories

• Contiguous:
   – Possible source on next door farm within specified period.
   – Over the fence or through the gate

• “Local”:
   – Disease within 3km in specified period, but not next door.
   – Through the gate

• “Long distance”:
   – No disease within 3km within a specified risk period.
   – Others have used 10km
   – Through the gate
                    Contiguous spread

• May be over the fence or through the gate

• Two sets of analysis suggest that around two thirds of
  contiguous farms that were infected were infected through the

• What was the source of this infection?
  – the neighbouring IP?
  – the weight of infection in the area?

• Biosecurity could have prevented some of it
Focus was on contiguous spread during 2001.
   • 17% chance of a CP becoming an IP

But most spread was “local” with no known source but infection within
    • Lower risk at farm level than CPs, but because more farms,
       equal or greater number of IPs in this category

Significant numbers of long distance transmission events
into clean country
     • Each formed a “spark” around which other cases
         often occurred
     • Origin of the clusters. No sparks, no clusters
      Control measures used during FMD UK2001: Others

• National movement controls. Introduced on 23-Feb-01

• Protection zone: 3km radius zone around an IP. Surveillance
  patrols of all premises, strict movement controls.

• Surveillance zone: 10km radius zone around an IP

• Biosecurity.
   – Informal at first with guidelines issued to farmers.
   – Restricted Infected Areas (“Blue boxes”) introduced from July
     with strict controls. First introduced in Cumbria on 07-Aug-01.

• These are VITAL control measures but hard to quantify
Location of the clusters
included in the analysis

       892 IPs

       1,134 IPs
                    All GB

Looks convincing?            Nick Honhold
All GB

         Nick Honhold
All GB

         Nick Honhold
GB without Cumbria

                     Nick Honhold
GB without Cumbria   Cumbria

                               Nick Honhold
                   What does this mean?

Killing cows and sheep in Devon controls disease in Cumbria?
     Accuracy of diagnosis by DVO
Percentage of farms from which samples were
  submitted that had positive lab result:

• Leeds 93%, Carlisle 89%

• Exeter 69%, Ayr/Dumfries 65%, Cardiff 59%

• Worcester 36%, Gloucester 31%

Source: NAO
Location of the clusters
included in the analysis

       892 IPs

       101 IPs

       185 IPs

   Largest clusters in
    Field data extracted from Disease Control System database

    For each IP:    First lesion date
                    Report date
                    Species with oldest lesions
                    Confirmation date
                    Slaughter completion date

For each DC cull:   Authorisation date
                    Category of DC cull: CP / NonCP
              Composition and overall culling intensity

                                    Cumbria               Settle
No of IPs                             892         185      101

% IPs, Cattle with oldest lesions    66.9%        52.4%   70.3%

% IPs, Sheep with oldest lesions     33.1%        47.6%   29.7%

Overall DC:IP ratio                   1.07         4.68   4.42
        Overall CP:IP ratio           0.74         3.19   3.84
        Overall NonCP:IP ratio        0.33         1.49   0.56
The clusters are not homogeneous

• Different composition of animals with oldest lesions on the IP
   – Cumbria and Settle mainly cattle
   – South-West 50:50 cattle and sheep

• Different delays from report to slaughter
   – Cumbria and South-West initially high
   – Settle low from the start

• Different CP:IP culling intensities
   – South-West and Settle high, Cumbria low

• Different NonCP:IP culling intensities
   – South-West high, Cumbria and Settle low
        Parameters calculated for weekly periods for each cluster

Estimated dissemination rate     No of IPs by first lesion date in this week
(EDR)                          No of IPs by first lesion date in previous week

First lesion to slaughter        Total days first lesion to slaughter for IPs
(FLtoS)                         No of IPs with first lesion date in that week

FLtoSLag                                FLtoS in the previous week

DC:IP ratio                           Total No. of DC premises culled
                                       No of confirmed report cases

CP:IP ratio                                   No. of CP Culls
                                       No. of confirmed report cases

NonCP:IP ratio                              No. of NonCP Culls
                                       No. of confirmed report cases
Linear regression analysis to show effect of control measures on EDR

      EDR = Constant + (Coefficient x Control measure)

      FLtoSLag: Rapid slaughtering affects EDR by
                preventing spread, so effect is delayed
                by incubation period, approx 7 days

           DC culling affects EDR by removing farms
           already incubating the disease, so effect is rapid

      CP:IP ratio:     CP culling intensity

      NonCP:IP ratio: NonCP culling intensity
            Regression statistics for EDR vs. FLtoSLag

             n     Constant      Coefficient       r2           p

Cumbria     31        0.05          0.26         0.48      <0.0001

            16        0.50          0.15         0.15          0.139

Settle      10       -0.23          0.32         0.53          0.017

Reducing FLtoS will reduce EDR significantly in Cumbria and Settle

Is South-West different because more IPs where sheep had oldest

Despite differing equations, EDR falls below 1 when FLtoS in
previous week is between 3-4 days in all three.
             Regression statistics for EDR vs. NonCP:IP

              n     Constant       Coefficient         r2            p

Cumbria      31        1.15           -0.29          0.05        0.251

             16        1.72           -0.44          0.27        0.038

Settle       10        1.35           -0.37          0.36        0.064

 Statistically significant effect in South-West and a biologically
 significant r-squared in Settle and South-West
                  Regression statistics for EDR vs. CP:IP

              n       Constant      Coefficient         r2           p

Cumbria      31         1.16           -0.08          0.08         0.130

             16         1.49           -0.09          0.07         0.325

Settle       10         1.82           -0.21          0.18         0.219

  No statistical or biologically (low r-squared) significant effect in
  any of the three clusters
                  Summary of analysis

• There is a statistically and biologically significant
  relationship between EDR and FLtoSLag in two of the
  three clusters

• In some circumstances there is a statistically and
  biologically significant relationship between NonCP:IP
  ratio and EDR

• There is no statistically or biologically significant
  relationship between CP:IP ratio and EDR in any of the
  three clusters
• An FMD outbreak on this scale cannot be analysed using
  aggregated data. Clusters must be analysed separately,
  not treated as if they are homogeneous

• Reducing time from first lesion to slaughter (FLtoS) is a
  key factor in controlling the disease

• Culling of premises with known dangerous contacts can
  contribute effectively to control

• There is no evidence from the analysis presented that
  culling of contiguous premises was a critical factor in
  bringing the outbreak under control
Rapid case finding and killing of IPs can control an
intense outbreak of FMD without recourse to extensive

Good biosecurity (with all that means) is vital in any
control policy

In 2001, when FLtoS was reduced below 3.5 in the
presence of reliable biosecurity, then the EDR would be
below 1 and the disease would be extinguished
Dissonance between field data and experimental results

   Experimental results leading to adoption of the contiguous
   culling policy came from mathematical models

 • Used early data
    – Incomplete
    – Possibly could not take into account the dynamic
      nature if the problem
    – Data was not accurate enough

 • All models contain assumptions
Dissonance between field data and experimental results

 • The models explicitly excluded any idea of an “on farm”
   epidemic i.e. no build up of infectivity over time. Once a
   farm became infectious to other farms, it was regarded as
   maximally infectious from that time

 • An infected farm became maximally infectious to other
   farms 1, 3, or 4 days after it became infected i.e. probably
   before any animal had lesions

 • This builds into the models a very high level (speed and
   amount) of farm to farm infectivity
Infectivity is related to area under the curve before culling
Model assumptions must be rigorously checked and
made very transparent. In 2001 we had to dig for them.

High farm to farm infectivity rates made CP culling
essential in the models, a self-fulfilling prophecy

“Witch ball epidemiology” cf. Norfolk elephant scarer

Epidemiology is the study of disease in populations not

It is better to be approximately right than precisely wrong
        Problems identified by Anderson Inquiry

• The Lessons to be Learned Inquiry highlighted the
  problems of data gathering, quality and analysis during
  the outbreak, particularly in the early stages

• Decision makers did not have access to timely, accurate
  and relevant information about the ground reality

• Recommends that where possible responsibility should be
  devolved to local centres
                      For the future

• Clusters should be analysed locally
   – by people who know the local situation

   – data can be cleaned close to the time and place of

   – preserve and highlight important differences

   – avoid inappropriate aggregation of data

• Key indicators can and should be monitored and
  reported in real time

• The value of contiguous culling must be re-examined
         Key indicators for local real time analysis
• EDR or case ratio (cases today/cases yesterday) indicates
  the course of the epidemic

• First lesion to slaughter (FLtoS)
   – how the epidemic is likely to behave in the following
   – how quickly affected animals are being found and

• Age of oldest lesion (First lesion date to report date): this
  indicates how quickly disease is being found

• Report to Slaughter. Speed of culling of affected animals

• Non-contiguous premise:IP (NonCP:IP) ratio can also be a
  useful indicator and should be monitored
   Evaluation of the application of veterinary
judgement in the pre-emptive cull of contiguous
premises during the foot and mouth epidemic in
                 Cumbria, 2001

  N. Honhold, N.M. Taylor, A. Wingfield, P. Einshoj,
  C. Middlemiss, L. Eppink, R. Wroth & L.M. Mansley
               Contiguous culling policy

• March 29: all susceptible stock on contiguous premises
  (CPs) to be culled

• April 04: Exceptions allowed where a strong case can
  be made that animals have not been exposed

• April 26: Veterinary judgement allowed in culling
  decisions based on risk assessment. In particular, the
  50m rule is introduced. Intention would still be that
  sheep should be culled…….
                  Carlisle DC Section

•   3-4 vets
•   Assessed all DCs including CPs
•   Information gathered from field by vets
•   Electronic information not sufficiently reliable
•   Previous intelligence used
•   Farming is a complex business
•   Co-located with IP section for speed and info exchange
•   Close co-operation with the epidemiology section
•   Mapping and recording of each decision
Numbers of IPs in Cumbria and numbers with a contiguous
assessment, by month from May 2001.

mainly South Penrith
Number of assessments made on a contiguous property,
by land classification and stocking status
Contiguous land parcels assessed, by land classification and
action taken
Contiguous land parcels assessed, by land classification and action
taken; <50m stock to stock separation
Contiguous land parcels assessed, by land classification and action
taken; >=50m stock to stock separation
      How successful was the use of veterinary

• Hard to prove a negative if everything is killed

• There was controversy about apparent breakdowns
  in the policy with CPs that were either not culled or
  partially culled becoming IPs “shortly” afterwards.

• Had they been missed?
                          Temporal risk window

To be attributable to the IP against which it was assessed, the CP
must itself become an IP in a given period, the “Temporal risk window”

Animals start shedding virus the day before the first lesions develop
and will do so until the end of slaughter

Incubation period is 2-14 days

So temporal risk window is
       from 2 days after first shedding (day after first lesion date)
       to 14 days after end of slaughter
Temporal risk window
Contiguous land parcels assessed that were part of a CPH which
became an IP within the temporal risk window, by land
classification and action taken
     Success rate for veterinary judgement

                   586 assessments
                    on stocked land

529 “successes”      57 “failures”    90% success rate
Contiguous land parcels assessed that were part of a CPH which
became an IP within the temporal risk window, by land
classification and action taken
     Success rate for veterinary judgement

                   586 assessments
                    on stocked land

529 “successes”      57 “failures”

  20 complete       37 where stock    94% success rate
 depopulations          was left
No of assessments where stock were spared and number that
became an IP within the temporal risk window, by separation
distance between contiguous stock
     Success rate for veterinary judgement

                   586 assessments
                    on stocked land

529 “successes”      57 “failures”

  20 complete       37 where stock
 depopulations          was left

 26 with stock      11 with stock     98% success rate
separation >50m    separation <50m
     Success rate for veterinary judgement

                   586 assessments
                    on stocked land

529 “successes”      57 “failures”    90% success rate

  20 complete       37 where stock    94% success rate
 depopulations          was left

 26 with stock      11 with stock     98% success rate
separation >50m    separation <50m
                   11 real “failures”

• 6 occasions there is a more convincing link than the
  neighbouring IP
   – Vehicle
   – Personnel

• 3 occasions the farmers did not give complete details
  of links with the IP

• 2 occasions relating to one farm a specific decision
  was taken to leave stock because there were no
  neighbouring farms to be affected
A) Risk of a contiguous premises
   becoming an IP was quoted as being:
         • “1 in 5”
         •     17%

B) Risk to a farm is composed of two routes:
   1) Direct = “across the fence”
   2) Indirect = “through the farm gate”

C) So, for a contiguous premises:

      Overall risk = Direct + Indirect = 17%
No of assessments where stock were spared and number that
became an IP within the temporal risk window, by separation
distance between contiguous stock

        Not affected by distance between stock
   Overall risk = Direct + Indirect = 17%

       Indirect risk = 12%
       Direct risk   = 17 – 12 = 5%

Most risk of infection, even for contiguous
premises, was through the farm gate at
this stage of the epidemic.
For contiguous premises, most risk was indirect
Other work has shown that at least 50% of IPs
were on non-contiguous premises i.e. they faced
indirect risk only
Over 75% of all risk was “through the gate” not
“over the fence” at this stage of the epidemic
and possibly earlier
Biosecurity was the key to stopping the disease
once report to slaughter times had been brought
down to under 24 hours
What was the role of contiguous culling?
Milk tankers as a possible means of spread
•“Clinical impression”. Field epidemiology investigations
often showed that milk tankers were the strongest known
link between IPs

•Commonest form of movement between farms, particularly
early in the outbreak before silaging and other field work
became common

•Milk tankers carry all the risks of any other vehicle plus
may be carrying contaminated milk as well as faeces,
saliva etc.

•When DEFRA staff rode “shot gun” on tankers, problems
of milk spillage and poor biosecurity were seen

•Imposition of biosecurity (“blue box”) was related to end of
outbreak and included heightened measures on milk
tankers including pressure washers
          Milk tanker movements are complex

• 17% of collections were from 6pm to 6am, collections
  in all 24 hours. How do you clean a tanker in the dark
  at a lane end?

• Mother tankers

• Changing runs because if changing restrictions.
  Farms visited by more than one tanker

• Drivers change from day to day
          When is there a risk of transmission?

• Tanker must visit a shedding farm
   – Between estimated First Lesion date minus 4 days to Form
     A date
   – risk increases closer to Form A date and certainly significant
     once vesicular lesions are present

• Subsequent visit to farms on the same run may pose
  a risk of transmission
   – risk decreased by increasing time after visit to shedding farm
   – transmission may have occurred if receiving farm becomes
     an IP within 2-14 days after the visit
Essential elements for a risk visit and possible transmission by a milk tanker
              Conclusion of milk tanker
             matched case control study:

No evidence that there was a relationship between the
number of risk visits that a farm “received” and whether the
farm became an IP or not

This was true for various levels of risk. No “dose response”
             Possible reasons for lack of effect

• Milk tankers do not spread FMD. How credible is this?

• Milk tankers pose a risk of spread but the effect is too weak to
  detect with the analytical methods used

• Most contamination is picked up on roads between farms: risk is
  not related to the status of farms visited earlier on the run. How
  likely is this?

• Milk tankers do have the potential to spread disease but this risk
  is mitigated by biosecurity measures taken by individual farmers
  and drivers. Most likely explanation?

• Any others?
                       Overall conclusions
• Dairy farms formed a disproportionately high percentage of IPs

• Field investigations often implicated milk tankers as the strongest
  link between IPs

• The analysis undertaken does not show a systematic relationship
  between visits by a milk tanker to a shedding farm and risk of
  subsequent transmission

• The most likely explanation would seem to be that the risk is
  ameliorated in some cases, probably by biosecurity at the farm
  gate, both leaving the shedding farm and entering subsequent

• Biosecurity only needs to fail once
                  Biosecurity is the key

In future epidemics, if dairy farms are allowed to continue
selling milk, they should be targeted for enhanced, and
enforced, biosecurity levels from the start of the epidemic
This should be taken into account in contingency planning
                   Amass et al (2003)
•   Infected pigs experimentally
•   Pigs were clinically examined and sampled
•   Different biosecurity protocols undertaken
•   Then handled uninfected pigs and sheep

• If washed hands, changed boots, outerwear and
  gloves washed hands, transmitted infection to sheep
  but not pigs

• If showered, changed boots, outerwear and gloves
  did not transmit infection to sheep or pigs

• So seemingly small changes in biosecurity van have
  significant effects
                Could it happen again?

• Yes
• Less likely because of swill ban, possibly much less
• But remember CSF in GB 2000

• Movement standstill should decrease spread
• But what if a farmer doesn’t report it for 14 days
• Shortening “silent spread” is vital
            What should we do next time?

• Speed of case finding and culling is essential
• Unburied bodies are bad publicity but don’t produce
• Biosecurity heightened and enforced from the start
• Tracing prioritised
• Culling of veterinary assessed Dangerous Contacts
• Don’t leave an information vacuum: real time analysis
  of key parameters

• Automatic contiguous or area culling?????
  Apart from model predictions, there is no evidence
  that this was material in controlling the outbreak

To top