Download 1. Background on HPAI H5N1 control policies in

1
Evaluating the control of HPAI H5N1 in Vietnam: virus transmission
2
within infected flocks reported before and after vaccination.
3
4
5
Technical Appendix
1. Background on HPAI H5N1 control policies in Vietnam
6
The two initial epidemic waves occurred during the period of December 2003 to May
7
2005 during which vaccination was not included as part of the national disease control
8
program. Vaccination-based control policy interventions are officially put in place in
9
September 2005 until this date. The infected premises (IPs) of the third HPAI H5N1
10
epidemic wave are reported in October 2005 until December 2005. The fourth and
11
fifth epidemic waves occurred during November 2006 to November 2007 when
12
countrywide vaccination-based disease containment policies were in place. During the
13
first epidemic wave, depopulation dominated the portfolio of disease control measures
14
and the extent of its application was not systematically documented [1]. Overall, the
15
implementation of depopulation during this period is expected to have varied in
16
nature, quality, time and space across the country in contrast to later epidemic waves
17
and this would have contributed to a higher heterogeneity of within-flock
18
transmission estimates compared with those for the second wave. In addition, at that
19
time, the laboratory diagnostic capacity of the country was insufficient, which
20
increased the confirmation time of suspected IPs. This in turn vindicated the
21
application of rapid depopulation of the flocks. Finally, the technical guidelines for
22
prevention and control of avian influenza were released 6 months after the first
23
outbreaks of the first wave were reported. During the third outbreak wave,
24
depopulation upon identification of an IP was applied and flocks within different
1
25
provinces were already being vaccinated. Surveillance data available for this period
26
did not allow unbiased ascertainment as to whether the IPs recorded in the dataset
27
were in an area where vaccination was being put in place.
28
During the second epidemic wave (2004-2005), depopulation was implemented at the
29
village level and poultry had not been vaccinated. IPs and pre-emptively culled farms
30
were allowed to repopulate their flocks after a period of 60 days but data on the
31
proportion of farms which repopulated are unreliable.
32
During the 2006-2007 epidemic (fourth epidemic wave) a protection zone (PZ) was
33
established which equals the area of the village where the IP was located. Within the
34
PZ the IP was depopulated and movement restricted. For this period, pre-emptive
35
depopulation of flocks in the PZ was no longer compulsory as was the case during the
36
first to third waves; poultry farms within the PZ were offered the option of
37
compensated depopulation. By the time the first outbreak of the fourth epidemic
38
wave was reported, three vaccination campaigns had taken place and the fourth
39
campaign was ongoing. This campaign focused on poultry flocks located in 33
40
provinces, which had a history of being high-risk areas for HPAI H5N1 outbreaks in
41
poultry. A H5N2 vaccine (Nobilis, Intervet) was made available at cost for
42
commercial breeders of great grandparent (GGP) and grandparent (GP) stocks. This
43
vaccine required primo-vaccination followed by a booster after a 4 week interval and
44
a second booster another 4 months later. An inactivated Re-1 strain, reassortant
45
H5N1 vaccine (Homologous Habin-Weike Reassortant Avian Influenza vaccine) was
46
used for vaccination of broiler ducks, the cost of which was fully borne by the
47
government. With this vaccine, ducks required a booster one month after primo-
48
vaccination and a second booster 6 months thereafter. The Homologous Habin-Weike
2
49
Reassortant Avian Influenza vaccine was also being used in chicken broiler flocks as
50
a single booster vaccine (i.e. two injections). In addition a H5N1 recombinant
51
vaccine (Trovac) had been introduced for use in chicken breeder hatcheries and
52
applied to day-old-chicks.
53
2. Data quality and data handling issues
54
Outbreak surveillance data generated during the period from December 2003 to
55
September 2008 is publically available at the World Organization for Animal Health
56
(OIE) which is officially notified by the Department of Animal Health of the Ministry
57
of Agriculture and Rural Development of Vietnam (DAH-Vietnam) [2]. The HPAI
58
H5N1 outbreak surveillance data considered for this study was kindly facilitated by
59
the Epidemiology Division of DAH-Vietnam. Data concerning flock type (i.e. species
60
reared), total poultry population initially at risk (i.e. flock size) and the number of
61
observed bird deaths were available.
62
Upon discussion with senior animal health officials and epidemiologists at DAH-
63
Vietnam we made the decision of excluding the data from the first (i.e. December
64
2003 to March 2004) and third (i.e. October 2005 to December 2005) epidemic waves
65
from the analysis. The criteria for exclusion were based on evidence about the lack of
66
consistency across the country in data recording for the first epidemic wave and in
67
disease control policies for the first and third epidemic waves[1]. Furthermore, data
68
aggregation was a caveat of the outbreak surveillance data generated during the first
69
epidemic wave while uncertainty with respect to regional vaccination status of IPs
70
recorded during the third epidemic wave were issues that supported the exclusion of
71
these data.
3
72
For the comparative investigation between periods of vaccine-based control versus
73
depopulation-based control we considered a dataset which included outbreak
74
surveillance data from the second (i.e. depopulation-based control; Period I) and
75
fourth (i.e. vaccination-based control; Period II) epidemic waves.
76
The dataset for Period I included infected premises (IPs) which were pre-emptively
77
depopulated as part of the control policy; however, these records were excluded from
78
the final dataset for analysis. None of the IPs of Period I had been vaccinated.
79
The official reporting data contained a total of 1005 IPs for the period 1st January
80
2005 to 1st May 2005 (second epidemic wave; Period I) and a total of 114 IPs for the
81
period from 16th November 2006 to 7th March 2007 (fourth epidemic wave; Period II)
82
with observed bird mortality due to H5N1 infection. During these epidemic periods a
83
total of 545 communes from 32 (50%) provinces of Vietnam were affected and in
84
90% (491/545) of affected communes the number of IPs with reported mortality
85
ranged from one to four. The majority of IPs during Period I (81%; n=818) and
86
Period II (96%; n=110) epidemic waves were located in provinces within the Mekong
87
River delta (in South Vietnam). For the Period II the dataset did not contain any IPs
88
located in provinces other than the Red River (North Vietnam) and Mekong River
89
deltas.
90
We were interested in obtaining unbiased estimates of the within-flock R0 therefore
91
the final dataset for analysis needed to contain unbiased records of estimates of
92
mortality and respective denominators for each IP. Therefore, and based on
93
discussions with the DAH-Vietnam officials, we anticipated recording biases of the
94
total mortality observed in some of the flocks particularly in those with flock sizes
4
95
greater than 1,500 heads. This lead to the inclusion in the final dataset for analysis of
96
924 flocks for Period I (i.e. 8% (81/1005) were excluded) and 106 flocks for Period II
97
(i.e. 7% (8/114) were excluded).
98
We could have introduced biases in our analysis for Period II (vaccination-based
99
control) if vaccinated flocks had been included; in this scenario, mortality in
100
vaccinated flocks would not be a good indicator of infection. According to
101
information provided by the DAH-Vietnam and information available at OIE (at the
102
World Animal Health Information Database (WAHID) Interface) none of the flocks
103
considered in the analysis for this period was listed as being vaccinated at the moment
104
when the outbreak was reported.
105
For the comparative investigation of the reproductive number of infection between
106
flocks of different type/sizes compositions we have classified infected premises of the
107
final dataset for analysis into discrete categories pertaining to their expected scale of
108
production. The Vietnam’s poultry sector has been historically sub-divided according
109
to different criteria [3-5] and the IP categorization used assumes 6 cut-offs based on
110
the frequency distribution of IPs by flock type and size composition (see Table 1).
111
This was also performed in order to take into account possible non-linearity of effects
112
of size during statistical analysis.
113
Finally, we argued whether mortality in duck flocks due to H5N1 might not be a good
114
indicator of virus transmission. Indeed, available evidence suggests that the initial
115
1997-2004 highly HPAI H5N1 viruses in domestic and wild waterfowl tended to
116
behave like low-pathogenic avian influenza viruses in other avian species. In addition,
117
experimental evidence specific to viruses isolated from Vietnam during outbreak in
118
2003 and 2004 show that those tended to replicate and transmit efficiently in ducks
5
119
but exhibited variable pathogenic potentials in ducks, ranging from the complete
120
absence of clinical disease (intravenous pathogenicity index [IVPI] 0.0 in ducks) to
121
severe neurological dysfunction and death (IVPI of 3.0 in ducks) [6]. However, recent
122
experimental evidence for viruses lineages circulating in Viet Nam from 2005 onward
123
– which corresponds to the epidemic periods analysed in the manuscript – suggests
124
that these were highly virulent in ducks , causing 100% mortality in 2-week old Pekin
125
ducks within four days of virus challenge [7, 8]. In addition, mortality figures caused
126
by these viruses are suggested to correlate well with mortality figures observed in
127
chicken and these viruses were much more virulent to ducks compared with viruses
128
circulating before 2005 [7]. Based on this recent evidence the decision was made to
129
include the outbreak data from duck flocks.
130
131
3. Estimation of the within-flock reproductive number of
infection, R0 based on the method of moments
132
The basic reproductive number R0 allows the quantification of the infection potential
133
of a population and is an averaged epidemiological property that summarises the
134
relationship between the effective contact rate (β) and the mean duration of
135
infectiousness (1/α) adjusted for the size of the initial population at risk. It is difficult
136
to estimate β when complete follow-up data for animals are not available [9]. In order
137
to do estimate the within-flock reproductive number, R0 we have assumed that the
138
same category of flock size/type would have the same degree of mixing as their
139
homologous flocks. In this scenario we estimated the within-flock R0 which will be
140
the average number of secondary cases created by a typical infectious bird from the
141
point it became infectious until the moment of flock depopulation.
6
142
The theory of moments of martingales applied to the general epidemic model allows
143
the estimation of the parameters for β and 1/α simultaneously [10]. This quantitative
144
approach takes into account the final size of an outbreak, the initial population at risk
145
and time. We have used the expression below to calculate the infection potential of a
146
flock up to the time point when it was culled, considering the following relationships,
147
ˆ f 

1 
s0f  0.5


ln
f
f
f

cT  N 0  cT  1  0.5 


148
where f is the infected flock.The parameter (θ) to be estimated is conceptually equal
149
to β/α which, as mentioned above, are the primary ingredients for the estimation of R0.
150
The necessary parameters for the estimation of θ were the total number of birds at risk
151
in the flock at the start of the outbreak ( s 0f  N 0f  1 ), the number of birds in the flock
152
at the start of the outbreak ( N 0f ) and the cumulative number of observed deaths (new
153
cases) when the infection chain was stopped by culling ( cTf ). Assuming there was a
154
single case introduced to the flock, the actual initial population at risk at the start of
155
the outbreak for a given flock was estimated as N 0f -1, with N 0f being the recorded
156
total population of the flock at the moment when infection was reported. Previous
157
studies have shown that mortality observed on a given day had a positive linear
158
association with time after initial infection and that without any intervention, entire
159
flocks would die within 12 days of introduction of the HPAI virus to the infected farm
160
[11]. However, the aim of our analysis was to estimate the efficacy of the control
161
policy and not to examine the biological properties of transmission.
162
The R0 of an IP was calculated by multiplying θ by the total number of birds initially
163
at risk in the flock. An estimator of the standard error of θ has also been previously
7
164
described [10] and was used for the estimation of confidence intervals around the
165
estimated R0 values.
166
4. References
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Pfeiffer DU, Minh PQ, Martin V, Epprecht M, Otte MJ: An analysis of the
spatial and temporal patterns of highly pathogenic avian influenza
occurrence in Vietnam using national surveillance data. Vet J 2007,
174(2):302-309.
Office International des Epizooties. Update of Highly Pathogenic Avian
Influenza in Animals (Type H5 and H7)
[http://www.oie.int/downld/avian%20influenza/A_AI-Asia.htm]
Delquigny T, Edan M, Nguyen DH, Pham TK, Gautier P: Impact of Avian
Influenza Epidemic and Description of the Avian Production in Vietnam.
In. Rome, Italy: Food and Agriculture Organization of the United Nations;
2004: 119.
Rushton J, Viscarra R, Bleiche EG, McLeod A: Impact of influenza
outbreaks in the poultry sectors of five South east Asian countries
(Cambodjia, Indonesia, Lao PDR, Thailand, Viet Nam) outbreak costs,
responses and potential long term control. In.: FAO, Rome, Italy; 2004: 25.
ACI: The Economic Impact of Highly Pathogenic Avian Influenza Related Biosecurity Policies on the Vietnamese Poultry Sector In.
Bethesda, Maryland: Prepared for the Food and Agriculture Organization of
the United Nations and the World Health Organization, Agrifood Consulting
International; 2007: 1013.
Sturm-Ramirez KM, Hulse-Post DJ, Govorkova EA, Humberd J, Seiler P,
Puthavathana P, Buranathai C, Nguyen TD, Chaisingh A, Long HT et al: Are
ducks contributing to the endemicity of highly pathogenic H5N1 influenza
virus in Asia? J Virol 2005, 79(17):11269-11279.
Pantin Jackwood MJ, Suarez DL, Swayne DE: Increased pathogenicity of
H5N1 Vietnam viruses in ducks. . In: Proceedings of the 110th United States
Animal Health Association Meeting: October 2006 2006; Minneapolis,
Minnesota.; 2006.
Pantin Jackwood MJ, Suarez DL: Pathogenicity of H5N1 HPAI viruses
from Vietnam in chickens and ducks [abstract]. In: Workshop on Avian
Influenza: Research to Policy, June 16-18, 2008. Hanoi, Vietnam; 2008.
Diekmann O, Heesterbeek JAP: Mathematical Epidemiology of Infectious
Diseases: Model Building, Analysis and Interpretation. Chicester, West
Sussex: John Wiley & Sons Ltd; 2000.
Becker N: Martingale methods. In: Analysis of Infectious Disease Data.
London: Chapman and Hall Ltd; 1989: 224.
Yoon H, Park CK, Nam HM, Wee SH: Virus spread pattern within infected
chicken farms using regression model: the 2003-2004 HPAI epidemic in
the Republic of Korea. Journal of veterinary medicine 2005, 52(10):428-431.
8