The Royal Veterinary College The Swiss Federal Veterinary Office Schweiz. Arch. Tierheilk. 2008
Vector-borne viral disease of ruminants 24 known serotypes, member of genus Orbivirus Transmitted by midges of genus Culicoides Abundant and widespread in Switzerland Vector-free period December to April (2007-8) Sheep: acute disease, high morbidity + mortality Cattle + goats: usually subclinical (viral reservoirs) BTV-8: distinct clinical signs + mortality in some cattle
1998 onwards: BT causing outbreaks in Europe Largest epidemic of BTV-8 in 2006 affecting: The Benelux, France, Germany Between September 2007 and January 2008: Denmark, Switzerland, Spain, UK, Czech Republic
2003: Nationwide serological survey Establishment of entomological trapping sites Sentinal herd early warning system 2006: Northern European Outbreak Confirmed BTV-8 Absence 2007: Bulk milk testing Passive clinical surveillance Determined via Scenario Tree Analysis
Bulk milk testing: Monthly sampling of 208 herds Higher risk zones (vector presence) July 2007- June 2008 Switzerland divided into 16 BT regions (EC 1266/2007) and 11 farms sampled monthly Randomly chosen + verified by CVS Liechtenstein an additional region
Passive Clinical Surveillance: Intensified FVO + CVS launched information campaign (August 2007) www.bluetongue.ch Clinical Sentinel Surveillance Initiated in sheep in November 2007 Spatially focused on cantons of Basel-Landschaft, Basel-Stadt, Solothurn, Aargau (previous cases)
BTV-8: spread rapidly + high seroprevalence What if low prevalence at start of epidemic? eg: few infected vectors introduced Regional differences in surveillance intensity may bias the observed spatial distribution of cases Also: Regional prevalence = spatial differences in expected infection pressure Therefore important to analyse surveillance systems + determine regional differences when determining spatial prevalences
Descriptive + spatial analysis of BTV-8 surveillance schemes in Switzerland Determine regional prevalence estimates of BTV-8 infection Nationwide study initiated for calves born in the autumn of 2008 (following vaccination scheme)
Clinical Suspects: ELISA Real-time RT-PCR (Martin Hofmann) +ve in both tests Bulk milk: Monthly samples from quality testing labs Indirect ELISA +ve samples tested 2x EDTA-bloods from all cows Positive farms replaced Sheep Surveillance: National Health Service for Small Ruminants Voluntary programme Monthly reporting of clinical signs 28 participating farmers
Never been used to examine the spatial distribution of surveillance intensity Rather than case + population data (to author s knowledge) Spatial variability in Swiss BTV-8 surveillance evokes use of such a technique
SaTScan Software: Assessment of spatial clustering using spatial scan statistic Independently All surveillance methods combined Alternative hypothesis: Elevated risk of a farm being included in BTV-8 surveillance within the window, compared to outside Poisson model: Cases = farms in surveillance method Population at risk = all cattle, sheep and dairy farms in Switzerland Circular scanning window; radius 0-50% of pop. at risk
2 different methods should be combined in any cluster analysis: reduce the likelihood of detecting false +ve applied Moran s I (global cluster test) to confirm the presence of clustering. GeoDa 0.9.5-1 Software: Evaluation of global spatial autocorrelation (county level) Observed test statistic compared to Monte Carlo Randomization distribution (999 simulations) = Monte Carlo p-value for significance Prevalence Estimates: Binomial predictive model using Beta function in @Risk Run using Monte Carlo simulation of 1000 iterations Used results from cluster analysis
October 2007 - February 2008: 12 cattle +ve from 3 different farms: 9 clinical suspect cases, 3 bulk milk serology. Four cases in the Cantons Basel-Stadt, Basel-Land and Solothurn occurred in a confined area, whereas two cases found by bulk milk serology were distant from this area. 207 cattle farms tested using the bulk milk surveillance programme 10 false positive farms were produced +ve for bulk milk antibodies, -ve follow-up tests.
Bulk Milk Clinical Suspects BTV-8 Positive Farm False-positive Farm Sheep Surveillance - 10 cantons with no sheep surveillance farms Negative Farm Sheep Surveillance Farm
Clinical suspects from a total of 122 farms were reported 119 farms negative, 3 positive for BTV-8 92 reports were for cattle, 26 for sheep, 2 for goats, and the 2 for buffalo. Highest reports were Bern (18%), Basel-Landschaft (13.9%) and Solothurn (11.5%) The Canton of Jura exhibited the lowest number of reports (0.8%) followed by Valais and Neuchâtel(1.6%) 1st case late October 2007 - clinical surveillance. High number of clinical suspects Oct-Nov: release of BT documentary 11 false +ve bulkmilk results Decreased reporting towards the end of 2007 start of vector free-season
Spatial scan statistic: Two statistically significant clusters for all surveillance methods combined: Cluster A(radius 22.8 km, P=0.001, RR=5.85) incorporated Aargau, Basel-Landschaft, Basel Stadt, Jura and Solothurn Cluster B (radius 4.9 km, P=0.002, RR=28.21) was based in Ticino. Global spatial autocorrelation: Moran s I statistic 0.164 (P = 0.001) Indicating significant degree of spatial dependence in the surveillance data.
Surveillance Farm Cluster A
Prevalence estimate within Cluster A: Significantly higher: 9.62% (95% CI= 3.25%-18.85%) compared to: 0.98% (95% CI=0.2%-2.22%) outside this cluster. 4 BTV-8 positive farms were included in Cluster A Estimated number of infected farms within Cluster A is 136 (95% CI= 46 267) out of a total of 1,415 farms. Estimated number of infected farms outside Cluster A is 494 (95% CI= 101 1,175) from a total of 50,365 farms.
Surveillance Method Species Included Moran s I Statistic 1 Obtained (I) p-value SaTScan Spatial Scan Statistic2 Radius (km) Relative Risk Surveillance Farms in Cluster Cantons Bulk milk All (Cattle) 0.0195 nsig 4.9/- 45.5/- 7 TI/- Sheep Surveillance Clinical Suspect Cases All (Sheep) 0.1409 sig 14.4/- 18.6/- 7 AG, BL, SO Cattle Only 0.1269 sig 25.7/* 6.4/- 21 BE, LU, NI, OB, UR, VS Sheep Only 0.0975 sig - - - - Goats Only -0.0100 nsig - - - - Bulk milk and Sheep Surveillance All Surveillance Types Cattle and Sheep All (Cattle, Sheep and Goats) All (Cattle and Sheep) All (Cattle, Sheep and Goats 0.1847 sig 18/8.3 14.2/97.3 26/4 AG, BL, BS, SO/ GE 0.2299 sig 18/8.3 15.4/86.3 31/4 AG, BL, BS, SO/ GE 0.0864 sig 4.9/- 42.8/- 7 TI 0.1639 sig 22.8/4.9 5.85/28.21 50/7 Ag, BL, BS, JU, SO/ TI
Swiss BT surveillance between October 2007 and April 2008 confirmed the presence of BTV-8 +ve cattle. 3 clinical cases occurred in a confined area, 3 cases found with bulk milk serology were widely distributed. Clinical surveillance in sheep on 28 farms no +ve No +ve goats (not technically true) Importance of underlying distribution of farms: Largest number of reports from Bern - cattle and sheep populations dictate this. Jura and Neuchatel (lowest and 2nd lowest reports) even when merged together, still contain the lowest population of sheep farms of all the BT-regions.
Expected clustering where increased surveillance activities were : purposefully initiated, in the case of clinical sentinel surveillance in sheep (or a proactive CVS) expected, due to the presence of BT cases within that region and aimed at testing this statistically.
Area of higher surveillance corresponding with the area of higher BT prevalence Cluster A incorporates 4 out of the 6 BT positive farms Sheep surveillance data only also confirmed a cluster in this area Increased intensity of surveillance in the area + corresponds with the original aim of the sheep surveillance All clinical suspect reports also produced a cluster in this area Clustering of clinical suspect data, as expected due to an increased disease awareness in the area, could be shown to be statistically significant. The Moran s I statistic for all surveillance methods combined Geographic variation in surveillance farms did not follow underlying distribution of farms in Switzerland - clustering of the surveillance within a region.
Cluster B - Canton of Ticino: Bulk milk testing data in the spatial scan statistic also resulted in the same cluster Role of bulk milk surveillance spatial distribution in cluster. 7 of the 13 bulk milk farms in Ticino included in cluster. These farms were selected by the CVS, as all randomly selected farms were replaced with farms believed to be at higher risk for BT. Moran s I statistic for bulk milk testing does not support the presence of clusters. Cluster is likely an artefact due to the high sensitivity of this method (Kulldorff et al., 2003).
Clinical suspect reports from cattle farms 6 significant clusters found All except 1 very small and comprised 2-7 surveillance farms Moran s I test was also significant for this dataset Spatial autocorrelation present in the dataset. Local effect of single veterinarians as well as relationship between neighbouring farmers reporting clinical suspects.
Significantly higher BTV-8 prevalence inside Cluster A compared to outside the distribution of surveillance methods, is more intense in regions of higher risk for BTV-8 positive farms. High level of surveillance intensity for BT all over Switzerland Surveillance was significantly higher where cases were detected early in 2007 Additionally, the prevalence estimates are higher for that area. In contrast with UK findings, where the prevalence of infected herds within 10 km of the first 4 cases of BT identified was determined to be 58.8% (DEFRA, 2007): Estimated prevalence of infected farms was much lower in Switzerland, even in the cluster region. Limited incursion of infectivity through vectors or hosts, or a less effective transmission prior to case detection in Switzerland compared to the situation in UK.
Swiss livestock owners Cantonal Veterinary Services Swisslab, Zollikofen Qualitas, Zug Milk quality laboratories Christian Griot and Eric Breidenbach Barbara Wieland and Martin Reist