Preventive Veterinary Medicine 78 (2007) 1223 www.elsevier.

com/locate/prevetmed

Reporting of suspect cases of foot-and-mouth-disease during the 2001 epidemic in the UK, and the herd sensitivity and herd specicity of clinical diagnosis
Melissa McLaws a,1, Carl Ribble a,*, Craig Stephen b, Bruce McNab c, Pablo Romero Barrios a,1
Department of Population Medicine, University of Guelph, Guelph, Ont. N1G 2W1, Canada b Centre for Coastal Health, 900 5th Street, Nanaimo, BC V9R 5S5, Canada c Ofce of the Chief Veterinarian for Ontario, Ontario Ministry of Agriculture, Food and Rural Affairs, 1 Stone Rd., W., Guelph, Ont. N1G 4Y2, Canada Received 25 August 2005; received in revised form 17 July 2006; accepted 17 September 2006
a

Abstract We described the clinical diagnostic process utilized during the 2001 epidemic of foot-andmouth-disease in the United Kingdom (UK), and considered it as a series of diagnostic tests. Premises were classied according to these diagnostic-test results and actual disease status, determined by the reference test, which in this case was one or more internationally accepted laboratory tests. The herdlevel sensitivity (HSe) and herd-level specicity (HSp) of the clinical diagnostic process were calculated directly, relative to these internationally accepted reference tests. In this process, the rst diagnostic test was routine monitoring, which resulted in the identication of suspect cases based solely on the clinical observations of farmers or veterinarians. 6762 suspect cases were identied, and the test had a HSe of 97.6% (95% C.I.: 96.7, 98.3) and a HSp of 95.2% (95% C.I.: 95.0, 95.3). Suspect cases were then subject to the second diagnostic test, termed declaration, which consisted of a review of a description of the clinical signs by government veterinarians. Premises that tested positive became clinical cases. The HSe of this test was 97.1% (95% C.I.: 96.2, 97.9), and the HSp was
* Corresponding author at: Faculty of Veterinary Medicine, University of Calgary, HS G 380, 3330 Hospital Drive NW, Calgary, Alta. T2N 4N1, Canada. Tel.: +1 403 220 4008; fax: +1 403 210 3919. E-mail address: cribble@ucalgary.ca (C. Ribble). 1 Present address: Alberta Agriculture, Food and Rural Development, O.S. Longman Building, 6909 116 St, Edmonton, Alta. T6H 4P2, Canada. 0167-5877/$ see front matter # 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2006.09.001

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

13

90.9% (95% C.I.: 90.1, 91.6). During the epidemic, these tests were combined and applied in series, with an overall HSe of 94.7% (95% C.I.: 93.5, 95.7) and an overall HSp of 99.6% (95% C.I.: 99.5, 99.6). We also examined the effect of a policy shift that prohibited delaying the diagnosis pending laboratory testing where the animals exhibited equivocal clinical signs. # 2006 Elsevier B.V. All rights reserved.
Keywords: Foot-and-mouth-disease; Clinical diagnosis; Clinical epidemiology; Disease surveillance; Herd sensitivity; Herd specicity

1. Introduction Foot-and-mouth-disease (FMD) is an extremely contagious viral disease that can affect cloven-footed mammals. Most developed countries are ofcially considered free of the disease, a status that is very desirable for trade purposes. On 20 February 2001, however, the UK conrmed that FMD had been identied in an abattoir in England. This was the beginning ofalargeepidemicthatresultedin2026premises(plusfourinNorthernIreland)beingdeclared infected and the loss of the UKs FMD-free without vaccination status for 11 months, until restored by the World Organisation for Animal Health (OIE) on 22 January 2002. Contingency plans to control an epidemic of FMD were in place in the UK prior to the 2001 outbreak. The plans were sufcient to cope with up to 10 simultaneous infected premises (outbreaks), which was the scale of epidemics experienced in Europe in the previous three decades. However, the 2001 epidemic was much larger than anticipated, and rapidly surpassed the scope of these contingency plans (Scudamore and Harris, 2002). Rapid detection of infection is critical to the control of a contagious pathogen such as the FMD virus. The strategy in place prior to the epidemic called for laboratory testing of the rst case of suspected FMD; thereafter, premises could be declared infected on the basis of clinical signs alone. On premises where the clinical picture was equivocal, the plans stipulated that diagnosis should await laboratory verication, which can take up to 5 days (Anderson, 2002). The need for rapid diagnosis during this epidemic led to a policy shift to declaration of FMD based solely on clinical ndings after 22 March 2001. Slaughter on suspicion of FMD was used for equivocal cases (Anderson, 2002). In this paper, we explored the effectiveness of the system used to detect and diagnose cases of FMD during the 2001 outbreak in the UK by calculating the herd-level sensitivity (HSe) and herd-level specicity (HSp) of the component tests of the clinical diagnostic process. To investigate the effect of permitting laboratory testing prior to diagnosis, we compared the diagnostic test parameters before and after the policy changed on 22 March 2001. We also considered the implications and usefulness of this analysis to the management of future epidemics. 2. Materials and methods 2.1. The diagnostic process during the course of the epidemic Premises, as opposed to individual animals, are usually the unit of analysis in FMD investigations because the virus is so highly contagious. Premises that contained FMD-

14

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

susceptible livestock in Great Britain in 2001 can be divided into four categories regarding their FMD status: (1) disease never suspected, (2) suspect case, (3) clinical case, and (4) laboratory-veried case. The latter three categories could be considered the result of a positive diagnostic test (Fig. 1). We avoid the commonly used terms infected premises and conrmation of disease in this paper because they led to confusion when describing the true implication of the different diagnostic steps. The rst diagnostic test was routine monitoring (RM) of FMD-susceptible stock, performed by farmers, veterinarians and others that checked stock for clinical signs of FMD. Under the Foot-And-Mouth-Disease Order of 1983, FMD is a notiable disease in the UK and all suspect cases must be reported to the national veterinary authorities (DEFRA, 2002). In response to every report of suspected FMD during the 2001 epidemic, the premises was visited by a government veterinarian from the nearest Local Disease Control Centre. If the report had originated from a government veterinarian, the same veterinarian continued the investigation. Regardless of the degree of suspicion of the attending veterinarian, the clinical and epidemiological ndings were relayed by telephone to a veterinarian at the National Disease Control Centre in London. The veterinarian in London recorded the ndings on a standard form and then consulted with one of three senior veterinarians. On the basis of the contents of the telephone report, the senior veterinarian decided whether or not to declare the suspect premises a clinical case; there was no ofcial case denition (Wilesmith, personal communication). All suspect cases were subject to this second diagnostic test, which we refer to as declaration in the remainder of the paper.

Fig. 1. Outline of the diagnostic process used to detect cases of FMD on the basis of clinical signs during the epidemic in the UK in 2001.

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

15

The procedure described above was followed throughout the epidemic, except for 2 weeks in April when the veterinarians answering the telephone in London were given authority to declare premises clinical cases of FMD without consulting a senior veterinarian (Wilesmith, personal communication). Prior to 22 March 2001, the diagnosis of FMD could be delayed pending laboratory verication for suspect cases when the clinical signs were ambiguous. After 22 March 2001, it was not permitted to await laboratory results. Regardless of the eventual laboratory results, all clinical cases of disease were subject to infection-control measures: all FMD-susceptible livestock on these premises were killed, as were those on contiguous premises and premises judged to be at increased risk of FMD through links with a clinical case, such as common workers. Approximately 85% of clinical cases had samples submitted for laboratory verication, the third test. The decision as to whether or not samples should be submitted was based primarily on available laboratory capacity at the time. Internationally agreed laboratory procedures, as described in the OIE Manual of Standards for Diagnostic Tests and Vaccines (Kitching et al., 2000), were used during the epidemic (Royal Society of London, 2002; DEFRA, 2002). An antigen-detection enzyme-linked immunosorbent assay (ELISA) was immediately performed, which takes about 3 h to complete. Samples negative to the ELISA were subjected to the more sensitive virus-isolation test, which can take up to 4 days to complete (Royal Society of London, 2002). A case was deemed laboratory-positive if either the ELISA or the virus-isolation test was positive, and laboratory-negative only if both tests yielded negative results. Some laboratory-conrmed cases were identied by the discovery of antibodies to FMD virus in blood submitted as part of routine procedures. Cases detected in this manner represented the failure of RM to detect FMD, because the clinical phase of disease had passed without any suspicion being reported. Because clinical signs are much more obvious in cattle than small ruminants, only the latter were subject to routine serological testing. A full description of the protocol for identifying premises subjected to serological testing may be found in the report submitted to the OIE following the outbreak (DEFRA, 2002). Briey, all premises with sheep and/or goats within a 3-km radius around each outbreak were tested prior to the lifting of restrictions. A random sample of premises within a 310-km radius around each outbreak were tested, using a protocol able to detect a 2% prevalence of seropositive ocks of sheep and/or goats with 95% condence. On each of the selected premises, the serological sampling was designed to detect a 5% prevalence of seropositive sheep and/or goats with 95% condence in each management group of small ruminants. Additional sero-surveillance to ensure freedom from FMD was conducted in counties with many cases of FMD, such that >95% of ocks in these counties were tested. Furthermore, throughout the UK, sheep and goats had blood taken for serological testing: (1) during pre-emptive culls of animals at high risk for FMD, (2) in conjunction with various epidemiological investigations conducted in areas considered to be at high risk of containing infected animals, and in areas where the FMD status of animals was unknown, and (3) as part of pre-movement inspection protocols. The competitive solid-phase ELISA (csp ELISA) test was the serological-screening test; blood samples giving an inconclusive result were resolved by the virus-neutralisation test (VNT) (DEFRA, 2002). The decision to designate a premises as infected with FMD or not on the basis of the serological-test results depended on the number of seropositive animals, the

16

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

location of the premises, any epidemiological links to other FMD cases, and if FMD virus was isolated from a sample of esophagealpharyngeal uid at the time of slaughter. 2.2. Herd-sensitivity and herd-specicity calculations Because a true gold standard is rarely available (Dohoo et al., 2003; pp. 9394), sensitivity and specicity are usually estimated in relation to an accepted reference test. Therefore, when referring to the results of the reference test, the terms reference-positive and reference-negative are preferable to true-positive and true-negative or diseasepositive and disease-negative, and are used in this paper. The entire population susceptible to FMD present in Great Britain at the time of the epidemic was used in the calculations. No single reference test was used in our calculations. Rather, the reference state of premises (reference-positive or referencenegative) was inferred from the reported laboratory results, as determined by one or more of several internationally accepted OIE-prescribed tests: virus isolation, an antigendetecting ELISA, an antibody-detecting ELISA, and virus neutralisation. Each agricultural holding (including livestock markets, abattoirs, and farms) in the UK is assigned a unique identier, termed a county-parish-holding (CPH) number. For suspect cases that were not declared clinical cases, the date of report and the CPH number were obtained from the Department for Environment, Food and Rural Affairs Disease Control System. This database was created during the outbreak to manage the epidemic. For each clinical case, the CPH number, date of report, origin of report (to determine whether the case was discovered by sero-surveillance or as a clinical case) and laboratory results were obtained from the database developed by the epidemiology team at the National Disease Control Centre during the outbreak. Both of these databases have been described previously (Gibbens et al., 2001). The herd sensitivity and herd specicity of RM and declaration were estimated directly (Tables 1 and 2). We assumed that clinical cases from which samples were not submitted to the laboratory had the same distribution of laboratory results as the laboratory-tested premises. In other words, because (over the course of the epidemic) 76% of laboratorytested clinical cases were laboratory-positive, we assumed that 76% of laboratory-untested
Table 1 Classication of premises for estimation of herd sensitivity and herd specicity of routine-monitoring test for FMD during the 2001 epidemic in the UK Reference-positive Test-positive (reports of suspected FMD) Lab-positive clinical cases +76% of lab-untested clinical cases plus false-negative declaration testsa Cases identied by serological testing Reference-negative True-negative declaration tests plus lab-negative clinical cases +24% of lab-untested clinical cases All premises with susceptible stock in Great Britain on which FMD was never suspected or identied

Test-negative (premises never reporting suspected FMD)

See text for detail.

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

17

Table 2 Classication of premises for estimation of herd sensitivity and herd specicity of declaration of FMD at the National Disease Control Centre during the 2001 epidemic in the UK Reference-positive Test-positive (declared clinical cases) Test-negative (declared disease-free)
a

Reference-negative Lab-negative clinical cases +24%a of lab-untested clinical cases Premises that reported suspect disease that were correctly declared free of diseaseb

Lab-positive clinical cases identied by routine monitoring +76%a of lab-untested clinical cases Reports of suspect disease wrongly declared disease-freeb

Of premises that reported suspect FMD prior to March 22nd, 82% of lab-tested premises tested positive, compared to 74% of lab-tested premises that reported on or after March 22nd. The weighted overall average was 76%. HSe and HSp calculations for these sub-groups were adjusted accordingly. b See text for details.

clinical cases were laboratory-positive and the remaining untested clinical cases were laboratory-negative. Suspected FMD was reported more than once on some premises; each report represented a positive routine-monitoring test. A premises may consist of either a single land parcel or several non-conuent land parcels owned by the same person; we considered that premises reported suspect disease more than once if the CPH number was repeated in the data. Some premises on which suspected FMD was reported more than once eventually became clinical cases. If FMD was actually present on these premises at the time of the initial report(s) of suspicion, then the rst report(s) had been incorrectly declared FMDfree. Such reports were true-positive routine-monitoring tests and false-negative declaration tests. We assumed that FMD was present at the rst report(s) if the nal result was laboratory-positive (plus 76% of untested clinical cases that t the criteria), and if the difference in time between the declaration test-negative and test-positive reports minus the age of oldest lesion was 3. This gave liberal allowance for the inaccuracy inherent in estimating the age of lesions (Gibbens et al., 2001). If there were multiple declaration testnegative reports associated with a premises, only those that t the above criteria were considered false-negatives. 2.3. HSe and HSp of routine monitoring The routine-monitoring test was positive if disease was suspected and reported to the authorities. The test was negative if disease was either never suspected and therefore not reported, or suspected but not reported; it was not possible to distinguish between these two possibilities. We assumed that all FMD-susceptible stock in Great Britain had RM to some extent, during milking, feeding, or inspection by veterinary patrols. We classied premises into a 2 2 table with respect to their RM and reference-test status (Table 1). The total number of premises in Great Britain with susceptible stock was estimated by subtracting the number of premises in N. Ireland with susceptible stock in 2001 (N = 26,287) (Department of Agriculture and Rural Development Northern Ireland, 2004) from the total number of such premises in the UK (N = 134,600) (European Commission, 2003). Because it is theoretically possible that some FMD cases remained undetected by the sero-

18

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

surveillance program, we repeated the calculations increasing the number of cases detected by sero-surveillance from 39 to 129. This is the number of cases that would be expected if the number of undetected cases was proportional to the number of untested premises. This is an extreme example, because the concentration of FMD cases varied widely by region with the most highly infected areas subjected to very extensive sero-surveillance. 2.4. HSe and HSp of declaration Routine monitoring was a screening test; only premises positive to this test were subject to the second diagnostic test of declaration. Performing HSe and HSp calculations on this population is valid because this is the only population to which this test would ever be applied. FMD cases identied by routine sero-surveillance were not included in the evaluation of the declaration test, because suspect clinical disease had never been reported on these premises. Premises were considered declaration test positive if they were deemed a clinical case by the National Disease Control Centre in London. Those authorities referred to these cases as infected premises. Premises were otherwise considered declaration test negative. The classication of premises used to evaluate the declaration test is shown in Table 2. Because the policy of laboratory testing on premises where stock had equivocal clinical signs changed after 22 March 2001, we calculated the HSe and HSp of declaration (HSeDec and HSpDec) considering: (1) only reports prior to 22 March, (2) only reports after 22 March, and (3) all reports. The assumed proportions of laboratory-untested clinical cases considered laboratory-positive and laboratory-negative were adjusted to reect the actual proportions of the laboratory-test results in each time period. 2.5. HSe and HSp of the clinical diagnostic process The routine-monitoring and declaration tests were interpreted in series. Overall, premises were considered test-positive only if they tested positive to both tests (i.e. were declared a clinical case following a report of suspected FMD). Premises were considered test-negative if either test had a negative result. HSe and HSp of the clinical diagnostic process were calculated with premises classied as follows: (1) test-positive and referencepositive: all laboratory-positive clinical cases (including 76% of laboratory-untested clinical cases) detected by RM; (2) test-positive but reference-negative: clinical cases that were laboratory-negative (including 24% of laboratory-untested clinical cases); (3) testnegative but reference-positive: FMD cases detected by results of serological tests plus suspect cases wrongly declared FMD-free; and (4) test-negative and reference-negative: all premises with FMD-susceptible stock in Great Britain on which FMD was never suspected plus suspect cases correctly declared FMD-free.

3. Results There were 6762 reports of suspected clinical FMD led in Great Britain in 2001, on 6182 different premises (Fig. 2). A total of 2026 premises became FMD cases. RM initially

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

19

Fig. 2. Breakdown of the FMD status of premises in Great Britain during the 2001 epidemic. a These occurred on premises that had a negative declaration test but became a clinical case of FMD shortly afterwards, at which time the age of FMD lesions indicated that the rst declaration test had been falsely negative. These premises, and associated laboratory testing, are also included within the 1987 clinical cases.

detected all but 39 of these cases. The remaining 39 cases were identied subsequent to the discovery of antibodies to FMD virus in blood from sheep or goats submitted as part of routine-sampling procedures. Samples were submitted to the Institute for Animal Health (Pirbright Laboratory) for laboratory conrmation of disease from 85% of clinical cases. Of all samples submitted from clinical cases, 76% tested positive for the FMD virus. Seventy premises had a negative declaration test but became clinical cases of FMD shortly afterwards, at which time the age of FMD lesions indicated that the rst declaration test had been falsely negative. Samples were submitted for laboratory analysis from 55 of these 70 clinical cases; 34 were found to be laboratory positive and 21 laboratory negative. We assumed 76% (11) of the untested cases were reference-test positive, and thus used a value of 45 false-negative declaration tests in the calculations. No single premises had more than one false-negative declaration test. Prior to 22 March 2001, 87% of premises declared infected had samples submitted for laboratory conrmation. Eighty-two percent of these samples were laboratory positive. With the available data, it was not possible to distinguish between suspect cases that had been declared FMD cases on the basis of the laboratory results, and those that had samples submitted subsequent to being declared clinical cases. After 22 March 2001, when the policy changed to prohibit laboratory testing prior to declaration, samples were submitted to the laboratory from 84% of clinical cases, and 74% of these were positive. The difference in the proportion of laboratory-positive samples before and after the change in policy was statistically signicant (Chi-square = 9.77, d.f. = 1, p = 0.002).

20

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

Table 3 Herd sensitivity and herd specicity of routine monitoring, declaration and the entire clinical diagnostic process during the 2001 epidemic of FMD in the UK Sensitivity (95% CI)a Routine monitoring Declaration Overall Suspicion reported prior to 22 March 2001 Suspicion reported on or after 22 March 2001 Clinical diagnostic process (both tests interpreted in series)
a

Specicity (95% CI) a 95.2 (95.0, 95.3) 90.9 (90.1, 91.6) 93.6 (92.2, 94.9) 89.9 (88.9, 90.8) 99.6 (99.5, 99.6)

97.6 (96.7, 98.3) 97.1 (96.2, 97.9) 97.3 (95.1, 98.6) 97.1 (95.9, 98.0) 94.7 (93.5, 95.8)

The condence intervals are exact based on the binomial distribution.

The HSe and HSp of routine monitoring, declaration and the overall clinical diagnostic process (i.e. RM and declaration tests applied in series) are presented in Table 3. Increasing the number of cases detected by sero-surveillance from 39 to 129 decreased the HSeRM to 92.35% (95% CI: 90.98 and 93.57%) and had no signicant effect on the HSpRM.

4. Discussion Our results suggest that the HSe and HSp of clinical diagnosis of FMD during the 2001 epidemic in the UK were high, relative to internationally accepted reference tests. This indicates that, after identifying the rst case in an epidemic, determination of FMD status based on clinical signs alone can be a fairly accurate diagnostic test. No other studies of this kind concerning FMD have been performed to our knowledge. For any surveillance program that relies heavily on farmers for detection of cases, the reporting of suspected disease must be actively encouraged. This might be accomplished by programs designed to increase awareness about the disease (e.g. media campaigns, veterinary visits), and also by ensuring that fair compensation is provided for infected stock. Both of these tactics were employed during the 2001 FMD epidemic. Another approach is to provide a nancial incentive for all reported suspect cases (Doherr and Audige, 2001). Although the HSe of RM (the screening test used during this epidemic) was very high, 39 premises were detected on the basis of seroconversion rather than clinical signs. Because cattle were not subject to testing as part of the sero-surveillance program, sheep or goats were the species affected on all of these 39 premises. Because clinical disease in sheep is often subtle and transient, the stock on these premises might not have exhibited obvious clinical signs; indeed, some might not have manifested any symptoms at all. In this situation, an effective awareness campaign should emphasize the inter-species variation so that all farmers are aware of the need for extra vigilance with small ruminants. It might be worthwhile to offer additional compensation for the labour that this vigilance requires. During this epidemic, some farmers co-grazed a few cattle with large ocks of small ruminants, on the basis that the cattle would act as sentinels for FMD infection. The effectiveness of this strategy during an epidemic is worthy of further investigation.

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

21

Both the HSe and HSp of declaration were important. Less-than-perfect HSeDec resulted in the delayed removal of infected animals, which potentially allowed further spread of disease. Less than 100% HSpDec resulted in the destruction of healthy animals, with important economic and social consequences. The discussion between the veterinarians in the eld and at the National Disease Control Centre was an important component of the declaration test during this epidemic. In this type of situation, excellent communication between the two parties is needed to achieve an accurate diagnosis from a remote location. Communication might be improved with the use of digital cameras, which would enable the remote veterinarian, or possibly an FMD expert, to view the lesions. To ensure that diagnostic errors are not repeated, both the veterinarian in the eld and at headquarters should be informed of the laboratory results of cases with which they have been involved. Developing and using a formal case denition might also improve the HSe and HSp of the declaration test; this is worthy of further research. However, the need for a nal diagnosis based on clinical signs might soon be eliminated by the development of penside diagnostic tests for FMD, which could be used as alternative to the declaration test to provide rapid diagnosis. According to a recent review, such tests have been developed but have not yet been adequately validated for use in the eld (Alexandersen et al., 2003). During this epidemic, the HSpDec dropped after it was prohibited to delay clinical diagnosis pending laboratory results on premises where stock had equivocal clinical signs. This suggests that the veterinarians making the diagnosis at headquarters responded to the change in policy by labelling some questionable cases infected, rather than using the slaughter-on-suspicion provision or risk declaring them FMD-free. The risk of falsepositive diagnosis (as determined by the reference test) was approximately 6% prior to 22 March 2001, and 10% after 22 March 2001. Further analysis is required to compare the benets gained by increasing the speed of diagnosis (faster implementation of control measures that eliminate the spread of infection from FMD cases) to the losses that resulted from increased risk of false-positive diagnosis of FMD. The change in policy did not appear to affect the risk of false-negative diagnosis. The sensitivity and specicity of the reference laboratory tests for FMD used during the UK epidemic are reported to be very good under ideal conditions (Alexandersen et al., 2003). However, to our knowledge, the precise values for these tests have not been published. Under eld conditions, false-negative reference-test results might have resulted from inappropriate sampling (non-acute disease, poor specimen quality) or improper handling of the sample (contamination with disinfectant, delayed testing). This would effectively decrease the sensitivity of the laboratory test. It was necessary to make assumptions in this study. Firstly, we assumed that all premises containing cases of FMD in the UK in 2001 were identied, and that the remaining premises were truly free from FMD. This assumption is supported by the extensive serosurveillance campaign carried out to prove national freedom from FMD (DEFRA, 2002). However, it is theoretically possible that a few, subtle dead-end cases were missed. These would have relatively little effect on the calculated HSeRM and essentially none on the HSpRM. Secondly, we assumed that the submission of laboratory samples (i.e. submitted: yes or no) was not related to the true FMD status of the animal.

22

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

We might have overestimated the number of false-negative declaration tests. If the number of these cases were indeed overestimated, all of the HSe and HSp parameters would also be overestimated, except for the HSeDec, which would be underestimated. However, even if the actual number of false-negative declaration tests were 50% lower than the number used in the analysis, this would result in changes of <1.5% in the HSe and HSp values because the number of false-negative declaration tests was very small relative to the other numbers in the calculations. In the evaluation of the RM test, the number of premises in Great Britain with susceptible stock where FMD was never suspected was approximated. However, even substantial changes in this number would result in only a minor effect on the calculated HSpRM. Increasing or decreasing the population of premises by 10,000 changed this value by <1%. The use of the total number of premises with FMD-susceptible to calculate the value of the reference-negative, test-negative cell for the RM test (Table 1) would be appropriate if, by denition, only one RM test was conducted on each premises throughout the epidemic. However, in our analysis, there were premises on which disease was suspected more than once, and therefore on which we considered more than one RM test to have been conducted. Alternatively, we could have considered each time anybody checked stock as a test. This would result in a much larger number in the reference-negative, test-negative cell (very approximately, the number of premises with susceptible stock multiplied by the number of days of the epidemic). Regardless of the denition of test, the number of reference-negative, test-negative subjects for monitoring was very large compared to the values of the other cells in the 2 2 table. Therefore, by the nature of the calculations, the HSpRM was very high. In reality, it is likely that HSe and HSp of clinical diagnosis varied with the characteristics of the population to which they were applied (e.g. because FMD is easier to diagnose clinically in cattle than in sheep, it is likely that HSeRM and HSpRM were higher in cases concerning cattle than in those concerning sheep). Further research is needed concerning the association of certain factors, such as species, with the outcome of these diagnostic tests. Because the species suspected of disease was only recorded in approximately 7% of RM tests that were not declared FMD-positive, the methodology described in this paper could not be applied to examine this issue. However, the relationship between the diagnostic test results and other factors is investigated in the paper by McLaws et al. (2006). Our results should be useful in the control of future FMD epidemics. For example, knowledge of the specicity of RM could enable the veterinary authorities to anticipate the proportion of suspect cases that will be declared clinical cases, and thereby assist in planning the deployment of veterinary and other resources. The effect of policies (such as the use of the laboratory to conrm equivocal cases) could be assessed by evaluating their impact on the HSe and HSp.

5. Conclusion By classifying premises directly according to their test results, we demonstrated that the HSe and HSp of clinical diagnosis during the 2001 FMD epidemic in the UK were high (94.7 and 99.6%, respectively).

M. McLaws et al. / Preventive Veterinary Medicine 78 (2007) 1223

23

Acknowledgements We gratefully acknowledge the assistance of Graeme Cooke with data collection. We thank John Wilesmith for his comments on the manuscript. The Ontario Veterinary College DVM/PhD Fellowship and a grant from the Department for Environment, Food and Rural Affairs (UK) provided funding for this project. The latter also supplied the data used in the analysis.

References
Alexandersen, S., Zhang, Z., Donaldson, A.I., Garland, A.J.M., 2003. The pathogenesis and diagnosis of foot-andmouth disease. J. Comp. Pathol. 129, 136. Anderson, I., 2002. Foot and mouth disease 2001: lessons to be learned inquiry report http://213.121.214.218/fmd/ documents/index.htm. DEFRA, 2002. Foot and mouth disease in the United Kingdom 2001: Report to the OIE. Part 1. Great Britain. Doherr, M.G., Audige, L., 2001. Monitoring and surveillance for rare health-related events: a review from the veterinary perspective. Phil. Trans. R. Soc. Lond. B 356, 10971106. Dohoo, I., Martin, W., Stryhn, H., 2003. Veterinary Epidemiologic Research. AVC Inc., Charlottetown, P.E.I, p. 706. European Commission, 2003. Farm structure - 1999/2000 survey. http://epp.eurostat.cec.eu.int/cache/ITY_ OFFPUB/KS-52-03-756-3A/EN/KS-52-03-756-3A-EN.PDF. Gibbens, J.C., Sharpe, C.E., Wilesmith, J.W., Mansley, L.M., Michalopoulou, E., Ryan, J.B.M., Hudson, M., 2001. Descriptive epidemiology of the 2001 foot-and-mouth disease epidemic in Great Britain: the rst ve months. Vet. Rec. 149, 729742. Kitching, R.P., Barnett, P.V., Donaldson, A.I., Mackay, D.K.J., 2000. Foot and mouth disease.In: Manual of Standards for Diagnostic Tests and Vaccines. fourth ed. OIE, Paris. McLaws, M., Ribble, C., Martin, S.W., Stephen, C., 2006. Factors associated with the clinical diagnosis of foot and mouth disease during the 2001 epidemic in the UK. Prev. Vet. Med. 77, 6581. Royal Society of London, 2002. Infectious Diseases in Livestock. Royal Society, London http://www.royalsoc. ac.uk/inquiry/index.html. Scudamore, J.M., Harris, D.M., 2002. Control of foot and mouth disease: lessons from the experience of the outbreak in Great Britain in 2001. Rev. Sci. Tech. 21, 699710.