- Email: [email protected]
Herd-level risk factors for antimicrobial demanding gastrointestinal diseases in Danish herds with finisher pigs
Preventive Veterinary Medicine 98 (2011) 190–197
Contents lists available at ScienceDirect
Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed
Herd-level risk factors for antimicrobial demanding gastrointestinal diseases in Danish herds with finisher pigs A register-based study G.K. Hybschmann a,∗ , A.K. Ersbøll b , H. Vigre c , N.P. Baadsgaard d , H. Houe a a b c d
Department of Large Animal Sciences, Faculty of Life Sciences, University of Copenhagen, Denmark National Institute of Public Health, University of Southern Denmark, Denmark Division of Microbiology and Risk Assessment, National Food Institute, Technical University of Denmark, Denmark Danish Agriculture and Food Council, Pig Research Centre, Denmark
a r t i c l e
i n f o
Article history: Received 13 November 2009 Received in revised form 16 September 2010 Accepted 9 October 2010 Keywords: Finisher pig herds Antimicrobial use Herd size SPF Herd type
a b s t r a c t Endemic gastrointestinal (GI) diseases have a substantial negative impact on pig production, because, when present, they reduce animal welfare, productivity and generate high antimicrobial (AM) demand. In Danish legislation, AM can be prescribed only for therapeutic purposes. The objective of the study was to estimate the association between herd-level risk factors and the amount of AM use (AMU) in connection with GI diseases in finisher herds. We conducted a register-based cross-sectional study with repeated measurements from 2004 to 2007. Data were extracted from databases in the Danish Register of Veterinary Medicine, the Central Husbandry Register and the Danish Agriculture and Food Council. In total, 3192 pig herds with 26,973 records (quarters with prescriptions) were included. The outcome was presented as average AM use (measured as Animal Daily Dosage) for GI diseases per finishing pig per quarter per herd. Three potential herd-level risk factors were evaluated: herd size (number of finishers delivered for slaughter); herd health status (herds in the Specific Pathogen Free (SPF) System, conventional herds); and herd type (herds including only finishers, integrated herds). Data were analyzed using general linear mixed models with repeated measurements. Smaller herds had a larger AMU per finisher than larger herds. Integrated herds had lower AMU as compared with herds with only finishers. Herds within the SPF System had a larger decrease in AMU with increasing herd size compared to conventional herds. Significant regional differences in AMU were seen. Additionally, the results showed that other herd factors and veterinarians were more influential than the investigated herd risk factors. This illustrates the difficulties of characterising AM-demanding GI diseases in herds by the use of register data only. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Endemic gastrointestinal (GI) diseases in finisher pigs have a considerable negative impact on pig production,
∗ Corresponding author at: University of Copenhagen, Faculty of Life Sciences, Department of Large Animal Sciences, Veterinary Epidemiology, Grønnegaardsvej 8, DK-1870 Frederiksberg C, Denmark. Tel.: +45 3528 3011; fax: +45 3533 3022. E-mail address: [email protected] (G.K. Hybschmann). 0167-5877/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2010.10.005
because they compromise animal welfare and reduce productivity and hence cause substantial economic losses (McOrist et al., 1997; Kroll et al., 2005; McOrist, 2005; Straw et al., 2006). In Denmark, the major infectious cause of GI diseases in finishers is Lawsonia intracellularis, the aetiological agent of porcine proliferative enteropathy. However, Brachiospira pilosicoli, Brachiospira hyodysenteriae, Escherichia coli and Porcine Circovirus Type 2 also cause diarrhoea and poor growth performance in finishers (Stege et al., 2000; Jacobson et al., 2003). Furthermore, these GI diseases trigger considerable use of antimicrobials
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
(AM), and this is believed to have an impact on the development of AM resistance within pig herds as well as in humans (Aarestrup, 1999). Better understanding both at the herd and at the regional level, of the risk factors at play here should, therefore, help to clarify the factors that can influence the presence of these multi-factorial GI diseases. Previous studies evaluating risk factors for infectious GI diseases have identified associations between the presence of the infectious agent/disease and management and feedrelated factors (Pearce, 1999; Stege et al., 2001; Chouet et al., 2003). Herd size is frequently studied as a risk factor for swine diseases, and there do appear to be associations between swine diseases and management, environmental factors and herd size (Gardner et al., 2002). Smith et al. (1998) found a strong association between proliferative enteropathy and herd size in herds with more than 500 sows. However, a Danish study by Stege et al. (2001) did not identify herd size as a significant factor. Pig herds in the Danish herd health system, the so-called SPF-system (SPF – Specific Pathogen Free) must comply with high standards of biosecurity and have to be free from at least one of the specified SPF diseases (Boklund et al., 2004). Therefore, herd health status is a relevant risk factor. Herds with only finishers continuously have to purchase new pigs and introduce them into the herd. This could mean that the risk of infectious GI agents being transmitted into the herd is higher in these herds compared to integrated herds. There is a growing tendency to use register data in epidemiological studies, as this is a manageable and affordable way to obtain access to large amounts of data describing longer periods of time. The objective of the present study was to determine the association between AM-demanding GI diseases in finishers and the following herd-level risk factors: herd size, herd health status and herd production type, adjusted by region, season and year. Danish legislation requires that AM only can be used if it is prescribed by a veterinarian and it must be strictly used for therapeutic purposes (Anonymous, 2007). Therefore, AM treatment can be used as an indirect measure of the presence of disease. The study uses general linear mixed models with repeated measurements of register data. 2. Materials and methods 2.1. Study design and population A cross-sectional study with repeated measurements was designed. Danish pig herds with finisher pigs were the target population. Finishers were defined as pigs from 30 kg until slaughter. The study population was indoor production herds with more than 50 finishers produced each quarter and with AM prescriptions for GI diseases to finisher pigs; see Section 2.4.1 for a description of the exclusion criteria. The study unit was the herd per year per quarter. 2.2. Data Data on AM consumption were obtained from the Danish Register of Veterinary Medicine (Vetstat). Since 2002,
191
it has been mandatory to record any AM prescribed to production animals in Vetstat. Information on prescribed AM use (AMU) for pigs is recorded in Vetstat by primarily pharmacies. For each prescription, the data encompass information on day of prescription, herd identity code (Central Husbandry Register (CHR) number), animal species, age group (sows and piglets, weaners, finishers) and disease category in terms of diseased organ (gastrointestinal diseases, respiratory diseases, joints and limbs diseases, reproductive diseases, udder diseases, metabolic diseases, and other diseases) as well as information on the drug (product identity (the Anatomical Therapeutic Chemical (ATC) classification system, specific number and name of the product); and quantity (amount of the product in specified unit) (Anonymous, 2009a, 2010a). In order to compare different AM types of different strengths and potency, the amount of AM prescribed is in Vetstat standardized by a standardized measure for drug use, Animal Daily Dosage (ADD). The measure is defined as the average daily dose for the main indication, using recommendations on doses approved by the Danish Medicine Agency and the Veterinary Pharmaceutical Producer Association (Jensen et al., 2004); on the main part in practice, these doses recommendations for animals are only specified on animal species. The disease category (indication) and ADD is included in the information from Vetstat. Data on herd characteristics (herd type, location) were extracted from the CHR and from the Danish Agriculture and Food Council. Abattoirs submit information on the number of delivered finishers to the Zoonosis Register, and these data were extracted. Furthermore, laboratories automatically send laboratory results to the SPF Company. This information had been applied in the SPF Company to classify the herds into the SPF system. These data were obtained for this study. The majority of Danish pig herds have a herd health contract, which implies monthly visits by a veterinarian. Therefore, using a month as the time unit was unsuitable and quarters of a year were selected as the unit of time. Data from the different sources were merged and aggregated at herd-year-quarter level. In the final data set, one observation (herd-year-quarter) included information on the amount and type of prescribed AM, the number of pigs delivered for slaughter, the number of sows in the herd, herd health status, prescribing veterinarian and region, and, additionally, derived variables. 2.3. Description of derived variables 2.3.1. Outcome variable Outcome was amount of prescribed AM for GI diseases per finisher per quarter at herd-level, quantified by ADD. We linked AMU with antimicrobial demanding disease as AM is only allowed for treatment purpose in Denmark. Compliance to this legislation is closely followed by the authorities, and therefore, we assumed that this legislation in general was followed. AM prescribed for GI diseases was included for all application types. AMU is prescribed to groups of finishers in the herds (population at risk). Vetstat does not give information on whether for example five pigs had been treated
192
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
once or one pig has been treated five days in a row. In Vetstat, information on AMU prescribed for pigs is recorded at herd-level. As the number of treated pigs was not recorded in Vetstat, a proxy for the number of pigs at risk of disease/treatment was defined. Since it takes approximately 3 months from the time of inclusion of pigs in the herd at 30 kg until slaughter, the number of finishers sent to slaughter within the quarter was used as a proxy for the population at risk. Hence, AMU was standardized by dividing ADD by the number of finishers sent to Danish Abattoirs in the actual quarter.
2.3.2. Herd-level risk factors Herd health status was a dichotomous variable: herds that were declared free from at least one of the selected SPF diseases and in compliance with stringent biosecurity rules (SPF), or not (conventional). The conventional herds were not as such obliged to follow the stringent biosecurity rules, and could either have (1) unknown SPF status or (2) a known status of porcine reproductive and respiratory syndrome, but unknown status regarding the selected SPF diseases. SPF herds are examined on regular basis so that they can be declared free of selected infectious agents: Actinobacillus pleuropneumoniae, Mycopacteria hyopneumoniae, haemolytic Brachyspira hyodysenteriae, toxin-producing Pasturella multocida, Haematopinus suis and Sarcoptes scabiei var. suis (Anonymous, 2009c). Herd type was defined as a dichotomous variable with either integrated herds (both sows and finishers) or herds with only finishers. Herd size was defined as the moving average of the quarterly delivery to Danish abattoirs each year, using the number of delivered finishers for the actual quarter and the previous three quarters.
2.3.3. Region Region was included in the study in order to adjust for regional differences in, for example, overall pig production, the prevalence of disease, and production patterns. Denmark was divided into the following six regions: Zealand (West Zealand, Storstrøms, Frederiksborg, and Roskilde Counties), Funen (Funen County), Northern Jutland (North Jutland and Viborg Counties), Western Jutland (Ribe and Ringkøbing Counties), Southeastern Jutland (South Jutland and Vejle Counties), and Eastern Jutland (Århus County). The island Bornholm was excluded, because it had too few herds.
2.3.4. Quarter Season was defined by the four quarters of the year, i.e. January–March, April–June, July–September and October–December.
2.3.5. Veterinarian In each herd, different veterinarians had prescribed AM during the period, 2004–2007. Hence, for each herd, the main veterinarian was identified as the veterinarian with most prescriptions in the period.
2.4. Statistical analysis 2.4.1. Data control and editing In order to detect missing values and extreme values, data control was performed for qualitative variables by means of frequency distributions, and for quantitative variables by means of maximum and minimum values and scatter plots. Herds with extreme values of CHR number, ADD, herd health status or herd size were checked using the original registers. Observations were excluded if: (1) herd characteristics (herd type, herd health status, herd size) were missing; (2) herds had irregular CHR numbers (<10,000 and >115,000); (3) observations included prescriptions, but with no information on type of AM; (4) observations had prescriptions for fluoroquinolones, penicillin for exterior use, intramamamaria, amphenicols, aminoglycosides, cephalosporins, colistin, combination-penicillins, and simple penicillins, thus leaving prescriptions on lincosamids, lincospectin, makrolids, sulfamethoxazole/trimethoprim, tetracyclines and tiamulines in the study; or (5) the 12-month moving average of delivered finishers in the quarter was less than 50 finishers. Some of the excluded AM were removed due to failure in recording (e.g. some AM types was not in agreement with indication, and colistin is used for weaners in Denmark). In total, the excluded AM types accounted for only 1.5% of the total prescribed ADD. 2.4.2. Descriptive analyses Descriptive analysis of quantitative variables was performed in terms of measures for location and spread. Due to the left-skewed distribution of the outcome, the description was given in median, min., max., Q1 and Q3 . The qualitative variables were described in frequency distributions. 2.4.3. Statistical analyses Data were hierarchically structured with herds nested within each region and with herd-level risk factors varying over time. A general linear model with repeated measurements was used to evaluate the effect of potential risk factors on the average amount of prescribed AM per finisher. Herd-level risk factors, region, year and season were included as fixed effects as well as two-way interaction terms (Model 1). Any correlation between repeated measurements was taken into account by including an autocorrelation structure in the model. Different autocorrelation structures (compound symmetry, firstorder ante-dependence, and first-order autoregressive structure with non-equidistant intervals) were evaluated using Akaike’s Information Criterion (AIC). In order to evaluate different autocorrelation structures, only herds with a minimum of seven quarters with AMU were included. A first-order autoregressive structure with nonequidistant intervals was selected. In the further analyses, all herds with prescribed AMU in at least one quarter were used. Model assumptions were evaluated using the Kolmogorov–Smirnov test for normality and visual inspection of residual plots. Additionally, Box-Cox analysis was performed in order to find the optimal transformation of the outcome (Weisberg, 1985). There was a non-linear rela-
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
tion between the outcome and the continuous explanatory variable, herd size. Consequently, herd size variable was log-transformed. Furthermore, the log-herd size was standardized by subtracting the mean of the log-value. Two observations with extreme residual values had a strong impact on the results of the analyses and were therefore deleted. However, visual inspection of QQ-plots of residuals used to evaluate the Gaussian distribution of the residuals showed that the assumption still was not fulfilled. Therefore, the analyses were also performed with ranktransformed outcome (Conover and Iman, 1981). Backward elimination was applied in order to identify any significant fixed effects using the Kenward–Roger’s test for overall significance (Brown and Prescott, 2006). Insignificant interactions, and subsequently main effects, were excluded, one by one, at a significance level of 1%. The results from the log and rank transformed analyses gave the same significant risk factors. Therefore, results from the model with log transformed outcome were applied in the subsequent analyses. Confounding between potential risk factors was examined by 2-by-2 tables, and by estimating relative difference in parameter estimates when including and then excluding the potential confounding variable in the analyses (Bruun et al., 2004). In order to assess the ability of the herd-level risk factors to account for the between herd variation for AMU a general linear mixed model was evaluated with a random effect of herds. The herd-level risk factors were excluded and herd was included as random effect (Model 2). Veterinarians were not included in the model (Model 1) in order to avoid confounding with herd effects, because the majority of veterinarians had prescribed AM to only a small number of herds. For example, 40% of the veterinarians had prescribed AM to only one (28%) or two herds (12%). A model (Model 3) was constructed without herdlevel risk factors and region, but with veterinarian included as a random effect in order to evaluate the variation due to veterinarians. The statistical analyses were performed using the MIXED procedure in the Statistical Analysis System (SAS vers. 9.1. Cary, NC, USA). 3. Results
193
ranging from 1 to 59 (Q1 = 1, median = 3, Q3 = 12), and with number of quarters with prescriptions ranging from 1 to 585 (Q1 = 6, median = 21, Q3 = 99). Among the six types of AM prescribed for GI diseases, tetracyclines, macrolides and tiamulin accounted for 94% of the use. Descriptive results of the data applied in the analysis (full study population) are presented in Table 1. Overall, the median AMU was 1.38 ADD per finisher per quarter per herd. 3.2. Herd-level risk factors The final model (Model 1), evaluating the herd-level risk factors, is presented in Table 2. The main effects – herd size, herd type, region and quarter – were significantly associated with the amount of prescribed AMU per finisher (p < 0.001). The main effect of herd health status was not significant; however, the interaction between herd size and herd health status was significant. Integrated herds had a lower adjusted AMU than herds having only finishers. SPF herds had a larger reduction in adjusted AMU than conventional herds with increasing herd size. There was a significant difference in AMU between regions. Herds on Zealand and in Western Jutland had the lowest adjusted AMU, whereas herds in Eastern and Southeastern Jutland and on Funen had the highest adjusted AMU. Although only with small differences, the highest AMU was in the first quarter of the year and the lowest was in the third quarter. The ability of herd-level risk factors (herd size, herd type and herd health status) to account for the between herd variation was evaluated. The total variation of the logtransformed adjusted ADD was 2.65. The residual variation in Model 1 with herd-level risk factors was estimated at 2.57. Therefore, the herd-level risk factors accounted for 3.0% of the total variation. If the between herd variation was modelled with a random effect of herds, the between herd variation was estimated at 1.06 and with a residual variation estimated at 1.58. Thus, the between-herd variation accounted for 40% of the total variation. The variation due to differences in AMU between veterinarians (Model 3 with veterinarian included as a random effect) was estimated at 0.28 with a residual variation at 2.41. Hereby, the random effect of veterinarians could explain 10.7% of the total variation.
3.1. Descriptive statistics 4. Discussion The distribution of all AM prescribed to pigs in 2006–2007 stratified by the group of pigs (sows and piglets, weaners and finishers) and by five diagnostic groups (GI, respiratory, joint, mastitis, and reproductive diseases) is illustrated in Fig. 1. As the figure shows, GI diseases in finishers resulted in the second largest use of AM; they were exceeded only by GI diseases in weaners. The study population consisted of 3192 herds (26,973 observations), each with recorded prescriptions within the range of 1–16 quarters (Q1 = 4, median = 9, Q3 = 12) in the study period, 2004–2007. In total, 897 veterinarians had prescribed AM to a number of herds, ranging from 1 to 124 herds (Q1 = 1, median = 2, Q3 = 7); of these, 369 were main veterinarians with prescriptions for a number of herds
4.1. Risk factors The herd-level risk factors in the study accounted for only 3.0% of the total variation. Obviously, this may be due to the fact that the applied registers were created with other purposes in mind. With the between-herd variation, accounting for 40% of the total variation, further analyses on herd risk factors will therefore require more detailed information than available in registers on for example farm management and housing. It was not possible to adjust for veterinarians. Therefore, a mixed model with veterinarian as random effect (Model 3) was constructed. The results indicated that the
194
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
40 Data used in study Other data
35
Percentage
30 25 20 15 10 5 0
GI
Joint Resp Finishers
GI
Joint Resp Weaners
GI
Joint Resp Sows/Piglets
Repro Mastit Sows
Fig. 1. Distribution (%) of the total antimicrobial use to all Danish pigs on combined age-diagnosis groups, 2006–2007. GI: GI diseases; Joint: joint diseases; Resp: respiratory diseases; Repro: reproductive diseases; Mastit: mastitis.
variation due to veterinarians could explain 10.7% of the total variation. Even though, it was not possible to adjust for herd risk factors in Model 3 (because many veterinarians had health consultancy appointments with one or few herds), veterinarians was found to be more influential than the herd risk factors investigated in the present study (Model 1); still, the between herd variation were almost 4 times higher than the variation between veterinarians. Vigre et al. (2010) found similar results, where the variation of the total amount of AMU (ADD) prescribed to
finishers was 4 times higher than the variation within the veterinarians themselves. A Danish questionnaire study by Johansen (2005) investigated swine practitioner’s strategy for treatment of gastro-intestinal diseases and found variation between the responding practitioners. A study on antibacterial drug use in Dairy cattle (Bruun et al., 2003) found that veterinary practice influenced the choice of broad- versus narrow-spectrum AM. However, to our knowledge there are no publications with studies looking more into detail about this. The difference in treatment
Table 1 Descriptive statistics of amount of prescribed AM for finishers for GI diseases measured by adjusted Animal Daily Doses (ADD) in 3192 Danish pig herds, 2004–2007. Variable
Level
Amount of prescribed AM for finishers (ADD per finisher) N
Overall
Min.
Q1
Median
Q3
26,973
0.00
0.32
1.38
3.15
207.79
Max.
Herd health status
Conventional SPF
16,583 10,390
0.00 0.01
0.33 0.31
1.43 1.34
3.18 3.09
127.46 207.79
Herd type
Integrated Only finishers
19,921 7052
0.00 0.01
0.28 0.47
1.28 1.79
2.87 4.19
160.50 207.79
Region
Eastern Jutland Funen Northern Jutland Southeastern Jutland Western Jutland Zealand
3468 3044 7547 5463 4385 3066
0.01 0.01 0.00 0.01 0.01 0.01
0.46 0.41 0.30 0.39 0.26 0.19
1.52 1.47 1.46 1.38 1.29 1.10
3.45 3.16 3.34 3.12 3.13 2.56
60.10 80.92 176.50 207.79 116.22 55.03
Year
2004 2005 2006 2007
6153 6872 7168 6780
0.01 0.01 0.01 0.00
0.33 0.30 0.33 0.32
1.44 1.38 1.36 1.37
3.27 3.08 3.14 3.13
207.79 188.68 127.46 159.82
Quarter
1 2 3 4
6949 6660 6575 6789
0.01 0.01 0.01 0.00
0.34 0.32 0.30 0.33
1.42 1.38 1.34 1.40
3.19 3.14 3.10 3.15
159.82 188.68 198.91 207.79
Herd sizea
a
N
Min.
Q1
Median
Q3
Max.
27,898,647
54
576
920
1385
9795
Herd size is for each quarter defined by the moving average of number of pigs delivered to Danish abattoirs in the actual quarter and the three preceding quarters.
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
195
Table 2 Parameter estimates, standard error and p-values for the final Model 1 with prescribed AM for finishers measured by log-transformed adjusted Animal Daily Doses (ADD) for GI diseases in 3192 Danish pig herds, 2004–2007. Fixed variable
Estimatea
Level
Intercept
99% CI
p-Value
−0.34
−0.49
−0.19
Herd type
Integrated Only finisher pigs
−0.23 0
−0.32 –
−0.14 –
<0.0001
Log herd size (herd health status)
Conventional SPF
−0.19 −0.37
−0.27 −0.47
−0.12 −0.28
<0.0001
Log herd size* log herd size Region
Eastern Jutland Funen Northern Jutland Southeastern Jutland Western Jutland Zealand
Quarter
1 2 3 4
0.12
0.06
0.18
0.52a 0.42abc 0.36bc 0.43ab 0.28c 0d
0.36 0.25 0.22 0.28 0.12 –
0.69 0.59 0.50 0.58 0.44 –
<0.0001
−0.03 −0.06 −0.12 –
0.06 0.04 −0.03 –
<0.0001
0.02a −0.01a −0.07b 0a
Random variable
Estimate
SE
Autocorrelation Residual
0.62 2.57
0.005 0.03
a
Levels within the same variable with different superscript letters are significantly different at a 1% significance level.
pattern might be due to veterinarians own experience and tradition. Economic interest cannot be the motivation as the veterinarians do not profit on prescribing specific types and amount of AM. The veterinarian receives a fixed honorarium for prescribing medicine in Denmark. Of the investigated herd risk factors, herd size had a significant effect on AMU for GI diseases in combination with herd health status, as herds with only finishers had a larger decrease in AMU with increasing herd size than integrated herds. The effect of herd size could be due to the different conditions in large and small herds, e.g. differences in purchasing patterns. Smith et al. (1998) found consistent positive associations between herd size and GI diseases, albeit with herd size defined by number of sows. In a Danish questionnaire survey returned by 116 participating swine producers, Boklund et al. (2004) found that effective and consistent all-in all-out production in either sections or the whole farm was more often achieved in large herds, and that large herds had better management procedures, housing facilities and biosecurity measures regarding contact or transport than small herds. Additionally, Stege et al. (2001) found that consistent all-in all-out production was protective against L. intracellularis and B-haemolytic spirochetes. The effect of herd type could be because herds with only finishers had purchased 7 kg or 30 kg pigs from different herds with different disease status. Herd health status was expected to be significant. Although the fact that most of the diseases used for classification of SPF status are not GI-diseases, herd health status was included in this study because of the assumption on that the high demands of good hygiene and biosecurity would influence the level of AM demanding GI diseases on the SPF farms. The apparent insignificance in this study could be due to misclassification bias. Such bias might be the result of labelled conventional herds with only finishers
that have biosecurity levels as high as those of SPF herds. The main incentive for farmers to join the SPF system is that these farmers get higher prices for selling weaners to other farms, whereas there is no difference in the price of pigs being delivered to slaughter between the two different herd health status groups. Therefore, it would not be as imperative for finisher herds to participate in the SPF system with the additional cost that involves. The significant difference in AMU between some regions indicates differences in the distribution of GI diseases, and in diseases influencing the development of GI diseases such as PRRS disease. Zealand and Western Jutland with the lowest AMU among the Danish regions investigated, had a lower pig density, and this might have influenced the distribution of GI diseases (Boklund et al., 2004). The difference in region could also be due to differences in treatment patterns. 4.2. Data quality and availability The study was based on register data. The coverage of the data included herds with finishers in Denmark annually producing at least 200 finishers. The study revealed that 33% of the herds did not use any antimicrobials during the study period 2004–2007. This figure might be overestimated due to, e.g. inconsequent registration. A limitation for using register data in the study is that it was not possible to identify bias such as erroneous diagnosis. Since AM-demanding GI diseases were the outcome variable, a syndromic approach was used. In this approach, not just actual disease (e.g. clinical diseases like diarrhoea or poor performance) but also the farmer’s perception of disease in the pigs, and the veterinarian’s diagnosis of clinical disease (e.g. as clinical diseases like diarrhoea or poor performance) were included. In the present study,
196
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
syndromic GI diseases were evaluated, using AMU for GI diseases as the outcome. Number of pigs sent for slaughter in Denmark was used as a proxy for number of pigs at risk of treatment. Only a limited number of finisher pigs were exported for slaughter (2–4%) during the study period; and the exported pigs came from all over Denmark (Anonymous, 2009b). Therefore, exportation did not bias the results in the present study. A potential bias by herd type could be that integrated herds are breeding gilts themselves and thus are sending fewer of their finishing animals to slaughter, appearing to have higher ADD than finisher herds. However, we considered that this would not influence the results considerately, because only approximately 2% of integrated herds were breeding gilts within the herd (calculation of the estimate is based on information from the references Anonymous, 2009d and Anonymous, 2010b). The number of finishers sent to slaughter was used as a proxy for the population at risk in order to standardize the AMU. In this way, it was implicitly assumed that there was a uniform treatment pattern. The possibility that disease in, and hence also treatment of, the finishers with GI diseases in large herds was limited to sections, rather than being spread across the whole group of finishers, is in conflict with this assumption, since a limited spread of this sort would entail that the defined population at risk (all finishers sent to slaughter) in large herds is larger than the actual population. It would have been relevant to include weaners, as this age group, followed by the finishers, had the largest prescribed use of AM for GI diseases (Fig. 1). However, suitable information on the risk factor herd size (considering number of weaners) is currently not available for the objective of this study; and at present, nothing indicates that the included herd risk factors could explain more of the variation in AMU for weaners compared to finishers. Further studies of herds with no AMU regarding their farming practice could be relevant. The fact that we are working with data on prescribed AM and not actually recordings on applied AM, will always give rise to some uncertainty on which age group actually got the medicine in integrated herds. However, we assumed that the risk of using AM to another age group than prescribed would be small. When it is prescribed to one age group, we assume that this was the group, which were diseased. Additionally, veterinarians make consultation visits every month, hereby giving rise to make prescriptions for other groups according to legislation. 5. Conclusion The present study offers a better understanding of the influence of herd size and herd type on AM-demanding GI diseases in finishers in Denmark. Smaller herds had a larger AMU per finisher than larger herds. Integrated herds had lower AMU as compared with herds with only finishers. Herds within the SPF System had a larger decrease in AMU with increasing herd size compared to conventional herds. There were significant regional differences in AMU. However, the study found that other herd factors and veterinarians were more influential than the investigated herd
risk factors. These results illustrate the difficulties of characterising AM-demanding GI diseases in herds by the use of solely register data. Acknowledgements We thank Bodil Ydesen from the Danish Agriculture and Food Council; Erik Jacobsen from the National Veterinary Institute and Vibeke Frøkjær Jensen from the National Food Institute (Technical University of Denmark); and the Danish Veterinary and Food Administration (Ministry of Food, Agriculture and Fisheries) for providing data. We are also grateful to Nils Toft (Faculty of Life Sciences, University of Copenhagen), Bjørn Lorenzen and Poul Bækbo (Danish Pig Production) for support. References Aarestrup, F.M., 1999. Association between the consumption of antimicrobial agents in animal husbandry and the occurrence of resistant bacteria among food animals. Int. J. Antimicrob. Agents 12, 279–285. Anonymous, 2007. Bekendtgørelse om dyrlægers anvendelse, udlevering og ordinering af lægemidler til dyr. BEK nr 482 af 29/05/2007. http://www.retsinfo.dk. Anonymous, 2009a. Bekendtgørelse om indberetning af oplysninger til Lægemiddelstatistik. BEK nr 816 af 28/08/2009 http://www. retsinfo.dk, May 15th 2010. Anonymous, 2009b. http://www.lf.dk/Vidensbank/Statistik/Aktuelle statistikker svin, September 1st 2009. Anonymous, 2009c. http://www.spf-sus.dk, September 1st 2009. Anonymous, 2009d. Dansk Årsberetning 2008, Genetisk forskning og udvikling, Dansk Svineproduktion. http://www.arkiv. dansksvineproduktion.dk/, 28th of May, 2010. Anonymous, 2010a. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atc, May 15th 2010. Anonymous, 2010b. http://www.statistikbanken.dk/statbank5a/default.asp?w=1755, 28th of May, 2010). Boklund, A., Alban, L., Mortensen, S., Houe, H., 2004. Biosecurity in 116 Danish fattening swineherds: descriptive results and factor analysis. Prev. Vet. Med. 66, 49–62. Brown, H., Prescott, R., 2006. Repeated Measures Data. Applied Mixed Models in Medicine, 2nd ed. John Wiley & Sons Ltd., West Sussex PO19 8SQ, England, p. 226. Bruun, J., Flensburg, M.F., Ersbøll, A.K., 2004. Bias and interaction. In: Houe, H., Ersbøll, A.K., Nils, T. (Eds.), Introduction to Veterinary Epidemiology. Bifolia, Frederiksberg, Denmark, pp. 241–245. Bruun, J., Ersbøll, A.K., Bennedsgaard, T.W., Larsen, P.B., 2003. Risk factors for different health measures in Danish dairy cows – with metritis, antibacterial drug use and udder health as examples. PhD Thesis. Department of Animal Science and Animal Health, Frederiksberg, Denmark (article 2). Chouet, S., Prieto, C., Mieli, L., Veenhuizen, M.F., McOrist, S., 2003. Patterns of exposure to Lawsonia intracellularis infection on European pig farms. Vet. Rec. 152, 14–17. Conover, W.J., Iman, R.L., 1981. Rank transformations as a bridge between parametric and nonparametric statistics. Am. Stat. 35, 124–133. Gardner, I.A., Willeberg, P., Mousing, J., 2002. Empirical and theoretical evidence for herd size as a risk factor for swine diseases. Anim. Health Res. Rev. 3, 43–55. Jacobson, M., Hård af Segerstad, C., Gunnarsson, A., Fellstrom, C., de Verdier Klingenberg, K., Wallgren, P., Jensen-Waern, M., 2003. Diarrhoea in the growing pig—a comparison of clinical, morphological and microbial findings between animals from good and poor performance herds. Res. Vet. Sci. 74, 163–169. Jensen, V.F., Jacobsen, E., Bager, F., 2004. Veterinary antimicrobial-usage statistics based on standardized measures of dosage. Prev. Vet. Med. 64, 201–215. Johansen, M., 2005. Lawsonia og andre mave-tarmlidelser behandlingstid og kontrolforanstaltninger. VetInfo nr. 0527. http://www.danishmeat.dk/Veterinaerfagligt/VetInfo/2005/2005/ VetInfo 0527.aspx. Kroll, J.J., Roof, M.B., Hoffman, L.J., Dickson, J.S., Harris, D.L., 2005. Proliferative enteropathy: a global enteric disease of pigs caused by Lawsonia intracellularis. Anim. Health Res. Rev. 6, 173–197.
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197 McOrist, S., Smith, S.H., Green, L.E., 1997. Estimate of direct financial losses due to porcine proliferative enteropathy. Vet. Rec. 140, 579–581. McOrist, S., 2005. Defining the full costs of endemic porcine proliferative enteropathy. Vet. J. 170, 8–9. Pearce, G.P., 1999. Epidemiology of enteric disease in grower-finisher pigs: a postal survey of pig producers in England. Vet. Rec. 144, 338–342. Smith, S.H., McOrist, S., Green, L.E., 1998. Questionnaire survey of proliferative enteropathy on British pig farms. Vet. Rec. 142, 690–693. Stege, H., Jensen, T.K., Moller, K., Baekbo, P., Jorsal, S.E., 2000. Prevalence of intestinal pathogens in Danish finishing pig herds. Prev. Vet. Med. 46, 279–292.
197
Stege, H., Jensen, T.K., Moller, K., Baekbo, P., Jorsal, S.E., 2001. Risk factors for intestinal pathogens in Danish finishing pig herds. Prev. Vet. Med. 50, 153–164. Straw, B., Zimmerman, J., D’Allaire, S., Taylor, D. (Eds.), 2006. Diseases of Swine, 9th eds. Blackwell Publishing, Ames, Iowa. Vigre, H., Dohoo, I.R., Stryhn, H., Jensen, V.F., 2010. Use of register data to assess the association between use of antimicrobials and outbreak of Postweaning Multisystemic Wasting Syndrome (PMWS) in Danish pig herds. Prev. Vet. Med. 93 (2–3), 98–109. Weisberg, S., 1985. Applied Linear Regression, 2th eds. John Wiley & Sons, New York.
Contents lists available at ScienceDirect
Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed
Herd-level risk factors for antimicrobial demanding gastrointestinal diseases in Danish herds with finisher pigs A register-based study G.K. Hybschmann a,∗ , A.K. Ersbøll b , H. Vigre c , N.P. Baadsgaard d , H. Houe a a b c d
Department of Large Animal Sciences, Faculty of Life Sciences, University of Copenhagen, Denmark National Institute of Public Health, University of Southern Denmark, Denmark Division of Microbiology and Risk Assessment, National Food Institute, Technical University of Denmark, Denmark Danish Agriculture and Food Council, Pig Research Centre, Denmark
a r t i c l e
i n f o
Article history: Received 13 November 2009 Received in revised form 16 September 2010 Accepted 9 October 2010 Keywords: Finisher pig herds Antimicrobial use Herd size SPF Herd type
a b s t r a c t Endemic gastrointestinal (GI) diseases have a substantial negative impact on pig production, because, when present, they reduce animal welfare, productivity and generate high antimicrobial (AM) demand. In Danish legislation, AM can be prescribed only for therapeutic purposes. The objective of the study was to estimate the association between herd-level risk factors and the amount of AM use (AMU) in connection with GI diseases in finisher herds. We conducted a register-based cross-sectional study with repeated measurements from 2004 to 2007. Data were extracted from databases in the Danish Register of Veterinary Medicine, the Central Husbandry Register and the Danish Agriculture and Food Council. In total, 3192 pig herds with 26,973 records (quarters with prescriptions) were included. The outcome was presented as average AM use (measured as Animal Daily Dosage) for GI diseases per finishing pig per quarter per herd. Three potential herd-level risk factors were evaluated: herd size (number of finishers delivered for slaughter); herd health status (herds in the Specific Pathogen Free (SPF) System, conventional herds); and herd type (herds including only finishers, integrated herds). Data were analyzed using general linear mixed models with repeated measurements. Smaller herds had a larger AMU per finisher than larger herds. Integrated herds had lower AMU as compared with herds with only finishers. Herds within the SPF System had a larger decrease in AMU with increasing herd size compared to conventional herds. Significant regional differences in AMU were seen. Additionally, the results showed that other herd factors and veterinarians were more influential than the investigated herd risk factors. This illustrates the difficulties of characterising AM-demanding GI diseases in herds by the use of register data only. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Endemic gastrointestinal (GI) diseases in finisher pigs have a considerable negative impact on pig production,
∗ Corresponding author at: University of Copenhagen, Faculty of Life Sciences, Department of Large Animal Sciences, Veterinary Epidemiology, Grønnegaardsvej 8, DK-1870 Frederiksberg C, Denmark. Tel.: +45 3528 3011; fax: +45 3533 3022. E-mail address: [email protected] (G.K. Hybschmann). 0167-5877/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2010.10.005
because they compromise animal welfare and reduce productivity and hence cause substantial economic losses (McOrist et al., 1997; Kroll et al., 2005; McOrist, 2005; Straw et al., 2006). In Denmark, the major infectious cause of GI diseases in finishers is Lawsonia intracellularis, the aetiological agent of porcine proliferative enteropathy. However, Brachiospira pilosicoli, Brachiospira hyodysenteriae, Escherichia coli and Porcine Circovirus Type 2 also cause diarrhoea and poor growth performance in finishers (Stege et al., 2000; Jacobson et al., 2003). Furthermore, these GI diseases trigger considerable use of antimicrobials
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
(AM), and this is believed to have an impact on the development of AM resistance within pig herds as well as in humans (Aarestrup, 1999). Better understanding both at the herd and at the regional level, of the risk factors at play here should, therefore, help to clarify the factors that can influence the presence of these multi-factorial GI diseases. Previous studies evaluating risk factors for infectious GI diseases have identified associations between the presence of the infectious agent/disease and management and feedrelated factors (Pearce, 1999; Stege et al., 2001; Chouet et al., 2003). Herd size is frequently studied as a risk factor for swine diseases, and there do appear to be associations between swine diseases and management, environmental factors and herd size (Gardner et al., 2002). Smith et al. (1998) found a strong association between proliferative enteropathy and herd size in herds with more than 500 sows. However, a Danish study by Stege et al. (2001) did not identify herd size as a significant factor. Pig herds in the Danish herd health system, the so-called SPF-system (SPF – Specific Pathogen Free) must comply with high standards of biosecurity and have to be free from at least one of the specified SPF diseases (Boklund et al., 2004). Therefore, herd health status is a relevant risk factor. Herds with only finishers continuously have to purchase new pigs and introduce them into the herd. This could mean that the risk of infectious GI agents being transmitted into the herd is higher in these herds compared to integrated herds. There is a growing tendency to use register data in epidemiological studies, as this is a manageable and affordable way to obtain access to large amounts of data describing longer periods of time. The objective of the present study was to determine the association between AM-demanding GI diseases in finishers and the following herd-level risk factors: herd size, herd health status and herd production type, adjusted by region, season and year. Danish legislation requires that AM only can be used if it is prescribed by a veterinarian and it must be strictly used for therapeutic purposes (Anonymous, 2007). Therefore, AM treatment can be used as an indirect measure of the presence of disease. The study uses general linear mixed models with repeated measurements of register data. 2. Materials and methods 2.1. Study design and population A cross-sectional study with repeated measurements was designed. Danish pig herds with finisher pigs were the target population. Finishers were defined as pigs from 30 kg until slaughter. The study population was indoor production herds with more than 50 finishers produced each quarter and with AM prescriptions for GI diseases to finisher pigs; see Section 2.4.1 for a description of the exclusion criteria. The study unit was the herd per year per quarter. 2.2. Data Data on AM consumption were obtained from the Danish Register of Veterinary Medicine (Vetstat). Since 2002,
191
it has been mandatory to record any AM prescribed to production animals in Vetstat. Information on prescribed AM use (AMU) for pigs is recorded in Vetstat by primarily pharmacies. For each prescription, the data encompass information on day of prescription, herd identity code (Central Husbandry Register (CHR) number), animal species, age group (sows and piglets, weaners, finishers) and disease category in terms of diseased organ (gastrointestinal diseases, respiratory diseases, joints and limbs diseases, reproductive diseases, udder diseases, metabolic diseases, and other diseases) as well as information on the drug (product identity (the Anatomical Therapeutic Chemical (ATC) classification system, specific number and name of the product); and quantity (amount of the product in specified unit) (Anonymous, 2009a, 2010a). In order to compare different AM types of different strengths and potency, the amount of AM prescribed is in Vetstat standardized by a standardized measure for drug use, Animal Daily Dosage (ADD). The measure is defined as the average daily dose for the main indication, using recommendations on doses approved by the Danish Medicine Agency and the Veterinary Pharmaceutical Producer Association (Jensen et al., 2004); on the main part in practice, these doses recommendations for animals are only specified on animal species. The disease category (indication) and ADD is included in the information from Vetstat. Data on herd characteristics (herd type, location) were extracted from the CHR and from the Danish Agriculture and Food Council. Abattoirs submit information on the number of delivered finishers to the Zoonosis Register, and these data were extracted. Furthermore, laboratories automatically send laboratory results to the SPF Company. This information had been applied in the SPF Company to classify the herds into the SPF system. These data were obtained for this study. The majority of Danish pig herds have a herd health contract, which implies monthly visits by a veterinarian. Therefore, using a month as the time unit was unsuitable and quarters of a year were selected as the unit of time. Data from the different sources were merged and aggregated at herd-year-quarter level. In the final data set, one observation (herd-year-quarter) included information on the amount and type of prescribed AM, the number of pigs delivered for slaughter, the number of sows in the herd, herd health status, prescribing veterinarian and region, and, additionally, derived variables. 2.3. Description of derived variables 2.3.1. Outcome variable Outcome was amount of prescribed AM for GI diseases per finisher per quarter at herd-level, quantified by ADD. We linked AMU with antimicrobial demanding disease as AM is only allowed for treatment purpose in Denmark. Compliance to this legislation is closely followed by the authorities, and therefore, we assumed that this legislation in general was followed. AM prescribed for GI diseases was included for all application types. AMU is prescribed to groups of finishers in the herds (population at risk). Vetstat does not give information on whether for example five pigs had been treated
192
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
once or one pig has been treated five days in a row. In Vetstat, information on AMU prescribed for pigs is recorded at herd-level. As the number of treated pigs was not recorded in Vetstat, a proxy for the number of pigs at risk of disease/treatment was defined. Since it takes approximately 3 months from the time of inclusion of pigs in the herd at 30 kg until slaughter, the number of finishers sent to slaughter within the quarter was used as a proxy for the population at risk. Hence, AMU was standardized by dividing ADD by the number of finishers sent to Danish Abattoirs in the actual quarter.
2.3.2. Herd-level risk factors Herd health status was a dichotomous variable: herds that were declared free from at least one of the selected SPF diseases and in compliance with stringent biosecurity rules (SPF), or not (conventional). The conventional herds were not as such obliged to follow the stringent biosecurity rules, and could either have (1) unknown SPF status or (2) a known status of porcine reproductive and respiratory syndrome, but unknown status regarding the selected SPF diseases. SPF herds are examined on regular basis so that they can be declared free of selected infectious agents: Actinobacillus pleuropneumoniae, Mycopacteria hyopneumoniae, haemolytic Brachyspira hyodysenteriae, toxin-producing Pasturella multocida, Haematopinus suis and Sarcoptes scabiei var. suis (Anonymous, 2009c). Herd type was defined as a dichotomous variable with either integrated herds (both sows and finishers) or herds with only finishers. Herd size was defined as the moving average of the quarterly delivery to Danish abattoirs each year, using the number of delivered finishers for the actual quarter and the previous three quarters.
2.3.3. Region Region was included in the study in order to adjust for regional differences in, for example, overall pig production, the prevalence of disease, and production patterns. Denmark was divided into the following six regions: Zealand (West Zealand, Storstrøms, Frederiksborg, and Roskilde Counties), Funen (Funen County), Northern Jutland (North Jutland and Viborg Counties), Western Jutland (Ribe and Ringkøbing Counties), Southeastern Jutland (South Jutland and Vejle Counties), and Eastern Jutland (Århus County). The island Bornholm was excluded, because it had too few herds.
2.3.4. Quarter Season was defined by the four quarters of the year, i.e. January–March, April–June, July–September and October–December.
2.3.5. Veterinarian In each herd, different veterinarians had prescribed AM during the period, 2004–2007. Hence, for each herd, the main veterinarian was identified as the veterinarian with most prescriptions in the period.
2.4. Statistical analysis 2.4.1. Data control and editing In order to detect missing values and extreme values, data control was performed for qualitative variables by means of frequency distributions, and for quantitative variables by means of maximum and minimum values and scatter plots. Herds with extreme values of CHR number, ADD, herd health status or herd size were checked using the original registers. Observations were excluded if: (1) herd characteristics (herd type, herd health status, herd size) were missing; (2) herds had irregular CHR numbers (<10,000 and >115,000); (3) observations included prescriptions, but with no information on type of AM; (4) observations had prescriptions for fluoroquinolones, penicillin for exterior use, intramamamaria, amphenicols, aminoglycosides, cephalosporins, colistin, combination-penicillins, and simple penicillins, thus leaving prescriptions on lincosamids, lincospectin, makrolids, sulfamethoxazole/trimethoprim, tetracyclines and tiamulines in the study; or (5) the 12-month moving average of delivered finishers in the quarter was less than 50 finishers. Some of the excluded AM were removed due to failure in recording (e.g. some AM types was not in agreement with indication, and colistin is used for weaners in Denmark). In total, the excluded AM types accounted for only 1.5% of the total prescribed ADD. 2.4.2. Descriptive analyses Descriptive analysis of quantitative variables was performed in terms of measures for location and spread. Due to the left-skewed distribution of the outcome, the description was given in median, min., max., Q1 and Q3 . The qualitative variables were described in frequency distributions. 2.4.3. Statistical analyses Data were hierarchically structured with herds nested within each region and with herd-level risk factors varying over time. A general linear model with repeated measurements was used to evaluate the effect of potential risk factors on the average amount of prescribed AM per finisher. Herd-level risk factors, region, year and season were included as fixed effects as well as two-way interaction terms (Model 1). Any correlation between repeated measurements was taken into account by including an autocorrelation structure in the model. Different autocorrelation structures (compound symmetry, firstorder ante-dependence, and first-order autoregressive structure with non-equidistant intervals) were evaluated using Akaike’s Information Criterion (AIC). In order to evaluate different autocorrelation structures, only herds with a minimum of seven quarters with AMU were included. A first-order autoregressive structure with nonequidistant intervals was selected. In the further analyses, all herds with prescribed AMU in at least one quarter were used. Model assumptions were evaluated using the Kolmogorov–Smirnov test for normality and visual inspection of residual plots. Additionally, Box-Cox analysis was performed in order to find the optimal transformation of the outcome (Weisberg, 1985). There was a non-linear rela-
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
tion between the outcome and the continuous explanatory variable, herd size. Consequently, herd size variable was log-transformed. Furthermore, the log-herd size was standardized by subtracting the mean of the log-value. Two observations with extreme residual values had a strong impact on the results of the analyses and were therefore deleted. However, visual inspection of QQ-plots of residuals used to evaluate the Gaussian distribution of the residuals showed that the assumption still was not fulfilled. Therefore, the analyses were also performed with ranktransformed outcome (Conover and Iman, 1981). Backward elimination was applied in order to identify any significant fixed effects using the Kenward–Roger’s test for overall significance (Brown and Prescott, 2006). Insignificant interactions, and subsequently main effects, were excluded, one by one, at a significance level of 1%. The results from the log and rank transformed analyses gave the same significant risk factors. Therefore, results from the model with log transformed outcome were applied in the subsequent analyses. Confounding between potential risk factors was examined by 2-by-2 tables, and by estimating relative difference in parameter estimates when including and then excluding the potential confounding variable in the analyses (Bruun et al., 2004). In order to assess the ability of the herd-level risk factors to account for the between herd variation for AMU a general linear mixed model was evaluated with a random effect of herds. The herd-level risk factors were excluded and herd was included as random effect (Model 2). Veterinarians were not included in the model (Model 1) in order to avoid confounding with herd effects, because the majority of veterinarians had prescribed AM to only a small number of herds. For example, 40% of the veterinarians had prescribed AM to only one (28%) or two herds (12%). A model (Model 3) was constructed without herdlevel risk factors and region, but with veterinarian included as a random effect in order to evaluate the variation due to veterinarians. The statistical analyses were performed using the MIXED procedure in the Statistical Analysis System (SAS vers. 9.1. Cary, NC, USA). 3. Results
193
ranging from 1 to 59 (Q1 = 1, median = 3, Q3 = 12), and with number of quarters with prescriptions ranging from 1 to 585 (Q1 = 6, median = 21, Q3 = 99). Among the six types of AM prescribed for GI diseases, tetracyclines, macrolides and tiamulin accounted for 94% of the use. Descriptive results of the data applied in the analysis (full study population) are presented in Table 1. Overall, the median AMU was 1.38 ADD per finisher per quarter per herd. 3.2. Herd-level risk factors The final model (Model 1), evaluating the herd-level risk factors, is presented in Table 2. The main effects – herd size, herd type, region and quarter – were significantly associated with the amount of prescribed AMU per finisher (p < 0.001). The main effect of herd health status was not significant; however, the interaction between herd size and herd health status was significant. Integrated herds had a lower adjusted AMU than herds having only finishers. SPF herds had a larger reduction in adjusted AMU than conventional herds with increasing herd size. There was a significant difference in AMU between regions. Herds on Zealand and in Western Jutland had the lowest adjusted AMU, whereas herds in Eastern and Southeastern Jutland and on Funen had the highest adjusted AMU. Although only with small differences, the highest AMU was in the first quarter of the year and the lowest was in the third quarter. The ability of herd-level risk factors (herd size, herd type and herd health status) to account for the between herd variation was evaluated. The total variation of the logtransformed adjusted ADD was 2.65. The residual variation in Model 1 with herd-level risk factors was estimated at 2.57. Therefore, the herd-level risk factors accounted for 3.0% of the total variation. If the between herd variation was modelled with a random effect of herds, the between herd variation was estimated at 1.06 and with a residual variation estimated at 1.58. Thus, the between-herd variation accounted for 40% of the total variation. The variation due to differences in AMU between veterinarians (Model 3 with veterinarian included as a random effect) was estimated at 0.28 with a residual variation at 2.41. Hereby, the random effect of veterinarians could explain 10.7% of the total variation.
3.1. Descriptive statistics 4. Discussion The distribution of all AM prescribed to pigs in 2006–2007 stratified by the group of pigs (sows and piglets, weaners and finishers) and by five diagnostic groups (GI, respiratory, joint, mastitis, and reproductive diseases) is illustrated in Fig. 1. As the figure shows, GI diseases in finishers resulted in the second largest use of AM; they were exceeded only by GI diseases in weaners. The study population consisted of 3192 herds (26,973 observations), each with recorded prescriptions within the range of 1–16 quarters (Q1 = 4, median = 9, Q3 = 12) in the study period, 2004–2007. In total, 897 veterinarians had prescribed AM to a number of herds, ranging from 1 to 124 herds (Q1 = 1, median = 2, Q3 = 7); of these, 369 were main veterinarians with prescriptions for a number of herds
4.1. Risk factors The herd-level risk factors in the study accounted for only 3.0% of the total variation. Obviously, this may be due to the fact that the applied registers were created with other purposes in mind. With the between-herd variation, accounting for 40% of the total variation, further analyses on herd risk factors will therefore require more detailed information than available in registers on for example farm management and housing. It was not possible to adjust for veterinarians. Therefore, a mixed model with veterinarian as random effect (Model 3) was constructed. The results indicated that the
194
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
40 Data used in study Other data
35
Percentage
30 25 20 15 10 5 0
GI
Joint Resp Finishers
GI
Joint Resp Weaners
GI
Joint Resp Sows/Piglets
Repro Mastit Sows
Fig. 1. Distribution (%) of the total antimicrobial use to all Danish pigs on combined age-diagnosis groups, 2006–2007. GI: GI diseases; Joint: joint diseases; Resp: respiratory diseases; Repro: reproductive diseases; Mastit: mastitis.
variation due to veterinarians could explain 10.7% of the total variation. Even though, it was not possible to adjust for herd risk factors in Model 3 (because many veterinarians had health consultancy appointments with one or few herds), veterinarians was found to be more influential than the herd risk factors investigated in the present study (Model 1); still, the between herd variation were almost 4 times higher than the variation between veterinarians. Vigre et al. (2010) found similar results, where the variation of the total amount of AMU (ADD) prescribed to
finishers was 4 times higher than the variation within the veterinarians themselves. A Danish questionnaire study by Johansen (2005) investigated swine practitioner’s strategy for treatment of gastro-intestinal diseases and found variation between the responding practitioners. A study on antibacterial drug use in Dairy cattle (Bruun et al., 2003) found that veterinary practice influenced the choice of broad- versus narrow-spectrum AM. However, to our knowledge there are no publications with studies looking more into detail about this. The difference in treatment
Table 1 Descriptive statistics of amount of prescribed AM for finishers for GI diseases measured by adjusted Animal Daily Doses (ADD) in 3192 Danish pig herds, 2004–2007. Variable
Level
Amount of prescribed AM for finishers (ADD per finisher) N
Overall
Min.
Q1
Median
Q3
26,973
0.00
0.32
1.38
3.15
207.79
Max.
Herd health status
Conventional SPF
16,583 10,390
0.00 0.01
0.33 0.31
1.43 1.34
3.18 3.09
127.46 207.79
Herd type
Integrated Only finishers
19,921 7052
0.00 0.01
0.28 0.47
1.28 1.79
2.87 4.19
160.50 207.79
Region
Eastern Jutland Funen Northern Jutland Southeastern Jutland Western Jutland Zealand
3468 3044 7547 5463 4385 3066
0.01 0.01 0.00 0.01 0.01 0.01
0.46 0.41 0.30 0.39 0.26 0.19
1.52 1.47 1.46 1.38 1.29 1.10
3.45 3.16 3.34 3.12 3.13 2.56
60.10 80.92 176.50 207.79 116.22 55.03
Year
2004 2005 2006 2007
6153 6872 7168 6780
0.01 0.01 0.01 0.00
0.33 0.30 0.33 0.32
1.44 1.38 1.36 1.37
3.27 3.08 3.14 3.13
207.79 188.68 127.46 159.82
Quarter
1 2 3 4
6949 6660 6575 6789
0.01 0.01 0.01 0.00
0.34 0.32 0.30 0.33
1.42 1.38 1.34 1.40
3.19 3.14 3.10 3.15
159.82 188.68 198.91 207.79
Herd sizea
a
N
Min.
Q1
Median
Q3
Max.
27,898,647
54
576
920
1385
9795
Herd size is for each quarter defined by the moving average of number of pigs delivered to Danish abattoirs in the actual quarter and the three preceding quarters.
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
195
Table 2 Parameter estimates, standard error and p-values for the final Model 1 with prescribed AM for finishers measured by log-transformed adjusted Animal Daily Doses (ADD) for GI diseases in 3192 Danish pig herds, 2004–2007. Fixed variable
Estimatea
Level
Intercept
99% CI
p-Value
−0.34
−0.49
−0.19
Herd type
Integrated Only finisher pigs
−0.23 0
−0.32 –
−0.14 –
<0.0001
Log herd size (herd health status)
Conventional SPF
−0.19 −0.37
−0.27 −0.47
−0.12 −0.28
<0.0001
Log herd size* log herd size Region
Eastern Jutland Funen Northern Jutland Southeastern Jutland Western Jutland Zealand
Quarter
1 2 3 4
0.12
0.06
0.18
0.52a 0.42abc 0.36bc 0.43ab 0.28c 0d
0.36 0.25 0.22 0.28 0.12 –
0.69 0.59 0.50 0.58 0.44 –
<0.0001
−0.03 −0.06 −0.12 –
0.06 0.04 −0.03 –
<0.0001
0.02a −0.01a −0.07b 0a
Random variable
Estimate
SE
Autocorrelation Residual
0.62 2.57
0.005 0.03
a
Levels within the same variable with different superscript letters are significantly different at a 1% significance level.
pattern might be due to veterinarians own experience and tradition. Economic interest cannot be the motivation as the veterinarians do not profit on prescribing specific types and amount of AM. The veterinarian receives a fixed honorarium for prescribing medicine in Denmark. Of the investigated herd risk factors, herd size had a significant effect on AMU for GI diseases in combination with herd health status, as herds with only finishers had a larger decrease in AMU with increasing herd size than integrated herds. The effect of herd size could be due to the different conditions in large and small herds, e.g. differences in purchasing patterns. Smith et al. (1998) found consistent positive associations between herd size and GI diseases, albeit with herd size defined by number of sows. In a Danish questionnaire survey returned by 116 participating swine producers, Boklund et al. (2004) found that effective and consistent all-in all-out production in either sections or the whole farm was more often achieved in large herds, and that large herds had better management procedures, housing facilities and biosecurity measures regarding contact or transport than small herds. Additionally, Stege et al. (2001) found that consistent all-in all-out production was protective against L. intracellularis and B-haemolytic spirochetes. The effect of herd type could be because herds with only finishers had purchased 7 kg or 30 kg pigs from different herds with different disease status. Herd health status was expected to be significant. Although the fact that most of the diseases used for classification of SPF status are not GI-diseases, herd health status was included in this study because of the assumption on that the high demands of good hygiene and biosecurity would influence the level of AM demanding GI diseases on the SPF farms. The apparent insignificance in this study could be due to misclassification bias. Such bias might be the result of labelled conventional herds with only finishers
that have biosecurity levels as high as those of SPF herds. The main incentive for farmers to join the SPF system is that these farmers get higher prices for selling weaners to other farms, whereas there is no difference in the price of pigs being delivered to slaughter between the two different herd health status groups. Therefore, it would not be as imperative for finisher herds to participate in the SPF system with the additional cost that involves. The significant difference in AMU between some regions indicates differences in the distribution of GI diseases, and in diseases influencing the development of GI diseases such as PRRS disease. Zealand and Western Jutland with the lowest AMU among the Danish regions investigated, had a lower pig density, and this might have influenced the distribution of GI diseases (Boklund et al., 2004). The difference in region could also be due to differences in treatment patterns. 4.2. Data quality and availability The study was based on register data. The coverage of the data included herds with finishers in Denmark annually producing at least 200 finishers. The study revealed that 33% of the herds did not use any antimicrobials during the study period 2004–2007. This figure might be overestimated due to, e.g. inconsequent registration. A limitation for using register data in the study is that it was not possible to identify bias such as erroneous diagnosis. Since AM-demanding GI diseases were the outcome variable, a syndromic approach was used. In this approach, not just actual disease (e.g. clinical diseases like diarrhoea or poor performance) but also the farmer’s perception of disease in the pigs, and the veterinarian’s diagnosis of clinical disease (e.g. as clinical diseases like diarrhoea or poor performance) were included. In the present study,
196
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197
syndromic GI diseases were evaluated, using AMU for GI diseases as the outcome. Number of pigs sent for slaughter in Denmark was used as a proxy for number of pigs at risk of treatment. Only a limited number of finisher pigs were exported for slaughter (2–4%) during the study period; and the exported pigs came from all over Denmark (Anonymous, 2009b). Therefore, exportation did not bias the results in the present study. A potential bias by herd type could be that integrated herds are breeding gilts themselves and thus are sending fewer of their finishing animals to slaughter, appearing to have higher ADD than finisher herds. However, we considered that this would not influence the results considerately, because only approximately 2% of integrated herds were breeding gilts within the herd (calculation of the estimate is based on information from the references Anonymous, 2009d and Anonymous, 2010b). The number of finishers sent to slaughter was used as a proxy for the population at risk in order to standardize the AMU. In this way, it was implicitly assumed that there was a uniform treatment pattern. The possibility that disease in, and hence also treatment of, the finishers with GI diseases in large herds was limited to sections, rather than being spread across the whole group of finishers, is in conflict with this assumption, since a limited spread of this sort would entail that the defined population at risk (all finishers sent to slaughter) in large herds is larger than the actual population. It would have been relevant to include weaners, as this age group, followed by the finishers, had the largest prescribed use of AM for GI diseases (Fig. 1). However, suitable information on the risk factor herd size (considering number of weaners) is currently not available for the objective of this study; and at present, nothing indicates that the included herd risk factors could explain more of the variation in AMU for weaners compared to finishers. Further studies of herds with no AMU regarding their farming practice could be relevant. The fact that we are working with data on prescribed AM and not actually recordings on applied AM, will always give rise to some uncertainty on which age group actually got the medicine in integrated herds. However, we assumed that the risk of using AM to another age group than prescribed would be small. When it is prescribed to one age group, we assume that this was the group, which were diseased. Additionally, veterinarians make consultation visits every month, hereby giving rise to make prescriptions for other groups according to legislation. 5. Conclusion The present study offers a better understanding of the influence of herd size and herd type on AM-demanding GI diseases in finishers in Denmark. Smaller herds had a larger AMU per finisher than larger herds. Integrated herds had lower AMU as compared with herds with only finishers. Herds within the SPF System had a larger decrease in AMU with increasing herd size compared to conventional herds. There were significant regional differences in AMU. However, the study found that other herd factors and veterinarians were more influential than the investigated herd
risk factors. These results illustrate the difficulties of characterising AM-demanding GI diseases in herds by the use of solely register data. Acknowledgements We thank Bodil Ydesen from the Danish Agriculture and Food Council; Erik Jacobsen from the National Veterinary Institute and Vibeke Frøkjær Jensen from the National Food Institute (Technical University of Denmark); and the Danish Veterinary and Food Administration (Ministry of Food, Agriculture and Fisheries) for providing data. We are also grateful to Nils Toft (Faculty of Life Sciences, University of Copenhagen), Bjørn Lorenzen and Poul Bækbo (Danish Pig Production) for support. References Aarestrup, F.M., 1999. Association between the consumption of antimicrobial agents in animal husbandry and the occurrence of resistant bacteria among food animals. Int. J. Antimicrob. Agents 12, 279–285. Anonymous, 2007. Bekendtgørelse om dyrlægers anvendelse, udlevering og ordinering af lægemidler til dyr. BEK nr 482 af 29/05/2007. http://www.retsinfo.dk. Anonymous, 2009a. Bekendtgørelse om indberetning af oplysninger til Lægemiddelstatistik. BEK nr 816 af 28/08/2009 http://www. retsinfo.dk, May 15th 2010. Anonymous, 2009b. http://www.lf.dk/Vidensbank/Statistik/Aktuelle statistikker svin, September 1st 2009. Anonymous, 2009c. http://www.spf-sus.dk, September 1st 2009. Anonymous, 2009d. Dansk Årsberetning 2008, Genetisk forskning og udvikling, Dansk Svineproduktion. http://www.arkiv. dansksvineproduktion.dk/, 28th of May, 2010. Anonymous, 2010a. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atc, May 15th 2010. Anonymous, 2010b. http://www.statistikbanken.dk/statbank5a/default.asp?w=1755, 28th of May, 2010). Boklund, A., Alban, L., Mortensen, S., Houe, H., 2004. Biosecurity in 116 Danish fattening swineherds: descriptive results and factor analysis. Prev. Vet. Med. 66, 49–62. Brown, H., Prescott, R., 2006. Repeated Measures Data. Applied Mixed Models in Medicine, 2nd ed. John Wiley & Sons Ltd., West Sussex PO19 8SQ, England, p. 226. Bruun, J., Flensburg, M.F., Ersbøll, A.K., 2004. Bias and interaction. In: Houe, H., Ersbøll, A.K., Nils, T. (Eds.), Introduction to Veterinary Epidemiology. Bifolia, Frederiksberg, Denmark, pp. 241–245. Bruun, J., Ersbøll, A.K., Bennedsgaard, T.W., Larsen, P.B., 2003. Risk factors for different health measures in Danish dairy cows – with metritis, antibacterial drug use and udder health as examples. PhD Thesis. Department of Animal Science and Animal Health, Frederiksberg, Denmark (article 2). Chouet, S., Prieto, C., Mieli, L., Veenhuizen, M.F., McOrist, S., 2003. Patterns of exposure to Lawsonia intracellularis infection on European pig farms. Vet. Rec. 152, 14–17. Conover, W.J., Iman, R.L., 1981. Rank transformations as a bridge between parametric and nonparametric statistics. Am. Stat. 35, 124–133. Gardner, I.A., Willeberg, P., Mousing, J., 2002. Empirical and theoretical evidence for herd size as a risk factor for swine diseases. Anim. Health Res. Rev. 3, 43–55. Jacobson, M., Hård af Segerstad, C., Gunnarsson, A., Fellstrom, C., de Verdier Klingenberg, K., Wallgren, P., Jensen-Waern, M., 2003. Diarrhoea in the growing pig—a comparison of clinical, morphological and microbial findings between animals from good and poor performance herds. Res. Vet. Sci. 74, 163–169. Jensen, V.F., Jacobsen, E., Bager, F., 2004. Veterinary antimicrobial-usage statistics based on standardized measures of dosage. Prev. Vet. Med. 64, 201–215. Johansen, M., 2005. Lawsonia og andre mave-tarmlidelser behandlingstid og kontrolforanstaltninger. VetInfo nr. 0527. http://www.danishmeat.dk/Veterinaerfagligt/VetInfo/2005/2005/ VetInfo 0527.aspx. Kroll, J.J., Roof, M.B., Hoffman, L.J., Dickson, J.S., Harris, D.L., 2005. Proliferative enteropathy: a global enteric disease of pigs caused by Lawsonia intracellularis. Anim. Health Res. Rev. 6, 173–197.
G.K. Hybschmann et al. / Preventive Veterinary Medicine 98 (2011) 190–197 McOrist, S., Smith, S.H., Green, L.E., 1997. Estimate of direct financial losses due to porcine proliferative enteropathy. Vet. Rec. 140, 579–581. McOrist, S., 2005. Defining the full costs of endemic porcine proliferative enteropathy. Vet. J. 170, 8–9. Pearce, G.P., 1999. Epidemiology of enteric disease in grower-finisher pigs: a postal survey of pig producers in England. Vet. Rec. 144, 338–342. Smith, S.H., McOrist, S., Green, L.E., 1998. Questionnaire survey of proliferative enteropathy on British pig farms. Vet. Rec. 142, 690–693. Stege, H., Jensen, T.K., Moller, K., Baekbo, P., Jorsal, S.E., 2000. Prevalence of intestinal pathogens in Danish finishing pig herds. Prev. Vet. Med. 46, 279–292.
197
Stege, H., Jensen, T.K., Moller, K., Baekbo, P., Jorsal, S.E., 2001. Risk factors for intestinal pathogens in Danish finishing pig herds. Prev. Vet. Med. 50, 153–164. Straw, B., Zimmerman, J., D’Allaire, S., Taylor, D. (Eds.), 2006. Diseases of Swine, 9th eds. Blackwell Publishing, Ames, Iowa. Vigre, H., Dohoo, I.R., Stryhn, H., Jensen, V.F., 2010. Use of register data to assess the association between use of antimicrobials and outbreak of Postweaning Multisystemic Wasting Syndrome (PMWS) in Danish pig herds. Prev. Vet. Med. 93 (2–3), 98–109. Weisberg, S., 1985. Applied Linear Regression, 2th eds. John Wiley & Sons, New York.