Administration of ZnO or colistin revealed strongest effects on the occurrence of PWD out of all 69 analysed explanatory variables. Similar to the results of an experimental study demonstrating that treatment with ZnO (2,500 ppm) reduced the incidence of diarrhoea by approximately 50% (57), in our study the percentage of farms with PWD not using ZnO was almost twice as high (58.9%) compared to farms with regular administration of ZnO (31.7%). However, 45 farmers reported to observe PWD regularly in nursery pens despite treatment of weaned piglets. Thus, ZnO does not prevent PWD completely, especially if other protective measures might be neglected. On the other hand, our data demonstrate that administration of other feed additives such as probiotics or horseradish can prevent PWD, although effects were less pronounced compared to ZnO. Therefore, we assume that a combination of different management practices along with the administration of other feed additives may compensate at least partially for the administration of ZnO to prevent PWD.
Besides treatment with ZnO and/or colistin, only neonatal diarrhoea was also associated with the occurrence of PWD in all splits, which was not surprising since ETEC expressing F4 fimbriae can cause both, neonatal diarrhoea and PWD (1). Furthermore, piglets which have already had neonatal diarrhoea might be more susceptible to PWD due to previous damage on the intestines and worse body condition compared to piglets without diarrhoea during the suckling period. Therefore, several approaches to prevent neonatal diarrhoea like sow vaccination against ETEC, Clostridium perfringens or rotavirus A, as well as treatment of piglets with anticoccidials might prevent PWD indirectly. However, those variables were excluded from statistical analysis due to collinearity with the occurrence of neonatal diarrhoea. Similar to PWD, PDS is a multifactorial disease with E. coli playing a key role in its pathogenesis (58). Thus, parameters increasing the risk for PDS might also increase the risk of PWD. Moreover, dysgalactia of sows leads to malnutrition and reduced growth rates of suckling piglets resulting in a higher susceptibility to disease.
Despite the fact that good feeding management and choice of beneficial ingredients in the feed are described to be very crucial to prevent PWD (20), the number of feeding phases until 21 days after weaning was the only variable on feeding management which was associated with the occurrence of PWD in at least 95% of all splits. This could be due to the lack of information on quantities of ingredients contained in purchased commercial products. While in most Austrian farms corn, barley and wheat are produced on-site and exact quantities used in feed were available, protein sources were predominantly supplemented with commercial products. Since information on exact quantities of all ingredients contained in each commercial product (n = 96) was not available, precise evaluation of feed ingredients was not feasible. Although we had to exclude most variables on creep feed from statistical analysis due to collinearity, information whether creep feed was offered or not, is still contained in the variable “number of feeding phases”. This could explain why farms with more feeding phases were less likely to have problems with PWD, since it was demonstrated before that administration of creep feed can decrease the risk of PWD (59). In addition, creep feeding containing fibre had preventative effects in 69% of all splits. While van Hees et al. did not observe significant differences in average daily weight gain (ADG) between piglets getting low-fibre creep feed and piglets with high-fibre creep feed after challenge with ETEC (60), Choudhury et al. demonstrated a reduced incidence and duration of PWD in piglets fed fibrous pre-weaning diet compared to the control group receiving only milk (61).
Despite the fact that reduction of crude protein and increase of crude fibre are widely known strategies to prevent PWD (36), we did not observe any differences of both variables amongst case and control farms. Our results also go in line with observations of another investigation (62). In most studies evaluating the effect of dietary protein level, incidence of diarrhoea was higher in piglets receiving feed with protein levels ranging from 20–26% compared to groups receiving 16–19% crude protein in feed (34–38, 63, 64). Therefore, we assume that the level of protein did not have an influence on the occurrence of PWD in our data, since protein levels of most analysed farms were already comparably low (16–17%). Hence, reduction of crude protein and increase of crude fibre in the diet of weaned piglets alone might not be sufficient to prevent PWD under those circumstances. Albeit in our study farms with automatic feeding in the nursery units had less problems with PWD compared to farms feeding piglets manually, no differences between automatic and manual feeding were observed in a similar study conducted in Finland (65).
In contrast to feeding management and feed composition, several variables on feed additives (8/18) were associated with PWD in at least 60% of all splits. Out of all analysed feed additives, supplementation of probiotics or horseradish had the strongest effects on the occurrence of PWD (Table 3). While strains of Lactobacillus spp., Bifidobacterium spp. and Streptococcus spp. are administered most frequently in pig production worldwide (66), Bifidobacterium spp. was not administered in a single analysed farm. Administration of probiotics might have multiple effects on the intestines, similar to treatment with ZnO, including an increase of villus heights in the jejunum (67) and decreased levels of tumour necrosis factor alpha (67, 68) and interleukin-1 beta (68). Horseradish (Armoracia rusticana) is a plant commonly grown in Styria and has not been described to have preventative properties against PWD so far. Potential explanations why farms administering horseradish were less likely to have problems with PWD could be provided by confirmed antimicrobial properties of mustard essential oils, which are contained in horseradish (69, 70). Unique effects of the administration of horseradish on the prevalence of PWD are yet to be evaluated in experimental trials or additional field studies.
Humic acids contained in peat (71) could explain why the majority of farms administering peat reported to have no problems with PWD (25/34). Administration of 2,000 ppm sodium humate to 24 piglets resulted in significantly higher ADG and decreased levels of pro-inflammatory cytokines compared to the untreated control group (72). Since offering pig peat permits expression of natural behaviour and reduces fighting at weaning, it might also decrease stress levels of piglets and further decrease the probability of PWD (73). Nevertheless, peat is a frequently described hazard for the introduction of pathogens (74, 75). Therefore, offering peat to prevent PWD should be reconsidered based on biosecurity protocols. Since whey protein and casein are well digestible for three to four week old piglets, supplementation of various dairy products, which was also negatively associated with the occurrence of PWD in 86% of all splits, could be considered as a preventative measure against PWD as well (76).
There is currently no evidence that coke could also prevent PWD. In contrast, our data demonstrate that farms administering coke to piglets were more likely to have problems with PWD. Phosphoric acid, an ingredient of coke is an inorganic acid and has never been described to prevent PWD (77). We assume that administration of coke to piglets might be linked to the fact that coke is still used as a home remedy against diarrhoea. This old wives’ tale could be related to the fact that citric acid was used instead of phosphoric acids in previous formulations of coke. In comparison to phosphoric acid, it has already been demonstrated that supplementation of citric acid can prevent diarrhoea evidently (41, 77, 78).
PWD was also less likely to occur in farms administering live yeast and electrolytes, but effects were weaker compared to administration of probiotics or horseradish. Capability of yeast (Saccharomyces cerevisiae) to reduce incidence of PWD and shedding of E. coli has already been demonstrated in experimental studies (53, 79). Additionally, similar to ZnO it was demonstrated that administration of yeast decreased serum concentrations of pro-inflammatory cytokines (80). Application of glucose or electrolytes could generally temper symptoms of PWD (81).
Pronounced effects of variables like all-in/all-out system in the nursery or the type of floor in most splits highlight the importance of general management measures in order to prevent PWD (Table 3). Consequent compliance of an all-in/all-out system in the nursery units can reduce the frequency of PWD, probably due to a reduction of pathogen transmission between piglets of different age groups (82). While implementation of an all-in/all-out system proved to reduce the prevalence of various enteropathogens (83–85), it had no significant effect on the prevalence of ETEC in a Canadian study (62). In addition to an all-in/all-out system, other measures reducing pathogen transmission between individuals might also be beneficial to prevent PWD. One example could be to avoid or reduce cross fostering, as implementation of this management practice was associated positively with the occurrence of PWD in 80% of all splits.
Despite discussions on fully slatted floors in pig production in the context of animal welfare, our data emphasize the advantages of fully slatted floors in the context of hygiene measures, since PWD was less likely to occur in farms keeping piglets on fully slatted floors. Due to easier cleaning and disinfection and the quick removal of faeces through the slats, bacterial load is considered to be lower in pens with fully slatted floors compared to solid floor pens (86, 87). On the other hand, diarrhoea can be observed more easily in pens with solid floor. Similar to our results, Lozano et al. demonstrated that piglets housed on fully slatted plastic floors were less likely to develop PWD than piglets housed on concrete floors (88). Berrocoso et al. demonstrated that husbandry of weaned piglets under optimal hygienic conditions including regular cleaning with detergents and disinfection resulted in significantly lower incidence of PWD and better ADG of weaned piglets (89). Since investigated farms which regularly cleaned and disinfected nursery units were less likely to have problems with PWD, our results emphasize that hygienic measures are crucial to decrease ETEC burden and shall not be neglected in order to prevent PWD.
Besides environment, drinking water and water pipes could be another reservoir of ETEC (90). Due to the fact that looped water distribution systems have less dead endings compared to branched water distribution systems, accumulation of sediments and biofilm formation is less likely (91). Thus, infections with ETEC via contaminated water coming from water pipelines with branched endings could be more likely, since ETEC are frequently recovered from biofilms in water pipes (90, 92).
In addition, rats and mice can also act as a vehicle for introduction of various Enterobacteriaceae like Salmonella spp. (93, 94). Even though there are currently no data on the relation between rodent burden and incidence of PWD, our results emphasize that high rodent burden could presumably lead to increased transmission rates of ETEC and higher bacterial loads. Thus, strict measures against rodents might potentially decrease the probability of occurrence of PWD in swine stocks.
Data collection through questionnaires heavily depends on perspectives and honesty of participants. Thus, bias due to dishonesty, subjective assessments of farmers and recall bias cannot be excluded. However, in order to reduce further bias related to subjectivity, data were always collected by the same person, who visited each farm to ensure consistent quality of collected data. Especially, recall bias for classification into case and control farms might have been high due to difficulties to estimate the number of weaned batches having problems with PWD. Additionally, farmers spending more time in the stables might be more likely to observe PWD. Furthermore, separate analysis for ZnO and colistin was not feasible, since products containing both substances were administered in numerous farms. Categorization of the variable treatment into “yes” or “no” for farms treating at least 50% of all batches of weaned piglets with antibiotics other than colistin, was done based on the results of a previous study with E. coli isolates recovered from Austrian swine stocks (16). However, antimicrobial resistance testing was not performed for E. coli from the visited farms in the current study.
Since an abundance of factors contributes to the pathogenesis of PWD, the number of variables could not be further limited without eliding pivotal parameters. Thus, elastic-net model was chosen to assess the influence of each variable for the output of PWD. While elastic-net is able to handle datasets with numerous predictors, it cannot deliver p-values. Therefore, we decided to renounce p-values. On the other hand, by using this method we were able to identify more variables with an influence on the output of PWD compared to previous similar analyses on PWD (62, 65). McFadden pseudo R² accounting for 0.47 implicates that our dataset was able to partially explain the observed variance. However, since PWD is a complex, multifactorial disease the unexplained variance was no surprise. Further, certain factors influencing the output of PWD might have been elided in the questionnaire or during data processing.