Annex 4: Final Report for Defra Project OD2006

Annex 4: Final report for Defra project OD2006.

Investigating the persistence of antimicrobial-resistant Escherichia coli in depopulated and cleaned broiler-chicken houses[a]

Eve Pleydell, Lourdes Migura, Sue Bedford, Simon Barnes, Rob Davies

Aims

To assess the potential for antimicrobial-resistant E. coli (AR E. coli) to carry-over between flocks via the house environment.
To identify potential problem areas within broiler houses in terms of the persistence of AR E. coli in the environment.
To compare the potential carry-over of AR E. coli in two contrasting management systems.
To assess the potential for multiply-antimicrobial-resistant E. coli to persist between flocks of birds. (See Annex 5)

Summary

Using laboratory methods involving overnight enrichment of samples, E. coli and antimicrobial-resistant E. coli were detected in cleaned broiler houses on one conventional and one organic broiler farm. The odds of detecting E. coli in cleaned houses were higher for the organic farm, implying that the more rigorous disinfection protocols used on the conventional farm were acting to decrease carry-over of bacteria between flocks. These results indicate that further work should be undertaken in collaboration with commercial broiler farms to investigate the most effective cleaning and disinfecting regimes, including the most suitable length of time for houses to remain empty between flocks. The effectiveness of the regimes should be assessed both in terms of decreased contamination of cleaned houses and with respect to levels of shedding of resistant organisms by subsequent flocks.

On the conventional farm ampicillin-resistant and chloramphenicol-resistant E. coli were significantly (P=0.009 and P=0.003 respectively) more likely to be detected in samples taken from cracks in the floors and walls than samples from other areas in the house. Thus implying that the presence of cracks in the structure of the house can act to maintain a reservoir of resistant E. coli.

On the organic farm chloramphenicol-resistant E. coli were isolated more frequently (P=0.005) from the mobile barns than the brooding houses. This corresponds with a rise in the median faecal count of chloramphenicol-resistant E. coli from 0 for birds up to 6 weeks of age to 1x104 cfu/g for birds of 10 weeks of age. The reasons for this proliferation of a resistant E. coli within the faecal flora of birds that have not consumed antimicrobial agents is worthy of further investigation.

To assess the significance of these results in terms of the risk of colonisation of incoming chicks with resistant bacteria remaining in the house, work is needed to ascertain practical methods for assessing the level of contamination within a house in terms of colony-forming units of bacteria remaining after cleaning.

Methods

Descriptions of study farms

This study was conducted on two broiler-chicken farms; one managed by a conventional broiler company and the other a certified organic farm.

The conventional farm finishes 300,000 birds per year over 6 crops.
The birds are permanently housed in six modern, climate-controlled houses in groups of 50,000.
Between flocks the site is depopulated, cleaned and disinfected.
Cleaning regime:
Blow down high dust, scrape out litter, sweep the floor.
Power-wash with quaternary ammonium sanitiser solution.
Disinfect using peracetic acid / H2O2 combination disinfectant.
Additional disinfectant step using a phenolic disinfectant (second visit only).
The houses remain empty of birds for a further 5-7 days after cleaning.

The organic farm finishes 2300 birds per week on a multi-age site.
The birds are reared in groups of 1150.
During days 1 to 21:
The birds are housed in heated brooding houses with access to natural light via a glass conservatory.
Between groups of chicks the brooder houses are cleaned and disinfected.
Scrape the litter out, sweep the floor.
Power-wash and immediately disinfect with iodophor disinfectant.
The houses are then bedded-up and remain empty of birds for 6-8 days.
At 22-days of age:
The birds are transferred to a mobile barn in a field with unrestricted access to range during daylight hours.
Between groups of growers the mobile barns are cleaned and disinfected
Barn dragged to a fresh section of the field.
Inner surfaces and equipment washed and then disinfected with iodophor disinfectant.
Houses are then immediately bedded-up prior to the arrival of the next batch of birds on the following morning.
The poultry-litter is composted at a separate site on the farm.

Sampling Strategies

On each farm the sampling was conducted in a structured manner.

Conventional Farm

From March until May 2004, a flock of birds was monitored for faecal shedding of antimicrobial-resistant E. coli twice weekly from day of arrival to day of departure.

Either side of the monitored flock, two post-cleaning environmental swabbing visits were carried out in two of the houses on-site.
A total of 150 samples were collected in each house on each visit.
The area of the floor (2528m2) was visually divided into 16 equally-sized sections and specified numbers and types of sample were collected from each area.
Surface samples, such as floor and walls, were collected by swabbing 50cm² using gauze cloth swabs and placing two swabs from the same virtual-area of the house into a single jar 200ml buffered peptone water.
Smaller areas, such as nipple drinkers and floor cracks, were sampled using wand swabs in charcoal transport media.

Organic Farm

From January until April 2005, four groups of birds were monitored weekly for faecal shedding of antimicrobial-resistant E. coli from day of arrival to day of departure.

Before the arrival of each group of chicks the brooder houses were sampled the day after cleaning and disinfecting.
The floor (120m²) was split into 6 areas of equal size.
95 samples were collected per house.
Before the birds were transferred to the field the mobile barns were sampled immediately after cleaning, disinfecting and bedding-up.
The floor (80m²) was split into 4 areas of equal size.
90 samples were collected per house.

Laboratory Methods

The samples were transported at ambient temperature.
The wand swabs were transferred into 20ml buffered peptone water (BPW) and all samples were incubated overnight at 37ºC in order to increase the sensitivity of detection of antimicrobial-resistant E. coli.
10μl bacteriological loop of incubated BPW was plated onto each of three types of agar media:
Chromagar ECC.
Chromagar ECC incorporating 8mg/l ampicillin.
Chromagar ECC incorporating 16mg/l chloramphenicol.
The streaked plates were incubated at 37ºC for 18-24 hours.
Presence/absence of E. coli was recorded based upon morphological identification.
The growth of colonies other than typical E. coli was also noted.
Isolates were prepared for storage by sub-culturing a single colony per plate onto media of the same specification, incubating at 37ºC, inoculating glycerol-coated beads and freezing at –80ºC.

Statistical Methods

The trellis-plot bar charts were produced using R 2.1.1(4). Preliminary analysis and odds ratios were performed using STATA 8.2(5). Generalised linear mixed models were fitted in R 2.1.1 using penalised-quasi likelihood.

Results and Discussion

Descriptive Statistics

When interpreting these results it must be remembered that this study has simply sought to detect the presence of antimicrobial-resistant E. coli within cleaned houses. To increase the sensitivity of detection, the methods used involved an enrichment step. Thus it is plausible that a positive sample may result from the presence of a very low number of bacteria. Therefore these results are not necessarily direct indications of the levels of bacterial contamination present in cleaned houses. To ascertain the actual level of contamination would require a quantitative protocol and thus sampling of this scale could not have been carried out. The primary aims of this study were to look for persistence of resistant bacteria in broiler houses and to ascertain if there are problem areas of persistence within a house.

In the light of this, using the methods detailed above, E. coli and antimicrobial-resistant E. coli were detected in cleaned poultry houses on both farms (see Table 1a). Furthermore, a higher percentage of samples were positive for both ampicillin- and chloramphenicol-resistant E. coli on the organic farm, compared with the conventional.

The odds ratios for each phenotype of E. coli detected, comparing the conventional farm to the organic unit, are shown in Table 1b. These results are significantly less than 1, the odds ratio of “no-effect”, thus signifying that the lower level of detection of E. coli on the conventional farm was unlikely to be due to chance alone.

The most likely explanation of these results is that the cleaning and disinfecting regime utilised by the conventional farm significantly reduced post-cleaning contamination of the broiler sheds to a greater degree than the regime used on the organic farm.

There are two factors that differ between the two systems and the sampling undertaken.

Firstly, the conventional farm used 2-3 different disinfecting products within its cleaning regime, whilst the organic farm relied on a single product alone.
Secondly, on the conventional farm the samples were collected 4,5,6 and 10 days after cleaning. Whereas on the organic farm the samples were collected 2 or 3 days after cleaning in the brooder sheds, and immediately to 24 hours after cleaning in the mobile barns. This occurred because there was a narrower window of opportunity to sample the organic houses between cleaning and re-bedding and/or the arrival of the next group of birds.

It is worth noting, that after the organic brooder houses had been sampled, the farm staff placed fresh sawdust on the floors and then the houses stood for a further 6 days prior to the arrival of the birds. Therefore the levels of contamination shown here may have altered by the time the chicks arrived. However, with respect to the mobile barns, the morning after the barns had been sampled the birds were transferred from the brooder sheds. Therefore the results from the mobile barns are likely to reflect the situation encountered by those newly arrived birds.

Table 1a. Descriptive statistics for the detection of E. coli from environmental swab samples collected from empty, cleaned broiler houses.

Positive Samples (%) / Odds[b] / 95% Confidence Interval
Conventional Broiler
Non-Specific E. coli[c] / 35 / 0.53 / 0.45 - 0.62
Ampicillin-Resistant E. coli / 37 / 0.59 / 0.50 - 0.70
Chloramphenicol-Resistant E. coli / 20 / 0.26 / 0.20 - 0.34
Organic Broiler
Non-Specific E. coli / 80 / 4.04 / 3.38 - 4.83
Ampicillin-Resistant E. coli / 82 / 4.60 / 3.81 - 5.54
Chloramphenicol-Resistant E. coli / 46 / 0.85 / 0.74 - 0.98

Table 1b. Odds ratios for three phenotypes of E. coli detected in cleaned broiler sheds comparing a conventional broiler farm with an organic broiler farm.

Conventional versus Organic

/ Odds Ratio / 95% Confidence intervals / P-Value
Non-Specific E. coli / 0.13 / 0.10 – 0.17 / <0.0001
Ampicillin-Resistant E. coli / 0.13 / 0.10 – 0.17 / <0.0001
Chloramphenicol-Resistant E. coli / 0.30 / 0.22 – 0.42 / <0.0001

The other point of note from Table 1a is that the statistics for non-specific E. coli and ampicillin-resistant E. coli (AmpR E. coli) are very similar on both farms. This result refers to a similarity in resistance at the level of the sample not the isolate. That is to say, that this does not necessarily mean that that the majority of the E. coli population are resistant to ampicillin, but rather that if a sample contained an E. coli it was also likely to contain an AR E. coli. This can be illustrated using the results from the resistance phenotyping study of E. coli isolates collected from these two farms (see Annex 5). From these results it was seen that, for the conventional farm, 70% isolates grown upon plain Chromagar ECC were found to be resistant to ampicillin. This contrasts with the results from the organic farm where 48% of the plain Chromagar ECC isolates were ampicillin-resistant, and this does represent a significantly lower percentage ampicillin-resistant isolates (Χ2=6.68, d.f.=1, P≈0.01).

In fact, the number of samples positive for AmpR E. coli was actually slightly higher for every post-cleaning visit than the number of positive samples obtained using plain Chromagar ECC. This slightly unusual result is likely to be an artefact due to the methods utilised. On both farms atypical pink colonies grew on many of the plates, denoting the presence of other coliforms in the samples. Biochemical identification of small numbers of these atypical colonies found them to be species of Enterobacter, Citrobacter and Klebsiella. Thus the slightly lower percentage-positive plain E. coli samples may be due either to competition between E. coli and the other bacteria present in the sample on the surface of the plate, or may reflect that it is simply more difficult to ascertain the presence of E. coli when there is heavy overgrowth of other bacteria on the same plate. The incorporation of ampicillin into the media may therefore act to decrease the level of intra-plate competition and thus the growth of the ampicillin-resistant E. coli is less inhibited than the growth of general E. coli on the antimicrobial-free plates. However pink colonies were also commonly observed on the ampicillin-plates and, albeit less commonly, on the chloramphenicol plates, implying that a proportion of these other coliforms were also expressing resistance to these two drugs.

In terms of trying to assess whether the use of disinfectants is also applying selection for antimicrobial-resistant bacteria on farms. The results from the resistance-phenotyping study found that only 7% of the non-specific E. coli isolates recovered from the cleaned conventional houses were sensitive to all 17 of the antimicrobial drugs against which they were tested. This contrasts with 38% all-sensitive post-cleaning isolates from the organic farm. Once again this is a significant result (Χ2=16.63, d.f.=1, P<0.001). However this is may simply represent the nature of the E. coli shed by the birds themselves on these two farms rather than the result of selection pressures exerted by the disinfectants (see Annex 5). Further work comparing the bacteria present after mucking-out with those present after cleaning and disinfection are needed to investigate this hypothesis.

Annex 4: Final report for Defra project OD2006.

Figure 1. A trellis plot of bar charts depicting the levels of detection of three phenotypes of Escherichia coli persisting in broiler houses after depopulation and cleaning on two farms; one conventional broiler and one organic broiler.

Annex 4: Final report for Defra project OD2006.

Areas of contamination

Figure 1 illustrates the percentage of samples from which E. coli were detected from each of the different areas that were sampled across and around the houses on both farms. These trellis-plots show the combined totals of positive samples from all visits and all houses on each farm. In order to take the analysis further than this, generalised linear models were used to fit the binomial response variable of detection/non-detection to conditioning variables such as visit, house and origin of sample. Due to the inherent differences in management systems, separate models were fitted to the data from each farm. Furthermore, separate models were fitted for each phenotype of E. coli within each farm.

1.Conventional farm

Due to the large size of the houses on the conventional farm only two houses were sampled and therefore house was incorporated within the models as a fixed effect. The lack of random effects meant that simple generalised linear models incorporating a binomial response distribution and a logit-link were used.

Referring to Table 2 shows that the strongest influence on the proportion of positive samples for each type of E. coli was the conventional house from which the samples were derived. On both post-cleaning visits to this farm House 3 had been cleaned before House 6. On the first visit the houses had been cleaned 6 and 4 days prior to sampling respectively, and on the second 10 and 5 days. Furthermore, on both visits whilst the overall amount of organic material remaining in the cleaned houses was negligible, a subjectively higher amount of organic material was noted within House 6 compared with House 3 at the time of the visit. Therefore there are two possible components to this higher level of bacterial contamination in House 6, firstly the reduced length of time since cleaning and secondly the less rigorous cleaning. A third possibility was raised by the managers of the farm when discussing these results. They commented that an independent company was used to clean the sheds between flocks. The farm supplies the company with the amount of disinfectant needed to cover the entire site. However recent monitoring of the concentration of disinfectant being applied suggested that the earlier sheds were receiving higher concentrations than the later sheds. Thus House 6 being the last house to be cleaned on site may have had sub-standard concentrations of the disinfectants applied.