Risk factors and spatial distribution of extended spectrum β-lactamase-producing- Escherichia coli at retail poultry meat markets in Malaysia: a cross-sectional study

Selangor is one of the 13 states of Malaysia; with a total area of 8,104 km², the Malaysia’s most populous state, also known as the industrial hub of Malaysia, with the largest economy in terms of gross domestic product (GDP), providing almost 23 % of Malaysian GDP; it is located at centre of Peninsular Malaysia [13‐15]. Almost 90 % of Malaysian poultry production is in peninsular Malaysia (while the remaining is in East Malaysia), which was reported to be among the world highest per capita consumption of poultry meat at 40 kg per year [16]. Poultry meat is the most stable protein source across all ethnic groups of the Malaysian population [16]. According to the USDA recent International Egg and Poultry review [17], Malaysia is self-sufficient in poultry meat production, which was almost entirely of broiler meat, with production forecast at 1.44 million tons; about 40 % are marketed to the consumers through wet-markets [17].

A total of 240 samples comprising 160 broiler chicken meat samples (breast, wing, thigh, and keel) and 80 swab samples from contact surfaces (weighing scales and cutting boards/instrument) were examined to detect the presence of ESBL-EC. The estimated sample size was calculated using G*Power 3.1.9.2, using A priori F-tests ANOVA: Fixed effects, omnibus, and one-way; given α = 0.05, Power = 0.80, Effect size = 0.25, Number of groups = 8. Informed consents were obtained from each participant prior to sample collection and questionnaire administration. The participants were informed about the purpose of the research, likewise they were assured of protecting their identities and received information would be highly confidential, as it was only used for research purpose.

A total number of 40 individual stalls were selected at random within eight districts in Selangor (Table 2). Three wet-markets were selected in every district, out of which five different stalls were recruited to represent a district area. At each stall, six representative samples were obtained, which consisted of four broiler chicken meats (breast, keel, wing and thigh) and two swabs (weighing scale and cutting board/instrument). The meat samples were bagged separately in a sterile plastic bag. The surfaces of the cutting board/instrument and weighing scale covering an area of 25 cm² were swabbed using pre-moistened sterile cotton swab, [18], and each swab was placed in sterile transport media containing 9 ml of peptone water (Oxoid). All samples were transported to the laboratory in ice cool box and processed within five hours of collection.

Each meat sample was processed by taking 25 g of the meat sample, vigorously shaken in a sterile plastic bag containing 225 ml of peptone water and homogenized in a stomacher for 1 min [19]. Five ml of each homogenized sample was subsequently placed in 45 ml of peptone water and vortexed gently to form a mixture of processed meat sample. Each of the bottles containing the swab was vortexed gently; then, 1 ml was taken and diluted in 9 ml of peptone water and was further vortexed gently.

A few loopfulls of each processed sample was directly inoculated on Chromocult Coliform agar (Merck, UK) and incubated at 37 °C for 24 h. The agar served as differential and selective media, it distinguished E. coli from other coliform colonies and inhibited gram-positive bacteria growth, which was further confirmed by using Kovac’s reagent (the blue-violet E. coli colony becomes red) as indicated by the manufacturer.

The non-duplicate representative of E. coli isolates were subcultured on CHROMagar ESBL (CHROMagar, Paris, France) to screen for the presence of ESBLs [20]. The plates were incubated aerobically overnight at 37 °C. The presumptively identified ESBL-EC colonies were selected (dark pink to reddish colonies) and further subjected to phenotypic confirmation.

The phenotypic confirmation was conducted by using the Combination Disk Method, as described by CLSI [21]. The bacterial suspension (turbidity of 0.5 McFaland standard) was spread on Mueller-Hinton agar (oxoid) impregnated with cefotaxime (CTX-30 μg), ceftazidime (CAZ-30 μg), cefotaxime/clavulanic acid (CTX/CV), and ceftazidime/clavulanic acid (CAZ/CV) disks, and incubated at 37 °C for 16 to 18 h. The interpretation was based on the zones of inhibition produced by cefotaxime (CTX-30 μg) and ceftazidime (CAZ-30 μg) disks against cefotaxime/clavulanic acid (CTX/CV), and ceftazidime/clavulanic acid (CAZ/CV) respectively [21]. A difference of ≥5 mm between the zones of CTX and CTX/CV or CAZ and CAZ/CV was considered as phenotypically confirmed ESBLs. Klebsiella pneumoniae ATCC 700603 and Escherichia coli ATCC 25922 were used as quality control strains [21].

The DNA of phenotypically confirmed ESBL-EC was isolated by using the boiling method, on 2–3 colonies grown on nutrient agar. The colonies were suspended in 100 μl of sterile water, and the suspensions were boiled for 15 min, cooled to 4 °C, and then, subsequently centrifuged for 30 s at 12 000 × g. The supernatant served as a source of DNA template for the PCR, while the specific oligonucleotide primers for bla _TEM, bla _SHV, bla _CTX-M, bla _OXA, and E. coli were used for amplifications of these bla-genes and E. coli. Furthermore, MyTaq Protocol (BIOLINE) was used as standard for the PCR reactions and amplification. The sizes of bla _CTX-M, bla _TEM, bla _SHV, bla _OXA, and E. coli amplicons are shown in Table 1, using K. pneumoniae ATCC 700603 and EC1003-1 as positive controls. The amplified products were separated by gel electrophoresis on 1.5 % agarose gels stained with ethidium bromide.

A questionnaire was developed as an instrument to gather information and data for the risk factors analysis. The items were derived from literature search, interviews, and focus group discussions with local veterinarians and representative of butchers. The questionnaire was further evaluated by the research team for face validation. In order to evaluate the adequacy of the instrument content, the questionnaire was further assessed by five experts in the field for content validity. On top of that, the instrument was refined and pilot tested. Unclear items were finally dropped, while items with overlapping meaning were merged together as single items.

Data were analysed by using SPSS version 20. chi-square test was used to compare the prevalence; whereas the relationship between potential risk factors and detection of ESBL-EC was explored by using univariate logistic regression at 95 % confidence interval (95 % CI). However, only factors that were found to be significantly associated with detection of ESBL-EC had been included in the multivariate logistic regression model for estimation of their odds ratio at 95 % CI.

A total of 240 samples were collected in wet-markets within eight districts of Selangor from July −2012 to February 2013. The overall prevalence of ESBL-EC was 48.8 % (95%CI, 42 – 55 %), although the difference in prevalence was not statistically significant between the districts (χ2 = 6.921, df = 7, p = .437) with chi-square test at 95 % confidence interval (95 % CI). Stalls in the district of Hulu Selangor were shown to have the highest prevalence at 66.7 %. while moderate prevalence was observed in Hulu Langat at 56.7 % and Kuala Selangor at 50 %. Nevertheless, the lowest prevalence was found in Klang at 46.7 %, Sepang at 46.7 %, Petaling at 43.3 %, Gombak at 40 %, and Kuala Langat at 40 % (Fig. 1; Table 2).

However, when the occurrence of ESBL-EC in meat was compared to that of the contact surfaces, the meat samples displayed higher occurrence rate of 53.8 % compared to contact surfaces at 38.8 % (Fig. 2; Table 3).

Furthermore, more than 50 % of the meats sampled were contaminated with ESBL-EC; the highest occurrence was in breast, wing, and thigh with proportions of 65, 52, and 50 % respectively. Keel and weighing scale had moderate rates, while cutting board/instrument had rather lower rates at 47.5, 40, and 37.5 % respectively (Table 4).

Univariable logistic regression model indicated four factors that had the likelihood of causing ESBL-EC contamination, at p-value less than 0.05, which included stall sanitation, type of counter top, source of cleaning water, and type of cutting board/instrument (Table 5).

However, based on the multivariable logistic regression model, only two factors significantly predicted ESBL-EC contamination, which were stall sanitation, and type of countertop (Table 5).