Avian Influenza Virus Detection Rates in Poultry and Environment at Live Poultry Markets, Guangdong, China

We report the use of environmental samples to assess avian influenza virus activity in chickens at live poultry markets in China. Results of environmental and chicken samples correlate moderately well. However, collection of multiple environmental samples from holding, processing, and selling areas is recommended to detect viruses expected to have low prevalence.

3.1% for H7, and 34.1% for H9; all p<0.02 by Mann-Whitney test) were much higher at the retail LPM (rLPM) than that at the wLPM (Figure 1), probably because of poultry mixing, aggregation, and extended stay at retail settings. Human H5 or H7 zoonotic infections clustered in winter, but we observed no correlation (p>0.215 for both) between temperature (14) and H5 or H7 detection rates in chickens or environmental samples at both markets. We did not assess other confounding factors, including market interventions and poultry holding duration.
We evaluated correlations between monthly AIV detection rates in chickens and environmental samples (moderate correlation for r s >0.5, at which point environmental samples are considered useful to monitor AIV in chickens). We observed a positive correlation for H5 (Spearman r s = 0.569, p<0.001) and H9 viruses (r s = 0.702, p<0.001) at the wLPM and for H5 (r s = 0.581, p<0.001), H7 (r s = 0.760, p<0.001), and H9 viruses (r s = 0.685, p<0.001) at the rLPM. We examined the use of environmental samples to assess AIV activity in poultry (Table 1). Environmental  †Sensitivity: probability that the environmental samples will test positive when the subtype of AIV is present in chickens on site (true positive rate). Specificity: probability that the environmental samples will test negative when the subtype of AIV is not present (true negative rate). Positive predictive value: probability that the subtype of AIV is present in poultry when environmental samples are tested positive. Negative predictive value: probability that the subtype of AIV is not present in poultry when the environmental samples are tested negative. Accuracy: probability that the presence or absence of AIV in poultry will be correctly determined based on the test results of environmental samples. ‡H9 was detected during every month during the study period in the environmental samples (monthly data can be found in Appendix Table 1, https://wwwnc.cdc.gov/EID/article/26/3/19-0888-App1.pdf). §H9 was detected during every month during the study period in both the poultry and the environmental samples (monthly data can be found in Appendix Table 2).
samples collected at the rLPM provided higher sensitivity in detecting H5, H7, or H9 viruses in poultry than those from the wLPM. Environmental samples were less likely to detect H5 and H7 viruses in poultry at the wLPM than at the rLPM (Appendix Tables  1, 2), possibly because of the low prevalence of infection in birds, a higher poultry turnover rate, and comparatively thorough daily cleaning practices at the wLPM. The lower specificity for H5 at the rLPM may be the result of carryover contamination at the poultry slaughtering area caused by processing birds of other species. The probabilities of accurately detecting the presence or absence of H5, H7, and H9 subtypes in poultry from environmental samples were comparable for the wLPM (68.8%-93.8%) and the rLPM (59.4%-100%) ( Table 1). This finding suggests that environmental samples provided a useful indication of AIV activity in chickens at LPMs. Nevertheless, for H5 and H7 viruses at the wLPM, in only 1 month did all environment samples test positive when bird samples were also positive, demonstrating the need to take a wide range of environment samples.
We investigated correlations between specific environmental samples and monthly H5, H7, and H9 detection rates in chickens (Figure 2 p = 0.003) in fecal droppings correlated moderately well with viral prevalence in chickens, whereas drinking water provided the best correlation for H5 (r s = 0.633, p<0.001) and H9 (r s = 0.702, p<0.001) ( Figure 2) and was more sensitive for H9 (Appendix Figure 3). At the rLPM, H9 detection rates in drinking water (r s = 0.593, p<0.001), poultry feed (r s = 0.550, p = 0.002), and fecal droppings (r s = 0.506, p = 0.003) best correlated with H9 prevalence in chickens; drinking water was most sensitive (Appendix Figure 3). H7 detection rates in drinking water (r s = 0.784, p<0.001), fecal droppings (r s = 0.663, p<0.001), defeathering machines (r s = 0.634, p<0.001), and air (r s = 0.585, p<0.001) best correlated with H7 prevalence in chickens. The H5 detection rates in fecal droppings (r s = 0.729, p<0.001), defeathering machines (r s = 0.556, p<0.001), and poultry feed (r s = 0.550, p = 0.02) best correlated with H5 prevalence in chickens. Collectively, these results suggest that fecal droppings may provide a good estimation for H5, H7, and H9 prevalence in chickens at LPMs and that drinking water can be more sensitive in some settings and useful for determining virus contamination in LPMs. For viruses present at low prevalence (e.g., H5), low sensitivity is expected.

Conclusions
AIV detection rates in environmental samples correlated moderately with AIV activity in chickens at LPMs. Environmental sampling at rLPMs provides greater sensitivity in detecting H5, H7, and H9 AIV in poultry than that at the wLPMs and should be included as routine surveillance to monitor AIV activity. At the rLPM, H5 and H7 viruses were most frequently detected from poultry selling and poultry slaughtering areas, whereas the highly prevalent H9 viruses were detected frequently at poultry holding, slaughtering, and selling areas. Environmental samples with the highest detection rate for H5, H7, and H9 viruses may not provide the best indication of virus activity in poultry, however. Some market stalls containing viruses with low prevalence would be misclassified if only environmental or bird samples were collected. To detect viruses expected to be present at low prevalence, environmental samples should be collected from multiple sites in each market stall, including samples from holding, processing, and selling areas.  ‡Environmental swab specimens were collected within the same poultry stall at LPMs but may not be from the same cage where the chickens were sampled. Fecal droppings were collected from the ground or cages, drinking water was collected from the water troughs, and poultry feed was sampled from the surface of the bowls or feeders. §A positive air sample may be positive for AIV by qRT-PCR in any of the 3 size fractions collected by a NIOSH sampler (11). Two to 6 NIOSH samplers were applied monthly to sample air at the retail markets.