When the live poultry market surveillance program began in September 1999, eight of a total of 80 live poultry markets were selected to represent the 3 major regions of Hong Kong: Hong Kong island, Kowloon, and the New Territories. Since then, the number of markets has declined, and by May 2011, only 5 of the 8 selected live poultry markets continued in operation (of a total of 39 operating live poultry markets). A total of 53,541 samples were collected during these 141 months of consecutive sampling.

We previously published data on the effect of introducing various interventions in live poultry markets, which included the ban on the sales of live quail and the introduction of rest days (7). In addition to collecting fecal droppings from the cage floors for virus isolation, we collected data on the total sales of chickens and minor poultry, the proportion of chickens imported as a ratio of the whole, the temperature and relative humidity, and the type of ventilation used, as described (7). Laboratory processing of the specimens was conducted as described (7). Samples collected in virus transport medium were inoculated into 9–11-day-old embryonated eggs, and allantoic fluid with positive hemagglutination was confirmed and subtyped using standard antiserum.

Because HPAI A virus (H5N1) is rarely detected in live poultry markets in Hong Kong, we used isolation rates of influenza A virus (H9N2) as an indicator of the effect of these interventions on avian influenza virus circulation. The median numbers of samples collected weekly from chickens and minor poultry were 107 (range 3–722) and 23 (range 1–397), respectively (see Technical Appendix for weekly numbers of samples). The Poisson generalized model (8) with influenza virus (H9N2) weekly isolation counts as the outcome variable was fitted as described (7) and adjusted for proportion of chickens imported; total sales of chickens and minor poultry; ventilation system; weekly average temperature; relative humidity; seasonal variations; sample size; and periods corresponding to the respective interventions: period I (no rest day), II (1 monthly rest day with quail being sold in the live poultry market), III (1 monthly rest day with elimination of live quail from the live poultry market), IV (2 monthly rest days), and V (ban on holding poultry overnight in live poultry market). These variables were considered potentially important confounders related to transmission efficiency of avian influenza virus (9), source, type, and volume of poultry.

Weekly virus isolation counts were analyzed from September 22, 1999, through May 31, 2011. A separate model for poultry and minor poultry was fitted, and all analyses were implemented by using R version 2.12.1 software (R Development Core Team, Vienna, Austria).

Figure

Figure. . . Weekly influenza virus A (H9N2) isolation rates for chickens (A) and minor poultry (B) in live poultry markets, Hong Kong, September 1999–May 2011. Vertical lines denote periods for different...

The Figure shows overall isolation rates by week for chicken and minor poultry from 1999 through 2011; the Table gives the parameter estimates for the final fitted models, which were adjusted for the effect of covariables that could affect the isolation of influenza in the study. For chickens and minor poultry, compared with the reference category of 2 monthly rest days, the ban on keeping live poultry overnight in live poultry markets was associated with dramatic and significant reduction of influenza virus (H9N2) isolation. The isolation rate of influenza virus (H9N2) among chickens declined 84% (adjusted relative risk 0.16; p = 0.006), and no influenza subtype H9N2 viruses were isolated from minor poultry after the ban on holding poultry overnight in live poultry markets was implemented. Higher volume of minor poultry sales was also significantly associated with higher isolation rate of influenza virus (H9N2).

A previous study that used a stochastic metapopulation model showed that frequent rest days in live poultry markets were effective for reducing transmission of avian influenza (H5N1) (10). Our findings show a large additional decline in the influenza virus (H9N2) isolation rate after implementation of a ban on keeping live poultry overnight, which suggests that this intervention has an even greater effect on reducing viral load in live poultry markets than the previous intervention of 1 or 2 rest days per month. While low pathogenic influenza virus (H9N2) was the indicator virus in our study, it is likely that these interventions would have comparable effects on highly pathogenic viruses such as avian influenza (H5N1); this effect has been demonstrated by mathematical modeling (10). Studies by others on social network analysis have shown that daily cage cleaning and disinfection of live poultry markets in southern China (11), and protective factors including removal of waste in Indonesia (12) contributed to a reduction of HPAI (H5N1) in live poultry markets. Taken together, these studies show that eliminating the carryover of live poultry in markets from one day to the next, in the form of rest days or a total ban, is highly effective for reducing viral amplification and persistence in live poultry markets and consequently minimizes zoonotic risk.