To our knowledge, this is the first report of influenza activity in Cambodia. It appears that there is a consistent pattern of influenza seasonality during these 3 years, with detection peaking during the rainy season from June to December. During the hot and dry seasons there was little to no detection of influenza, apart from an unusual peak during March 2007, a period where increased influenza activity was detected at all 5 sentinel sites. This overall pattern contrasts with that of Northern Vietnam, Thailand or Singapore in which influenza viruses circulate year-round and appears more similar to that of countries of southern hemisphere (Australia, New Zealand) [10–13]. The reasons for the seasonality of influenza are not clearly known and are probably the result of a combination of multiple factors as, for instance, climate conditions, epidemiology (population's age, density, migration, etc.), host susceptibility and virus characteristics . Others also reported that influenza seasonality coincides with rainy season [15–18] and that temperature and humidity play a more important role than population factors in driving seasonality of transmission [18, 19]. Importantly, the role of influenza transmission in schools (and subsequently to households) is probably very limited in Cambodia because they are closed from July to September .
This study provides data useful for accurate recommendations on influenza vaccination. People living in Cambodia would ideally receive vaccines by April - May in order to develop protective immunity before the peak season of transmission.
We were able to also identify severe influenza infections among hospitalized patients paralleling the same seasonality as observed within the NIC's ILI surveillance. To our surprise, few cases were detected among hospitalized patients with ALRI. We believe that this low frequency of cases may not be a reflection of the extent of influenza-related severity, rather that it is explained by the delay in hospital admission of Cambodian patients presenting with severe respiratory infections. Indeed, most bacterial lung infections included in our hospital study were diagnosed among patients who attended hospital too late after the onset of respiratory symptoms for viral detection (average, 6 days; IPC's unpublished data). It is plausible that many of these bacterial infections were secondary infections from an initial influenza disease. Further testing for influenza using serology should be performed among patients with bacterial respiratory infections during the seasonal epidemics to establish this hypothesis. In addition, the average age of the patients presenting to hospital with ALRI was significantly higher (p < 0.001) than patients with ILI symptoms since influenza is mainly found in children, it may explain the rate differences between both studies. In 2006, few influenza viruses were detected, as the NIC and its sentinel sites were just established in August 2006. In this year, all influenza-positive isolates belonged to influenza A/H3N2 sub-type which is consistent with influenza activity reported worldwide by WHO. Indeed, in 2006, a low influenza activity was observed and influenza A/H3N2 virus was the predominant subtype in many European and Asian countries . In 2007, both influenza A subtypes as well as influenza B viruses co-circulated at comparable level. Among others, outbreaks of influenza B virus were also reported in 2007 in China, Hungary and in the United States . In 2008, influenza A and B viruses were co-circulating as well but influenza A viruses represented two thirds of the strains detected. Similar general circulation patterns were observed worldwide in 2008 and, as in Cambodia, H3N2 subtype was responsible for most outbreaks in some European countries or in Japan .
Since the establishment of the NIC and the ILI surveillance system, we observed a regular increase in the influenza positivity rates in the samples that were tested. This could be linked to a higher influenza activity but also to an improvement of the system from patient recruitment (with adjustment of the case definition in 2008 for the fever criteria), to sample collection and transport to testing. Significant differences in the influenza positivity rate were observed between the sentinel sites. It was expected that paediatric hospitals will have a higher positive rate than adult hospitals, but interestingly, comparing the two paediatric hospitals, the one based in Phnom Penh produced more influenza-positive samples than the hospital located in Siem Reap, perhaps emphasizing the important role of the physician in the successful recruitment of patients who are likely to have influenza infection.
There are some limitations to the sentinel surveillance data. Sentinel data cannot be extrapolated precisely to the rest of the population, as the outpatients clinic serving sentinel sites were not truly representative. Random selection of patients is not yet well standardized and depends on the goodwill of physicians in participating in influenza surveillance. In addition, ILI consultation rates were not provided, and data is lacking on the catchment areas of the sentinel sites. As a result, little inferences can be made on disease burden or severity in Cambodia in general. Despite these problems, the system has been useful in meeting the purposes of influenza surveillance, identifying the predominant circulating strains in the community and allowing the formulation of guidelines for influenza vaccine composition for the subsequent year.
Since 2005, a total of eight human cases of influenza A/H5N1 have been identified in Cambodia . Interestingly, H5N1 human cases (e.g. in March 2007 and December 2008) were detected while there was evidence of background human influenza activities. As in other Asian countries, there is some evidence of endemicity of avian influenza H5N1 circulation in poultry in Cambodia . There is little evidence for human-to-human transmission of the H5N1 virus. Additional mutations or re-assortment events are probably required to permit efficient human-to-human spread which could trigger a pandemic [26, 27]. Thus, if H5N1 virus infection in humans occurs during these seasonal epidemics, the risk of re-assortment increases.
Even though there is little evidence of mild or asymptomatic H5N1 human infections, the increasing incidence of cases with influenza-like symptoms have been reported more frequently since 2005 . In most countries, H5N1 testing is usually limited to symptomatic cases with a recent history of exposure to dead or sick poultry. Thus, if the H5N1-infected patients deny contact with the usual animal source of contamination, hospitalization and testing would be unlikely and the symptoms could be mistaken for seasonal flu. Others may not seek medical attention, assuming that the H5N1 infection is mild and does not require medical attention, leading to exposure of others. Hence, H5N1 virus detection by NICs should be systematic.