Avian influenza virus (AIV) can cause severe outbreaks in the poultry population. Nevertheless, it may occasionally infect humans exposed to infected poultry. The term outbreak refers to a number of cases of a specific disease in excess of normal endemicity. However, an outbreak is not always defined by a specific number of cases. The relative occurrence varies upon the infectious agent, the composition of the population exposed and prior exposure or lack of exposure of the population to the certain infectious agent (immune system status). Moreover, the time and the place of occurrence also play a significant role in identifying an outbreak. Therefore, an outbreak refers to a specific population (community), at a specific time point (season) and in a specific place (geographic area) [1–4]. On the other hand, a case is considered as the occurrence of a single individual (i.e. wild bird) with a disease, and its detection is of great importance as an outbreak typically depends on the detection of individual cases [5].

Avian influenza is an A-type influenza virus and belongs to the Orthomyxoviridae family. Type-A influenza theoretically contains thousands of different antigenic subtypes, due to the combination between the main virion antigens, haemagglutinin (HA) and neuraminidase (NA) [6]. Until recently, 16 HA subtypes and 9 NA subtypes have been recognised, while 2 additional HA and NA subtypes have been identified in bats [7, 8]. The term “highly pathogenic avian influenza” (HPAI) generally refers to the strains that may induce “intravenous pathogenicity index” (IVPI) greater than 1.2 or mortality rate over 75% in a defined chicken population during the specified interval of 10 days. Using this definition, all the HPAI strains isolated to date are of H5 and H7 subtype. However, viruses of these subtypes can also be of low pathogenicity.

According to the World Organisation for Animal Health (OIE), AIV is defined as “an infection of poultry caused by any influenza A virus with high pathogenicity (HPAI) and by H5 and H7 subtypes with low pathogenicity (H5/H7 LPAI)” [9]. Moreover, OIE requires notification for all H7 and H5 subtypes, regardless of their pathogenicity, as they have the potential to mutate into HPAI viruses [10]. In other words, non-H5 and non-H7 low pathogenic avian influenza (LPAI) are not determined as notifiable.

The symptoms of AIV infection vary widely, depending on the infected species, the age, the sex, the strain, the subtype involved, concurrent infections and, of course, many predictable and non-predictable environmental factors. The clinical manifestations vary from mild to severe respiratory, nervous, reproductive and gastrointestinal system disorders. There are also cases of no clinical signs and sudden deaths [11].

Generally, the economic consequences of AIV in poultry are severe, since not only the production of eggs is affected, but also the quality of the egg influenced. In addition, a high mortality in birds is observed, as well. Moreover, given that AIV determines, to a certain extent via genetic reassortment [12], the development of advanced human strains, AIV is a major subject to be taken into account as far as public health is concerned [13].

The year 2010 was the fifth year when measures and surveillance programmes were imposed by proper directives, and legislation was enforced by all proper authorities. Therefore, 2010 is considered as the time when all procedures were routinely established. Comprehensive epidemiological analysis of global avian cases of AIV is scarce, and a few studies have presented in detail the changing epidemiology of AIV cases [14–16]. To improve the understanding of AIV epidemiology, we conducted a systematic review of the AIV outbreaks to describe the distribution and the magnitude of all avian cases which occurred globally during the period 2010–2016.

An a priori protocol was performed, based on the PRISMA statement [17], to specify the search strategy, the eligibility criteria, the objectives and the methods of this systematic review. However, this systematic review was not registered with PROSPERO (International prospective register of systematic reviews).

In this systematic review, a global dataset, spanning 7 years, was systematically collated to investigate the epidemiological profile of AIV and its’ subtypes. Our findings suggest that AIV still exists and expands. H5N1 remains the most dominant AIV subtype. Moreover, regarding the geographical extent of outbreaks, Asia is the prevalent continent. China, Viet Nam, India, Taiwan, Israel, Japan and South Korea have reported some of the largest outbreaks during 2010–2016. This can be attributed to the enhanced laboratory and clinically programmes implemented in all these countries over the last years or to the fact that these countries dispose a unique ecosystem with numerous lakes, ponds, creeks and rivers, constituting wintering areas for migratory birds. Also, Africa (Egypt, Nigeria, Ghana, Cameroon, Togo, South Africa, Tunisia, Cote d’Ivoire, Burkina Faso, Libya) reported a great proportion of cases. The expand of the virus in Europe, especially the spread of HPAI H5N8 in Germany, strengthens the close relation between the spread of the virus via wild bird migration and their habitats. Our results, combined with the data from all the past 15 years, show that influenza activity may change from year to year and season to season. Although the drivers/reasons are not fully understood, a correlation with certain climatic conditions is proposed [27, 28]. Our results, also, demonstrate a difference in reported cases based on serologically tested samples when compared with those confirmed by molecular testing. This difference seems to be expected, as it is due to cases serologically identified, but not confirmed and characterised molecularly. This could be both attributed to possible differences in validity (sensitivity and specificity) of all methods of testing and to the fact that none of the diagnostic tests are validated for all species or specimen types [29].

Concerning the age factor, we should consider that in commercial species, in particular broilers, the rearing period is highly standardised worldwide. In some cases, age was specified as a week before reaching slaughter or at week 4 of the rearing period. In all of these cases, we considered that information on “age” was not given. Also, there were outbreaks referring to a number of flocks within a region, where it was likely that several ages were involved, i.e. one outbreak involving 20 farms in the same region each with birds of a certain age. This should be also taken into account regarding the validity of AIV distribution by age.

Overall, our evidence robustly reports that AIV exists and is disseminated worldwide. Our systematic review has several limitations that could be divided into two separate categories: firstly, there are limitations of the methodology of the review per se and secondly, limitations related to the variable quality of the source-articles. The identification of numerous references obtained through the lists of the studies distinguished by search engines, i.e. 8% of all citations included, indicates that the search strategies and the eligibility criteria sets may have been very specific and restrictive. Concerning the source limitations, the population of the avian species, the surveillance and diagnostic procedures and the recording system applied in each country, as well as the state of the country, are not the same among the studies evaluated. Analyses and data referring to cases that reported the application of vaccination programme could not be confirmed, as flock histories (percentage of vaccination, doses of vaccination applied, etc.) are often not available [30]. For example, vaccine composition in Viet Nam generates an immune response that cannot be distinguished from natural infection [31]. It has been proven to be difficult to achieve the ideal balance between sensitivity and specificity, as time, keywords and resource constraints are set in this paper. Publication bias might account for this observed effect. The keywords set in our search strategy were “avian influenza” and “avian influenza virus” and “avian flu” and “outbreak” and concerned the article title, abstract or keywords. Thus, papers missing to mention the above key terms in their titles, abstracts or keywords, albeit reporting them in their full text, may have also been missed. Another point that had to be clarified was whether AIV cases were related to human or avian species. A great number of studies which were identified were referring to human infections of AIV. We, therefore, believe and suggest that future reviews should follow a more comprehensive search strategy and approach. Furthermore, taking into account that none, except for 13 of the included studies, scored the maximum on the quality assessment; there is a concern for the methodological quality of this systematic review and the risk of bias of the included studies. This may be attributed to various factors, such as sample heterogeneity, variety of statistical analyses followed and others.

The cases identified in this systematic review are likely to be underestimates of the real incidence of AIV, due to the limiting timeline of 2010–2016 set by us and difficulties of precisely scrutinising and including data referring to years 2010–2016. There were outbreaks reporting massive numbers of cases between 2007 and 2012, which did not separate cases per year and were rejected since they did not allow us to “isolate” the number of cases concerning the timeline set by us (2010–2016). Moreover, it has been proven very difficult to make comparisons among the outbreaks included, as there were vast differentiations in avian influenza subtypes, environmental and geospatial conditions and key characteristics of affected populations. Wherever mentioned, there was, also, a solid variation in case definitions among the outbreaks identified in this systematic review, e.g. an outbreak within a flock in which a number of infected birds is reported, an outbreak in a region involving flocks and with no information about number of cases in each flock (i.e. the unit is the flock), an outbreak that in fact is not an outbreak, but an endemic situation that just happens to be reported. The issue of the definition of the outbreak, namely the level at which the outbreak is defined and the extent to which it is reported as an outbreak, should be considered carefully. Hence, the calculation of the attack rate is likely to vary by study due to methodological differences.

This review, also, suffers from a lack of details provided by the primary papers. For example, it was very difficult to assess the degree of association between the vicinity of water mentioned and AIV cases, because of the limited amount of relevant information that was available, particularly in the ProMED mail reports.

Deficiencies emerging from our review draw attention to key areas that future outbreak reports should try to address. Future studies in this area should aim to record the presence of wild birds that are thought to be involved, while effort should, also, focus on possible routes of transmission. This could allow a greater number of studies to be assessed and included in reviews like this one. Generally, anything that is considered to be implicated (weather conditions, water collections, etc.) should be clearly stated. The case definition should be clearly remarked, as well. It is of great importance to highlight the contingent role of such factors and to encourage researchers to explicitly investigate whether such factors occurred prior to the outbreak or after the outbreak.

The global distribution of avian influenza outbreaks or cases as mentioned in the scientific literature is very likely to be subject to considerable publication bias. A greater proportion of these outbreaks identified through ProMED mail was in Asia compared to those reported in peer-reviewed journals. Outbreaks in countries that apply preventive measures and follow surveillance and vaccination programmes are often thought to be easily controlled [32]. The scientific literature is, also, likely to originate from countries with greater academic, financial and surveillance capacity. Therefore, the number and type of studies published may not be proportionate to the extent and consequences of an outbreak. ProMED mail is a passive reporting tool, and this must be taken into account, as discrepancies are really possible to appear. For instance, outbreaks which occurred in countries of international interest, with potential public health risks (i.e. China) and minor commercial effects, were more likely to be reported [33].

The scientific literature confirms that all haemagglutinin and neuraminidase influenza A subtypes of possible combinations have been identified from avian species and affect all type of domestic or captive birds, worldwide [34]. Nevertheless, our results demonstrate a difference in H5 and H7 rates, which are regarded as common virulent subtypes that are characterised as highly pathogenic avian influenza viruses (HPAI) [34]. These subtypes, particularly H5N1, H5N8, H5, H5N6, mixed, H5N2 and H7N9, were also found to be more dominant in the wild bird cases of this systematic review. Moreover, our findings agree with those of Bui et al., according to whom H7N9 is not highly pathogenic in wild birds, whereas we also concluded that H5N1 cases remain consistent worldwide, via wild bird migration and poultry trade activities [35]. Bui et al. found that H7N9 in wild birds occurred largely in a great number of contiguous provinces of China, and our results support this evidence, as all H7N9 cases of our systematic review took place in China [35]. Bui et al. also assume that species with affinity to water collections and coastal places are considered to be primary reservoirs of AIVs. Cases of H5N1 affected a greater area of China compared to those of H7N9 infection [35]. Van Kerkhove et al. recognise endemicity of H5N1 subtype in several parts of Asia and Egypt [36]. Wan et al. concluded that, indeed, China is an AIV epicentre, as minimum two large pandemics (1957 and 1968) originated from there, spreading across Asia, Africa and Europe [37].

This systematic review suggests that AIV outbreaks do occur in both developed and developing countries, and this, per se, constitutes an important burden on public health. Further epidemiological studies may contribute to identify possible risk factors and understand the extent and routes of the spread of AIV. In terms of consistent reporting, proper case definitions (farm, flock, bird, etc.) should also be designated when outbreaks are reported. Systematic surveillance and prevention programmes shall continue to be enhanced in “risky” countries (e.g. China, Viet Nam), while auditing should be imposed in wet and live bird markets in order to ensure safe and appropriate animal practices.