Influenza in Tropical Regions

Viboud and colleagues discuss the implications of a new study by Peiris et al. that assesses the burden of influenza in Hong Kong.


Perspectives
April 2006 | Volume 3 | Issue 4 | e89 T he global burden of infl uenza on morbidity and mortality is considerable, with an estimated 1 million annual deaths worldwide [1]. In temperate regions, there are clear seasonal variations in the occurrence of infl uenza, with a marked peak in cold winter months [2]. In contrast, seasonality is less defi ned in tropical regions-here, there is high background infl uenza activity throughout the year, on top of which epidemics occur at intermediate months between the infl uenza season in temperate countries of the Northern and Southern hemispheres. Infl uenza seasonal patterns in the Americas are summarized in Figure 1, which illustrates the transition between the Northern and Southern hemispheres. Very little is known about exactly how and why the seasonality of infl uenza varies with latitude. No link with environmental conditions has been convincingly shown, although infl uenza activity appears to coincide with the rainy season in some tropical countries [3].
Until very recently, the impact of infl uenza on health was believed to be insignifi cant in tropical regions. But this impression may be changing, in part thanks to the recent studies by Joseph S. M. Peiris and his colleagues from Hong Kong, one of which appears in this issue of PLoS Medicine [4]. Hong Kong is a subtropical city located within the likely epicenter of pandemic infl uenza in Southeast Asia [5].

Methods for Estimating the Burden of Infl uenza
Estimating the burden of infl uenza on hospitalizations and deaths is not straightforward in temperate countries, and it is even less so in tropical countries [2]. Severe clinical outcomes are often caused by secondary bacterial infections, and a primary infl uenza infection may be unrecognized. Additionally, laboratory confi rmation of infl uenza infection is rarely conducted. As a result, most infl uenza-related hospitalizations and deaths are not attributed to infl uenza on discharge forms and death certifi cates, and statistical methods are necessary to assess the true burden of infl uenza on health.
Traditionally, for temperate countries, the impact of a particular infl uenza epidemic is estimated as the number of "excess deaths" or "excess cases" above a seasonal baseline. This approach relies on identifying periods without infl uenza activity to generate a model baseline of expected winter deaths (or cases) [6][7][8]. But the approach fails in tropical regions, where there is no well-defi ned non-infl uenza period ( Figure 1). An alternative method recently developed by the United States Centers for Disease Control uses virus surveillance data to guide models of burden, assuming a proportional relation between virus activity and the number of hospitalization or deaths attributable to infl uenza [9,10]. For the US, both traditional and new methods produced very similar estimates of infl uenza mortality burden (Table  1 and [11,12]). In their Hong Kong study, Peiris et al. appropriately used a variation of the virus-guided modeling approach by linking weekly hospital discharges with weekly prevalence of infl uenza virus in respiratory specimens sampled in the community [4].

Infl uenza Burden is Broadly Similar in Hong Kong and the US
Peiris et al. found that the annual infl uenza-related hospitalization rates are quite similar in Hong Kong and temperate countries (Table 1), confi rming their earlier studies [13,14]. In Hong Kong, as in many other countries, the impact of infl uenza is seen not only on pneumonia and infl uenza outcomes [12,15], but also on a wider range of chronic health conditions such as diabetes and cardiovascular diseases [12,15]. Peiris et al. also report that infl uenza-related hospitalization rates vary with age as a U-shaped curve, where young infants and elderly are at highest risk of severe disease outcome, reminiscent of the age pattern of interpandemic infl uenza in the US and other temperate countries. However, one difference between Hong Kong and US children stands out: these authors had shown in a previous hospital cohort study that young children in Hong Kong had a hospitalization rate that was roughly 10-fold higher than in the US (Table 1 and [16]). The reasons for this substantial discrepancy are elusive, but may in part be attributed to the notorious diffi culty of assessing with accuracy the impact of infl uenza on children. There is a particularly high background rate of acute respiratory illnesses caused by other respiratory pathogens in this age group. Unfortunately, the authors did not reexamine this apparent discrepancy in the new, larger citywide study published this month [4]-so the question of a true regional difference in pediatric rates of infl uenza-related pneumonia hospitalizations remains unresolved.

Many research questions about infl uenza in tropical regions remain
The Perspectives section is for experts to discuss the clinical practice or public health implications of a published article that is freely available online.

unanswered. For instance, Hong
Kong is a wealthy subtropical city, with a higher per-capita gross domestic product than Australia or New Zealand. Thus, the present Hong Kong study cannot inform us about the impact of infl uenza in poor tropical countries, where the severity of infections is likely compounded by malnutrition, limited supplies of antibiotics to treat secondary bacterial infections, and limited access to hospital care.
Modeling the infl uenza burden in tropical countries relies on goodquality virus surveillance data. And because of large year-to-year variations in the impact of infl uenza, modeling also depends on the availability of several years' worth of data. Since infl uenza surveillance efforts were only recently initiated in tropical countries, studies of the infl uenza burden are generally of short duration (fi ve years in the study by Peiris et al. [4]). The information generated so far during a period mostly dominated by more pathogenic infl uenza A (H3N2) viruses may not refl ect the average infl uenza experience in Hong Kong [13]. An additional shortcoming of the virus-driven modeling approach is that it cannot study past epidemics and pandemics for which no viral surveillance data was collected. Furthermore in this approach, deaths or hospitalizations attributed to infl uenza are directly proportional to the intensity of virus activity, and therefore, week-to-week variations in virus isolation rates should truly refl ect changes in disease prevalence and not sampling error. Laboratory surveillance needs to be strengthened before modeling of disease burden becomes feasible in most tropical countries [1].  The latitude of the capital city is indicated for each country in the legend. (A) Weekly distribution of infl uenza laboratory isolates (from week 1 to week 52; weekly frequency ( y -axis) is calculated as the weekly number of isolates divided by the annual number of isolates). (B) Weekly numbers of isolates were aggregated over four-week periods to show a (nearly) monthly distribution of infl uenza circulation. Each color represents a different month (color bar on the right). Note the transition in seasonal patterns from north to south, ranging from marked seasonal winter activity centered around January in the US, to uniform circulation throughout the year in Columbia and again, strong winter epidemics center around July in Argentina. Infl uenza viral surveillance data compiled from WHO Flunet.
Finally, more studies are needed to elucidate infl uenza seasonal patterns across a large range of latitudes, representing several tropical countries in both hemispheres. Of particular interest is the long-standing question of persistence: does infl uenza persist in temperate countries all year long (through undetectable chains of transmission in summer) or is it reintroduced from the Tropics each year at the beginning of winter? A combination of enhanced virus surveillance and sequence analysis of the viral genome of circulating strains is an important avenue for future research to address this question, and Hong Kong may be an essential sampling site in this respect.
Resolving this question is of paramount importance to vaccine composition and infl uenza control in tropical regions. Because of yearround infl uenza activity, it is not clear whether the Northern Hemisphere or Southern Hemisphere recommended composition should be used, and what the optimal timing for the vaccination campaign would be. Indeed, the infl uenza vaccination strategy may have to be tailored to each individual tropical region, based on local prevalence, timing, cycling of infl uenza strains, and how they match the strains circulating in temperate areas.

Who Should Be Vaccinated?
Peiris and colleagues' study implies that infl uenza is an underappreciated public health issue in tropical regionsindeed, infl uenza vaccine coverage is far lower in Hong Kong than in most developed countries with a temperate climate [17]. The authors argue for wider use of infl uenza vaccines in the elderly and chronically ill; however, this strategy has not been associated with any measurable reduction in infl uenzarelated mortality in the US [8], likely in part because of reduced responsiveness to the vaccine in the elderly [18]. In the US and Europe, moving toward universal vaccination is now being discussed as a strategy to further reduce the impact of infl uenza in high-risk groups by indirect means. Vaccinating schoolchildren may be especially benefi cial in this regard, since they are the main spreader of infl uenza in schools and households [19][20][21]. If the unusually high impact of infl uenza in young Hong Kong children is real, then this would be an additional argument for vaccinating schoolchildren in this city, as preschoolers are likely to contract infl uenza from their older siblings [22].  Model guided by viral surveillance data [4]. f Cohort study [23]. g Cohort study [16].