Global Influenza Seasonality: Reconciling Patterns across Temperate and Tropical Regions

1Few seasonal disease patterns have generated as much interest, or have so confounded researchers, as the distinct winter epidemics of the influenza virus in temperate populations. Influenza is one of the most significant diseases in humans, considered to be associated with approximately 250,000–500,000 deaths globally each year.

2It remains to be shown what mechanisms are responsible for global patterns of influenza seasonality. Influenza epidemics in temperate populations occur in the northern and southern hemispheres during their respective winters. Consequently, it has long been speculated that cold temperatures, low indoor humidity, and minimal solar radiation are causally linked to influenza epidemics. Yet in the tropics, where humidity and temperatures remain high year-round, and where solar radiation does not vary strongly (relative to temperate regions), mounting evidence suggests that many of these populations also experience marked seasonal influenza epidemics. Further, recent evidence points to a link between increased influenza activity and the rainy season in several tropical populations when humidity is typically greatest, in contrast to low-humidity (indoor) temperate epidemics. Also in opposition to temperate regions, many tropical populations have significant influenza activity year-round, and some tropical populations are characterized by two distinct influenza seasons, such as Singapore (Figure 1) and Hong Kong.

3A variety of explanations have been proposed to account for the seasonal nature of influenza. Although many of these explanations have been discussed in past literature (e.g.,;;;;; ), here we review and organize the evidence available for each putative mechanisms in order to identify research gaps (summarized in Tables 1–3). We followed a framework where putative seasonal stimuli drive seasonal influenza incidence through three primary mediating mechanisms: seasonal variations in host contact rate, virus survival, and host immunity (Figure 2). Within this framework, we examined the available evidence regarding each seasonal stimulus, assessed its potential to explain influenza seasonality in both temperate and tropical regions, identify key outstanding questions, and provide recommendations for future research.

4Increased proximity between susceptible and infected hosts is frequently suggested to be an important driver of influenza seasonality (Table 1). Observations of the rapid dissemination of influenza at the Hajj in Saudi Arabia and on passenger aircraft and the possibility that crowding among soldiers in World War I hastened the development of the 1918 influenza pandemic suggest that contact rates are important in influenza virus transmission.

5A long-held hypothesis suggests that crowding indoors during cold weather causes wintertime temperate epidemics. A similar mechanism potentially explains the coincident timing between tropical influenza epidemics and the rainy season because individuals may move indoors to escape precipitation. Using a human activity database for various locations in the United States, demonstrated that individuals spend on average 1–2 hr more indoors during cold weather and spend about 0.5 hr more time indoors during rainy weather. Although these differences are minimal relative to the 21–22 hr individuals spend indoors on average, it is possible that even a relatively small change in contact and transmission rates could be sufficient to cause epidemics. Accordingly, seasonal variability in contact rates related to school schedules, such as children returning to school from holiday, could drive the seasonal nature of influenza. This is consistent with findings in, which showed that holidays reduce transmission among children in France by 20–29%. In addition, geographical variation in the timing of the fall wave of the 2009 pandemic across the United States has been shown to coincide with geographical variation in school schedules.

6However, although school closures likely have a significant effect on transmission, it has yet to be explained why influenza peaks during the winter in temperate locations, and not during the fall or spring when children are also in school. In addition, it is difficult to imagine how school schedules could result in influenza seasonality in both temperate and tropical locations, because tropical epidemics are sometimes characterized by bimodal seasonality and year-round transmission. Furthermore, crowding also occurs year-round at festivals, sporting events, and conferences without consistent outbreaks of infection. In addition, in locations such as the deserts of the southwestern United States, hot temperatures regularly drive individuals indoors during the summer, yet these locations are still characterized by winter influenza epidemics.

7Therefore, and as has been noted previously, no empirical data link increased contacts rates due to weather conditions and increases in influenza transmission. Nevertheless, it is a strong possibility that variability in contact rates interacts with other seasonal stimuli to determine the precise timing of influenza epidemics. Thus, these mechanisms warrant further investigation. A population-based prospective study using time diaries to investigate the contact characteristics of 7,290 individuals over the course of a single day was a significant step forward in this regard.

8The effect of contact rates at larger spatial scales (e.g., state and continental) on the spread of the influenza virus must also be considered. For instance, demonstrated that human mobility patterns in the United States synchronize epidemics among highly connected populations and that the virus tends to spread from populous to less populous locations. It is possible that contact rates at these larger scales may be particularly important to the seasonal patterns of influenza in smaller populations, or for locations that are characterized by minimal environmental variability. In such cases, seasonal variability in the volume of infected individuals entering a population may exceed any environmentally mediated seasonal forcing. Analysis of seasonality in large-scale human mobility patterns—for instance, using global database on air passenger flows—may shed light on this possibility. In addition, it would be extremely interesting to screen year-round incoming air passengers at selected destinations (e.g., islands), collect respiratory samples, and test for seasonal variations in the prevalence of influenza importations.

9Influenza can be transmitted through several distinct mechanisms, including large droplets, aerosols, and direct contact (including contact with contaminated hosts and surfaces). To survive during transport among hosts, the influenza virus must be able to endure a variety of environmental conditions. Thus, the effects of the ambient environment on virus survival have long been considered an important factor related to the seasonality of influenza (Table 2).

10Several classical experiments performed in past decades directly measured the tolerance of aerosolized influenza virus to humidity and temperature. Of the six studies that have investigated the issue, four found that influenza virus survival increases monotonically with a decrease in relative humidity (RH). The two remaining studies observed a bimodal relationship, with virus survival greatest at low RH, minimal at mid-RH, and moderate at high RH.

11Examining the effect of temperature and RH on aerosol transmission among guinea pigs, showed that the efficiency of aerosol transmission decreases as temperature increases from 5°C to 20°C and is completely prevented at 30°C. Furthermore, transmission among guinea pigs was inversely related to RH, with high transmission when RH was 20–35% and completely absent at 80%. In all, the experiments provided evidence that virus survival is important to aerosol transmission, at least in some laboratory settings.

12Reexamining available data from the virus survival studies and, demonstrated that absolute humidity (AH) is a better predictor (vs. RH and temperature) of influenza virus survival and transmission among guinea pigs. Specifically, virus survival and transmission among guinea pigs increased monotonically with a decrease in AH. This distinction is important because AH is a measure of the amount of water vapor in a volume of air, whereas RH is a measure of the amount of water vapor in the air relative to the amount of water vapor in saturated air of the same volume and air temperature. Further, showed that the relationship between AH and virus survival is consistent with observations in the United States where anomalously low AH conditions generally precede the onset of influenza epidemics by approximately 2 weeks. However, temperature and AH are strongly correlated, thereby making it difficult to exclude a confounding effect. It is unlikely this mechanism can explain influenza seasonality in the tropics because those regions are typically humid year-round, and epidemics tend to occur during the rainy season, when AH is typically at locally maximal levels. However, as previously stated, there is some evidence that the effect of RH (and potentially AH) on virus survival is bimodal, possibly explaining dry-temperate and rainy-tropical epidemics.

13Fewer studies have investigated the effects of humidity on the survivability of influenza on surfaces. However, demonstrated that influenza survival on steel surfaces is also inversely related to AH.

14The recent findings regarding influenza survival and AH have—at least temporarily—supplanted seasonality explanations regarding the effects of other factors on influenza survival. However, laboratory experiments have shown that the influenza virus becomes more susceptible to inactivation as envelope lipids become increasingly disordered above 21°C, suggesting an independent relationship between temperature and virus survival.

15Further, several laboratory studies have demonstrated the sensitivity of influenza viruses to ultraviolet radiation. Accordingly, it is hypothesized that viruses expelled into the environment incur lower inactivation rates during seasons with reduced sun activity, consistent with temperate epidemics occurring during the winter and tropical epidemics during the rainy seasons. Hypotheses based on virus survival have to account for the fact that indoor environments, where a great amount of interaction among hosts occurs in modern society, are significantly insulated from external conditions. Because environments often vary considerably between the indoors and the outdoors, the primary location of influenza transmission needs to be determined so that the conditions the virus is exposed to during transmission can be better specified.

16Here we broaden the technical definition of immunity to include all physiological functions that enable a host to avoid or mitigate infection after exposure to influenza viruses. The hypothesis that seasonal variation in immunity explains influenza seasonality is supported by observations that humans are less likely to experience influenza-like symptoms after manual inoculation with influenza viruses during interepidemic periods relative to common epidemic months.

17Temperature and humidity can affect host immunity through a number of processes (Table 3). For example, the inhalation of cold air causes vasoconstriction in the nose and respiratory tract, resulting in reduced blood flow, diminishing the supply of leukocytes and phagocytic activity in these areas. Dry conditions can result in moisture losses in the nasal mucosa and reduce mucociliary clearance. Finally, reactions of host physiology to temperature may alter viral shedding. For instance, indicated that peak viral shedding lasted 40 hr longer for guinea pigs housed at 5°C relative to those exposed to 20°C.

18Ultimately, although there is sufficient evidence to indicate that human immune function is negatively affected by temperature and humidity, the magnitude of these factors on influenza-specific susceptibility will remain unclear until more laboratory and clinical experiments are performed (Table 3). An example of progress in this regard is a study by, which indicated that the antiviral and proinflammatory responses of guinea pigs housed at 5°C and 20°C were similar, suggesting that the innate immune response is not impaired at low temperatures. Although this experiment could have been more complete by testing the first line of host defenses, such as mucociliary function, it is a useful example of how laboratory experiments can shed light on seasonal variations in immunity and their impact on virus replication and transmission.

19The duration of human exposure to solar radiation, or “photoperiod,” may also modulate immunity at seasonal time scales related to host vitamin D status. Human vitamin D levels are generally dependent upon exposure to ultraviolet B radiation, and in turn, deficiencies of this vitamin are common in temperate populations during the winter when solar radiation is lowest. Vitamin D has been demonstrated to stimulate innate immunity. Two studies have shown that individuals with lower vitamin D levels are significantly more likely to report respiratory infections. Furthermore, a recent randomized, controlled study to test the effect of vitamin D supplementation on influenza A and B incidence in school children indicated that the controls were significantly more likely to become infected with influenza A than the experimental group. However, another randomized, controlled study found no significant association between vitamin D supplementation and respiratory infections, and reported no significant effect of vitamin D supplementation status and influenza B infection in school children. Less is known about vitamin D levels in tropical latitudes where solar radiation varies less substantially over a year, but seasonality in vitamin D levels has been reported in subtropical Hong Kong. The lowest levels of solar radiation in the tropics typically coincide with the local rainy seasons due to cloudiness. This appears consistent with the relationship between tropical epidemics and the rainy season. Clothing, skin pigmentation, age, behavioral habits, and other factors also affect vitamin D status, providing opportunities to investigate how individual levels of vitamin D in the same localities and season are correlated with influenza occurrence.

20Other nutrients are also required for proper immune function and can modify the pathology of infection. This opens the possibility that seasonal variations in nutrient availability and/or requirements on diet might also have a role on influenza seasonality. Experimental studies in mice suggest that the supplementation of vitamin C and vitamin E can diminish the severity of influenza infection. Selenium deficiency has also a negative effect on host immune response and the severity of infection in human airway epithelial cells. indicated a statistically significant association between general micronutrient deficiency and respiratory disease in elderly individuals in Quito, Ecuador. In all, this hypothesis may be most viable in low-income communities because seasonal effects on local food availability and diet are more difficult to compensate for with other nutritional sources. Variations in dietary intake of poorer individuals in developed communities may occur because of seasonal financial pressures, such as those caused by the cost of heating during temperate winters. Ultimately, it is unlikely that the seasonal variability of diet is sufficiently widespread and severe to account for global influenza patterns; however, it may be important in some populations.

21Several understudied hypotheses regarding seasonal variations in human immunity also warrant mentioning. For instance, there is an abundance of information on the effects of photoperiod and dark/light cycles on physiology, affecting the immune systems of mammals, including humans, possibly mediated through secretions of melatonin. Seasonal changes in airborne particulate matter and pollutants may also have a detrimental effect on immune function. Increases in the energy required for thermoregulation may limit the energy available for immune function. This may explain influenza epidemics during temperate winter conditions or during damp conditions related to the rainy season in the tropics. The effects of abrupt changes in temperature may also be detrimental to immune function.

22Another immune mechanism that needs consideration is the interaction among influenza and other pathogens. Host cells produce and release interferons in response to many viral infections, resulting in a decrease in susceptibility to subsequent viral challenges. This heightened state of immune activity during and after a viral epidemic increases herd immunity, making it difficult for additional viruses to become established in a population. In temperate locations, the influenza season overlaps with periods of infection by other viral pathogens, such as rhinoviruses and respiratory syncytial viruses. Observational studies suggest that the timing of non-influenza epidemics can modify the timing of influenza epidemics. For instance, during the fall of 2009, a rapid decline of laboratory-confirmed H1N1 influenza cases in Sweden and an unexpectedly slow start to the H1N1 epidemic in France both coincided with rhinovirus outbreaks. It is less clear how influenza covaries with other viruses in tropical regions. Yet, although pathogen interactions may explain subseasonal variations of influenza incidence, it likely does not explain why influenza viruses (and other co-circulating viruses) preferentially spread during temperate winters and tropical rainy seasons.

23A major outstanding question regarding influenza seasonality is whether a single seasonal stimulus (or set of stimuli) accounts for universal patterns in influenza virus activity, or whether the key mechanisms underlying the seasonality of influenza in temperate regions differ from those in the tropics. Currently, the most accepted hypotheses explaining influenza seasonality, such as AH and virus survival, attempt to explain influenza seasonality only in temperate regions. Less attention has been given to influenza seasonality in the tropics, likely because of the lack of information about seasonal signals in the past. To advance our understanding, it is important either that hypotheses explaining temperate influenza seasonality are amended so that they include the tropics, or that hypotheses are generated specifically for the tropics that will work in harmony with temperate explanations. To this end, identifying where temperate (cold and dry) and tropical (rainy) relationships with influenza epidemics break down geographically and how this relates to relevant environmental variables (e.g., temperature, humidity) is a short-term and practical goal.

24Although numerous mechanisms have been proposed to account for global patterns of influenza seasonality, distinguishing the causal relationships from colinear and confounding associations has proved overwhelming. To overcome this challenge, rigorous observational and experimental studies will be necessary. Laboratory studies are crucial, because confounding factors can be isolated and controlled, and their effects on host immunity and/or virus survival can be specifically addressed and properly measured. Ecological studies should be used to assess the consistency between the results of experimental and observational studies and influenza seasonality observed globally. When possible, it is important that studies use laboratory-confirmed influenza data so that influenza-specific factors are identified, versus those related to other infections with similar symptomology.

25Finally, no understanding of influenza seasonality is complete without consideration of the intrinsic aspects of the disease and the ability of the virus to continually reinfect humans by escaping recognition by host antibodies via frequent mutation, a process of selectively driven evolution termed “antigenic drift”. According to, antigenic drift and the subsequent waning immunity of a population provide a seasonal forcing that is nearly sufficient to stimulate seasonal influenza epidemics each year (“dynamic resonance”), requiring only relatively minor changes in extrinsic seasonal forces to “lock” epidemics into phase. However, the mathematical model used to support this hypothesis relies upon unrealistically large rates of reproduction for the influenza virus and oversimplifies important evolutionary questions related to antigenic drift and partial cross-immunity among strains. Moving forward, more studies, conducted in parallel with a robust description of antigenic changes in circulating viruses, are needed that explore changes in population immunity and virus transmissibility over time, both between seasons and during the course of an epidemic. These studies would allow a better understanding of the intrinsic dynamics of influenza viruses and would help determine the contribution of dynamic resonance to influenza seasonality patterns.

26In addition to increasing our capacity to design effective public health prevention and control interventions, understanding the seasonal processes associated with influenza seasonality will potentially inform us about the processes related to the seasonality of other infectious diseases. Further, the understanding of the environmental factors that cause influenza seasonality should also be useful to forecast novel dynamics in regional contexts due to changing environmental and climatological scenarios. Thus, this is an important scientific and public health endeavor. Ultimately, the more systematically we proceed in our investigation of influenza seasonality, the more likely we are to unravel this age-old epidemiologic mystery.