Seasonality of Respiratory Viruses at Northern Latitudes

Key Points Question What is the seasonal pattern of respiratory viruses at northern latitudes? Findings In this cohort study, a simple mathematical model was fitted to temporal data for 37 719 infections with respiratory syncytial virus, human metapneumovirus, or human coronaviruses 229E, NL63, OC43, or HKU1 and showed a marked biennial pattern. The same pattern was observed for 10 212 respiratory syncytial virus hospitalizations in young children. Meaning These findings suggest that alternating severe and mild winter peaks occur with striking regularity for multiple virus species, providing a pattern of health care utilization and possibly anticipating the evolution of the SARS-CoV-2 pandemic.


Properties of Respiratory syncytial virus (RSV)
RSV infection is nearly ubiquitous in infancy, and most children are infected by the age of 2 years (1). In Alberta, approximately 1.6% of children have one or more RSV hospitalizations before the age of 5 years, with most infections occurring in the first year of life (2).
RSV shows striking seasonality, which is accentuated in northern latitudes (3). In temperate zones, most RSV cases occur in the cold season and in tropical zones, most cases occur in the wet season (4). The mean duration of viral shedding is 6.7 days (range 1-21) (5). Re-infection is common, with annual re-infection rates of 20%, 17%, 10%, and 3-6% among children 5-9, 10-14, 15-19 years of age, and adults, respectively (6). Immunity wanes over time, as demonstrated experimentally in adults with previous natural immunity to RSV re-challenged with a homologous RSV strain. By 2 months after natural RSV infection, 50% of healthy adults became re-infected, and by 8 months, two thirds became re-infected (7). Within 26 months, 76% had two or more infections and 47% had three or more infections (7). Most RSV infections are mild, even in young infants. The rate of hospitalization among infected infants is 1.6 to 3.3% (1,8).

Properties of human metapneumovirus (hMPV)
Nearly all individuals are infected with hMPV early in life. One quarter of Dutch children between 6 and 12 months of age and nearly all five-year-old children had antibodies to hMPV (9). Similarly, in a Japanese cohort, 77% and 100% of children 2 to 5 and >10 years of age had hMPV antibodies, respectively (10). In a study from Thailand, 99.7% of children 6-16 years old had positive hMPV serology (11). Four genetic subgroups of hMPV (A1, A2, © 2021 Hawkes MT et al. JAMA Network Open. B1, and B2) are recognized, with variable incidence from year to year (12). They are antigenically similar and induce a variable cross-neutralizing antibody response (12,13).
Reinfection occurs throughout childhood. In Thailand, 5% of children had evidence of reinfection over a four-year period, as defined by a 4-fold rise in immunoglobulin titres.
Many children with acute hMPV exhibit both a positive IgM and IgG at presentation, followed by a 4-fold rise in IgG, indicating that they were previously infected, but that circulating IgG does not confer complete immunity (10). In non-human primate models, antibodies can mediate protection but titers wane over time (14). Taken together, these results suggest that transient protective immunity to hMPV allows recurrent infection over the lifetime of an individual.
Seasonality of hMPV is well documented in temperate zones. A biennial pattern has been described previously, with strong peaks of activity in late spring-summer months every second year (15).

Properties of seasonal coronaviruses
The duration of viral shedding was 5.6 days after experimental challenge with HCoV 229E in healthy volunteers (16).
Infection with HCoVs does not confer long-lasting sterile immunity. In a hospital-and community-based study of three coronaviruses (HCoV NL-63, OC-43, or 229E) among Kenyan children, 4-21% had documented re-infection within 47 and 98 days of the initial episode (17). In a serologic study from the Netherlands, ten healthy adults followed for 35 years experienced 101 infections (3 to 17 per individual) with HCoV NL-63, 229E, OC43, or HKU1, often repeatedly with the same strain (18). In an experimental challenge model with HCoV 229E in 15 healthy volunteers, 10 were infected, 8 were symptomatic, and 6 were infected a second time with repeat challenge 12 months later, although none of the second infections were symptomatic and the duration of viral shedding was reduced (16). Natural immunity to other severe and epidemic coronaviruses, SARS-CoV and MERS-CoV, has not been reported.
On the other hand, short-lived immunity and associated antibody responses are well documented. Volunteers resistant to initial HCoV 229E challenge had significantly higher levels of specific serum IgG and nasal IgA, suggesting that prior natural infection conferred temporary immunity (16). Although the precise duration of immunity is not known, these same volunteers were susceptible to re-infection after a period of 12 months (16). In another study, antibodies levels following natural infection decayed by 50% and 75% within 6 and 12 months of infection, respectively (18). In another modelling study, immunity to HCoV OC43 and HKU1 lasted approximately 45 weeks (19).
Whether temporary or longer-term immunity to SARS-CoV-2 exists is unknown, but has important implications for natural and vaccine-induced immunity. Using a non-human primate model, resistance to re-infection with SARS-CoV-2 has been demonstrated one month after initial infection, with associated SARS-CoV-2-specific memory B-cells (22). PCR positivity in patients with COVID-19 can return following repeated negative results (23, 24); however, the relatively short period before repeat positivity suggests they may not have cleared the initial infection. Reinfection with SARS-CoV-2 several months after clearing an initial infection has been reported (25). It is thus too early to say whether natural immunity to SARS-CoV-2 will contribute to herd immunity.  The dependent (time-to-event) variable was the interval from birth to the age at first RSV admission. The independent variables were the month of the year (e.g., January, coded as a 12-level categorical variable) and the RSV season (even or odd, coded as a binary variable).
Members of the birth cohort who were not hospitalized for RSV were censored at age 5 years.

eAppendix 4. Qualitative Analysis of Model
To simplify the qualitative analysis, we assumed that the birth rate and death rate were equal, such that the population was of constant size. Without loss of generality, we assumed a population size N=1, such that S, I and R were interpreted as proportions of the total population. The SIRS model has two stationary solutions: a disease free equilibrium (DFE):  The following race analogy has been used to explain the model's solution (27). The trajectory can be thought of as a hopeless 'pursuer' in a race to catch the EE, which plays the role of the fast 'leader' that cannot be caught (27). While ℛ 0 ( ) > 1, the leader is deemed as stable EE.
As soon as ℛ 0 ( ) < 1, the leader is deemed unstable and the solution's trajectory pursues the DFE. At that moment, the trajectory 'switches its objective,' pursues the DFE instead of the EE, and continues to do so while ℛ 0 ( ) < 1. The sudden change of objective gives rise to a period doubling bifurcation of the limit cycle (eFigure 1). Under what conditions will the stable limit cycle demonstrate a biennial pattern? We varied key model parameters 1 and and determined the peak infected fraction (eFigure 3). As predicted by the bifurcation analysis, a biennial pattern emerged for 1 > 1 − (<160 days) results in rapid decay of the recovered (immune) fraction before the onset of the next season. On the other hand, long-lasting immunity (>380 days) results in the accumulation of the recovered (immune) fraction over multiple years. When immunity wanes on the order of 1 year, a severe season results in a marked wave of infected individuals who enter the recovered (immune) fraction at a time coinciding with the following seasonal peak.
The large fraction recovered (immune) individuals blunts the seasonal peak, generating the observed biennial pattern.
In summary, the following conditions were associated with a biennial seasonal pattern (numerical values were computed from plausible model parameters): (1) The basic reproduction number of the model without seasonality must be greater than unity, in order for a non-zero stable endemic equilibrium to exist.
(2) The seasonal variation must generate a period doubling bifurcation.