Prior immunity helps to explain wave-like behaviour of pandemic influenza in 1918-9

Abstract

Background: The ecology of influenza may be more complex than is usually assumed. For example, despite multiple waves in the influenza pandemic of 1918-19, many people in urban locations were apparently unaffected. Were they unexposed, or protected by pre-existing cross-immunity in the first wave, by acquired immunity in later waves, or were their infections asymptomatic?

Methods: We modelled all these possibilities to estimate parameters to best explain patterns of repeat attacks in 24,706 individuals potentially exposed to summer, autumn and winter waves in 12 English populations during the 1918-9 pandemic.

Results: Before the summer wave, we estimated that only 52% of persons (95% credibility estimates 41-66%) were susceptible, with the remainder protected by prior immunity. Most people were exposed, as virus transmissibility was high with R0 credibility estimates of 3.10-6.74. Because of prior immunity, estimates of effective R at the start of the summer wave were lower at 1.57-3.96. Only 25-66% of exposed and susceptible persons reported symptoms. After each wave, 33-65% of protected persons became susceptible again before the next wave through waning immunity or antigenic drift. Estimated rates of prior immunity were less in younger populations (19-59%) than in adult populations (38-66%), and tended to lapse more frequently in the young (49-92%) than in adults (34-76%).

Conclusions: Our model for pandemic influenza in 1918-9 suggests that pre-existing immune protection, presumably induced by prior exposure to seasonal influenza, may have limited the pandemic attack-rate in urban populations, while the waning of that protection likely contributed to recurrence of pandemic waves in exposed cities. In contrast, in isolated populations, pandemic attack rates in 1918-9 were much higher than in cities, presumably because prior immunity was less in populations with infrequent prior exposure to seasonal influenza. Although these conclusions cannot be verified by direct measurements of historical immune mechanisms, our modelling inferences from 1918-9 suggest that the spread of the influenza A (H1N1) 2009 pandemic has also been limited by immunity from prior exposure to seasonal influenza. Components of that immunity, which are measurable, may be short-lived, and not necessarily correlated with levels of HI antibody.

Figure 1

Model for unobserved processes. Before wave 1, people are either susceptible (S₁) or resistant (R₁) because of prior immunity. Persons exposed to the virus either express symptoms (E₁), or have an asymptomatic infection (A₁). Others are unexposed (U₁). After exposure, persons become immune (E₁R or A₁R), and a proportion become susceptible again prior to wave 2 (E₁S₂, A₁S₂ and R₁S₂, plus the susceptible persons who escaped exposure in wave 1 (U₁S₂). All susceptible persons are at risk of exposure in wave 2 (see figure), and either express symptoms, have an asymptomatic infection, or remain unexposed. The extension to wave 3 adds an additional layer of complexity, but there are no new principles invoked.

Figure 2

Observations over three waves. For each population there was information on N individuals, of whom N₁ reported symptoms in wave 1 and N₀ did not; in wave 2, N₁₂ had a repeat attack, while N₁₀ of those affected in wave one did not have a repeat attack; N₁₂₃ individuals reported symptoms in all three waves...;;etc. The total number reporting symptoms in wave 2 was N₁₂ + N₀₂, while the total number for wave three was N₁₂₃ + N₁₀₃ + N₀₂₃ + N₀₀₃. The observed numbers N₁₂₃, N₁₂₀, N₁₀₃, N₁₀₀, N₀₂₃, N₀₂₀, N₀₀₃, N₀₀₀ for each population can be recovered from the proportions SAW, SA-, S-W, S- -, -AW, -A-, - -W, - - - shown in Table 1 by multiplication by the corresponding N.