Impacts of Zika emergence in Latin America on endemic dengue transmission

Abstract

In 2015 and 2016, Zika virus (ZIKV) swept through dengue virus (DENV) endemic areas of Latin America. These viruses are of the same family, share a vector and may interact competitively or synergistically through human immune responses. We examine dengue incidence from Brazil and Colombia before, during, and after the Zika epidemic. We find evidence that dengue incidence was atypically low in 2017 in both countries. We investigate whether subnational Zika incidence is associated with changes in dengue incidence and find mixed results. Using simulations with multiple assumptions of interactions between DENV and ZIKV, we find cross-protection suppresses incidence of dengue following Zika outbreaks and low periods of dengue incidence are followed by resurgence. Our simulations suggest correlations in DENV and ZIKV reproduction numbers could complicate associations between ZIKV incidence and post-ZIKV DENV incidence and that periods of low dengue incidence are followed by large increases in dengue incidence.

Fig. 1. Estimated dengue incidence in Brazil and Colombia (per 100,000).

Note that Zika incidence was not systematically reported prior to 2016 in Brazil (a) or late 2015 in Colombia (c) and that Chikungunya was not systematically reported prior to 2015 and late 2014 respectively in Brazil (a) and Colombia (c). In Brazil, updated data from the following year’s bulletin for 2014 to Epiweek 42, 2017 is used. States in Brazil (b) and departments in Colombia (d) are arranged by region and then by latitude from North to South.

Fig. 2. Comparison between predicted and observed dengue incidence.

Results for 2014–2017 are shown for Brazil (a, b) and Colombia (c, d) (see Supplementary Fig. 5 and 6 for full time series). b, d Red or blue indicate that the observed incidence fell above or below the median of 500 draws from the posterior of predicted values for that biweek. Medium or dark shading indicates that the observed incidence fell outside of the 90 or 95% prediction interval (PI) for that biweek. a, c The number of biweeks with observations falling below (blue) or above (red) the 90% PI are displayed with a quantile of the observed number of significant biweeks out of a distribution generated by 10,000 bootstrapped replicates. In these replicates, year labels were randomly re-assigned for each location before counting the biweeks in each year that were above or below the 90% PI (see “Methods” for further details).

Fig. 3. Spatial hierarchical biweekly dengue incidence models.

Shared coefficients from the year model are highlighted in yellow. Shared effect coefficients for Brazil (a) and Colombia (c). Zika and chikungunya coefficients are estimated from autoregressive dengue models. Positive (negative) coefficients indicate increases (decreases) in expected dengue incidence for the year model and indirectly as effects on transmission for the Zika and chikungunya models. Shared year multipliers for expected dengue incidence are shown for Brazil (b) and Colombia (d). The top row of panels b and d are translated from coefficients in (a) and (c). Other rows display subnational effects (combined shared and location-specific effects). Mean and 95% credible intervals (CrI) are shown. Intervals that overlap zero are displayed in gray.

Fig. 4. Simulation results incorporating immune-mediated interactions between DENV and ZIKV.

Mean and 95% inter-quantile range from stochastic simulations spanning 10 years post ZIKV-introduction. 100 simulations per scenario (see “Methods” for further details). ZIKV introduced after a 100 year burn-in period for four DENV serotypes (see Supplementary Fig. 11 for results when ZIKV is introduced 40 years after DENV). DENV and ZIKV reproduction numbers are assumed to be 4 and 2, respectively. Other reproduction number combinations are considered in Supplementary Fig. 10). The dashed line indicates one-half of the average incidence in (a) which we use to define the start and end of DENV prevalence troughs (see Methods and Supplementary Table 1). a–d Individuals with previous dengue exposure experience 20% of the DENV force of infection (FOI) that a fully susceptible person would. e–h Individuals with previous ZIKV exposure experience 80% of the FOI that a fully susceptible person would. i–l Individuals with ZIKV exposure experience 20% of the DENV FOI (same amount of cross-protection between dengue and Zika than between dengue serotypes).