Projections of dengue epidemics attributable to climate change in Peru

Discussion We estimated the future impact of climate change on dengue cases attributable to temperature suitable for dengue transmission in Peru under various climate change scenarios. Based on exposure–response curves derived from the observational period, dengue transmission risk increased with temperature, peaking at 24.6°C and displaying a near-unimodal pattern. Temperature-attributable dengue cases are projected to increase beginning in the 2030s, with variations in the magnitude, timing and rate of increase across different states. These findings highlight the importance of risk-informed surveillance and early warning, alongside effective control strategies, to mitigate the expected rise in dengue burden. The exposure–response curve identified in this study aligned with previous research. The RR increased with temperature up to 24.6°C and then declined. A slight, though not statistically significant, increase in RR was also observed at temperatures lower than the reference temperature of 15.3°C (the fifth percentile of the observed temperature distribution), possibly reflecting associations in states encompassing mountainous regions with lower annual mean temperatures. Because data are relatively sparse and uncertainty is greater in this lower-temperature range, this pattern should be interpreted cautiously as a potential manifestation of regional heterogeneity rather than as evidence of a consistent low-temperature effect. Previous studies conducted across multiple countries have suggested that the optimal temperature range for dengue transmission lies between 21.0°C and 31.0°C.3 4 12 34 35 Although the specific peak temperature may vary depending on environmental conditions and human behavioural patterns, most studies have observed a unimodal association: dengue transmission risk rises with increasing temperatures before declining beyond a certain point. This pattern is likely explained by several temperature-dependent biological processes that jointly shape vectorial capacity, including faster mosquito development and reproduction, improved adult survival within a moderate range, shortened extrinsic incubation periods, increased biting and accelerated viral replication.7 8 However, adult mosquito survival can decline at higher temperatures (eg, above approximately 30°C), thereby constraining onward transmission.6 In our study, the estimated peak risk occurred below these upper thermal-stress thresholds, suggesting that the population-level exposure–response association may reflect not only mosquito physiological constraints but also, in part, temperature-driven changes in realised mosquito–human contact rates,36 including shifts in host-seeking/biting activity and human behaviours (eg, heat avoidance and increased time spent indoors). Importantly, the estimated peak at 24.6°C should not be interpreted as a universal physiological ‘optimal’ temperature; rather, it is the data-supported maximum of the DLNM-estimated, lag-cumulative exposure–response surface pooled across regions, reflecting the net balance of temperature-dependent transmission processes. In addition, because this estimate integrates effects over a 2–6-week lag window, the observed association should be interpreted as reflecting the cumulative downstream impact of temperature on transmission over subsequent weeks, rather than an instantaneous effect of temperature alone. Our projections suggested that dengue transmission risk will increase in Peru as climate change progresses. Rising average temperatures are expected to more frequently reach optimal levels for dengue transmission, enhancing mosquito breeding conditions and accelerating viral replication. These climatic shifts are, therefore, likely to amplify dengue epidemics in the future. Although direct comparisons are limited by differences in model assumptions and outcome definitions, our findings were generally consistent with previous global estimates.4 17 For instance, a study applying a multi-model framework integrating mechanistic and statistical models projected increases in both the duration of dengue transmission seasons and the population at risk in the Americas.4 Similarly, a statistical mapping study predicting environmental suitability for dengue transmission suggested that, although the results were not statistically significant, the land area suitable for dengue could expand across the Americas.17 Despite the projected steady rise in annual average temperature from the 2020s onward, temperature-attributable dengue cases are expected to remain relatively stable during the 2020s, followed by a marked increase from the 2030s to 2050s. This pattern can be interpreted from two complementary perspectives. Biologically, it suggests that significant shifts in the physiological traits of Aedes mosquitoes and the dengue virus may occur only after temperature increases surpass specific thresholds. Previous studies have shown that Aedes larvae require temperatures around 15°C or higher to complete development, and adult mosquitoes typically become active at temperatures between 13°C and 16°C.37 Similarly, dengue virus replication is slow and largely confined to the mosquito midgut at temperatures below 18°C; however, once this threshold is exceeded, the virus can invade the salivary glands, enabling potential transmission to humans.37 These findings highlight the sensitivity of vector and viral physiology to temperature, with significant biological changes occurring once critical thermal thresholds are crossed. From a statistical perspective, an increase in temperature-attributable dengue cases reflects a higher frequency of temperatures associated with elevated RRs along the exposure–response curve. In our study, the RR remained around 1 between 15.3°C and 17.5°C but increased sharply from 17.6°C to 24.6°C. Thus, although warming shifts the overall temperature distribution, initial increases within the 15.3°C–17.5°C range may have limited effects on dengue transmission. However, sustained warming will eventually lead to more frequent occurrences of temperatures within the higher-risk range (17.6°C–24.6°C), resulting in a subsequent surge in dengue cases. We observed substantial regional variation across states in the magnitude, timing and rate of increase in temperature-attributable dengue cases under climate change scenarios. These heterogeneous projections are consistent with state-specific temperature–dengue associations, which likely reflect differences in baseline climate and seasonality as well as several contextual factors (eg, urban infrastructure and constraints on vector habitat) that modulate how temperature translates into realised transmission.38 39 In particular, states with relatively low dengue burden during the baseline period are projected to experience some of the largest relative increases in temperature-attributable cases, suggesting a potential geographic shift of dengue risk. This pattern underscores the importance of proactive surveillance and preparedness in areas that may have less prior dengue experience and limited capacity for rapid response. Moreover, the consistent upward trends in temperature-attributable dengue cases across all SSP scenarios highlight the critical need for effective disease intervention and control strategies. Although the SSP scenarios differ in carbon dioxide emission trajectories, their temperature pathways remain relatively similar through mid-century (figure 3), and the divergence becomes more pronounced in the second half of the 21st century.31 This helps explain why projected temperature-attributable burdens differ only modestly across SSPs within our 2021–2050 horizon. In addition, our results are summarised as 5-year aggregates and multi-model ensemble means (with a common baseline reference), which can smooth short-term variability and attenuate small between-scenario differences in estimated burdens. To manage the anticipated increase in dengue risk by the mid-21st century, adaptable, region-specific policies that integrate local environmental conditions and distinct transmission patterns will be essential. Dengue prevention efforts, such as vector control strategies (eg, use of insecticides and repellents, elimination of mosquito breeding sites) and enhanced public education and community engagement, have demonstrated effectiveness.1 Accordingly, in high-baseline endemic areas (eg, northern coast and parts of the Amazon), priorities include strengthening year-round integrated vector management and household water-storage control, with rapid outbreak response capacity during peak seasons.40 In regions with the largest projected relative increases, especially Andean/highland-including states, preparedness for seasonal and geographic expansion is critical, including expanded entomological and clinical surveillance in historically low-incidence districts, strengthened clinician awareness and training41 in settings with limited prior dengue experience. Across settings, integrating climate information with routine surveillance and leveraging subnational short-term forecasting tools could support earlier resource pre-positioning, targeted risk communication and more proactive planning and allocation of limited public health resources.42 Long-term future projections inherently rely on multiple assumptions and are subject to certain limitations. Therefore, the projections of temperature-attributable dengue cases presented in this study should not be interpreted as definitive predictions, but rather as potential scenarios based on predefined assumptions. First, precipitation was included in the association model for adjustment, but projected changes in precipitation were not propagated in the attribution calculations. Similarly, our scenarios assumed no changes in other key drivers such as public health policies, climate adaptation efforts or population dynamics, thus enabling isolation of the influence of altered temperature distributions on dengue risk. Accordingly, the projected burdens should be interpreted as changes attributable to temperature under an attribution framework, rather than as joint impacts from simultaneous changes in temperature and other determinants. To minimise the uncertainties introduced by these assumptions, projections were limited to the period up to 2050. Nevertheless, socioeconomic development and demographic changes are known to be key determinants of dengue risk.1 34 Increased access to air conditioning, improvements in housing infrastructure and behavioural changes that reduce exposure to mosquito bites may mitigate the projected increases in dengue transmission risk. Furthermore, future advances, such as expanded deployment of dengue vaccines or innovative vector control strategies (eg, Wolbachia-infected mosquitoes), could significantly reduce the future dengue burden beyond the estimates presented here.1 Several additional limitations should be noted. First, our estimates were based on temperature–dengue associations observed during the study period and may not fully capture future risks in areas where dengue was previously rare, particularly high-altitude regions. Future research incorporating mosquito vector dynamics and distribution data would help address this gap. Second, we applied average reporting delays at the state level during model fitting due to the absence of detailed symptom onset data, precluding the use of more precise backward estimation methods. This likely introduced additional model uncertainty. Third, as dengue diagnosis requires differentiation from other infectious diseases such as Zika virus and leptospirosis,3 case counts, including probable diagnoses, may have been affected during periods of concurrent outbreaks. If such outcome misclassification is largely non-differential with respect to temperature, the estimated temperature–dengue association would be expected to be attenuated towards the null. By contrast, differential misclassification that varies by season, region or outbreak context could bias estimates in more complex and less predictable ways. However, our sensitivity analysis restricted to laboratory-confirmed cases yielded results that were broadly consistent with the main analysis, supporting the robustness of our findings. Therefore, any outcome misclassification arising from the inclusion of probable cases is unlikely to materially change our main conclusions. Finally, due to data limitations, this study did not account for dengue virus serotypes (DENV-1 to DENV-4), each of which exhibits distinct epidemiological and clinical characteristics.1 The cocirculation of multiple serotypes can accelerate transmission and increase the risk of severe dengue through secondary infections. Failure to incorporate serotype dynamics may have introduced random errors. Future studies should integrate serotype-specific data to more accurately characterise the complexity of dengue epidemics.