First Evidence into the Association of Warming with Stroke and Myocardial Infarction Mortality in the Brazilian Amazon

Abstract

Rising temperatures intensify thermoregulatory stress, leading to increased cerebrovascular and cardiovascular morbidity and mortality. In this study, we aimed to examine the associations between rising temperatures and cerebrovascular and cardiovascular mortality in the Brazilian Amazon. Deaths from stroke and myocardial infarction were analyzed alongside monthly mean temperatures from 2000–2023. Poisson regression models were used to assess temporal trends and age-specific differences, whereas time series and distributed lag models were used to evaluate the influences of temperature and the cumulative effects of extreme heat. Increases in mean temperature were significantly associated with higher mortality for both outcomes, with older adults showing greater vulnerability, particularly those aged over 50 years. Prolonged exposure to extreme heat increased mortality risk, and the trend became more evident after 2015. These findings are the first to demonstrate that warming is associated with increased cerebrovascular and cardiovascular mortality in Amazonian populations, underscoring the urgent need for mitigation and adaptation strategies such as early warning systems, climate-resilient infrastructure, and improved healthcare access.

1. Introduction

Climate change is a major environmental determinant of public health, influencing cerebrovascular and cardiovascular morbidity and mortality [1]. Rising temperatures intensify thermoregulatory stress, leading to dehydration, hemodynamic alterations, and inflammatory responses that may exacerbate chronic conditions [2]. Stroke and myocardial infarction are among the most critical outcomes, accounting for a substantial number of deaths in tropical countries [3,4].

Older adults and individuals with chronic diseases are particularly vulnerable, as their adaptive capacity to thermal stress is reduced [5]. This vulnerability is further amplified in settings of social inequality and unplanned urbanization, where access to infrastructure and health services is limited [6]. This situation is especially concerning in tropical regions, where deforestation and land-use change have altered the microclimate, resulting in higher mean temperatures, more frequent heatwaves, and reduced relative humidity [7]. Deforested areas can experience temperature increases of up to 2 °C, thereby intensifying population exposure to excessive heat and aggravating health risks [2].

In this study, we investigated the associations between rising temperatures and mortality from stroke and myocardial infarction in the Brazilian Amazon between 2000 and 2023. Monthly time series data were analyzed to estimate the effects of extreme heat and cumulative exposure with lags of up to two months. Although global cardiovascular mortality has been rising, earlier evidence up to 2013 did not reveal an association between temperature and cardiovascular mortality in the region [8]. However, given recent warming trends and the increasing frequency of extreme events, we hypothesized that this association has become evident in more recent years. These mortality trends likely reflect concurrent demographic and socioeconomic transitions, including population aging, urban expansion, and improved reporting [6]. Distinguishing these non-climatic influences from temperature effects remains essential to quantify the true contribution of warming to cardiovascular mortality.

2. Materials and Methods

2.1. Study Area

The study was conducted in the Brazilian Amazon (Figure 1), a region with a humid equatorial climate characterized by mean annual temperatures of 24–28 °C, high relative humidity, and marked rainfall seasonality concentrated between January and July [9]. The study area includes the states of Acre, Amapá, Amazonas, Pará, Rondônia, Roraima, Tocantins, northern Mato Grosso, and the western Maranhão. According to the 2022 national census, the area hosts more than 29 million residents across over 770 municipalities.

2.2. Data Collection

Mortality data, including death from stroke and myocardial infarction categorized by place of residence, were obtained from DataSUS, the official public health information system of the Brazilian Ministry of Health. Monthly time series were constructed from 2000–2023 and stratified by age group; 2024 was excluded because of data unavailability. Population data were extracted from the 2000 and 2023 national censuses, with intercensal estimates provided by the Ministry of Health, enabling the calculation of mortality rates. The mean temperature data were retrieved from the Brazilian National Institute of Meteorology (INMET) from 45 automatic stations across the Amazon, and the monthly averages were aggregated for the study period.

2.3. Statistical Modeling

All analyses were performed in GNU R version 4.4.1 [10]. Mortality rates per 1000 inhabitants were calculated for stroke and myocardial infarction by dividing the number of deaths by the estimated population. To assess temporal trends by age group, we fitted Poisson regression models with mortality rate as the outcome and the year-age group interaction as predictors. This approach tested whether trends differed across age groups and identified the most vulnerable strata. Predictor significance was evaluated via likelihood ratio tests based on the chi-square distribution.

We modeled mortality rates in relation to the mean temperature via the autoregressive integrated moving average with exogenous variables (ARIMAX), which accounts for serial dependence while incorporating temperature as an external predictor [11]. To explicitly control for long-term trends in mortality unrelated to short-term temperature variability, such as demographic aging, shifts in cardiovascular risk profiles, and gradual improvements in death registration, the ARIMAX framework inherently incorporated differencing and autoregressive components selected to capture long-term temporal structure and minimize information criteria. Model selection was performed via the auto.arima function in the forecast package [11], and model adequacy was verified through residual diagnostics.

A distributed lag nonlinear model was applied to estimate the cumulative effects of high-temperature exposure on mortality, with lag structures of up to two months using the dlnm package [12]. Long-term temporal trends and seasonality were controlled by including a smooth function of time based on natural splines in the regression framework, allowing the separation of short-term and interannual temperature effects from underlying mortality trends. The models were fitted with Poisson regression corrected for overdispersion, and the results were expressed as relative risks with 95% confidence intervals (CIs).

In addition, a LOESS-based scatterplot was constructed showing the relationship between mean monthly temperature and mortality rates. Each point represents a monthly observation colored by year to visually capture temporal warming trends. Months of extreme heat were identified across the 24-year period, defined as those with mean temperatures at or above the 90th percentile of the distribution. Mortality rates were then compared between extreme and nonextreme months via Student’s t test. A bar chart illustrates mortality trends over time, contrasting deaths during extreme-temperature months with those in normal months, providing clearer insight into temporal and thermal patterns.

Sensitivity analyses were performed to evaluate model robustness and potential biases related to temporal structure and mortality reporting. Models were re-estimated after excluding years with extreme temperature or mortality values (2000 and 2023), varying the temporal lag structure between 1 and 3 months, and uniformly adjusting mortality counts by ±10% across the entire study period to simulate possible under- or overreporting in the mortality registration system. This approach assessed whether the temperature-mortality associations were sensitive to moderate, non-differential reporting uncertainty over time.

2.4. Ethical Aspects

All the data are secondary, publicly available, and anonymized, ensuring that there is no possibility of individual identification. Therefore, the study was not submitted for research ethics committee review.

3. Results

Between 2000 and 2023, the annual mean temperature increased by approximately 1.7 °C. Over the same period, mortality from stroke and myocardial infarction showed marked long-term upward trends. After accounting for long-term temporal trends and serial autocorrelation, ARIMAX models indicated that interannual temperature variability remained significantly associated with mortality from stroke (β = 0.18; p = 0.001; Figure 2a) and myocardial infarction (β = 0.26; p = 0.002; Figure 2b), with adequate residual diagnostics (p = 0.25 and 0.29, respectively). These estimates reflect temperature-related deviations from underlying mortality trends rather than the increase in mortality over time. For stroke, the year-age interaction was significant (D_6,154 = 751; p < 0.001), with the strongest effect among individuals aged >60 years (Figure 2c). For myocardial infarction, the interaction effect was also significant (D_6,154 = 494; p < 0.001), particularly in those aged >50 years (Figure 2d).

Cumulative exposure to elevated temperatures over the two preceding months was associated with increased mortality from stroke (D = 3.17; p < 0.001; Figure 3a) and myocardial infarction (D = 4.49; p < 0.001; Figure 3b). These effects became evident above mean monthly temperature thresholds of approximately 29 °C for stroke and 30 °C for myocardial infarction. LOESS-based scatterplots (Figure 4) showed a nonlinear positive association between temperature and mortality for both outcomes, with relatively stable mortality rates up to approximately 27–28 °C followed by a marked increase at higher temperatures, consistent with the distributed lag results.

Comparisons between extreme and non-extreme temperature periods further supported this pattern, with higher mortality observed during hotter years (stroke: t = 3.57; myocardial infarction: t = 3.83; both p < 0.01). Bar plots (Figure 5 and Figure 6) illustrated the temporal concentration of deaths during high-temperature months. Although overall mortality exhibited a long-term increasing trend, the relative contribution of deaths occurring during extreme heat months increased recently, particularly after 2015, coinciding with intensified regional warming and a higher frequency of extreme temperature events. This pattern was consistent across both outcomes, reinforcing the robustness of the temperature-mortality association after adjustment for long-term trends.

Sensitivity analyses confirmed the robustness of the temperature-mortality associations for both stroke and myocardial infarction. Effect estimates remained stable after excluding extreme years, varying temporal lags, and simulating potential under- and overreporting (

Table A1. Sensitivity analysis of the associations between mean temperature and mortality from stroke and myocardial infarction in the Brazilian Amazon (2000–2023). The table summarizes robustness analyses conducted to evaluate the stability of the temporal models. Estimates are presented for models excluding extreme years (2000 and 2023), applying alternative temporal lag structures, and adjusting mortality rates by ±10% to simulate potential under- and overreporting. Reported values include regression coefficients (β) and corresponding p-values.

Model	Stroke		Myocardial Infarction
	Β	p	β	p
excluding extreme years (2000 and 2023)	0.02	0.03	0.02	0.002
temporal 1 month lag	0.06	0.02	0.04	0.02
temporal 2 months lag	0.03	0.001	0.01	0.01
temporal 3 months lag	0.04	0.0001	0.02	0.001
underreporting (10%)	0.02	0.01	0.02	0.003
overreporting (10%)	0.03	0.02	0.02	0.003

).

4. Discussion

We observed for the first time in the Brazilian Amazon that rising temperatures are associated with increased mortality from stroke and myocardial infarction, particularly among older adults. This effect was evident above the thresholds of 29 °C for stroke and 30 °C for myocardial infarction. Similar thresholds have been reported in other tropical regions, where temperatures between 28 °C and 31 °C were linked to significant increases in mortality [13,14]. In such settings, the health impacts of heat are often intensified by high humidity, unplanned urbanization, and socioeconomic vulnerability [6,15] conditions are also present in the Amazon.

While long-term increases in stroke and myocardial infarction mortality may be partly associated with demographic aging, shifts in cardiovascular risk profiles, and gradual improvements in death certification and reporting over time, these patterns should not be interpreted as being driven by temperature alone. Importantly, our findings indicate that both interannual and intra-annual variability in temperature remain significantly associated with mortality, even after accounting for long-term temporal trends and seasonal structure. Although overall mortality increased across the study period, the proportion of deaths occurring during months of extreme heat increased recently, particularly after 2015, indicating that climate warming exacerbates preexisting demographic and epidemiologic processes rather than merely paralleling them. Earlier studies reported no association between cardiovascular mortality and high temperatures in the region between 1996 and 2013 [8]. In contrast, our analysis shows that this association has strengthened since 2015, coinciding with a higher frequency of months with elevated temperatures. This pattern is consistent with recent increases in mean temperature and heat extremes across tropical regions, driven by deforestation and more frequent wildfires [7]. Urban expansion and reduced vegetation cover further contribute to heat exposure through the intensification of urban heat islands [1].

Periods characterized by increased cardiovascular mortality are likely associated with the co-occurrence of elevated regional temperatures and multiple concurrent stressors [13,16]. In the Brazilian Amazon, positive thermal anomalies and drought-related conditions may be linked to increased physiological cardiovascular strain [2], while broader contextual factors, such as disruptions in healthcare access, delays in emergency care, and changes in health-seeking behavior during public health emergencies, may coincide with heightened vulnerability [17]. Additional stressors, including air pollution from wildfires and elevated psychosocial stress, may also contribute to short-term mortality fluctuations [18]. Although these overlapping factors may contribute to observed variability, our analyses consistently indicate that temperature patterns show a strong association with mortality risk, supporting the robustness of the temperature-mortality relationship beyond long-term structural changes in health systems or data registration.

Although populations in tropical regions are adapted to high baseline temperatures, additional increases may exceed physiological adaptation limits, increasing mortality risk [19]. Heat-induced responses, including water and electrolyte loss, reduced plasma volume, and circulatory overload, may lead to cerebrovascular and cardiovascular decompensations [20]. Heat stress also increases blood viscosity and thrombotic propensity [3]. In temperate regions, mortality is triggered at lower thresholds (23–26 °C), as observed in southern Brazil and Europe [15,21]. These differences likely reflect contrasts in thermal amplitude and exposure patterns, as tropical populations experience prolonged periods of high temperature rather than short, acute heat episodes. Similar associations between low thermal amplitude and cardiovascular mortality have been documented in cities across Latin America, Asia, and Europe [22,23,24].

Older adults were consistently the most vulnerable group, aligning with findings from Europe, North America, and Asia [25]. Reduced thermoregulatory capacity and limited physiological reserves increase the impact of extreme heat in this population [15]. Although thermal thresholds vary geographically, age-related vulnerability to heat is a global phenomenon [4]. In tropical regions, this vulnerability is further compounded by socioeconomic inequalities and limited access to health services [6], in contrast to temperate countries, which, despite lower adaptation to heat, often benefit from better infrastructure and mitigation measures such as air-conditioned environments and early warning systems [19]. Our findings underscore the urgency of implementing heat mitigation and adaptation strategies, particularly for vulnerable populations in tropical regions.

Despite the robustness of the associations observed, some limitations should be acknowledged. The use of routinely collected mortality data may be influenced by gradual improvements in registration over time, and the monthly temporal resolution limits the assessment of short-term effects observed at daily scales. Nevertheless, adjustment for long-term temporal trends and seasonality, combined with multiple sensitivity analyses, supports the stability of the temperature-mortality associations under alternative assumptions.

5. Conclusions

Rising temperatures are associated with increased mortality from stroke and myocardial infarction in tropical regions, with the strongest associations observed among older adults. Mortality risks increase beyond specific thermal thresholds, highlighting the importance of accounting for local climatic conditions and limited thermal variability in tropical settings. These associations occur within a broader socio-environmental context, in which factors such as unplanned urbanization, deforestation, and socioeconomic inequality may amplify population vulnerability to heat exposure. Together, these findings support the need for public policies that integrate climate adaptation, urban planning, and health surveillance to mitigate the impacts of extreme heat and protect vulnerable populations in tropical regions.