Association between ambient temperature and risk of stroke morbidity and mortality: A systematic review and meta‐analysis

Abstract Background Previous studies have suggested that ambient temperature is associated with the morbidity and mortality of stroke although results among these investigations remained unclear. Therefore, the purpose of present meta‐analysis was to summarize the evidence of the relationship between ambient temperature and stroke morbidity and mortality. Methods A systematic search of the PubMed, Embase, and Web of Science databases was from inception to April 13, 2022. The pooled estimates for heat ambient temperature and cold ambient temperature, which were defined as comparison between extreme hot or cold conditions and the reference or threshold temperature, were calculated utilizing a random‐effects model. A total of 20 studies were included in the meta‐analysis. Results The pooled estimated show that the heat ambient temperature was significant associated with 10% (relative risk [RR], 1.10; 95% confidence interval [95%CI]: 1.02–1.18) and 9% (RR, 1.09; 95%CI: 1.02–1.17) increase in the risk of stroke morbidity and mortality, respectively. In addition, the pooled estimated show that the cold ambient temperature was significant associated with 33% (RR, 1.33; 95%CI: 1.17–1.51) and 18% (RR, 1.18; 95%CI: 1.06–1.31) increase in the risk of stroke morbidity and mortality, respectively. Conclusion Integrated epidemiological evidence supports the hypothesis that both heat and cold ambient temperature have positive association with the risk of stroke morbidity and mortality. Targeted measures should be promoted in public health to reduce this risk.


INTRODUCTION
Stroke is characterized by brain tissue damage due to a blockage of a blood vessel or a sudden rupture of a blood vessel, including ischemic stroke and hemorrhagic stroke, with high morbidity, mortality, and disability rates (Collaborators GCoD, 2018;Collaborators GDaIIaP, 2018).
Stroke is currently still the second leading cause of death worldwide according to the World Health Organization 2020 report (Rana et al., 2021). Stroke not only is a public health issue but also contributes to a massive economic burden (Katan & Luft, 2018). Up till the present moment, there were approximately 34% of the global total health care expenditure spent on stroke. The economic burden of stroke per person in terms of cost of illness method was approximately equal to USD 1809.51-325,108.84, including hospitalization, rehabilitation exercises, and follow-up long-term care (Rochmah et al., 2021). With the growing problems of population aging around the world, the global health care and economic burden of stroke are expected to continue to multiply in the coming years (Feigin et al., 2014). Moreover, previous studies from the Global Burden of Disease Study showed that approximately 90% burden of stroke could be attributed to lifestyle and environmental risk factors Feigin et al., 2016;Finucane et al., 2011). Therefore, we need a clearer comprehension of the various pathways from these risk factors to stroke, and every measurement must be taken to minimize the outcomes of stroke.
In the context of global climate change, the increased frequency, intensity, and duration of extreme temperature due to changes in ambient temperature and global warming have emerged as a public health threat, especially for populations with scant physiological ability or socioeconomic status to cope with or adapt to these changes. It contributes to a burden of disease and is set to move up in the coming years (Collaborators et al., 2018;McMichael, 2013). At the moment, the factors that trigger stroke have become the focus of attention and discussion in recent years to cope with changes that pose a fundamental threat to human well-being and health, with published studies supported an association between extreme ambient temperature and an increased risk of stroke morbidity or mortality (Roye et al., 2019;Wang & Lin, 2014).
Although several studies have investigated the relationship between ambient temperature and the risk of stroke, the evidence has been conflicting (Breitner et al., 2014;Guo et al., 2017;Hong et al., 2003;Ikefuti et al., 2018;Wang & Lin, 2014). Although previous meta-analysis explored the pooled relative risk (RR) of stroke for temperature changes, there were some limitations in the results (Wang et al., 2016). First, this studies only reported the relationship among daily mean temperature, daily maximum and minimum temperature, and stroke with only eight studies included but did not investigate the relationship between specific ambient temperature changes and stroke, such as above or below a threshold temperature or reference temperature. Equally important, they did not separately report the effect of temperature changes on the stroke morbidity and mortality, due to a limited number of original studies were included. In addition, there were many new studies published, but the results have been reported inconclusive (Schulte et al., 2021;Zhan et al., 2022).
Considering the results on the influence of ambient temperature upon stroke have been inconsistent, we update those studies with a focus on ambient temperature connection with and stroke by pooling evidence from relevant epidemiological studies. In view of the increasingly serious climate change and the heavy burden caused by stroke, the aim of present study is to provide valuable information for primary prevention of stroke and reduction of global health care expenditure.

METHODS
The present meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and the Meta-Analysis of Observational Studies in Epidemiology Group (MOOSE) guidelines (Page et al., 2021;Stroup et al., 2000).

Search strategy
Literature searches were conducted in the PubMed, Embase, Web of Science databases. We conducted a literature search from the establishment of the database to April 13, 2022, and limited to studies on humans, published in English, with the following keywords: "temperature" or "heat effect" or "cold effect" or "hot exposure" or "cold exposure" or "extreme weather" or "climate" and "ischemic stroke" or "hemorrhagic stroke" or "cerebrovascular accident" or "cerebrovascular disorders." The detailed search strategy terms are provided in Table   S1.

Eligibility criteria
In present meta-analysis, we used the following inclusion criteria that met the PECOS approach: (1) population: total population; (2) exposure: heat ambient temperature was defined as extreme heat con-

Study selection
We screened and identified the studies which met the eligibility criteria in a step-by-step manner. First, we screened the title and abstract

Data extraction
Eligible articles were extracted independently in duplicate, and all reconciled by two authors (J.W. and Y.L.

Quality assessment
The Newcastle-Ottawa scale (NOS) was used to evaluate the quality of time-series and case-crossover studies, respectively. The scale score was calculated based on the three sections including study group selection (4 points), comparability of groups (2 points), and the exposures or outcomes (3 points) (Higgins et al., 2003). Quality was graded as low (0-3 points), moderate (4-6 points), and high (7-9 points), with higher points indicating higher quality.

Data analysis
We performed a meta-analysis to pool the estimates of RR from all included original epidemiological studies to measure the association between heat or cold ambient temperature effect and risk of stroke.
As the absolute risk of stroke is less than 10%, the reported OR was considered approximately as RR (Egger et al., 1997). Since the heterogeneity from the study design, different methods of ambient temperature measure, different temperature threshold or reference, geographical location, and lag pattern between original studies, the pooled estimates were evaluated using the random-effects model when the index of heterogeneity (I 2 ) was >25%; or otherwise, the fixedeffects model was applied (Begg & Mazumdar, 1994). We calculated the pooled maximum estimate from each original study. When different separate risk estimates were given for different geographic locations in a study that did not provide an overall risk estimate, we combined the results of these estimates to calculate the overall risk estimates for meta-analysis.
Heterogeneity across each study was valued by the Q statistic quantified and the I 2 statistic. In present study, statistical heterogeneity was expressed by a significant Q statistic at the p < .10. In addition, I 2 values were used to describe the degree of heterogeneity, with 25%, 50%, and 75% indicating low, moderate, and high heterogeneity, respectively (Higgins et al., 2003).
We performed subgroup analysis, sensitivity analysis, and publication bias assessment to investigate the possible heterogeneity and robustness of our meta-analysis. We conducted subgroup analyses according to the potential influence of some factors, such as study design, socioeconomic status, latitude, and whether the studies controlled for relative humidity, air pollution, day of week, holiday, and long-term and seasonal trend. The impact of individual studies on the pooled estimates was confirmed by conducting a sensitivity analysis using the leave-one-out method. Additionally, Egger's regression test, Begg's funnel plot, and the trim-and-fill method were used to test potential publication bias (Begg & Mazumdar, 1994;Duval & Tweedie, 2000;Egger et al., 1997).
All statistical analyses were conducted using Stata 16.0 (Stata Corporation, College Station, TX, USA), and p < .05 (two-sided) was deemed statistically significance.

Study selection
According to the systematic search, a total of 9847 articles were identified, of which 3326 were removed because of duplication. Of those, 6459 were removed after review of the title and abstract, because they were animal studies, clinical trials, letters, commentaries, reviews, biomechanics, or not relevant studies. Then, a total of 62 articles were excluded after a full-text screening because exposure was not heat/cold effect, the outcomes were not stroke, or information were The PRISMA flowchart is shown in Figure 1.

TA B L E 2 Characteristics of studies included in the review
Author

Study descriptions
A total of 20 articles were included, and 12 included studies were conducted in Asian, 6 included studies were conducted in Europe and North American, and only one was conducted in South American and Australia, separately. Of these, 15 used a time-series study design, and 5 used a case-crossover study design, which published between 2003 and 2022. In addition, the average quality scores of time-series studies were 8, ranged from 7 to 9, and the average quality scores of case-crossover studies were 7.2, ranged from 7 to 8. In terms of included temperature exposure type and health outcome, 10 reported the heat ambient temperature effect of stroke morbidity, 11 reported the cold ambient temperature effect of stroke morbidity, 7 reported the heat ambient temperature effect of stroke mortality, and 7 reported the cold ambient temperature effect of stroke mortality. The daily mean temperature and daily maximum and minimum temperature were most commonly used as measures in the included studies, although some studies used apparent temperature. Temperature for all included studies ranged from −33.1 to 40.9 • C. The lag effects of temperature were deemed among most studies ranged from 0 to 28 days, as heat ambient temperature exposures were immediate effect, and cold ambient temperature exposures were delayed effect. The most common method used to examine the relationship between ambient temperature and stroke morbidity or mortality was time-series using distributed lag nonlinear models, whereas some studies were case-crossover design using conditional logistic regression models.

Ambient temperature and risk of stroke morbidity and mortality
In the categorical meta-analysis, the pooled estimates of heat ambient temperature effect on stroke morbidity and mortality and cold ambient temperature effect on stroke morbidity and mortality were shown separately (Figures 2-5). We found that increase or decrease in temperature was positively correlated with the risk of morbidity and mortality from stroke compared to the reference or threshold value for temperature. For heat ambient temperature effect, the pooled RR of stroke morbidity was 1.10 (95%CI: 1.02-1.18; I 2 = 74.2%), and the pooled RR of stroke mortality was 1.09 (95%CI: 1.02-1.17; I 2 = 85.6%).

Subgroup analyses
Aim to evaluate the stability of the primary pooled estimates results and identify the possible resources of heterogeneity, subgroup analyses were conducted according to study design, socioeconomic status, latitude, and controlling for relative humidity, air pollution, day of week, holiday, long-term, and seasonal trend, respectively. In general, the results were similar across the subgroups in that heat or cold ambient temperature increased the risk of stroke. When stratified by socioeconomic status, we found that stroke morbidity and F I G U R E 2 Forest plot of stroke morbidity for the heat ambient temperature.
F I G U R E 3 Forest plot of stroke morbidity for the cold ambient temperature.

F I G U R E 4
Forest plot of stroke mortality for the heat ambient temperature.

F I G U R E 5
Forest plot of stroke mortality for the cold ambient temperature. for cold ambient temperature effect. The heterogeneity was 0.0% and 10.0% for heat and cold ambient temperature effect that did not control the long-term and seasonal trend, respectively. The result of the subgroup analysis is in Table S2.

Sensitivity analyses
Sensitivity analyses suggested that neither the study was likely to have a significant impact on the result for heat ambient temperature effect on stroke mortality, nor cold ambient temperature effect on stroke morbidity and mortality ( Figure S1). For heat ambient temperature effect, the pooled RR ranged from 1.

Publication bias
For heat ambient temperature effect on stroke morbidity and mor-

DISCUSSION
This meta-analysis supplies an updated examination of the effects of various ambient temperature conditions. To the best of our knowledge, this is the first study to assess not only the association between heat and cold ambient temperature effects and stroke morbidity, but also stroke mortality, which fill the gaps in previous studies. More specifically, present meta-analysis found that the risk of stroke morbidity and mortality significantly increased in relation to heat and cold ambient temperature effect. Our findings have important implications for developing the policy of public health and reducing the burden of stroke.
The combined estimate of heat exposure on elevated risk of stroke morbidity and mortality, which is accordant with the previous metaanalysis in which each 1 • C increase was associated with 1.1% increase in occurrence and mortality of stroke (Lian et al., 2015). However, another meta-analysis indicated no significant association between ambient maximum temperature and ischemic stroke occurrence (RR: .98, 95%CI: .94-1.02). The inconsistent results may be due to the fact that this meta-analysis study reported the relationship between maximum temperature and ischemic stroke occurrence, which temperature as a continuous variable, but in our study ambient temperature as a categorical variable, heat ambient temperature was compared to reference or threshold temperature. Moreover, there were only two original studies with greater chance in previous meta-analysis (Wang et al., 2016).
The increased risk of stroke morbidity and mortality with heat ambient temperature effect may be explained by some potential mechanisms. One possible explanation is that in heat conditions the body needs to increase heat diffusion through sweating, which may lead to dehydration and a reduction in central blood volume. This further leads to increased blood viscosity, altered serum osmolality, and altered BUN/cr, which may increase the risk of microvascular thrombosis and subsequent ischemic stroke (Bahouth et al., 2018;Donaldson et al., 2003;Keatinge et al., 1986;Qi et al., 2021). In addition, high temperatures may damage the endothelial cells of blood vessels and lead to coagulation dysfunction, and consequently to consumptive coagulation. This consumptive coagulation leads to excessive and prolonged multiple tissue hemorrhages resulting in multiple organ dysfunction syndrome, including cerebral hemorrhages (Lavados et al., 2018). At In our meta-analysis, the combined estimate of cold ambient temperature exposure on elevated risk of stroke morbidity and mortality was significant. Our assessment of the relationship between cold ambient temperature effect and stroke was in-line with similar previous meta-analyses (Lian et al., 2015). For example, the meta-analysis showed that a 1 • C decrease in temperature resulted in a 1.2% increase in stroke incidence and mortality. However, our results are inconsistent with the results of another meta-analysis pooled effect size of only two original studies (Wang et al., 2016), possibly due to different types of temperature variables and chance from fewer original studies. Therefore, more epidemiology studies are required to provide more evidence on the association between ambient temperature and stroke morbidity or mortality.
Several physiological mechanisms by which cold ambient temperature effect might act as a trigger for stroke have been suggested to explain this phenomenon. Exposure to cold ambient temperatures may lead to activation of the sympathetic nervous system and the reninangiotensin system and an increase in blood pressure, which may lead to plaque rupture and thus an increased risk of ischemic stroke (Barnett et al., 2007;Ryti et al., 2017;Sun, 2010). In addition, cold-induced vasoconstriction leads to reduced blood flow to the skin, whereas increased diuresis leads to concentrated and hyperviscous blood and increased fibrinogen and cholesterol levels, which are also a risk factor for the formation of blood clots (Keatinge et al., 1984;Liu et al., 2015;Neild et al., 1994). Moreover, hypothermia-mediated changes in hematological properties can enhance forces favoring wall deformation and increase frictional and shear stresses on the inner surface of the vessel, which may increase the risk of hemorrhagic stroke (Vitale et al., 2015).
As substantial heterogeneity was found among the included studies, we conducted subgroup analyses by study design, socioeconomic status, latitude, and controlling for relative humidity, air pollution, day of week, holiday, long-term and seasonal trend. In subgroup analyses, we found whether air pollution and long-term and seasonal trends were controlled may be sources of heterogeneity. Besides, we found that heterogeneity remained substantial which may be related to numerous factors. First, exposure measurement of studies was from different sources with varied definitions. For example, some studies use apparent temperature, including daily average air temperature, relative humidity, wind speed, and vapor pressure, which is different from ambient temperature (Roye et al., 2019;Zhan et al., 2022 Aim to access the robustness of our result, we conducted a publication bias assessment and a sensitivity analysis. Publication bias among studies which accessed the association between heat and cold ambient temperature effect and stroke morbidity and mortality were found in our study. Therefore, we used the trim-and-fill method to assess publication bias. We found that the trim-and-fill method could change our results, probably due to the small number of original studies. According to the principle of symmetry, multiple non-existing small sample size studies were supplemented, and the combined effect size was calculated on this basis, resulting in the appearance of overcorrection (Egger & Smith, 1998). However, we verified that the association between heat and cold ambient temperature effect and stroke morbidity and mortality was positive and statistically significant by sensitivity analysis, even though the results of some studies were not statistically significant (Goggins et al., 2012;Guo et al., 2017;Luo et al., 2018;Schulte et al., 2021).
The review highlighted the need for research to identify and study interventions in public health to reduce the problem of increased stroke morbidity and mortality due to extreme ambient temperatures, particularly in vulnerable populations. First of all, all countries are advised to take environment-related actions, such as reducing greenhouse gas emissions. Developed countries in particular should pay more attention to the public health issues arising from greenhouse gas emissions. The subgroup analysis showed that the risk of increased stroke morbidity and mortality associated with heat effect is higher in developed countries than in developing countries, which may be related to the higher greenhouse gas emissions in developed countries. Second, the health care system should also take action to deal with the health problems caused by extreme temperatures. For example, an early warning system for extreme temperatures should be set up, with extreme temperature warnings issued through television, radio, and the internet, so that the public, especially vulnerable people, can receive timely weather information and take measures to protect themselves. In addition, special adaptation measures and health promotion methods can also be developed for the characteristics of different vulnerable groups, which can lead to more effective results.
For example, when extreme temperature warnings are issued, health care system such as community hospitals can take the initiative to give health guidance and medication advice to elderly people with chronic illnesses, especially those living alone and suffering from neurological disorders or a history of stroke, using telephone visits or home visits to ensure that they stay in air-conditioned environments as much as possible during extreme temperature and seek prompt medical attention if they have any health problems. During a COVID-19 pandemic, the health care system can take measures for smart care, such as telemedicine and smart homes, to reduce the health risks associated with COVID-19-related isolation.
There are several highlights in our study. First, we provided updates on the associations of ambient temperature with the risk of stroke, including more than one times as many studies as for a meta-analysis of pooled effect sizes than previous meta-analysis. Moreover, the selection of studies was based on strict criteria for eligibility and exclusion, which helped provide more reliable evidence. Second, in contrast to previous study, this is the first studies covered four outcomes, allowing comparisons among these outcomes and a more comprehensive assessment of ambient temperature changes effect on stroke morbidity and mortality separately. Third, despite heterogeneity among the included studies, consistent results from sensitivity analyses and various subgroup analyses suggest that our findings are reliable and robust.
However, in our studies there were some limitations should be of concern. First, we included both time-series studies and case-crossover studies. Although we used NOS for assessing the quality of included studies, which is considered the most effective standard. There are some limitations in proving causation of disease, which may influence the validity of our results (Mendoza & Yeh, 2021). Future multicenter, long-term longitudinal studies are necessary to further investigate the effect of ambient temperature on the risk of stroke. Second, there was a high degree of heterogeneity between the included studies. Future investigations on a more precise temperature measurement standards or reference values are merited, and research in countries of different socioeconomic status would be worthwhile, particularly in developing countries.

CONCLUSIONS
The inspiration of our meta-analysis clearly suggests that heat or cold ambient temperature are associated with an increased risk of stroke morbidity and mortality. This study provides a potential mechanism for the association of ambient temperature changes on stroke, and an understanding of the broader effects of extreme ambient temperature on stroke has the potential to suggest enhancements to public messaging efforts. As well as authorities should be aware of the seriousness and provide interventions that might mitigate the adverse effects of extreme ambient temperature, which everyone can benefit from these.

CONFLICT OF INTEREST STATEMENT
The authors have no relevant financial or nonfinancial interests to disclose.

DATA AVAILABILITY STATEMENT
Some of the data and materials used in this study can be found in the Supporting Information section. More details can be obtained from the corresponding author (Jing Mao).