Associations between short-term exposure to ambient temperature and renal disease mortality in Japan during 1979–2019: A time-stratified case-crossover analysis

Abstract

Background:

Previous studies have indicated that renal disease mortality is sensitive to ambient temperatures. However, most have been limited to the summer season with inconclusive evidence for changes in population vulnerability over time.

Objective:

This study aims to examine the association between short-term exposure to ambient temperatures and mortality due to renal diseases in Japan, and how this association varied over time.

Methods:

We conducted a two-stage, time-stratified case-crossover study from 1979 to 2019 across 47 prefectures of Japan. We obtained the data of daily mortality counts for all renal diseases, acute renal failure, and chronic renal disease. We fitted a conditional quasi-Poisson regression model with a distributed lag nonlinear model. A random-effects meta-analysis was applied to calculate national averages. We performed additional analyses by four subperiods, sex, and age groups.

Results:

We analyzed 997,590 renal mortality cases and observed a reversed J-shaped association. Lower temperatures were associated with increased mortality in all renal disease categories. The cumulative relative risks at 2.5th percentile compared to the minimum mortality temperature percentile were 1.34 (95% confidence interval [CI] = 1.29, 1.40), 1.51 (95% CI = 1.33, 1.71), and 1.33 (95% CI = 1.24, 1.43) for all renal, acute renal failure, and chronic renal disease mortality, respectively. The associations were observed in individuals of both sexes and aged 65 years and above. The associations of kidney mortality with low temperature remained consistent, while the associations with high temperature were pronounced in the past, but not in recent periods.

Conclusions:

Protection for individuals with impaired renal function from exposure to low temperatures during cold seasons is warranted.

What this study adds

This study reveals a clear link between low temperatures and kidney disease mortality in Japan, analyzing 997,590 mortality cases over 41 years period. Unlike previous research, this study focuses on year-round temperature distribution and shows that the risk for cold is consistent over time. Furthermore, this study analyzes the major disease subcategories of renal disease mortality, namely acute renal failure and chronic renal disease. The findings emphasize the need to protect those with kidney issues from cold temperatures. This research provides essential insights that could guide health care policies and personal care strategies for people living with kidney diseases.

Introduction

Renal diseases are among the leading causes of death globally with an estimated 1.43 million deaths in 2019.¹ The burden of renal diseases has increased over time in comparison to other noncommunicable diseases. According to the recent Global Burden of Disease study, the ranking for chronic kidney diseases has increased from the 29th in the 1990s to the 18th in 2019 as one of the leading causes of disability-adjusted life years.² In Japan, a similar trend is observed where the prevalence of age-adjusted chronic kidney disease has increased from 57.8 per 1000 population in 2005 to 71.8 in 2017.²

Ambient temperature is a well-identified risk factor for morbidity and mortality.^3,4 High temperatures have been associated with an increased risk of hospitalization from renal diseases, suggesting a sensitivity to heat.^5,6 However, the association of temperature with mortality from renal diseases is less understood. Existing evidence remains inconclusive owing to the limited number of epidemiological studies and heterogeneous results. Most studies thus far have focused on the warm seasons and summer heat,⁷ while very few have examined the effects of low temperatures in cold seasons. For example, two independent studies focusing on heat exposure in the United Kingdom and Italy reported an increased risk of mortality due to renal causes.^8,9 On the other hand, a time series study conducted in Hong Kong did not observe any associations of mortality with heat and cold.¹⁰ These mixed findings suggest possible regional variations.¹¹ The studies also did not distinguish between the two subcategories of renal disease, namely acute renal failure and chronic renal disease, which have different clinical features and disease progression that warrant further investigations.¹²

In Japan, the effects of daily temperature on renal disease mortality are not well understood. Although heat-related mortality has decreased substantially in the country over the last few decades,^13–16 it is not clear if renal disease mortality follows the same temporal decline. It is also unclear if low temperatures have an impact, and if they do, the magnitude in comparison to high temperatures, as well as the temporal change of the risk over time, remain uncertain. In addition to these reasons, the growing disease prevalence, an aging population, and the continuous warming of the climate underscore an urgency to fully characterize the temperature–mortality associations for renal diseases in this population.

The present study aims to examine the association between short-term exposure to ambient temperature and renal disease mortality in Japan. The study examines both high and low temperatures including their delayed effects by applying a flexible method based on the distributed lag nonlinear model. In addition to all renal diseases, the investigation is extended to include mortality from acute renal failure and chronic renal disease. The study also examines the variation of the associations over 4 decades from 1979 to 2019 and explores effect modifications by age and sex.

Methods

Data

We obtained daily count of mortality from all renal diseases (International Statistical Classification of Diseases and Related Health Problems 9th and 10th Revision [ICD-9]: 580–599; ICD-10: N00–N39), acute renal failure (ICD-9: 584; ICD-10: N17), and chronic renal disease (ICD-9: 585; ICD-10: N18) from the Ministry of Health, Welfare and Labor, Japan. ICD-9 and 10 codes were coded based on the renal diseases as the primary cause of death.¹⁷ The data covered the period from 1 January 1979 to 31 December 2019 and spanned all 47 prefectures in Japan. We stratified the data by three age groups (0–64, 65–84, and ≥85 years old) and sex. Individual information was unidentifiable as the observations were aggregated by day and across prefectures.

We collected daily 24-hour average temperatures from the Japan Meteorological Agency. Data for the study period (1979–2019) were obtained from a weather station located in the capital city of each prefecture.

Statistical analysis

We conducted a two-stage analysis. First, we estimated the temperature–mortality association in each prefecture. Next, the prefecture-specific estimates were pooled in the second stage via a meta-analytical approach to obtain a national average effect estimate.

First-stage analysis

We used a time-stratified case-crossover study design^18,19 and fitted a conditional quasi-Poisson regression model with distributed lags nonlinear model to assess the temperature–renal mortality association in each prefecture.²⁰ This design allows each case to act as its own control, with control periods selected from the same day of the week of the same month and year. The exposure status of a case at the time of the event (case period) is then compared to its own exposure during control periods. The self-control approach adjusts for the time-invariant characteristics by design. Adjustments for long-term trends, seasonality, and the day of the week were achieved through time stratification.¹⁸ The model has the general form:

E(Y_t,s) = exp{α_s + β^T (cbtemp_t,l) + holiday} (1)

where E(Y_t,s) indicates the expected number of deaths on day t in stratum s, α_s is a vector denoting strata 1 to s, which represents an interaction term between year, month, and the day of the week for time stratification,¹⁸ β is a vector of coefficients for cbtemp_t,l, which represents a cross-basis function of temperature percentile on day t with l number of lag days, and holiday is an indicator for the public holiday. The cross-basis function allows for the estimation of two nonlinear functions at the same time, namely the temperature–mortality function and the temperature–lag–mortality function.²¹ For the former, we applied a quadratic B-spline with two equal-spaced internal knots, while for the latter, we constructed a natural cubic spline with three equal-spaced internal knots on a log scale using a lag period of 21 days. The selection of knots and lag periods in the final model was guided by the Quasi-Akaike Information Criterion¹⁹ via sensitivity analysis. We analyzed temperature percentiles instead of their absolute values to accommodate the slightly lower temperature range in northern Japan.²² To evaluate the effects of high and low temperatures, we calculated the cumulative relative risks (RRs) over 21 lag days and the corresponding 95% confidence intervals (CIs). We defined heat and cold based on a previous literature.¹¹ For heat, the RR compares the risk of mortality at the 97.5th percentile of daily mean temperature to the minimum mortality temperature percentile (MMP), the baseline where risk is the lowest. For cold, the 2.5th percentile was compared with MMP. We repeated the above analyses by age and sex subgroups, and for two subcategories of renal diseases (acute renal failure and chronic renal disease).

Second-stage analysis

We used a random-effects multivariate meta-analytical method to pool the prefecture-specific associations to obtain a national average estimate.²³ Then, for each prefecture, we generated the best linear unbiased prediction of the overall cumulative association from the fitted meta-analytical model. The MMP in each prefecture was derived from their corresponding best linear unbiased prediction.^11,24 We restricted the MMPs between the 10th and the 90th percentiles and used them to recenter the exposure–response curves for baseline comparison. The heterogeneity of effect estimates was assessed using Cochran Q test and I² statistics (Table S7; http://links.lww.com/EE/A259).²⁵

We calculated the attributable fractions of renal disease mortality due to cold and heat to examine the mortality burden attributable to non-optimal temperatures. The attributable fraction was obtained by dividing the number of deaths attributable to cold or heat by the total number of renal disease deaths.²⁶ Empirical CIs were computed using Monte Carlo simulations with 1000 replicates.²⁶

Temporal variation

We stratified the data into four subperiods with 10 years each (except for the last subperiod, which has 11 years) and repeated the first- and second-stage analyses to examine changes in the association between temperature and renal disease mortality across the study period. We checked the statistical significance of pairwise comparisons by comparing the estimate in each subperiod to the first.²⁷

Sensitivity analyses

We checked the robustness of our results to changes in the duration of the lag period (7, 14, or 28 days), the positioning, and the number of knots in the temperature–mortality spline function (at the 10th, 50th, and 90th percentiles, or at the 25th, 50th, and 75th percentiles), and the additional adjustment for relative humidity represented as a natural cubic spline with 3 degrees of freedom.

Results

This study examined a total of 997,590 renal deaths over 41 years. Table 1 shows the distribution of daily mean temperature in Japan, while the summary by prefectures is shown in Table S1; http://links.lww.com/EE/A259, and the distribution by case and control days is shown in Table S8; http://links.lww.com/EE/A259. Among all renal disease deaths, 13.8% were classified as acute renal injury and 38.5% as chronic renal disease (Table 2). There were slightly more females (54.8%), and most cases were 65–84 years of age (48.0%), followed by 85 years or older (43.1%).

Table 1.

Descriptive statistics of daily mean ambient temperature across 47 prefectures of Japan

Year	Ambient temperature (°C)
	Mean ± SD	Minimum	2.5th	25th	75th	97.5th	Maximum
1979–2019	15.3 ± 2.3	−3.7	1.8	7.8	22.4	28.9	32.1
Subperiods
1979–1988	14.6 ± 2.4	−3.7	0.9	7.0	21.8	28.0	30.7
1989–1998	15.3 ± 2.3	−1.8	2.4	8.0	22.1	28.7	31.7
1999–2008	15.5 ± 2.4	−1.8	2.2	8.0	22.8	28.9	31.5
2009–2019	15.7 ± 2.3	−2.1	2.0	7.9	22.9	29.4	31.9

SD indicates standard deviation.

Table 2.

Overall cumulative relative risk and attributable risks of renal disease mortality for cold and heat over 47 prefectures in Japan

Mortality outcome	Total number of deaths, N (%)	MMP (°C)	RR cold (95% CI)	RR heat (95% CI)	Attributable fractions
					Cold	Heat	Total
All renal	997,590 (100.0)	87.2 (25.6)	1.34 (1.29, 1.40)	1.01 (0.99, 1.03)	11.11 (9.95, 12.13)	0.10 (0.02, 0.18)	11.21 (10.05, 12.21)
Acute renal failure	137,700 (13.8)	82.0 (24.2)	1.51 (1.33, 1.71)	1.06 (0.97, 1.16)	18.90 (14.90, 21.16)	1.70 (1.00, 2.16)	20.60 (16.71, 22.91)
Chronic renal disease	383,592 (38.5)	90.3 (26.5)	1.33 (1.24, 1.43)	0.99 (0.96, 1.02)	14.91 (12.70, 16.43)	0.11 (−0.01, 0.22)	15.02 (12.90, 16.48)

Figure 1 illustrates the overall cumulative RRs of mortality from all renal diseases, acute renal failure, and chronic renal disease. We observed a reverse J-shaped association of mortality due to all renal diseases and acute renal failure with ambient temperature. The MMPs were 24.2 °C (acute renal failure) to 26.5 °C (chronic renal disease) (Table 2). Below the MMP, the risk of mortality increases as temperature decreases. For temperatures higher than the MMP, the associations were less evident. The estimated associations by prefecture are shown in Figures S1, S8–S10; http://links.lww.com/EE/A259.

Figure 1.

Overall cumulative relative risks of (A) all renal, (B) acute renal failure, and (C) chronic renal disease mortality during 1979–2019 in Japan. Vertical dotted lines from left to right indicate the temperature at the 2.5th percentile, minimum mortality temperature percentile, and the 97.5th percentile.

Figure 2 shows the lag response curves for cold and heat at the 2.5th and 97.5th percentile of temperature distribution, respectively, for all renal diseases, acute renal failure, and chronic renal disease. Association with low temperature was delayed for approximately 2 weeks for all three disease categories. However, a risk of all renal disease mortality increased 1 day after exposure to high temperatures. The corresponding lag-specific RRs are shown in Table S2; http://links.lww.com/EE/A259.

Figure 2.

Predictor-specific lag response at 2.5th percentile of mean temperature for cold effect for (A) all renal, (B) acute renal failure, and (C) chronic renal disease mortality and predictor-specific lag response at 97.5th percentile of mean temperature for heat effect at (D) all renal, (E) acute renal failure, and (F) chronic renal disease mortality.

The overall cumulative RRs for cold and heat and their corresponding attributable fractions are shown in Table 2. Cold was associated with a higher risk of all renal disease-related mortality with an RR of 1.34 (95% CI = 1.29, 1.40). The same was observed for the other two subcategories of renal disease (RR: 1.51 [95% CI = 1.33, 1.71] for acute renal failure; RR: 1.33 [95% CI = 1.24, 1.43] for chronic renal disease, respectively). High temperature was not associated with the risk of mortality in any disease categories. An estimated 11.21% (95% CI = 10.05%, 12.21%) of all renal disease-related deaths were attributable to nonoptimal temperatures, with low temperatures accountable for most of the burden.

Subgroup analysis by age and sex showed comparable results for low temperatures, except for the youngest population aged 0–64 years old who showed no evidence of association. High temperatures did not show any association in general (Figure S2; http://links.lww.com/EE/A259).

Figure 3 illustrates the overall cumulative exposure–response curves by four subperiods for all renal disease-related mortality. The estimated association of low temperatures was comparable over time, except for mortality due to chronic renal disease, which showed a risk increase in the recent subperiods (Figure 4A and Table S3; http://links.lww.com/EE/A259). High temperature was associated with a higher risk of mortality for all renal diseases in the earliest subperiod 1979–1988. The same was observed for chronic renal disease, but not for acute renal failure (Figure 4B and Table S3; http://links.lww.com/EE/A259). The estimated association of heat disappeared in more recent subperiods, while MMPs had increased in general over time (Table S3; http://links.lww.com/EE/A259). Changes over time in the corresponding attributable fractions are presented in Table S6; http://links.lww.com/EE/A259.

Figure 3.

Associations between ambient temperature and all renal disease mortality in four subperiods as (A) overall cumulative relative risks of all renal disease mortality with 21-day lag period and predictor-specific lag response curve at (B) the 2.5th percentile mean temperature for cold effect and (C) the 97.5th percentile mean temperature for heat effect.

Figure 4.

Predictor-specific overall cumulative relative risks at (A) the 2.5th percentile mean temperature for cold effect and (B) the 97.5th percentile mean temperature for heat effect with 21-day lag period by all renal and disease subcategories, age, and sex groups for four different subperiods. All indicates all renal mortality; ARF, acute renal failure; CRD, chronic renal disease mortality.

Results by sex and age show that the risk of cold had increased among females in the more recent period (2009–2019), while the risk of high temperature had reduced among individuals ≥65 years old and females (Figure 4 and Table S4; http://links.lww.com/EE/A259).

The results of the sensitivity analyses based on all renal disease-related mortality were consistent despite the different knot placements and lag periods and are robust to humidity adjustment (Figures S3–S7; http://links.lww.com/EE/A259 and Table S5; http://links.lww.com/EE/A259).

Discussion

We examined the association between short-term exposure to ambient temperature and mortality related to all and cause-specific renal diseases in Japan. To our knowledge, this is the first countrywide study to utilize 4 decades of data to examine this association and its changes over time. Cold temperatures were associated with an increased risk of mortality from renal diseases, including acute renal failure and chronic renal disease. This increased risk had a delayed effect of approximately less than 2 weeks and was observed in both sexes and among older populations aged 65 years and above. The effect of cold remained relatively unchanged over time, except for females who recorded a higher risk of renal disease mortality in more recent periods. On the other hand, the effects of heat changed over time; heat was associated with an increased risk in the early study period, notably among females and older populations aged 65 years and above, but the risk diminished in more recent times.

The observed effect of cold temperatures on kidney is consistent with a previous study that analyzed data from the Global Burden of Diseases and found that the age-standardized mortality rate for chronic renal disease attributable to cold temperatures has been increasing over time.²⁸ A multicountry study also reported that, compared to high temperatures, the risk of all-cause mortality was higher for low temperatures worldwide.¹¹ Despite these findings, the existing literature has mainly focused on the biological mechanisms of heat-related kidney injury, while little is known about the underlying biological processes that contribute to renal disease mortality in cold climates. There could be three possible explanations for the association between cold and renal disease mortality. First, dehydration during the cold season could provide further understanding. A study conducted in Seoul investigated the impact of outdoor temperature on dehydration markers associated with cardiovascular dysfunction.²⁹ The results showed a U-shaped relationship between the markers and temperature, similar to the association observed between cardiovascular mortality and both high and low temperatures, suggesting dehydration could also occur in the cold season. Since kidney function is regulated by renal blood flow and is closely linked to the cardiovascular system, the role of dehydration on kidney disease mortality during the cold season as observed in our study should be further investigated. Second, seasonal variation in electrolyte imbalance could also offer a possible explanation. A single-center study in Japan found that the incidence of hyperkalemia was higher in winter, particularly among individuals with reduced glomerular filtration rates.^30,31 This can result in sudden death among kidney disease patients. This is consistent with previous studies that have demonstrated that organ failures, particularly acute kidney injury, are more prevalent in winter,³² and among older populations.³³ Mortality rates for individuals with end-stage kidney diseases are also higher during the winter season.³⁴ Third, hypertension has been reported to be more common in winter periods.^35,36 For individuals with renal disease, the kidneys’ compromised function struggles to regulate blood pressure effectively. Cold weather induces vasoconstriction, narrowing blood vessels, and raising blood pressure levels, which can further strain the kidneys and accelerate disease progression, and can result in adverse mortality outcomes.

Throughout our 4-decade study period, we did not observe significant changes in the risk associated with cold temperatures, except for an increased risk of chronic renal disease mortality and among the female population. This finding is consistent with previous studies on all-cause and cardiovascular mortality, which also reported that the risk of cold-related mortality remained constant over time.¹⁵ Despite the population in Japan acclimatizing to higher temperatures in recent years, there was generally no significant reduction in the risk associated with cold temperatures.^14,15 Our results suggest that, even though the MMP increased, the population’s vulnerability to cold weather persisted despite the warming climate. The increased risk for chronic renal disease mortality when exposed to cold temperatures also suggests that older populations with chronic conditions are at risk. In addition, our findings also indicate that there may be a maladaptation to cold temperatures among the female population. This finding may be owing to different physiological constitutions considering the hormonal factors.^37,38 Furthermore, it has been reported that females with concurrent disease conditions such as diabetes, cardiovascular, and chronic renal diseases were at higher risk of mortality compared to males.³⁹ Due to demographic changes in the Japanese population over time, elderly females may have become more vulnerable to the risks associated with exposure to cold temperatures in recent decades. However, we did not observe significant changes in cold risk for acute renal failure suggesting that chronic conditions are more susceptible to cold temperatures.

We observe a significant association between heat and renal disease mortality only in the early decades of our study period, specifically from 1979 to 1988 for mortality from all renal diseases and chronic renal diseases, and from 1989 to 1998 for mortality from acute renal failure. However, this significant association diminished in more recent periods. This trend was observed among females and the older population aged 65 years and above. These findings support previous evidence that acute kidney injury is susceptible to high temperatures.⁴⁰ The kidney is a vascular organ that maintains the body’s water and electrolyte balance, as well as its acid–base balance, by filtrating nitrogenous waste. Kidney function is maintained by renal blood flow, which accounts for 20% of total cardiac output. When exposed to high temperatures, the body regulates water and electrolyte balance by reducing renal blood flow, which can have a detrimental effect on the kidneys.^41–44 Under high temperatures, the renin–angiotensin–aldosterone system is activated to retain sodium in order to balance the body’s water and electrolyte.^41,42 However, increased production of renin–angiotensin–aldosterone can lead to metabolic acidosis and acute kidney injury.⁴⁵ In South Korea, it has been reported that the risk of hospitalization for acute kidney injury increases with each degree increase in temperature above 28.8 °C.⁴⁶ In patients with chronic renal disease, this mechanism can result in disease progression.⁴⁷ Dehydration can also lead to renal acidosis resulting in renal stone formation, which is one of the triggering factors for chronic renal disease.^48,49 The diminished heat risk in recent periods suggests that there has been a population adaptation to high temperatures for mortality from all renal diseases and chronic renal disease. Additionally, we observed that the MMP gradually shifted to higher temperatures in 2009–2019. This finding was in line with the previous studies in Japan examining the changes in minimum mortality temperature over time, where regional climate and socioeconomic factors were found to influence these temporal changes.¹⁶ Our findings might also be influenced by improvements in medical treatment and healthcare facilities over time.

This study is the first nationally representative study, focusing on renal disease mortality, which is a significant health burden in the Japanese population. The substantial sample size and 40-year time series provide a comprehensive assessment of the temperature–renal mortality associations across Japan and over time. However, our study has several limitations that should be acknowledged. First, we used the daily average temperatures across a prefecture to represent individual exposure, which can lead to Berkson-like measurement errors. This type of error can result in an inflated standard error, although the estimate is typically unbiased.⁵⁰ Second, our Japanese mortality dataset only contains the primary cause of death, so we may not have captured all renal disease-related mortality. Further research examining the comorbid conditions of renal diseases including hypertension and diabetes may provide additional insights. Third, this is an ecological study, and due to data limitations, we could not account for indoor conditions and individual-level temperature variations associated with air-conditioning use and the built environment throughout the entire study period. However, the self-control approach applied within the time-stratified case-crossover design was implemented to address, to some extent, individual variations in these factors. Future studies to investigate the associations among indoor conditions, the built environment, and mortality related to renal diseases will contribute to a more comprehensive understanding. Finally, our study specifically examined the short-term associations between ambient temperatures and renal disease mortality in the Japanese populations, including individuals with preexisting diagnoses of chronic renal disease. For a comprehensive understanding of the link between ambient temperature and the development of chronic renal diseases, a longitudinal study will be essential.

Since our study findings suggest the consistent cold risk of renal disease mortality in all analyses, targeted policy interventions should be implemented during the winter season. In addition to the existing weather alert system to warn the general population, more personalized interventions should be implemented to increase awareness for patients with renal diseases, particularly, close monitoring of fluid management in extreme weather conditions, preparedness of medications in advance, and ensuring sustained access to health care during heavy snowfall. Furthermore, patients undergoing home dialysis should also prepare in case of temporary blackouts due to high-energy demand during heatwaves or cold spells periods.

Conclusions

Our study found an association between daily ambient temperature and renal disease-related mortality in Japan. The evidence suggests that cold temperatures could increase the risk of mortality among populations with impaired renal function. We found that the association of renal disease-related mortality with low temperature in Japan was constantly significant over time, whereas the association with high temperature diminished in recent years. Implementing preventative measures and increasing awareness among individuals with kidney diseases, particularly among the older population, could help minimize the mortality risk of cold exposure during winter.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.

Funding

This work was supported by the Japan Science and Technology Agency (JST) as part of SICORP, grant number JPMJSC20E4 and the Japan Society for the Promotion of Science (Kakenhi) Grant-in-Aid for Scientific Research (B) (grant number: JP19H03900). The authors would also like to acknowledge the Environment Research and Technology Development Fund (JPMEERF23S21120) of the Environmental Restoration and Conservation Agency provided by Ministry of the Environment of Japan.

Supplementary Material

ee9-8-e293-s001.docx ^{(1.6MB, docx)}