In this study, we investigated the seasonality of mortality under a warming climate in Japan. We found that PTRs in all-cause, circulatory, and respiratory mortality for Japan as a whole decreased after adjusting for temperature. Furthermore, seasonality tended to be flattened for respiratory and all-cause mortality between 1972 and 2015. The changing seasonal amplitude was negatively associated with annual mean temperature. These negative associations remained significant after adjusting for other time-varying prefectural characteristics for all-cause and circulatory mortality, but the association with respiratory mortality tended toward to null. Although confirmation is required, our results suggest that a warming climate may lead to a flattening of seasonality when other potential confounders remain similar over time.
This seasonal pattern of mortality in Japan is consistent with findings from other studies conducted in regions with four distinct seasons [7, 9, 13, 15, 20, 21]. Similar seasonal patterns were reported for other health indicators, such as levels of C-reactive protein, a marker of inflammation . Various seasonal factors have been proposed to explain these patterns, including temperature, exposure to sunlight, human activity patterns, and the incidence of influenza [7, 12, 15]. Of these, temperature has gained the most attention from researchers. Gasparrini and colleagues  collected data on temperature and mortality in over 600 cities worldwide and observed that increased mortality was associated with both hot and cold temperatures. Relationships between temperature and key cardiac risk factors have also been reported . Cold temperatures can lead to peripheral vasoconstriction to reduce thermal conduction, increase metabolic heat production, trigger platelet activation, and provoke cumulative increases in inflammatory markers . Conversely, heat stress can lead to generalized peripheral vasodilatation and over-sweating, provoking increases in heart rate and hyperthermia and even syncope and myocardial ischemia .
We found that the amplitude of seasonality was reduced after controlling for temperature. However, the amplitude reduction varied between all-cause, circulatory, and respiratory mortality, with the largest reduction observed for circulatory mortality and the smallest for respiratory mortality. Using aggregated monthly data on mortality at the national level from 1970 to 1999, Nakaji and colleagues  identified a similar seasonal pattern and reduced seasonal amplitude when monthly temperature was considered in the analysis, despite having used a linear function for temperature. Furthermore, consistent with our findings, no notable reduction was observed for respiratory mortality after adjusting for temperature in that study . Our findings suggest that temperature is potentially a crucial contributor to seasonal variations in mortality, and especially to circulatory mortality.
The reduced seasonal amplitudes observed after adjusting for temperature in our analysis, coupled with the existing extensive evidence of temperature effects on mortality, led us to hypothesize that the amplitude of seasonality may be undergoing reduction as temperatures increase under conditions of climate change. Over the study period, annual mean temperature in Japan increased from 14.6 °C in 1972 to 15.9 °C in 2015, while PTRs decreased for all-cause and respiratory mortality. We found that annual mean temperature was negatively related to PTR for all-cause, respiratory, and circulatory mortality. This association persisted for all-cause and circulatory mortality even after adjusting for potential confounders. Warmer winters and fewer cold periods may result in lower mortality peaks in winter, whereas warmer summers and more frequent and intense heat waves may lead to a rise in mortality in summer, thereby flattening the seasonality of mortality over time. In addition, fewer cold-related deaths in warmer winters may also lead to a shift in the population susceptible to warmer summers and may result in more deaths in the warm seasons, translating to a flatter seasonality [24, 25]. Therefore, the flattening of seasonality in Japan may be related to a warming climate. A recent global projection  of temperature-related mortality predicted a decrease in cold-related mortality and an increase in heat-related mortality in East Asia including Japan, without considering potential changes in demographics and adaptation. A warming climate could produce a flatter seasonality of mortality in Japan in the future, unless confounded by other factors.
It is important to emphasize that our findings should be interpreted cautiously. Other than the warming climate, we also found that changing seasonal amplitude was associated with lower relative humidity, population aging, economic development, and increasing prevalence of air conditioning. Populations are likely to adapt to a changing climate. Heat-related mortality in the warm seasons may be reduced, and seasonal amplitude could remain consistent over time. In addition, efficient protection measures against seasonal risks, such as vaccination against infectious diseases, may prevent excess winter mortality, also in turn reducing the seasonal amplitude.
To our knowledge, so far only two studies [9, 10] have investigated the relationship between warming climate and changes in the seasonality of mortality, and the conclusions of these studies are similar to ours. Bennett and colleagues  reported that the ratio of summer-to-winter deaths in those aged 55 years and above in Australia increased from 0.71 to 0.86 between 1968 and 2007 in tandem with rising annual temperatures. McGregor and colleagues  found a decrease in seasonal amplitude for mortality from ischemic heart disease between 1974 and 1999 in five English counties, and this trend was positively correlated with the amplitude of the annual temperature cycle. However, these two studies relied on a relatively simple method of using monthly aggregated data to assess seasonality, and their results are limited to a specific population, region, or cause of mortality; moreover, they did not include recent years. We included recent time-varying annual data on prefecture-specific characteristics in the meta-regression analysis, which enabled us to gain a better understanding of the impacts of climate change on the seasonality of mortality by taking into account potential confounders.
Some limitations should be noted. First, we focused on the amplitude of seasonal variation in mortality and did not consider the changes in the shape of seasonal patterns (i.e., peak and trough). Although the shape of seasonal patterns did not appear to change substantially during our study period, it is possible that a warming climate may impact the timing of peak and/or trough. Second, climate change includes rising temperatures, shortening winter seasons, increasing extreme weather events, etc., but we only considered the increasing annual mean temperature in this study. Third, our investigation was conducted in Japan, where the seasons are distinct in most prefectures. Hence, future investigations in other locations with different climates are required to confirm our findings.