Unwrapping the Surprisingly High Risk of Dying from Cardiovascular Disease at Christmas: The HUNT Prospective Population-based Cohort Study in Norway


 BackgroundAlthough it is known that winter inclusive of the Christams holiday period is associated with an increased risk of dying compared to other times of the year, very few studies have specifically examined this phenomenon within a population cohort subject to baseline profiling and prospective follow-up. In such a cohort, we sought to determine the specific characteristics of mortality occuring during the Christmas holidays. MethodsBaseline profiling and outcome data were derived from a prospective population-based cohort with longitudinal follow-up in Central Norway - the Nord-Trøndelag Health Study. From 1984-1986, 88% of the target population comprising 39,273 men and 40,353 women aged 48±18 and 50±18 years, respectively, were profiled. We examined the long-term pattern of all-cause mortality and specific causes of death according to season, month and individual days of the year to determine the number of excess (cause-specific) deaths occuring at key timepoints (including the Christmas holidays). ResultsDuring 33.5 (IQR 17.1-34.4) years follow-up, 19,879 (50.7%) men and 19,316 (49.3%) women died at age-adjusted rate of 5.3 and 4.6 deaths per 1000/annum, respectively. Each winter, there were 44 (95% CI 43-45) more all-cause deaths compared to summer, with 21 (95% CI 20-22) more deaths attributable to cardiovascular disease. Compared to any other time of the year, December 25th-27th was the deadliest; being associated with an excess of 1.3 (95% CI 1.1-1.5) all-cause and 1.0 (95% CI 0.7-1.3) cardiovascular-related deaths per day each year. Compared to the pre-Christmas/Winter period (1st-21st December), the incidence rate ratio of all-cause mortality increased to 1.22 (95% CI 1.16-1.27) and 1.17 (95% 1.11-1.22) in men and women, respectively, in the next 21 days (Christmas/New Year holiday period). All observed differences were highly significant (P<0.001). A less pronounced pattern of seasonally-linked deaths attributable to respiratory illnesses (but not cancer) was also observed.ConclusionChristmas in Central Norway is characterised by a distinctive change and increase in cardiovascular-related mortality over and above that observed between winter (more deaths) and summer (fewer deaths). This distinctive pattern contrasted with cancer-related deaths. Further research to address vulnerability to the darker consequences of winter and, more specifically Christmas, is required.

deaths occur in winter compared to summer. [1] Paradoxically, seasonal variations in cardiovascularrelated mortality are not simply explained by exposure to environmental provocations such as cold temperatures, reduced daylight hours, infections, or increased pollution. [2][3][4][5] Rather, they appear to re ect a more complex interplay between the environment and an individual's physical and psychological condition, their behaviours and the culture/society in which they live. [4,6] In Scandinavia, for example, an individual-to-societal adaptation to extremely cold temperatures undoubtedly mitigates the cyclic exposure and physiological responses to seasonally driven provocations to cardiovascular health. [7] Previous studies have sought to link clusters of increased mortality to large earthquakes [8] and the FIFA World Cup. [9] Beyond these exceptional events, there is an event that has strong potential to be detrimental to an individual's cardiovascular health on an annual basis. [10,11] At Christmas, people around the world engage in potentially stressful social interactions and provocative behaviours they would not normally expose themselves to. In those already at risk of seasonal patterns of mortality (i.e. where Christmas coincides with winter), these factors may act as additional, short-term triggers for a broad range of cardiovascular-related events. 12 A number of studies based on administrative data have previously demonstrated increased rates of mortality, [12,13] hospitalisation [11] and acute myocardial infarction (AMI) in Sweden during the Christmas holidays. [10] Beyond these studies, however, this phenomenon remains poorly characterised. [1] We hypothesised that over and beyond long-term seasonal trends within a population periodically exposed to cold winters, we would nd an additional risk of dying over the Christmas holidays. We also hypothesised that cardiovascular disease (CVD) would be the major contributor to this phenomenon and that we would nd sex-speci c differences in this regard.

Study context
Norway (population ~5.5 million people) has a long tradition of undertaking insightful, longitudinal population cohort studies; including the Tromsø Study in Northern Norway, [7,14] and the focus of this report, the HUNT Study. [15] Although the warm currents of the Gulf Stream moderate its weather, given its northerly latitude, Norway still experiences extreme weather conditions. Central Norway's Köppen Climate Classi cation subtype is Continental Subarctic Climate. [16] The coldest month is January (mean temperature -3°C) and the warmest month is July (~13°C) with a mean annual temperature of ~ 4.8°C overall. Although Norway enjoys relatively clean air, the winter solstice and darkest days of the year coincide with Christmas.

Study design
We examined the long-term pattern of mortality within the prospective, longitudinal, population-based Nord-Trøndelag Health (HUNT) Study cohort living in Central Norway. [15,17] The present study was approved by the Regional Committee for Ethics in Medical Research (REK-midt, no. 2018/1509).

Data collection
The original wave of population screening (HUNT1) was undertaken during 1984-1986, with 88% of eligible inhabitants aged ≥20 years in Nord-Trøndelag County recruited. Here, we include the 79,626 men and women who attended a clinical examination and lled out detailed questionnaires about their health and lifestyle. [15] Speci cally, data on socio-economic status, perceived levels of health and life satisfaction, lifestyle behaviours, and self-reported cardiovascular health CVD were derived from validated questionnaires. [15,17] We used a previously developed index of physical activity to categorise levels of leisure-time physical activity. [18] Study outcomes The unique personal identi cation number of all Norwegian citizens allows linkage of each participant's record in the HUNT Study to information from the national Cause of Death Registry on primary cause of death. These are classi ed according to the International Classi cation of Disease (ICD) -with precise data available until 1 st January 2018. Based on the listed causes of death and the pre-speci ed hypotheses, the main codes of interest were -CVD (i.e., ICD-9: 390-459 and ICD-10: I00-I99t inclusive of the speci c codes for coronary artery disease (CAD), AMI, cerebral infarction, and sudden cardiac death), as well as cancer/malignancy and respiratory disease/illness.

Data Analyses
This study conforms to the STROBE guidelines for the reporting of observational studies. [19] Deaths were initially grouped according to their occurrence in winter (December, January and February), spring (March, April and May), summer (June, July and August) or autumn (September, October and November). Data were then grouped into progressively smaller periods (months and 3-day rolling totals) to identify speci c periods of increased mortality. Three, 21-day periods were then purposefully selected for more granular analyses and comparison -1) the period in which, on a statistical basis, the least number of all-cause deaths occurred (17 th May-6 th June); 2) the 1 st -21 st December (the winter /pre-Christmas reference period for all comparisons) and; 3) the subsequent 21-days inclusive of the Christmas holiday period (22 nd December-11 th January) in which mortality rates were elevated above the winter average. We modelled excess mortality by adjusting for sex, age at death, day and month xed effects as well as year trend.
This approach helps to isolate unobservable characteristics that vary across time from the observed seasonal patterns in mortality. The number of lower/excess deaths per period was then estimated using ordinary least squares (OLS). Mortality was later estimated using a Poisson approach to estimate the increased/decreased risk of mortality (incidence rate ratio [IRR] with 95% CI's) due to exposure to the Christmas holiday period. Using the pro ling data presented in Table 1, we generated adjusted hazard ratios (HR) for all-cause mortality during the median study period of 33.5 (IQR 17.1 to 34.4) years followup using Cox-Proportional Hazards models (entry models using only those cases with full pro ling data). All analyses were performed using SPSS v26.0 and STATA v13. Statistical signi cance was accepted at a 2-sided alpha of P< .05.

Cohort characteristics
The study cohort comprised 40,353 women (50.1%) and 39,273 men aged 50±18 and 48±18 years, respectively. Two-thirds were married and just over half had <10 years of formal education. Most participants reported generally positive health and life-satisfaction levels. Alternatively, many had relatively high levels of risk for CVD and other chronic diseases, including elevated baseline levels of blood pressure (BP) and smoking combined with relatively high levels of sedentary behaviours and overweight status (Table 1).

Seasonal patterns of mortality
A striking pattern of seasonal uctuations in all-cause mortality was evident throughout the study period ( Figure 1). Overall, 1,707 more deaths occurred in winter (10, [13][14] being the main contributors to this differential. Moreover, the winter-to-summer differential in mortality for cancer-related deaths was only 10 during the entire study period -see Figure 2. Monthly patterns of mortality All-cause mortality peaked in the winter months of December (3,675 deaths) and January (3,592 deaths). The lowest mortality occurred in June (2,920 deaths). The annual excess all-cause mortality occurring during each of the peak months of December and January versus the low of June accounted for 22 (95% CI 21-22) more deaths. Cardiovascular-related deaths were the main contributors to this phenomenon in both December (11, 95% CI 9-10 more deaths) and January (8, 95% CI 7-9 more deaths) -Supplementary Figure S1. Both respiratory disease and a range of other causes of death (6-8 more deaths/month) also contributed to this phenomenon -Supplementary Figure S2.

The Christmas Holiday Effect
Regardless of the season, accumulative 3-day mortality mainly uctuated between 90-110 deaths.
However, a clear increase in mortality commencing the 22 nd December was evident. The subsequent 3day period over Christmas was the deadliest of the year (Figure 3) with 439 all-cause deaths. This was not a random phenomenon and was largely driven by an increase in cardiovascular-related and, to a lesser extent, cancer-related deaths (Figure 4).  Figure S3. When compared to the preceding 21 days, the Christmas period was also notable in respect to within and between differences among men and women in respect to fatal AMI (78 versus 16 more deaths, respectively), strokes (13 fewer versus 32 more deaths) and heart failure (1 more versus 12 more deaths). Similarly, in men and women, the number of cancer-(18 and 29 more deaths, respectively) and respiratory-related (19 and 33 more deaths, respectively) deaths also increased.

Winter and Christmas vulnerability
Overall, except for cancer-related mortality (both sexes) and respiratory disease in men, compared to the rst 21 days of December/winter, the risk of dying in the late spring/early summer period of 17 th May to 6 th June was signi cantly lower -Supplementary Figure S4. Alternatively, except for an increased risk of dying from respiratory illnesses/disease among women, men had a higher risk of dying over the equivalent 21-day Christmas period; the major contributor to this increased mortality risk (from 6% to 22% higher overall) being CVD -Supplementary Figure S5. .001 -- Beyond advancing age, a combination of baseline demographic, health perceptions and clinical factors were independently correlated with dying during -1) late spring/early summer versus early winter, and then 2) early winter versus the Christmas holiday period. Whilst these factors were broadly similar for both sexes, including a 30% reduced risk during the Christmas holidays associated with being married at baseline, there were some notable differences. For example, consistent with an excess number of strokes among women, but not men, during the Christmas holidays, a pre-existing history of stroke conferred a 2fold risk of dying during this period among women. Educational status among women also appeared to modulate the additional risk of dying during this period -see Table 2.

Sensitivity analyses
We conducted sensitivity analyses by estimating four different models to test if the phenomenon of Christmas-related excess mortality is a reliable and consistent observation. All four models supported the ndings of a signi cant increase in mortality over the Christmas period -Supplementary Table S1.

Discussion
We investigated the seasonal pattern of mortality within the HUNT Study cohort living in Central Norway. This population cohort is regarded as representative for the Norwegian population as a whole, except for a lower proportion of non-whites and the absence of large cities. Our analyses revealed a striking longterm difference in mortality occurring in winter compared to summer. CVD accounted for half of this seasonality. Although not the coldest, December proved to be the deadliest month, with 22 more people dying each year compared to June. Overall, the 3-day period of 25th-27th December proved to be the deadliest time of the year with CVD as the major contributor. Critically, both the frequency and cause of death in men and women appeared to change over the Christmas period. Compared to the same pre-Christmas/wintry period, men were 22% and 17% more likely to die from all-causes and CVD (particularly AMI), respectively. In women, the equivalent risk increases were 17% and 15%, with the contribution of CVD (particularly stroke) even more prominent. Although previous studies have also identi ed a speci c Christmas effect on mortality [10], [11][12][13]20], we are unaware of any studies and ndings equivalent to those reported here. There is pre-existing evidence to support the hypothesis that Christmas can be harmful to some individuals. A study of the overall pattern of mortality in the US during 1973-2001 revealed a "holiday effect" during Christmas, with ~5% excess deaths, after adjustment for the winter season. [12] Similarly, data from a nationwide coronary care unit registry in Sweden revealed a 15% increase in AMI cases during the Christmas holidays. [10] A higher risk of 30-day mortality or readmission among those hospitalised at Christmas in Ontario, Canada has also been found. [11] From a Southern Hemisphere perspective there is both supportive [13] and contrary evidence [21] of an equivalent phenomenon occurring in summer conditions.
In the (understandable) absence of prospective studies, it is challenging to delineate between the overall impact of winter and a Christmas-speci c effect. As shown by the Tromsø Study [22], there is evidence of winter peaks in blood pressure, heart rate, body weight, total cholesterol and overall CVD risk. Seasonal variation in physical activity may also be an important consideration for cardiovascular-related mortality.
[23] Aerobic exercise, especially with high intensity, can acutely lower systolic BP in the hours following exercise.25 As in many parts of the world, life in Central Norway during the Christmas holiday period is characterised by festive celebrations, travel away from home/central services, and reduced health services. Reduced access to follow-up health care was noted to contribute to 26 excess deaths (and 188 hospital readmissions) per 100,000 patients in Canada during the Christmas holidays. [11] However, this phenomenon does not fully explain the size of the phenomenon we observed within our cohort and the contributory reasons are likely to be multifactorial. Consuming a high-fat diet for only 3 days exacerbates insulin resistance and glycolipid metabolism disorders in obese men. [24] Even among healthy men, decreasing physical activity for 1-3 weeks decreases insulin sensitivity and attenuates postprandial lipid metabolism. [25] Vascular stiffness, due to impaired endothelial function of the conduit vessels, is an important factor in the development of hypertension and an independent risk factor for a fatal cardiovascular event. [26] After a high-fat meal, which is typically consumed during Christmas in Norway, endothelial function decreases substantially postprandially. [27] The potential negative impact of increased emotional stress associated with dealing with loneliness and family tensions [28] with the potential for seasonally triggered depression [29], also cannot be ignored. As suggested by our sexspeci c ndings, any or all of these "stressors" may affect men and women differently. For example, it has been demonstrated that diabetes, high-density lipoprotein levels and triglyceride levels have more impact on cardiovascular health of women compared to men. [30] The emerging literature around Takotsubo cardiomyopathy with a predominance of women affected [31] is notable when considering the small, but intriguing, increase in deaths due to heart failure in women, but not men, at Christmas.
Unfortunately, in the absence of speci c interventions, expert clinical guidelines rarely mention or address seasonality. Beyond ensuring appropriate vaccination against in uenza [32], there is a strong justi cation for more proactive screening and management of high-risk patients by general practitioners leading up to

Study limitations
To robust test our primary hypothesis we examined patterns of long-term mortality within the HUNT cohort [15,17] in Central Norway. Although this is a speci c population and relies upon historical outcomes, ndings from the HUNT cohort have consistently supported previous novel observations as well as discover new ones that have been likewise externally validated. To maintain the size of outcome data for analyses, we relied upon baseline pro ling of the original cohort.
As of yet, we have not examined linked hospitalisation data, to improve and inform our ndings. However, even with such data, the HUNT Study was not speci cally designed to examine this speci c issue. For example, whilst we will be able to determine the cause and timing of any hospital admissions, we will not be able to ascertain the quality of care and extent of outpatient follow-up. However, the timing of death (unless a sudden cardiac death) is not indicative of exactly when a person becomes unwell and/or is admitted to hospital. [11] This has relevance to both the speci c timing of the observed increase in mortality and how long it took for that increase to subside. As such, our current analyses of the correlates of the speci c timing of mortality can only inform future research directions rather than identify speci c causality. For example, in the future we do plan to access all data from additional pro ling "waves" subsequently performed on some HUNT participant. In our statistical analyses at least, we were able to robustly test the timing of mortality and the Christmas period; our ndings consistently con rming a strong temporal relationship that warrants further investigation.

Conclusions
During long-term follow-up of the HUNT population cohort, there was a distinctive pattern of a seasonal increase in mortality during winter when compared to summer months. Over and above this broad pattern, a distinctive pattern of excess mortality predominantly, but not exclusively linked to CVD, was evident over the Christmas holiday period. The number of excess deaths over Christmas was substantial.
If present at the same levels within the entire age-matched population in Norway (a minimum of 3 million adults alive in 1980), there would have been more than 11,000 deaths (around 350 per annum) over the Inspectorate to store and handle these data. The key identi cation in the data base is the personal identi cation number given to all Norwegians at birth or immigration, whilst de-identi ed data are sent to researchers upon approval of a research protocol by the Regional Ethical Committee and HUNT Research Centre. To protect participants' privacy, HUNT Research Centre aims to limit storage of data outside HUNT databank, and cannot deposit data in open repositories. HUNT databank has precise information on all data exported to different projects and are able to reproduce these on request. There are no restrictions regarding data export given approval of applications to HUNT Research Centre. For more information see: http://www.ntnu.edu/hunt/data     EXCESS CHRISTMAS MORTALITY Legend: The number of deaths occurring on the 25th -27th of the remaining months (January to November) each year above and below the reference 3-day period 25th -