SEARCH

SEARCH BY CITATION

Keywords:

  • Cardiovascular disease;
  • seasonality;
  • heart failure;
  • hypertensive diseases;
  • vitamin D

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Objectives: To examine the seasonal pattern in heart failure (HF) and cardiovascular disease (CVD) by climate and cause of death in Australia.

Methods: A retrospective analysis of a national database of deaths in the eight Australian State and Territory capitals between January 1997 and November 2004. We examined the seasonal pattern in HF and CVD deaths, we identified variations in the pattern by: sex, age, time, climate and cause of death (for total CVD using seven groups determined by ICD-10 code).

Results: Deaths in all seven groups of CVD significantly increased in winter. The largest increase in mortality rates was 23.5% observed for HF. The magnitude of this increase varied greatly between cities, with the lowest winter mortality rates in the coldest (Hobart) and warmest (Darwin) cities. The pattern in CVD deaths showed a clearer correlation with climate than HF deaths.

Conclusion and implications: Winters in Australia are mild but winter increases in HF and CVD are a significant problem. Increased blood pressure and lack of vitamin D in winter are the most likely causes of the increase. Reducing exposure to cold, particularly in the elderly, should reduce the number of winter CVD deaths in Australia.

Cardiovascular disease (CVD) shows a strong seasonal pattern in many parts of the world, with greatly increased rates in winter compared to summer.1 Similar strong seasonal peaks in admissions to hospital have also been observed in relation to heart failure.2–5 Counter-intuitively, the seasonal pattern is strongest in regions with warmer climates.6,7 For example, a strong association between relatively cold temperatures and cardiovascular disease was found in Newcastle, Australia, but no association was found in three populations in Finland.1

Winter increases in heart failure (HF) admissions have been found in France,2 Spain,3 Scotland,4 Japan,5 and Argentina.8 In France, HF deaths were 20% above average in January and 15% below average in August.2 In Madrid, HF hospitalisations were 25 % above average in January and 33% below average in August.3 In Scotland, HF hospitalisations were 8% to 12% above average in December and 7% to 8% below average in July, and the seasonal increase was 23% to 35% larger in the elderly4 In Japan, HF admissions were highest in spring and lowest in summer and autumn.5

The cause of the winter increase in HF and CVD remains in doubt. Colder temperatures are not the only possible cause, as a number of other CVD risk factors change over the seasons.9 A summary of the putative pathways between winter and cardiovascular disease is given in Figure 1 (adapted from Scragg10). There are accepted strong associations (in any season) between CVD and the five risk factors of body mass index (BMI), cholesterol, blood pressure, inflammation (e.g. fibrinogen) and heart rate. However, it is not known which risk factor (or factors) are the cause(s) of the seasonal change in CVD, or what seasonal exposures are driving the seasonal change in risk. A number of exposures have been proposed, those shown in Figure 1 are physical activity,11 sunlight and vitamin D,12 temperature,11,13 influenza14 and air pollution15 (which may also interact with temperature16).

image

Figure 1. The putative pathways between winter and cardiovascular disease. Adapted from Scragg.10

Download figure to PowerPoint

The number of possible pathways between winter and increased CVD has made it difficult to identify the cause of the seasonal change in CVD. In this paper, we aim to add to the evidence by looking at seasonal patterns in CVD by cause of death and climate. Although previous studies have examined the effect of climate on seasonal changes very few have examined cause of death. We also examine the seasonal pattern in HF, and compare the results to CVD.

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Cardiovascular death data

We examined the monthly counts of cardiovascular and heart failure deaths in the eight Australian State and Territory capital cities for the years 1997 to 2004 (eight years). The eight cities studied are shown in Table 1.

Table 1.  Basic characteristics of the eight Australian State and Territory capital cities ordered by population size.
Capital Cardiovascular diseaseHeart FailureLatitudeAverage temperatured (°C)
State/TerritoryPopulationaDeathsbIncidencecDeathsbIncidencec(degrees south)JanuaryJuly
  1. Notes:

  2. (a) In 2000 (midpoint of years studied)

  3. (b) January 1997 to November 2004

  4. (c) Incidence per year per 100,000 people

  5. (d) Historic data from:http://www.worldclimate.com/

Sydney, NSW3,007,16274,306312406317.133.922.111.8
Melbourne, Vic2,570,36457,878284389419.137.819.59.0
Brisbane, Qld1,171,36728,639309110211.927.425.014.8
Perth, WA1,000,27320,511259107513.632.024.112.8
Adelaide, SA817,37524,472378141421.934.922.611.1
Hobart, Tas139,946422438121319.242.916.77.9
Canberra, ACT222,274342419524013.635.320.25.3
Darwin, NT62,260631128275.512.528.524.9
Total8,991,021214,08530112,02816.9   

Because of a delay in registration of deaths we excluded the last month of the study period (December 2004). The monthly counts of deaths were stratified by: month and year of death, sex, age (18-64 years, 65+ years) and the International Classification of Diseases (ICD-10) code of underlying cause of death.17 We classified the cause of death into seven groups using the subsections of the ICD classification.

Estimating seasonality

We estimated the seasonality in HF and CVD using a sinusoidal model.18 An example of a sinusoidal fit to the total CVD deaths is shown in Figure 2. The amplitude in total deaths is the difference from the mean to the peak value, and is 2,700 deaths in this example (or 17% of the mean). The peak in seasonal deaths in this example is 7.5 (mid-July). Our main goal was to estimate if the amplitude or peak of the seasonal pattern varied by: sex, age, climate (i.e., city), ICD-10 code and time.

image

Figure 2. Total cardiovascular disease deaths in Australian capital cities (1997–2003) and fitted seasonal sinusoid.

Download figure to PowerPoint

To correctly estimate the seasonality in HF and CVD an adjustment for long-term trends was required. This was achieved by using a Bayesian model19 which decomposed the observed data into three parts: trend, season and noise. The model was applied using the Bayesian WinBUGS software version 1.4.3.20 We used 18,000 Markov chain Monte Carlo iterations with a burn-in of 6,000 thinned by three.

The dependent variable in the model was the monthly number of deaths. These counts were standardised to account for the uneven number of days between months.18 We used a generalised linear model assuming a Poisson distribution for the monthly counts.21 We adjusted for changes in the population over time by including annual city population as an offset (using population estimates from the Australian Bureau of Statistics).

The seasonal patterns in HF and CVD may vary by city, sex, age group, ICD code and time and by interactions between these variables. As an example, previous studies have shown larger seasonal changes in warm compared to cold climates.1,6 In order to find the optimal models we examined a number of different models using combinations of these variables and their interactions. Each model was based on evidence from the literature concerning seasonal patterns in CVD. We selected the optimal models based on the deviance information criterion (DIC).22 This is an estimate of model fit that compromises between a close fit to the data and over-fitting from including unnecessary parameters. The optimal model had an amplitude that varied by city, time, ICD code and the interactions between sex and city and age and city. The peak for the optimal model varied by city, sex and ICD code.

We checked the residuals of the optimal model for any remaining seasonal pattern using the cumulative periodogram test.23 We performed this check for each city, age group and sex.

We looked for climatic patterns in seasonality by plotting the estimated amplitudes for each city against four indicators of climate: latitude, average January temperature, average July temperature, difference between January and July temperature. As the patterns appeared non-linear we fitted a generalised additive model to the estimates using a penalised spline with three degrees of freedom.24

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

The amplitudes and peaks of the seasonal change in cardiovascular disease by ICD group are shown in Table 2. These results are averaged across cities, age groups and sex. The largest amplitude was for heart failure (23.5%) and this peak occurred in early August (8.0 months).

Table 2.  Seasonality in cardiovascular disease deaths by ICD group ordered by amplitude in Australian capital cities, 1997–2004.a
 Amplitude (%)Peak (monthsb)
ICD group (ICD-10 codes)Mean95% PIMean95% PI
  1. Notes:

  2. (a) December 2004 excluded due to registration delays

  3. (b) 1=January, 12=December

  4. ICD=International Classification of Diseases; PI=posterior interval

Heart failure (I50)23.515.9,29.98.07.6, 8.3
Hypertensive diseases (I10-I13)16.08.3, 22.87.97.5, 8.3
Ischemic heart diseases (I20-I22, I24-I25)12.36.1, 17.27.57.3, 7.9
Electrical conduction system (I44-I49)10.73.5, 17.18.27.7, 8.7
Pulmonary and pulmonary circulation (I26-I28)10.72.3, 18.47.87.3, 8.3
Cerebrovascular diseases (I60-I64, I67, I69)9.23.0, 14.17.87.4, 8.1
Myocardium/cardiomyopathy (I40, I42)8.71.4, 15.87.87.3, 8.4

Although Table 2 shows a variation in the seasonal amplitude by ICD group, the amplitude in every group was statistically significant as the lower limit of the 95% posterior interval was always greater than zero. The range in mean amplitudes was from 8.7% to 23.5% and such figures also represent a clinically significant increase in deaths.

The peak months were broadly similar across the seven ICD groups, and all occurred in July to August (Table 2). The earliest peak was for ischemic heart disease (7.5 months) and the latest for the electrical conduction system (8.2 months).

The differences in the seasonal pattern by age, sex and time are shown in Table 3. The table shows the baseline (intercept) results and the changes due to age, sex and time. The winter increase in CVD deaths was 8.5% larger in the 65+ year age group compared to the 18-64 age group. A similar sized increase was observed in HF (7.7%) but with a much larger posterior interval. The seasonal increases in CVD and HF deaths did not change greatly by sex. For both CVD and HF there was a steady decrease in the seasonal amplitude over time. The peak time for both CVD and HF was 7.7 months. For HF deaths this peak was slightly later in women.

Table 3.  Differences in the seasonal pattern of heart failure and cardiovascular disease deaths by age, sex and time. Intercept rows show the baseline effect, and other rows show the positive and negative changes.
 All CVDHF
 Mean95% PIMean95% PI
  1. Note: CVD=cardiovascular disease; HF=heart failure; PI=posterior interval

Amplitude, per cent
Intercept (Men aged 18-64)7.11.2, 11.522.90.1,55.2
Age 65++8.5+0.9, +14.8+7.7−10.5, +29.3
Women+ 1.1−2.5, +5.1−3.1−8.2, +2.2
Time (per year)−0.9−1.7, −0.2−1.2−2.2, −0.2
Peak time, months Intercept (Men)7.77.3, 8.17.77.4,7.9
Women+0.1−0.1, +0.2+0.4+0.1, +0.8

There was a large variation in the seasonal amplitude of CVD and HF by city. In an attempt to explain this difference the mean CVD amplitude (averaged across age and sex) is plotted against average January and July temperature in Figure 3, and the mean HF amplitude in Figure 4. Using January and July temperature gave a clearer pattern between amplitude and climate compared to using either latitude or the difference between January and July temperature (plots not shown). The highest seasonal change in CVD mortality was in Perth (15.2%) and the lowest in Darwin (7.6%). The highest seasonal change in HF was in Sydney (40.5%) and the lowest in Hobart (15.4%). The seasonal amplitudes for total CVD showed a quadratic pattern when plotted against summer temperature (Figure 3). The amplitudes were generally smaller in the warmest and coldest cities. The quadratic pattern was less clear when the amplitudes were plotted against winter temperature. For heart failure there was also a quadratic pattern between season and climate, but the pattern was much noisier (Figure 4).

image

Figure 3. Cardiovascular disease seasonal amplitudes by city against January and July temperatures (degrees C). Smooth line and shaded area are the mean and 95% confidence interval from a non-linear model.

Download figure to PowerPoint

image

Figure 4. Heart failure seasonal amplitudes by city against January and July temperatures (degrees C). Smooth line and shaded area are the mean and 95% confidence interval from a non-linear model.

Download figure to PowerPoint

The residual checks for any remaining seasonal pattern were statistically significant for total CVD in older women in both Sydney and Brisbane, and for HF in older women in Sydney and young women in Hobart (four statistically significant results out of 64 tests). Further investigation revealed that the sinusoidal model did not fully capture the sharp winter peak in deaths for the older age groups. At the other end of the scale the observed results from young women in Hobart showed no seasonality as there were no HF deaths in young women.

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Australia has a large variation in climates across its capital cities (Table 1) but similar health systems. This makes it an ideal place to study seasonal patterns in cardiovascular disease and heart failure. Despite Australia's predominantly mild climate we found a large seasonal change in CVD-related deaths. All categories of cardiovascular disease were significantly seasonal and the peaks in deaths occurred at a similar time (between July and August, Table 2). Differences did occur in the size of the seasonal peak and by far the most seasonal was heart failure. The average winter increase in heart failure of 23.5% in Australia is comparable to other increases found in France, 20%;2 Madrid, 25%;3 and Scotland 12%.4

Congestive heart failure (CHF) is the end stage of many cardiac disease processes, with hypertension and ischaemic heart disease being the two largest causes of CHE. A patient with CHF has little physiological reserve to deal with an increase in cardiac workload. With a temperature drop comes a compensatory vasoconstrictive response, particularly to the skin. This is associated with an increased after-load for the failing heart, and is achieved by up-regulation of the neuro-humoral cascade and increased levels of vasoconstrictors. Hence, cardiac work increases to overcome the rise in after-load and at the end-stage the failing heart is unable to cope with this increased demand. There may be associated increases in exacerbations in minor respiratory tract infections, which will compound work load for the failing heart.

Magnitude of temperature change may correlate with degree of change in systemic vascular resistance (via release of circulating catecholamines) and hence cardiac workload. A heart that is already sick would cope poorly with this change and may fail. Blood pressure changes may also directly correlate with catecholamine levels, and previous studies have shown that blood pressure changes are greatest with large temperature changes.11,13

There is strong evidence to show that vitamin D insufficiency can precipitate heart failure.12 Australia has a very sunny climate, but vitamin D levels in adults have been shown to be very low in winter.25 People with pre-existing heart failure often have a sedentary lifestyle, which may result in few hours outdoors with minimal exposure to sunlight. For these people a vitamin D supplementation may reduce their event risk, although a specific trial of this intervention needs to be carried out before supplementation can be recommended.

Differences in the elderly and over time

The seasonal change in total CVD was strongly dependent on age (Table 3). In the elderly, the seasonal increase was 8.5% higher. A similar, but non-statistically significant, increase in deaths was found for HF A study in Scotland found that the winter increase in HF was 15-16% in those aged over 85, and 7-8% in those aged 65-74.4 The elderly are frailer and less well-physiologically equipped to deal with the winter drop in temperature. Many may also have pre-existing cardiovascular disease, and so an increase in blood pressure (due to exposure to cold) may have more serious consequences. The elderly may also be more susceptible to infection.

During the study period (1997-2004) the seasonal change in CVD and HF deaths became gradually smaller. For all CVD deaths there was a decrease of 0.9% per year and for HF deaths a decrease of 1.2% per year. This suggests that the seasonal pattern in deaths may disappear in time. Although in Sydney where the seasonal increase in HF deaths is 40.5%, this would take roughly 34 years. The reduction in seasonal deaths over time may be due to: gradual improvements in housing and treatment, an increased uptake of flu vaccinations, or possibly even higher ambient temperatures associated with the process of global warming.

Seasonality and climate

We found large differences in the seasonal pattern of CVD dependent on climate. Darwin is in the tropics with distinct wet and dry seasons (May to September). It showed the smallest average seasonal change in CVD deaths (Figure 3). Hobart is the coldest city on average, and it had the next smallest seasonal change. The largest seasonal changes in CVD deaths were in the cities in the middle of the Australian temperature range. This pattern of increasing seasonal change in CVD with milder climates has been previously found in Europe6 and using worldwide data.1 The results shown here suggest that it is summer temperature, rather than winter temperature, that most strongly dictates the degree of seasonality in CVD in Australia. This could be because there is no perceived need to trap heat in warm Australian cities, and building design is geared towards cooling.

The pattern between climate and HF (Figure 4) was not as clear as the pattern between climate and CVD. This was a surprise as the strongest candidates of the winter increase in HF are a lack of vitamin D,12 or increase in blood pressure due to lower skin temperature.13,26 It is possible that these factors are still important but that they are modified by local factors (e.g. housing type). Another possibility is that multiple factors are involved (Figure 1). Without detailed individual level data it is difficult to separate the many different seasonal exposures. Future studies would ideally collect information from new heart failure patients on recent physical activity, housing type, and their blood pressure and vitamin D levels. A future study could also collect information on country of birth and residential history. People born in cold climates who move to Australia may take better precautions to deal with cold temperatures compared with lifelong Australian residents.

Preventing seasonal deaths

It seems likely that reducing exposure to cold temperatures could prevent many CVD and HF deaths in Australia. Strategies for reducing exposure to cold need not be costly. One option is to encourage people (particularly the elderly) to wear warmer clothes on relatively cold days. A more expensive alternative is to insulate homes. Only 58% of Australian homes had roof insulation in 2002.27

The financial burden of heart failure in Australia is high, and is estimated to cost over $1 billion each year and affect more than 500,000 Australians.28 The cost per admission with a primary diagnosis of CHF is approximately $10,500 per admission (at The Prince Charles Hospital, Brisbane) and many patients require several admissions. Given this high financial cost, it would appear that further studies delineating methods of minimisation of temperature-related morbidity and mortality are needed. Such studies may reduce future health care costs, as the age of the population increases and with it the incidence of cardiac disease.

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  • 1
    Barnett AG, Dobson AJ, McElduff P, Salomaa V, Kuulasmaa K, Sans S, et al. Cold periods and coronary events: an analysis of populations worldwide. J Epidemiol Community Health. 2005;59:5517.
  • 2
    Boulay F, Berthier F, Sisteron O, Gendreike Y, Gibelin P. Seasonal variation in chronic heart failure hospitalisations and mortality in France. Circulation. 1999;100:2806.
  • 3
    Martínez-Sellés M, García Robles J, Prieto L, Serrano J, Muñoz R, Frades E, et al. Annual rates of admission and seasonal variations in hospitalisations for heart failure. Eur J Heart Fail. 2002;4:77986.
  • 4
    Stewart S, McIntyre K, Capewell S, McMurray J: Heart failure in a cold climate. Seasonal variation in heart failure-related morbidity and mortality. J Am Coll Cardiol. 2002;39:7606.
  • 5
    Ogawa M, Tanaka F, Onoda T, Ohsawa M, Itai K, Sakai T, et al. A community based epidemiological and clinical study of hospitalization of patients with congestive heart failure in Northern Iwate, Japan. Circ J. 2007;71:4559.
  • 6
    The Eurowinter Group. Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions of Europe. Lancet. 1997;349:13416.
  • 7
    Donaldson GC, Rintamäki H, Näyhä S. Outdoor clothing: its relationship to geography, climate, behaviour and cold-related mortality in Europe. Int J Biometeorol. 2001;45:4551.
  • 8
    Díaz A, Ferrante D, Badra R, Morales I, Becerra A, Varini S, et al. Seasonal variation and trends in heart failure morbidity and mortality in a South American community hospital. J Am Coll Cardiol. 2007;13:2636.
  • 9
    Pell J, Cobbe S. Seasonal variations in coronary heart disease. QJM. 1999;92:68996.
  • 10
    Scragg, R Seasonal variation of mortality in Queensland. Community Health Studies. 1982;6:1208.
  • 11
    Ulmer H, Kelleher C, Diem G, Concin H, Ruttmann E. Estimation of seasonal variations in risk factor profiles and mortality from coronary heart disease. Wien Klin Wochenschr. 2004;116:6628.
  • 12
    Zittermann A, Schleithoff S, Koerfer R. Vitamin D insufficiency in congestive heart failure: Why and what to do about it? Heart Fail Rev. 2006;11:2533.
  • 13
    Barnett A, Dobson A, Salomaa, Y Kuulasmaa K, Sans S, the WHO MONICA Project. The effect of temperature on systolic blood pressure. Blood Press Monit. 2007;12:195203.
  • 14
    Reichert T, Simonsen L, Sharma A, Pardo S, Fedson D, Miller M. Influenza and the winter increase in mortality in the United States, 1959–1999. Am J Epidemiol. 2004;160:492502.
  • 15
    Bai N., Khazaei M., Van Eeden S, Laher I. The pharmacology of particulate matter air pollution-induced cardiovascular dysfunction. Pharmacol Ther. 2007;113:1629.
  • 16
    Ren C, Williams G, Tong S. Does particulate matter modify the association between temperature and cardiorespiratory diseases? Environ Health Perspect. 2006;114:16906.
  • 17
    Australian Bureau of Statistics. Deaths, Australia, 2004. Canberra ( AUST ) : ABS; 2004. Catalogue No.: 3302.0.
  • 18
    Barnett AG, Dobson AJ. Estimating trends and seasonality in coronary heart disease. Stat Med. 2004;23:350523.
  • 19
    Gelman A, Carlin J, Stern H, Rubin D. Bayesian Data Analysis. 2nd ed. Boca Raton ( FL ) : Chapman & Hall/CRC; 2004.
  • 20
    Spiegelhalter D, Thomas A, Best N, Lunn D WinBUGS [user manual]. Version 1.4.31. Cambridge ( UK ) : MRC Biostatistics Unit; 2007.
  • 21
    Dobson AJ, Barnett AG. An Introduction to Generalized Linear Models. 3rd ed. Boca Raton ( FL ) : Chapman & Hall/CRC; 2008.
  • 22
    Spiegelhalter D, Best N, Carlin B, Van Der Linde A. Bayesian measures of model complexity and fit (with discussion). Journal of the Royal Statistical Society. Series B. 2002;64:583640.
  • 23
    Fuller WA. Introduction to Statistical Time Series. 2nd ed. New York ( NY ) : Wiley; 1996.
  • 24
    Ruppert D, Wand M, Carroll R. Semiparametric Regression. New York ( NY ) : Cambridge University Press; 2003.
  • 25
    Kimlin M, Harrison S, Nowak M, Moore M, Brodie A, Lang C. Does a high UV environment ensure adequate vitamin D status? J Photochem Photobiol B. 2008;89:139147.
  • 26
    Milo-Cotter O, Adams K, O'Connor C, Uriel N, Kaluski E, Felker GM, et al. Acute heart failure associated with high admission blood pressure–a distinct vascular disorder? Eur J Heart Fail. 2007;9:17883.
  • 27
    Australian Bureau of Statistics. Environmental Issues: People's Views and Practices. Canberra ( AUST ) : ABS; 2005 Mar. Catalogue No.: 4602.0.
  • 28
    Clark R, McLennan S, Dawson A, Wilkinson D, Stewart S. Uncovering a hidden epidemic: A study of the current burden of heart failure in Australia. Heart Lung Circ. 2004;13:26673.