Background Cerebrovascular disease (stroke) is the second most common cause of death and among the top five causes of morbidity in many developed and developing countries. The aim of this study was to investigate patterns of increase and decrease in stroke mortality in 48 different countries.
Methods The mortality curves of stroke for 48 countries that had reliable data and met other selection criteria were examined using age standardised death rates for 35–74 years from the WHO. Annual mortality rates for individual countries from 1950 to 2005 were plotted and a table and graph were used to classify countries by magnitude, pattern and timing of stroke mortality. Male and female trends were plotted separately.
Results The secular trend of stroke mortality varied markedly among countries. Different stroke patterns were distinguishable, including ‘declining’, ‘rise and fall’, ‘rising’ and ‘flat’. Furthermore, epidemic peaks per 105 (M/F) were higher in Asia, in particular in Japan (433/304), Russian Federation (388/221) and Bulgaria (301/214), and were lowest in Canada and Australia (29/18). There were considerable differences among some continental and regional geographic areas. For example, Japan, Belgium, Portugal and Eastern Europe exhibited a rise and fall pattern while other countries fell into the other three categories.
Conclusions In many countries, stroke mortality has decreased, between 1950 and 2005, often very considerably. The different dates of mortality downturn likely are consistent with the implementation of various prevention strategies. This could be translated to policy interventions for stroke control in countries with a rising trend of the disease.
Statistics from Altmetric.com
The Global Burden of Disease has estimated that cerebrovascular disease (that corresponds to stroke, as defined in this paper) is the second most common cause of death in the world, contributing to 9.5% of mortality in low and middle income countries and 9.9% in high income countries.1 It is also the second most common cause of morbidity in high income countries, responsible for 6.3% of loss of disability adjusted life years, and the fifth cause of morbidity in low and middle income countries, corresponding to 4.5% of disability adjusted life years.1
Stroke is a multifactorial disease where a combination of risk factors operate.2 Risk factors for stroke are of two types: non-modifiable and modifiable risk factors.3 Non-modifiable risk factors include increasing age, male gender and ethnicity. Major modifiable risk factors include hypertension, high serum cholesterol and tobacco smoking.4 5
The large variation in stroke mortality worldwide is likely to be due to differences in environmental factors.6 Overall, 26% of the world's adult population in 2000 had hypertension and this figure is projected to increase to 29% by 2025.7
Our recent study demonstrated the 1950–2000 epidemic curves of coronary heart disease (CHD) for 55 countries.8 Most countries had rising mortality in the early years, reached a peak and then subsequently declined in mortality. The height (deaths per 105) and timing (calendar year) showed interesting and important differences between countries.
In correspondence, Lotufo and Bensenor suggested that stroke can be a major competing cause of death in death certificates.9 They noted that in Brazil, a peak of CHD death was first observed in states where stroke mortality was lower. In the USA, UK and Australia, however, stroke mortality did not move in the opposite direction to CHD mortality.10 Stroke mortality declined continuously from 1950, or soon after, in these three countries. We undertook to examine and report stroke mortality in all countries whose CHD epidemic curves we had published earlier. In this paper, we explore trends in stroke mortality in 48 countries from 1950, or inception of data, to 2005. Our aim was to explore different patterns of stroke mortality in men and women in individual countries and to classify them.
Data and sources
The WHO mortality data are based on official national statistics compiled and transmitted to the WHO by health authorities in member countries. For those countries where vital registration systems cover only part of the country, the WHO applies demographic techniques to adjust the mortality data in its calculation of death rates.11 Official death certification numbers for stroke for 48 countries were derived from the WHO database.11 Only countries with more than 2 million inhabitants in 2000, with less than 5 years of a gap in their data, and with age and sex specific mortality and detailed population figures were included in the present study. All population and mortality data were used as reported by the WHO. For the Russian Federation, Germany and the Czech Republic, data of their former parts were used to plot the secular trend line. The periods at which names changed for these countries were added to the legend of the corresponding graphs.
The international trends in stroke, as seen in death certification, were derived primarily from the WHO Statistical Information System (WHOSIS) online database.11 Due to the significant lag time in availability of stroke data from China in the WHO database, stroke mortality data were extracted from Chinese Ministry of Health Vital Registration (1973–2003), as reported by Yang,12 which is the source of the WHO mortality data.
Over the study interval (1950–2005), there were five revisions of the International Classification of Diseases (ICD6–10: table 1). To minimise problems of classification and data comparability over time, all deaths recorded as ‘vascular lesions affecting central nervous system’ (330–334) in the sixth and seventh revisions of the ICD, and ‘cerebrovascular’ disease (430–438) in the eighth and ninth revisions were combined with those of the 10th revision (I60–I69) of the ICD.
Data for men and women were collected and analysed separately.
For years with missing stroke mortality data, up to 5 years' weighted linear interpolations were made if there was no ICD change in the period. To account for possible mortality discontinuities related to ICD revisions, changes in rates in subsequent calendar periods and between sexes were checked; apart from Belgium, none was relevant and no adjustments were made. The stroke mortality for Belgium increased by 60/105 in 1967 in both sexes, simultaneous to introduction of ICD 8, which is probably due to the ICD change. Stroke data for the former Soviet Union (FSU) were available in WHOSIS from 1980 to 1990. Further data for the period 1965–1980 were published in World Health Statistics Annual 1991.13 The data were reported 5 yearly and linear interpolation between points was made as described in the methods section. Data from Germany and the Czech Republic were aggregated with data from the former German Democratic Republic (before 1990) and Czechoslovakia (before 1992) to plot the trend for the study interval.
The WHO standard population, 2000, was used for direct age standardisation of rates for the population aged 35–74 years.14
Stroke mortality rates were plotted separately for men and women for individual countries by single year for the period (1950–2005); 4–6 countries in the same region or subregion were plotted together in contrasting colours. For some countries, data did not become available until after 1950. Stroke mortality rates were aggregated into the following broad country groups: Anglo-Celtic countries (excluding North America) and Europe, countries of North, Central and South America, the Former Soviet Union states, and Asian and Middle Eastern countries.
In summarising height and timing of peak mortality curves for comparing all 48 countries, male data are presented because they were usually at a higher level. The shape of the female mortality curves was usually similar (figures 1–4). Patterns were classified in a summary matrix based on the shape of the mortality rate curves. These patterns were categorised as: ‘declining’ only, ‘rise and fall’, ‘rising’ only and ‘flat or stable’. Countries were divided by peak male mortality rates per 105 and presented based on the following categories: over 300, 200–299, 100–199 and <100.
Joinpoint regression analyses is an objective method of identifying peak points where the linear slope of a trend changes significantly.15–17 Joinpoints were analysed with the Joinpoint software V.3.2.0 from the US National Cancer Institute (Bethesda, Maryland, USA; http://www.srab.cancer.gov/joinpoint). The software uses a Monte Carlo permutation test to identify points when the trend line changes in magnitude or direction are statistically significantly.18 Up to three joinpoints for each series were considered, assuming a Poisson distribution, as is conventional with mortality data. However, minimum numbers of significant joinpoints were selected. The joinpoints corresponding with a significant decline in the stroke mortality rate were chosen as the peak epidemic points. The peaks or highest numbers for stroke mortality in 13 countries between 1950 and 2000 are plotted in figure 5.
Stroke mortality was plotted for individual countries. To help comparisons, male and female mortalities are shown in separate graphs side by side. The same colour for each country is used for men and women. In general, stroke mortality in women was lower than that for men for each country and showed a similar pattern.
Australia, New Zealand and the UK showed a continuous decline in stroke mortality from 1950. However, the Ireland male figure peaked at 135/105 in 1973. Australia had the highest 1950 stroke mortality in this group, which declined to its lowest in 2004 at 29/105 (figure 1). The decline in Australia, New Zealand and the UK was steeper in all three countries between 1970 and 1985. This is more clearly seen in the male data.
Among Western continental European countries (figure 1), mortality was similar to Anglo-Celtic countries. Among Nordic countries, Finland stood out; at its peak, it had the highest stroke mortality in the Western world, followed by a steep fall in mortality after 1970 to a level similar to other Nordic countries (figure 1). Three other Nordic countries had similar patterns, but even their higher mortality was only about half of that observed in Finland. Denmark, Norway and Sweden had similar stroke mortality. Belgium showed a rise and decline. In Austria, there was a slower decline in stroke mortality in women.
Two European Mediterranean countries, Spain and Italy (figure 1), had stroke mortality rates similar to other Western countries. There was little change in stroke mortality in Greece, with a slight decrease after 1980. Mortality in Portugal, in contrast, rose up to a peak at around 1970 of (M/F) 275/200 which was twice as high as that in other Western countries. It then declined to a level near to neighbouring countries.
The stroke patterns for the USA and Canada were similar, despite the two countries having different healthcare systems (figure 2). The decline was steeper in the USA between 1968 and 1980 and is seen more clearly in men. From 1978 to 2000, the lines for the USA and Canada almost coincided.
Among central and South American countries (figure 2), two patterns were distinguishable. First, a decline since the inception of data was observed in Chile and Panama in men and possibly in Argentina and Cuba. Second, a stable mortality ‘flat pattern’ at <100/105 was observed in Mexico, El Salvador and Peru. In women, declining mortality was seen in Chile and Argentina.
Stroke mortality rose in Eastern European countries until the late 1980s to a level twice as high as that in Western European and Anglo-Celtic countries. Since then, mortality has clearly declined in Bulgaria, Hungary and later in Romania. Male stroke mortality in Poland also declined, but rates for men and women were much lower than those reported in Bulgaria, the Czech Republic, Romania and Hungary. In Romania, rates reached a late peak in both sexes after they were declining in neighbouring countries.
The European and Central Asian republics that emerged from the FSU demonstrated actual epidemic patterns, with increases in all countries to a high level of around 400/105, and in some cases showed subsequent declines as well. Stroke mortality in European states of the FSU, including the Russian Federation, Latvia and Ukraine, showed a pattern similar to Eastern Europe, although peak mortalities were higher in the FSU countries. In the Russian Federation, Ukraine and Belarus, female stroke mortalities were near levels in Western Europe and much lower than rates in men (figure 3). A late decline in stroke mortality (1990) was evident in Russia, Latvia, Ukraine, Azerbaijan and Georgia, while stroke mortality increased in Belarus, Kyrgyzstan, Kazakhstan, Tajikistan and Armenia. There was diversity in recent trends in Central Asian republics (figure 3).
In Asia, Japan had the highest recorded level of stroke mortality in the world, which was 433 per 105 in men in 1957 (figure 4). From 1964 there was a steep decline to less than 100/105 by 2004. Female rates in Japan showed the same pattern and timing but the maximum was lower at 304 in 1957. Stroke mortalities in Singapore and Hong Kong were relatively low compared with Japan but have been declining since the 1970s. An almost parallel decline was observed in South Korea since 1985 (the inception of data), but the present mortality in men and women is double that of Japan, Singapore and Hong Kong.
Rural China had lower stroke mortality compared with urban China. Urban mortality rates in China declined after 1998. Mortality rates have remained very low in Thailand for both sexes (figure 4).
In Israel, a steep decline in stroke was observed from 1975 (inception of data) per 105 from (M/F) 117/110 to 34/19 in 2003. Stroke mortality in Egypt was low during the 1960s and 1970s (<50/105); the data were incomplete thereafter. Stroke mortality in Kuwait per 105 declined steeply from (M/F) 123/102 in 1977 to around 34/27 in 2002. Mortality remained almost the same with some fluctuation thereafter (figure 4).
The information is condensed in table 2 which is a matrix of male stroke epidemic patterns in the 48 countries. The matrix shows how countries from different regions can have similar epidemic patterns. The magnitude of stroke mortality ranges from 433/105 male deaths per year in Japan in 1957 to only about 29/105 deaths per year in Canada and Australia in 2004. The timing of peaks, calculated by joinpoint regression, started in 1964 (Japan) and extended to 1999. Presumably, peaks were before 1950 in Australia, the USA and the UK.
Other patterns were seen in countries with less complete records, including: (a) gradual continuing decline—for example, in Argentina, Panama and Cuba; (b) falling, but records started 1980: Azerbaijan and Tajikistan; (c) little change in mortality: Mexico; (d) apparent late fall: urban China; and (e) very slowly rising (but incomplete records): El Salvador, Thailand and rural China.
The peaks or highest numbers for stroke mortality in 18 countries between1950 and 2000 were plotted. On the vertical axis, the magnitude ranges from around 100/105 to over 433/105 deaths per year in Japan. Timing of the peak, on the horizontal axis according to joinpoint regression, ranged from 1964 to 1999, a span of 35 years (figure 5). Countries not in figure 5 had their highest mortality at 1950. Their peak was either 1950 or earlier.
No peak could be plotted for countries whose stroke mortality has not yet clearly declined. Stroke mortality of some FSU Asian states are now close to the highest previous peak. The last available stroke mortality rates for Kyrgyzstan, Kazakhstan, Uzbekistan and Armenia were added to figure 5 to demonstrate this phenomenon.
In nearly all of these countries, stroke mortality had decreased between 1950 and 2005, often very considerably. In many countries it had reduced by 50% or more; in Japan by 85%.
Decline was continuous from 1950 in the USA, Canada, the UK, Sweden, Denmark, Finland, Italy and Australia, and (from 1980) in Hong Kong. In another group of countries, mortality did not change until it decreased in the early 1970s: Ireland, Norway, Spain, Chile, New Zealand and (from 1963) Singapore. In several of the first group of countries, a faster decline in mortality was seen in the 1970s.
A third group of countries experienced the rise and fall pattern of an ‘epidemic’. In Japan, the peak was 1964,19 in Portugal20 it was around 1970 and in the countries of Eastern Europe and former countries of the Soviet Union, the peak was between 1988 and 2000. In only three countries did stroke mortality clearly rise at the end of the 20th century: Kyrgyzstan, Kazakhstan and Armenia. In other countries the records were incomplete and the trends relatively small.
The magnitude of peak epidemics (where a downturn is recognisable) varied between 136/105 in Ireland to 433/105 deaths per year in Japan, and the timing varied from 1964 in Japan to 1999 in Hungary. In the ‘rising’ group (first part of the epidemic), the highest stroke mortality was in Kyrgyzstan (388/105 in 2005). In the ‘declining’ group, the lowest stroke mortality was in Canada and Australia (24/105 in 2004).
It is instructive to compare stroke mortality patterns with CHD patterns.8 For CHD, it is appropriate to use the term ‘epidemic’. In some major countries in North America and Western Europe, CHD mortality rates after 1950 rose to be very high, then fell, earlier in the West, later in Eastern Europe. They are still rising in parts of the FSU.
However, stroke mortality declined throughout the second half of the 20th century in North America and most Western European countries.15 Ostfeld reported a constant decline in stroke mortality in the USA in all ethnic groups and in both sexes from 1940.21 CHD was a new diagnosis that appeared and increased from1900 to 1950, but stroke had been diagnosed for hundreds of years. For example, in Ontario, stroke mortality stayed about level at 200/105 from 1901 to 1951, and then fell a little to 1961.22
According to the WHO Monitoring of Trends and Determinants in Cardiovascular Disease (WHO-MONICA) project, trends in systolic blood pressure at population level showed a strong association with trend in stroke, particularly in women, when the 3–4 year time lag was considered. A small fraction of the variation in trend for stroke event was explained by a risk score of combining trends in systolic blood pressure, daily cigarette smoking, serum cholesterol and body mass index.23
The high rate of stroke mortality in FSU countries has been linked to alcohol consumption, particularly binge drinking, which leads to high blood pressure, followed by stroke, especially in men.24–26
The reduction in stroke mortality in Japan was much greater than that in any other country and started in 1965. The different dates of downturns or peaks must reflect the advent of environmental conditions. As well as a phenomenal effect of blood pressure, there are reports of a large reduction in salt intake in Japan. Salt (NaCl) intake in Japan reduced from 360 mmol/day in 1950 to 187 mmol/day in 1988.27 Radical early treatment of acute cerebrovascular accidents was not available until recent years, and only in the technologically advanced communities.28
In Belgium men, the prevalence of systolic blood pressure above 159 mm Hg decreased from 51% to 21% between 1967 and 1986. Mean systolic blood pressure decreased from 159 to 142 mm Hg and the proportion of subjects receiving treatment for hypertension increased from 10% to 36%. Mean standardised 24 h sodium excretion decreased from 265 to 188 mmol. The decrease in stroke mortality in Belgium was related to the combined effects of treatment for hypertension and a decrease in sodium intake.29 This is in contrast with Portugal where the diet is traditionally very rich in salt. Although the salt content of the Portuguese diet has been declining in recent decades, Portugal still has the highest rates of stroke mortality in Western Europe. Higher rates of stroke in lower socioeconomic groups and in rural areas suggest a link between stroke mortality and adherence to the traditional Portuguese diet.30
The different dates of mortality downturn could reflect the period when pharmaceutical treatment of hypertension started to be effective in the at risk section of the population. This started in the 1950s with the introduction of ganglion blocking drugs—for example, hexamethonium.31 By the 1970s, more drugs were available, with fewer side effects. Thiazides, β-blockers and methyldopa were all introduced in the 1960s, but general uptake takes time, even in affluent countries, and is likely to have taken longer in Eastern Europe and Central Asia.28
The ICD codes of stroke have been considered comparable since the eighth revision.32 33 Apart from Japan which has had a well described epidemic, the decline in stroke mortality started in some countries in the 1970s when ICD 8 was in place. Furthermore, increasing trends in mortality from stroke in Eastern Europe were not isolated. Thus it is unlikely that the peak mortality, mortality pattern and the present classification were affected systematically across different countries by ICD changes.34
Misclassification and incomplete vital registration in some countries may result in under-enumeration of stroke mortality. The extent to which this has changed over time, within countries, will affect how interpretable secular trends are. Increases in mortality might reflect changes in vital registration coverage, and not real increases in mortality.35 Comparing the results of this study with our previous study on the trend of CHD in various countries,8 there is no clear evidence of CHD and stroke being competing causes of death.
To minimise the risk of misclassification, specific age groups were used and the elderly were excluded from analyses. Also, broad cause of death categories were used to minimise transference to adjacent similar codes. The data were inspected for short term changes potentially attributable to artefact or coding changes.
It would be desirable to compare the presented data with secular changes in the incidence of stroke. Population stroke mortality rates obviously depend on population incidence and population case death but the latter two are difficult to measure in populations such as countries.
Two previous papers explored trends in cardiovascular disease, including stroke.15 32 They reported annual changes and reported similar trends but for fewer countries and shorter periods of time. Our paper covers a longer period and more countries and provides a dynamic picture of a moving phenomenon and classifies the trends according to their patterns.
Unfortunately, the WHO international data for stroke mortality do not differentiate between cerebral haemorrhage and ischaemic stroke. Although hypertension is directly related to cerebral haemorrhage, it is also a risk factor for cerebral thrombosis.36 It is possible that artefactual between country differences may be apparent in the attribution of cause of death for these cases due to ‘cultural’ or ‘historical’ factors.
In conclusion, apart from a few exceptions, the trend for stroke is not parallel to the trend for CHD in various countries. Different patterns of stroke were observed over the past five decades. It is important for policy makers to know whether a mortality rate is increasing, declining or static, and whether CHD and stroke patterns are competing with each other or are independent. The experience of Japan and Belgium could be informative for other countries with rising stroke mortality.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.