Q J Med 2003; 96:45–52 doi:10.1093/qjmed/hcg005
Changes in seasonal deaths from myocardial infarction V.L.S. CRAWFORD, M. M CCANN and R.W. STOUT From the School of Medicine, Queen’s University Belfast, Belfast, UK Received 22 January 2002 and in revised form 10 October 2002
Summary Background: Cardiovascular disease is the major contributor to excess morbidity and mortality in winter. With the rise in temperatures through global warming, and the use of central heating and air conditioning, this seasonal variation may be declining. Aim: To study possible changes in seasonal variation in case-fatality rates of myocardial infarction (MI), in men and women, over a 20-year period and compare this with possible environmental influences. Design: Retrospective analysis of death certificate and climatological data. Methods: We analysed all monthly death certificate data from Northern Ireland, for death caused by MI from 1979 through 1998 (n = 68 683). Mortality data were standardized to a single reference group
for the whole period. Seasonal variation in mortality and in environmental variables was described using the cosinor model. Results: A total of 29 875 women and 38 808 men died from MI during the 20-year period. A significant decrease in mortality from MI was observed in both sexes, accompanied by a non-significant decline in the amplitude of the seasonal rhythm over the study period. Low temperature was associated with higher mortality rates from MI. Discussion: We have documented an overall decline in cardiovascular mortality from 1979 to 1998, together with a small but non-significant decrease in seasonal variation. While improvements in medical care, lifestyle, housing and diet may have contributed to the observed decline in mortality rate, seasonal fluctuations remain a significant problem.
Introduction In 1926. the first correlation was noted between seasons and attacks of coronary thrombosis.1 Since then, seasonal variations in cardiovascular mortality have been documented in both northern and southern hemispheres, normally with an increase in winter.2–5 Peaks have also been observed in autumn, spring and in summer, although this is less common6–8 and the seasonal rhythm is low or absent near the equator and in sub-polar regions.9 A seasonal trend in specific acute myocardial infarction (MI), peaking in winter and lowest in summer, has been observed extensively in the US and Europe.10–12 Both low and high temperatures increase MI mortality, therefore climatic trends and changes to indoor environments should impinge upon the observed seasonal variation.13 With global warming,
a decrease in the amplitude of the circannual rhythm would be expected.14 Additionally, increased use of central heating and air-conditioning should alleviate temperature stress in winter and summer respectively. A study in the US reported a decline in seasonality of coronary heart disease mortality from 1939–1970, after which time the trend reversed.15 The authors suggest that this is in keeping with the gradual increase in adequate heating and the subsequent increase in use of air-conditioning. In developed countries, death rates for cardiovascular disease (CVD) have decreased by nearly 60% from their peaks in the 1960s and 1970s. Northern Ireland, with one of the highest mortality rates in the world for ischaemic heart disease, has shown a late start in this decline.16 The MONICA
Address correspondence to Dr V.L.S. Crawford, School of Medicine, Department of Geriatric Medicine, Queen’s University Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7BL. e-mail:
[email protected] ß Association of Physicians 2003
46
V.L.S. Crawford et al.
(National Monitoring of Trends and Determinants of Cardiovascular disease) project, a ten-year study of cardiovascular disease between 29 populations in 19 countries, showed that most of the variance in CVD mortality was unexplained.17 Campaigns to address factors that contribute to CVD such as smoking, hypertension, hyperlipidaemia together with improved treatment can explain, at most, 50% of the decline in myocardial infarction.18,19 The two temporal attributes of MI mortality under consideration are its secular trend and its seasonal variation, and the two may be closely linked. While the first is carefully monitored, the second is less so. This epidemiological investigation reports the changes in the circannual rhythm in mortality from myocardial infarction (MI), over the period 1979–1998 in Northern Ireland. The influences of temperature, sunshine and air pollution are investigated in relation to the mortality rhythms.
Methods Mortality and environmental data Mortality data for the years 1979–1998 were obtained from the Registrar General’s Office, Northern Ireland. The International Classification of Disease (9th edition) was used to select deaths from MI (code 410). Mean minimum temperature and hours of sunshine for each month (1979–1998) were supplied by the Meteorological Office, from a central monitoring station in Aldergrove, Co Antrim. The environmental data therefore covered the same geographical area as the mortality data. Mean monthly levels of PM10 (particulate matter with diameter -10 mm) for the period 1992–1997 were provided by the National Atmospheric Emission Inventory for the central monitoring station in Belfast City Centre.
Cosinor models20 were fitted to each annual cycle to generate estimates from which the mesor, acrophase and amplitude could be calculated using a program developed in house. The mesor (M) (midline-estimating statistic of rhythm) or annual mean is where the fitted sinusoidal curve intersects with the y-axis, giving a rhythm-adjusted mean. The amplitude (A) is the distance from the mesor to the highest value on the fitted sinusoidal curve, or half the difference between the highest and lowest values. The amplitude represents half of the seasonal variation. The acrophase (w) is the time when the peak occurs, measured in degrees from 08. 0– 308 therefore represents the month of January, 30–608 February, etc. The acrophase for mean minimum temperature and for hours of sunshine represents the peak value and therefore 1808 were subtracted to obtain the time of the lowest mean temperature and minimum hours of sunshine. The acrophase for the mean level of pollution (PM10) needed no adjustment, since it is peak levels that are thought to affect cardiovascular mortality. The statistical significance of each seasonal rhythm was determined by F-test (F1-a (2,n-3)) of the amplitude A/0. The critical value of F is 4.256 when a = 0.05 with (2,9) degrees of freedom. A p value )0.05 or rejection of the zero amplitude (no rhythm) that is A/0 assumed that the fitted curve approximates the data more closely than does a straight line with zero slope. Acceptance of the hypothesis A = 0 implies that either the data is essentially constant with time, or the single cosinor model is otherwise inappropriate. A peak-to-trough ratio was calculated for each annual cycle. This expresses the seasonal variation as a percentage of the mean or mesor, therefore taking into account any change in the rate of mortality. Linear regression was used to measure the change in seasonal variation over the period of study. A quadratic term was also used to test for curvature in the relationship between peak-to-trough ratio and year.
Rhythm analysis Cosinor models were used to quantify seasonal rhythms in mortality and the environmental variables over the 20 years studied. Mortality data were standardized to a single reference population from the whole period prior to cosinor analysis. Mortality data were thus adjusted for any change in the age and sex structure of the population. Month and year length were also accounted for in determining the standardised mortality ratio (SMR). SMR ratios for each month throughout the 20 years were used for the cosinor analysis.
Results A total of 68 683 deaths from MI (38 808 males; 29 875 females) occurred during the 20 years studied. Both men and women showed significant seasonal variation in mortality for most years studied, with a decline in mesor over the study period (Table 1, Figure 1). Males declined from an annual mean SMR of 132 (1979) to 66 (1998), with the most rapid decline occurring in the period 1990–1998. The mean annual decline from
Seasonal deaths from myocardial infarction
47
Table 1 The seasonal parameters determined from cosinor modelling: mesor, amplitude, acrophase and peak-to-trough ratios are indicated by year and sex, for deaths from myocardial infarction Year
Males 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Females 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Mesor (SMR)
95%CI for mesor
Amplitude (SMR)
Acrophase w (8)
Month of peak
F
p
Peak/trough ratio
132 131 119 114 115 122 120 117 113 112 114 97 91 89 82 80 79 70 69 66
126–139 126–136 113–126 108–119 109–123 116–127 113–127 112–123 108–118 107–117 98–130 88–105 84–98 83–95 78–86 76–84 74–84 66–73 65–73 60–72
22 19 17 19 21 24 21 24 8 15 3 14 14 11 9 9 11 10 12 8
64 36 47 35 60 68 53 58 56 72 37 40 58 63 46 98 56 37 55 128
Early March Early February Mid-February Early February Late February Early March Late February Late February Late February Early March Early February Mid-February Late February Early March Mid-February Early April Late February Early February Late February Early April
15.2 19.7 9 14.3 10.5 22.9 10.9 26.2 3.6 12.1 5.4 4.1 4.9 4.8 5.8 6.9 6.3 12.6 12.4 2
0.01 0.001 0.01 0.01 0.01 0.001 0.01 0.001 NS 0.01 0.05 NS 0.05 0.05 0.05 0.05 0.05 0.01 0.01 NS
1.400 1.339 1.333 1.400 1.447 1.490 1.424 1.516 1.152 1.309 1.054 1.337 1.364 1.282 1.247 1.254 1.324 1.333 1.421 1.276
124 118 113 110 111 111 121 114 112 113 111 98 93 91 91 90 88 78 74 65
114–134 111–125 106–121 106–114 105–118 103–120 115–127 108–120 106–118 106–121 97–125 90–107 88–98 87–95 86–96 83–97 83–93 73–83 68–80 57–73
25 15 16 23 21 20 21 22 17 15 18 18 19 11 15 11 17 9 15 11
68 48 44 48 40 38 39 42 33 33 12 12 49 63 34 53 29 1 33 98
Early March Mid-February Mid-February Mid-February Mid-February Early February Early February Mid-February Early February Early February Early January Early January Mid-February Early March Early February Late February Late January Early January Early February Early April
8.3 5.9 5.9 48 13.6 6.7 15 15.4 10.1 4.5 2 5.8 20.4 8.5 11.5 3.2 16.1 3.7 8.6 2.435
0.01 0.05 0.05 0.0001 0.01 0.05 0.01 0.01 0.01 0.05 NS 0.05 0.001 0.01 0.01 NS 0.01 NS 0.01 NS
1.503 1.300 1.337 1.537 1.472 1.426 1.413 1.468 1.346 1.299 1.379 1.450 1.497 1.272 1.398 1.270 1.472 1.245 1.492 1.390
1979–1989 was 1.8 (b = 1.59, SE = 0.422, R2 = 31%, F = 14.21, p-0.001) compared to 5.3 (b = 3.84, SE = 0.374, R2 = 81%, F = 105.5, p-0.0001) for the period 1989–1998. Females too showed a steady decline in SMR from 124 (1979) to 65 (1998) with a mean annual decline of 1.3 (NS) in the first decade of the study compared with 5.1 (b = 3.66, SE = 0.476, R2 = 70%, F = 59.02, p-0.0001) in the second decade (Table 1, Figure 1).
Figure 2 shows the 20 data points depicting the annual ratio of peak-to-trough occurrence of MI deaths in Northern Ireland during the study period. For no year or sex does the ratio fall below one, indicating that seasonality is present. A linear trend line through these data points gave a slope of 0.0058 (SE = 0.004, R2 = 10%, t = 1.4, p = 0.179) for males and a slope of 0.0030 (SE = 0.003, R2 = 4%, t = 0.87, p = 0.394) for females. Neither slope was significant, nor was a
48
V.L.S. Crawford et al.
Figure 1. Cosinor models with 95% confidence bands for mortality from myocardial infarction in a males and b females, 1979–1998.
quadratic trend, which did not provide a significantly improved fit to the data. The actual trend is more complex, as demonstrated by the R2 values. Thus while Figure 1 depicting SMRs of seasonal deaths indicates an apparent decline in seasonal variation over the study period, this is shown to be statistically insignificant when interpreted in the context of the fall in the mean, i.e. when peak-to-trough ratios are examined. Figure 3 shows the changes over 20 years in absolute values for ambient mean minimum temperature and hours of sunshine for Northern Ireland. The mean annual minimum temperature increased by 0.04 8C per year over 20 years (SE = 0.02, R2 = 17%, t = 3.8, p = 0.07). The annual hours of
sunshine increased by 0.25 h (SE = 0.37, R2 = 2%, t = 0.47, p = 0.50), with 1989 and 1995 having the greatest amount of sunshine. The mortality peaks for both sexes occurred predominantly in February, coincident with lowest temperatures, but approximately one and a half months after the fewest hours of sunshine (Figure 4). The acrophase or timing of the peak for significant rhythms for MI varies within the range 358 to 988 (early February–early April) for males and for females within the range 128 to 688 (midJanuary to early March). In general, females showed less fluctuation in amplitude from year to year, and tended to peak slightly earlier in each annual cycle than did males (Table 1, Figure 1, Figure 4).
Seasonal deaths from myocardial infarction
49
Figure 2. Annual ratio of peak-to-trough occurrence of deaths from myocardial infarction in a males and b females, 1979– 1998. Fitted lines represent linear regression.
There was no significant change over the 20 years studied in the time of peak for the lowest temperature (b = 0.28, SE = 0.29, R2 = 5%, t = 0.92, p = 0.35); peaks occurred consistently in mid-February (Figure 4). However, the fewest hours of sunshine have occurred later each year (b = 1.1, SE = 0.27, R2 = 46%, t = 15, p = 0.001), changing from a late December peak to a mid-January peak (Figure 4). Examination of the 1988/1989 cycle transition for mean minimum temperatures showed a difference compared with other such cycle boundaries. The December of 1988 was the warmest over all cycles examined, and the mildest since 1934. The subsequent 1989 cycle was also the warmest year seen
in Northern Ireland since 1959. However, this year had a December temperature 0.7 8C lower than normal—the coldest December since 1981. The mean level of air particulate matter (PM10) has decreased in the period 1992–1997. Only 1996 had a significant seasonal rhythm, with other years showing poor fit of the cosinor model. The 1996 peak for PM10 occurred after the peak mortality in MI.
Discussion We report a seasonal rhythm in the mortality from MI, with a consistent peak in winter and similar amplitudes in both sexes, in agreement with other
50
V.L.S. Crawford et al.
Figure 3. Plot showing the absolute values for mean minimum temperature and hours of sunshine for each year 1979–1998 in Northern Ireland. Regression lines indicate the trend over the two decades.
Figure 4. Plot comparing the time of peak MI mortality, with time of lowest temperature and fewest hours of sunshine. Significant annual rhythms in each annual cycle from 1979–1998 are presented.
investigations.11,21All but one statistically significant rhythm peaked within the first 3 months of the annual cycle, with the majority peaking in February. The conformity of this rhythm in MI over the 20 years studied suggests the existence of endogenous biological or exogenous factors, which either work alone, or together, to effect the seasonal rhythm. The importance of one or the other may be difficult to separate. Seasonal deaths from MI may be due in part to endogenous physiological rhythms in cardiovascular risk factors. For example, large annual blood
pressure amplitude is observed, higher in winter than in the summer,22,23 and seasonal variation is seen in blood measurements such as fibrinogen,24–26 haemocrit, white cell count, platelet count27,28 and hormones.29 Fibrinogen levels are associated with cardiovascular disease, and are known to be a major risk factor for both myocardial infarction and stroke,30 and to increase during infections. Low body temperature is also associated with increased blood viscosity and platelets.27,31These effects are associated with thrombus formation and may contribute to cardiovascular deaths in winter.
Seasonal deaths from myocardial infarction
Exogenous factors such as climate, diet, activity levels, heating and air conditioning may react directly or indirectly with endogenous rhythms such as blood pressure, fibrinogen levels and platelets, to make the body more vulnerable to the occurrence of a cardiac event. Despite the changes in many of these exogenous factors over recent decades, publications on the secular trend in seasonal variation in cardiovascular disease are few. One significant study by Seretakis et al. (1997) reported a sharp decline in peak-to-trough ratio (seasonal variation) from 1937–1970 (b = 0.006) for coronary deaths in the entire US, followed by a reversal of this trend in the period 1970–1991 (b = 0.0026).15 Overall from 1937–1991, a linear trend gave a slope of 0.0024 (SE 0.0005). Our study period equates approximately to the second time frame of this US study. However, we report a negative slope similar to that calculated for the first time frame of the US study. Linear regression shows this decline in NI to be non-significant; significance or otherwise of the US slopes are not reported. Nonetheless, it is possible that we are seeing the same response to the same triggers in both the US and NI, with the changes in the triggers in NI lagging behind the same changes in the US. The US study proposed that the increase in second time frame of their study resulted from the increased use of air conditioning, which blunted the effects of heat waves on coronary mortality. The enhancement of the seasonal pattern was therefore due to the lack of summer deaths.15 In NI, the use of air-conditioning is minimal, and central heating is only now prevalent. One European study reported similar findings to both the US and NI studies, but for all-cause mortality. They found a steady decline in amplitude of the seasonal rhythm over a period of 50 years, coincident with the increased use of central heating and improvements in the public health system.32 We found that mean monthly minimum temperatures were negatively correlated with MI mortality, as observed in other studies.31We also formulated Poisson models to calculate the percentage change in death rate associated with a unit increase in temperature (unreported). Mortality data used was 1979– 1990. After controlling for season with respect to temperature, the death rate for both sexes for MI decreased by 2.2% (95% CI 3.3–1.1, p-0.001) for each 1 8C increase in temperature. In the twentieth century, the average annual temperature in most of Europe increased by about 0.8 8C.15Deaths affected by the cold weather would therefore be expected to decrease. Temperatures in Northern Ireland are considered moderate, and average monthly temperatures rarely exceed 17 8C. Over the period
51
1979–1998, the mean minimum temperature has risen by 2 8C (from 5 8C to 7 8C). We report a significant decline in MI fatalities coincident with the rise in temperature. The increased temperature, together with the expansion of central heating, may have contributed to the small nonsignificant decline in seasonal mortality rates in Northern Ireland. The lowest number of hours of sunshine per month occurred from the end of December to midJanuary, approximately 1–1.5 months before the lowest minimum temperature and highest mortality rates. The time lag between the two events suggests that minimal sunshine is unlikely to trigger MI. This study cannot therefore confirm any significant temporal relationship between the level of sunshine and mortality. There have been many studies that have related increased mortality rates with short- and long-term exposure to air pollutants such as PM10, especially in cardiovascular and respiratory diseases.33,34 Seaton et al. proposed that air pollutants provoke an inflammatory response in the lungs, which causes changes in the blood coagulability and increases susceptibility to cardiovascular death.35A detailed study did not support this hypothesis, but suggested that the increased frequency of mortality from cardiovascular disease during episodes of high particulate air pollution may be related to sequestration of red blood cells in the circulation.36 Air pollution is linked to elevated levels of blood viscosity and fibrinogen. The present study examined six years of PM10 data and found only one significant rhythm in PM10 levels, which occurred in 1996. The highest peak of PM10 in 1996 occurred less than a week after the peak for mortality from MI. This information is insufficient to comment on an association between PM10 and MI mortality. We therefore conclude that in NI, while the rate of mortality from MI has declined from 1979 to 1998, the seasonal variation of the deaths remains, emphasizing the importance of this phenomenon. The impact of climate change and other environmental changes upon these trends requires further study.
References 1. Wolff L, White PD. Acute coronary occlusion. Boston Med Surg J 1926; 195:13. 2. Rosenwaike I. Seasonal variation of deaths in the United States 1951–1960. Am Stat Assoc J 1966; 9:706–19. 3. Marshall RJ, Scragg R, Bourke P. An analysis of the seasonal variation of coronary heart disease and respiratory disease mortality in New Zealand. Int J Epidemiol 1988; 17:325–30.
52
V.L.S. Crawford et al.
4. Seto TB, Mittlemann MA, Davis RB, Taira DA, Kawachi I. Seasonal variation in coronary artery disease mortality in Hawaii: observational study. Br Med J 1998; 316:1946–7.
21. Spencer FA, Goldberg RJ, Becker R, Gore JM. Seasonal distribution of acute myocardial infarction in the second National registry of myocardial infarction. J Am Coll Cardiol 1998; 31:1226–33.
5. Kloner R, Poole WK, Perritt RL. When throughout the year is coronary death most likely to occur? A 12-year population-based analysis of more than 220,000 cases. Circulation 1999; 100:1630–4.
22. Woodhouse PR, Khaw KT, Plummer M. Seasonal variation of blood pressure and its relationship to ambient temperature in an elderly population. J Hypertens 1993; 11:1267–74.
6. Douglas AS, Dunnigan MG, Allan TM, Rawles JM. Seasonal variation in coronary heart disease in Scotland. J Epidemiol Community Health 1995; 49:575–582.
23. Brennan PJ, Greenberg G, Miall WE, Thompson SG. Seasonal variation in arterial blood pressure. Br Med J 1982; 285:919–23.
7. Peters RW, Brooks MM, Zoble RG, Liebson PR, Seals AA. Chronobiology of acute myocardial infarction: Cardiac Arrhythmias Suppression Trial (CAST) experience. Am J Cardiol 1996; 78:1198–201.
24. Stout R, Crawford V. Seasonal variations in fibrinogen concentrations among elderly people. Lancet 1991; 338:9–12.
8. Enquselassie F, Dobson AJ, Alexander HM, Steele PL. Seasons, temperature and coronary disease. Int J Epidemiol 1993; 2:632–6.
25. Woodhouse PR, Khaw K, Plummer M, Foley A, Meade TW. Seasonal variations of plasma fibrinogen and factor VII activity in the elderly: Winter infections and death from cardiovascular disease. Lancet 1994; 343:435–9.
9. Douglas S, Rawles J. Latitude-related changes in the amplitude of annual mortality rhythm. The biological equator in man. Chronobiol Int 1999; 16:199–212.
26. Van der Bom JG, de Maat MPM, Bots ML, Hofman A, Kluff C, Grobbee DE. Seasonal variation in fibrinogen in the Rotterdam study. Thromb Haemost 1997; 78:1059–62.
10. Marchant B, Ranjadayalan K, Stevenson R, Wilkinson P, Timmis AD. Circadian and seasonal factors in the pathogenesis of acute myocardial infarction: the influence of environmental temperature. Br Heart J 1993; 69:385–7.
27. Elwood PC, Beswick A, O’Brien JR, Renaud S, Fifield R, Limb ES, Bainton D. Temperature and risk factors for ischaemic heart disease in the Caerphilly prospective study. Br Heart J 1993; 70:520–3.
11. Eng H, Mercer JB. Seasonal variations in mortality caused by cardiovascular diseases in Norway and Ireland. J Cardiovasc Risk 1998; 5:89–95.
28. Keatinge WR, Coleshaw SR, Cotter F, Mattock M, Murphy M, Chelliah R. Increases in platelet and red cell counts, blood viscosity, and arterial pressure during mild surface cooling: factors in mortality from coronary and cerebral thrombosis in winter. Br Med J 1984; 289:1405–8.
12. Peckova M, Fahrenbruch CE, Coff LA, Hallstorm AP. Weekly and seasonal variation in the incidence of cardiac arrests. Am Heart J 1999; 137:512–15. 13. Eurowinter Group (Keatinage WR, Donaldson GC, Bucher K, Jendritsky G, Cordioli E, Martinelli M, Dardanoni L). 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; 349:1341–6. 14. Kovats RS, Haines A, Rosalind SS, Martens P, Menne B, Bertollini R. Climate change and human health in Europe. Br Med J 1999; 318:1682–5. 15. Seretakis D, Lagiou P, Lipworth L, Signorello LB, Rothman KJ, Trichopoulos D. Changing seasonality of mortality from coronary heart disease. JAMA 1998; 278:1012–14. 16. Thom TJ, Epstein F. Heart disease, cancer and stroke mortality trends and their interrelations: an international perspective. Circulation 1994; 90:574–82. 17. Luepker RV. Population versus clinical views in coronary disease: can epidemiological data be useful for clinicians? Circulation 1997; 96:3836–7. 18. Wilmshurst P. Temperature and cardiovascular mortality. Br Med J 1994; 309:1029–30. 19. Lenfant C. Task force on research in epidemiology and prevention of cardiovascular diseases, a revisit. Circulation 1996; 93:1605–7. 20. Nelson W. Methods for Cosinor rhythmometry. Chronobiologia 1979; 6:305–23.
29. Nicolau GY, Haus E, Popescu M, Sackett-Lundeen L, Petrescu E. Circadian, weekly and seasonal variations in cardiac mortality, blood pressure and catecholamine excretion. Chronobiol Int 1991; 8:149–59. 30. Qizilbash N, Jones L, Warlow C, Mann J. Fibrinogen and lipid concentrations as risk factors for transient ischaemic attacks and minor ischaemic strokes. Br Med J 1991; 303:605–9. 31. Lloyd EL. The role of cold in ischaemic heart disease: A review. Public Health 1991; 105:205–15. 32. Lerchl A. Changes in the seasonality of mortality in Germany from 1946 to 1995: the role of temperature. Int J Biometeorol 1998; 42:84–8. 33. Prescott GT, Cohen GR, Elton RA, Fowkes FGR, Agius RM. Urban air pollution and cardiopulmonary ill health: A 14.5 year time series study.Occup Environ Med 1998; 5:697–704. 34. Ostro BD, Hurley S, Lipsett MJ. Air pollution and daily mortality in the Coachella valley, California: A study of PM10 dominated by coarse particles. Environ Res 1999; 81:231–8. 35. Seaton A, Macnee W, Donaldson K, Godden D. Particulate air pollution and acute health affects. Lancet 1995; 345:176–8. 36. Seaton A, Soutar A, Crawford V, Elton R, McNerlan S, Cherrie J, Watt M, Agius R, Stout R. Particulate air pollution and the blood. Thorax 1999; 54:1027–32.