The Arbroath Multiple Disease Study

The remainder of this chapter uses the format of the template previously outlined for describing an investigation about a suspected point source of pollution.

Initial Response/Background

Previous surveys of patterns of environmentally sensitive respiratory disease in central Scotland had demonstrated a high mortality from lung cancer in residential areas of towns downwind of foundries.22-24 The town of Arbroath in eastern Scotland contained a central industrial area which housed a foundry.

Complaints of dust and fumes in the neighbourhood had been reported during the 1960s and 1970s. At that time, the fumes from the foundry emanated from a hole in the roof of the building. The absence of a fumestack would have exacerbated the tendency of the pollution to accumulate nearby when winds were light or absent. During the late 1970s and throughout the 1980s, the pollution was diminished, first because of an industrial recession and later because pollution control technology had been installed. Thus we collected data for the years 1966-76. The 10-year period allowed also for a reasonable occurrence of disease in the area.

Hypothesis

Due to concerns about the effects of this centrally located foundry on the health status of the surrounding community, it was decided to examine the spatial distribution of mortality arising for a range of diseases in the town area. Specifically the study aimed to test the hypothesis that deaths from bronchitis, gastric, oesophageal and lung cancer would be raised in areas affected by pollution from the foundry.

Selection of Diseases to Monitor

For all residents in the town, information was extracted from the death certificates for the years 1966-76 on age, sex, address, occupation, and the causes of death. Certificates with any mention of lung cancer were used unless the cancers were secondary to a primary cancer in another tissue. The address of each death was plotted on a map. The addresses of deaths with non-respiratory cancers were also extracted from the death certificates and the locations mapped. Two categories of non-malignant disease were selected from the death certificates: ischaemic heart disease where no other significant disease was mentioned and bronchitis. From the list of cases of ischaemic heart disease control deaths were selected which matched with the respiratory cancer deaths by 10-year age-group and by sex, to the closest year of death and to the closest death certificate number if within the same year.

Cancers of the lower body and deaths from ischaemic heart disease were used as control diseases.

Identification of the Population at Risk

The town was divided into four geographical areas by aggregating the enumeration districts of the 1971 census, which were geographically coherent. Area 1 consisted of the residential zones directly to the east and west of the foundry. This area was considered, a priori, to have been potentially at risk from the air pollution because of the prevailing wind directions in Scotland. The remainder of the town was divided into comparison areas 2 (north and south), 3 (west) and 4 (east). There are no particular topographic effects relating to this study.

Rates of Disease

The expected numbers of deaths from respiratory cancer in the four areas were calculated using indirect standardization. The Scottish rates for 1971 provided the standard population. Standardized mortality ratios were calculated for the age-groups 15 and over. The mean standardized mortality ratios for the years 1966-76 of the town's four areas were derived from the values for the observed and expected numbers of deaths within those areas. The statistical significance of the odds ratio of each standardized mortality ratio was calculated.

Mapping the Distribution of Disease

The locations of all cases dying from ischaemic heart disease in 1966, 1971 and 1976 and of bronchitis 1966-76 were mapped. The grid coordinates of these deaths, and of the deaths from respiratory cancer, all non-respiratory cancer, gastric cancer, colorectal cancer, female breast cancer, prostate cancer and cancer of the bladder and kidney were determined. The distributions of these categories of death were illustrated using three-dimensional mapping.25 The method involves producing surfaces of the local density of cases where elevated areas relate to high density and low areas to low density. The resulting surface maps are produced by an 'optimal' smoothing method (cross-validation).26

Results

There were 168 death certificates with the diagnosis of respiratory cancer with addresses within areas 1 -4 of Arbroath between 1966 and 1976, and 538 with the diagnosis of other cancers. A cluster of deaths from respiratory cancer was found in area 1, and a smaller cluster in part of area 3. When the SMRs were calculated for the areas (Table 5.3), only area 1 showed a high

Table 5.3. Standardized mortality ratios for lung cancer, non-respiratory cancer and coronary heart disease in the four areas of Arbroath, 1966-76

Areas OBS

EXP

SMR

95% CI

Respiratory cancer

1 29

19

153

103 to 220

2 44

47

94

68 to 126

3 36

41

88

62 to 122

4 59

58

101

77 to 130

Non-respiratory cancer

1 102

56

183

149 to 222

2 145

133

109

92 to 128

3 122

117

104

87 to 125

4 169

159

106

91 to 123

Coronary heart disease

1 17

19

90

52 to 143

2 43

47

92

66 to 123

3 51

41

125

93 to 164

4 56

58

96

73 to 125

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

Figure 5.3. Bronchitis case address location map: Arbroath

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

Figure 5.3. Bronchitis case address location map: Arbroath

to 40

T3 X

>*

x*

_

x

Xx

X XX, y *

i , X

* ^

V* y X

■..¿yy

•X

X

V

XX

>

X x

X

k VX x

X

I 1

0 10 20 30 40 50 60 70 80 90 x distance (km tenths)

0 10 20 30 40 50 60 70 80 90 x distance (km tenths)

Figure 5.4. Gastric and oesophageal cancer address location map: Arbroath value. Area 1 contained the highest male age-specific death rates for the age-groups 55-64 and 65-74, and the second highest for the age-group 75+.

Area 1 also contained the highest value for the SMRs for non-respiratory cancer (Table 5.3); the value of 183 was statistically significant. The standardized mortality ratios for the matched controls, who had died of coronary heart disease, did not show significant elevation of the values in area 1 compared to the other areas (Table 5.3).

Figures 5.3 to 5.7 show the spatial distributions of residential locations for deaths from bronchitis, gastric and oesophageal cancer, lung cancer, cancer of the lower body, and ischaemic heart disease.

To act as comparison the spatial distributions of ischaemic heart disease and cancers of the lower body were used as control diseases for bronchitis, gastric and oesophageal cancer, and respiratory cancer. With one control sample we can apply some simple statistical models to assess the relation between the disease location and the foundry location. These models are discussed more fully in Chapter 6. Briefly, we can specify a model for the

-

y * Xv x

:

^ xx y :x y v,*x x

^ X

' 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1

1 1 1 1 1 1

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

Figure 5.5. Lung cancer address location map: Arbroath

"U

"U

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

Figure 5.6. Lower body cancers control address location map: Arbroath

0 10 20 30 40 50 60 70 80 90 100 x distance (km tenths)

Figure 5.6. Lower body cancers control address location map: Arbroath excess risk around a source by defining a local density of cases (A(x)) at location x, which is a function of the local population at risk (represented by the spatial distribution of the control disease, cancers of the lower body) and a function of exposure variables related to the source location. For our example, the function of exposure variables could be simply based on distance from the source. A simple form of relation is given by 1 + a1 exp(—air(x) + ao, log r(x)) where r(x) is the distance of a case location from the source location. Here the three parameters control the degree of association with the source: a1 represents the overall association, while ai and ao represent the degree of evidence related to distance and peak-decline effects (see Chapter 6 for greater detail on these models).

It should be noted that the spatial distribution of the two controls is similar, and this adds weight to the appropriateness of their use in this study. The diseases most likely to show an association with the foundry (respiratory cancer and bronchitis) peak to the southwest and north of the source. Gastric and oesophageal cancers, which were also hypothesized to show an association with the foundry, peak on a northeast-southwest axis. But the

< distance (km tenths)

Figure 5.7. Ischaemic heart disease mortality address location map: Arbroath

< distance (km tenths)

Figure 5.7. Ischaemic heart disease mortality address location map: Arbroath pronounced peak only occurs to the northeast of the foundry. Thus, while respiratory cancer and bronchitis have similar spatial forms, gastric and oesophageal cancer show different spatial forms.

A complete analysis of these data would consider a possible exploratory analysis of smoothed maps of relative risk (see Chapter 4) and further analysis could be based on tests or statistical modelling of the relation of distance and direction from the source to the case locations. These methods are discussed more fully in Chapter 6.

Nicotine Support Superstar

Nicotine Support Superstar

Stop Nicotine Addiction Is Not Easy, But You Can Do It. Discover How To Have The Best Chance Of Quitting Nicotine And Dramatically Improve Your Quality Of Your Life Today. Finally You Can Fully Equip Yourself With These Must know Blue Print To Stop Nicotine Addiction And Live An Exciting Life You Deserve!

Get My Free Ebook


Post a comment