Associations Between Data for Male Lung Cancer and Female Breast Cancer Within Five Countries

FIONA L. R. WILLIAMS, MSC, MPH, AND O W E N L. LLOYD, MD. PHD

Mainly on the basis of associations noted between international rates of male lung cancer and female breast cancer, passive smoking has been suggested recently as a major risk factor for female breast cancer.' In this report, the authors describe the associations between data for female breast cancer and male lung cancer within five countries. For one country (Scotland) the authors examined the mortality rates of these cancers over time, and their relationship to trends for tobacco consumption.The correlations between female breast cancer and male lung cancer were significantly positive in Italy, USSR, and West Germany, weakly positive in Canada, absent in Scotland, and significantly negative in England and Wales. In Scotland, the mortality rates of these cancers over time were significantly correlated. There was a clear relationship between male lung cancer and tobacco consumption, but only a weak relationship between female breast cancer and tobacco consumption.

Cancer 64:1764-1768, 1989.

RADITIONALLY CIGARETTE SMOKING has not been T linked to an increased risk of breast cancer, and the endogenous sex hormones (such as estrogen, prolactin, and progesterone) and diet have been considered the most important causal factors. In a recent report. however, in- door tobacco smoke pollution has been suggested as a major risk factor for both female breast cancer and male lung cancer.' That study was based on analyses of asso- ciations between rates for these diseases from several data sources (including one data set comprising 5 1 countries), and concentrated largely on time trends and on differences between countries; rates within a single country were not examined in detail.

To test the hypothesis that passive smoking was a risk factor in female breast cancer, we investigated the rela- tionship between the frequency of breast cancer and lung cancer within several countries. For Scotland, we also ex- amined the mortality rates for these cancers over time, and related them to tobacco consumption. (Following the approach of Horton,' male lung cancer was used as a proxy for smoking.)

174) were analyzed for the provinces, regions or districts within five countries: Scotland, England and Wales, Can- ada, Italy, USSR, and West Germany.

For Scotland, mortality data for each of the 56 local government districts were obtained from the Annual Re- ports of the Registrar General (ARRG); the expected numbers of deaths were derived from the product of the appropriate mean Scottish cause-specific death rates (by 5-year age group) and the relevant populations at the cen- sus year of I98 1. Mean standardized mortality ratios (SMR) were calculated for the period 1979 to 1983. For the same districts, incidence rates for the two cancers (per 1 00,000, age-standardized to Segi's world population2) were available for the years 1975 to 1980.3 For the 23 standard regions of England and Wales, the SMR (stan- dardized to the population of England and Wales) were obtained for the years 1978 and 1 985.4 For the nine prov- inces of Canada, incidence rates (per 100,000, standard- ized to Segi's world population) were obtained for the

SCOT LAND

Methods Mortality and/or morbidity data for male lung cancer

(ICD8 & 9: 162) and for female breast cancer (ICD8 & 9:

rho 0.07

120 . . . ... . 9 . . . . . . - . .

I. .: . From the Department of Community Medicine, Ninewells Medical

School and Hospital, University of Dundee, Dundee, Scotland. Supported by the Economic and Social Research Council (ESRC award

no. G00232346). Address for reprints: Fiona L. R. Williams, MSc, MPH, Department

of Community Medicine, Ninewells Medical School and Hospital, Uni- versity of Dundee, Dundee DDI 9SY, Scotland.

Accepted for publication April 20, 1989.

40 ! I , 40 60 80 100 120 140

FEMALE BREAST CANCER

FIG. I . Scottish mortality data (SMR) 1979-1983.

I j0

1764

No. 8

120 -.

MALE LUNG A N D FEMALE BREAST CANCER

rho -0.152 .

I765 - Williams and Lloyd

80

70 -.

60 -.

50 -.

40 -.

30

rho 0.233

.=

’ . I I

. . - . m e . ..

l o o $ 80 . .-a i : # Y W

MALE LUNG CANCER

I .. . I : . . ..-. ... - . m - . I

40 ! 1 I

40 50 60 70 80 90

FEMALE BREAST CANCER

FIG. 2. Scottish morbidity data (rate per 100.000) 1975-1980)

years 1973 to 1977.’ For the 95 provinces of Italy, SMR (standardized to the Italian population) were obtained for the years 1975 to 1 977.6 For the 15 republics of the USSR, incidence rates (per 100,000, standardized to Segi’s world population) were obtained for the years 1969 to 1971.’ And for the 326 administrative districts of West Germany, mortality rates (per 100,000, standardized to Segi’s world population) were obtained for the years 1976 to 1980.’

The mean SMR and associated standard error were calculated for each disease. Spearman’s rank order coef- ficient of correlation (rho) was calculated for male lung cancer and female breast cancer, for the subdivisions of each country. (Spearman’s procedure was preferred to Pearson’s product moment correlation because mortality rates, for any disease, are not normally distributed.)

The Scottish mortality rates per 100,000 population, for female breast cancer and male lung cancer, were ob- tained from the ARRG for the years 1930 to 1932, to 1980 to 1982. The association between these cancers and the annual UK tobacco consumption9 (for males and fe- males combined) was shown graphically.

Results

No association was found between male lung and fe- male breast cancer within the 56 districts of Scotland in

MALE LUNG CANCER

FIG. 4

ENGLAND AND WALES

MALE LUNG CANCER

I 90t :. 70 !

85 90 95 100 105 110

FEMALE BREAST CANCER

Fic. 3. English and Welsh mortality data (SMR) 1985.

Canadian morbidity data (rate per 100,000) 1973-1977.

either the mortality data (Fig. 1) or the morbidity data (Fig. 2).

In England and Wales, significant negative correlations were found in both 1978 and 1985. In both years. the data formed two distinct geographical groups: the regions in the north of England constituted the group with high lung cancer and low breast cancer, whereas the remaining (more southern) regions constituted the group with lower lung cancer and high breast cancer. (Both the correlation and the distribution of SMR were similar in 1978 and 1985: Figure 3 shows, as an example, the association in 1985.)

The weak positive association found with the Canadian data (Fig. 4) was slightly strengthened after reanalysis of the data after exclusion of the two “outliers”: Northwest Territories and Newfoundland.

In Italy, the USSR, and West Germany, male lung can- cer and female breast cancer were positively and signifi- cantly correlated (Figs. 5-7).

With the two exceptions of West Germany and England and Wales, the mean rates for lung cancer and breast cancer were clearly dissimilar (Table I ) . Scotland (mor- tality data), Canada and Italy showed lower rates for lung cancer, whereas Scotland (morbidity data), England and Wales, the USSR, and West Germany showed lower rates for breast cancer.

ITALY 200

rho 0.761

O ! 1 1 I 40 60 80 100 120 140 160

FEMALE BREAST CANCER

FIG. 5. Italian mortality data (SMR) 1975-1977.

1766

60

50 -. MALE 40-. LUNG CANCER 30-.

20 -.

1 0 7 I I

rho 0.796

. -.. 9 .

CANCER October 15 1989

80 .. rho 0.455 ;:5. .. .. . 70 -.

60 -. .. ' 5 '

MALE .. LUNG 50 -. CANCER . *

40- . ' ;. 30-* . 1

20 I I

Vol. 64

FEMALE BREAST CANCER FEMALE BREAST CANCER

FIG. 6. USSR morbidity data (rate per lO0,OOO) 1969-1971. FIG. 7. West Germany mortality data (rate per 100,000) 1976-1980.

In Scotland, mortality from male lung cancer rose steeply from around five per 100,000 in the 1930s to around 120 per 100,000 in the 1980s, when it reached a plateau (Fig. 8). The mean annual tobacco consumption (per adult) rose from 2.2 kg in 1930 to 1932, peaked at 3. I kg in 1960 to 1962, and fell to under 2.7 kg in 1970 to 1972. Mortality from female breast cancer in Scotland doubled since the 1930s, rising from around 25 per 100,000 in 1930 to 1932, to around 50 per 100,000 in the 1980s (Fig. 9). Only in the two pre-war periods, 1930 to 1932 and 1940 to 1942, did the rates for breast cancer exceed those for lung cancer. The correlation over time between these cancers was +0.905, P < 0.002.

Discussion

In this report we have concentrated mainly on the as- sociation between male lung and female breast cancer within various countries. The very strong, positive cor- relations between these cancers in Italy and the USSR, and to a lesser extent West Germany, might be considered consistent with the hypothesis described by Horton,' linking passive smoking and an increased risk of female breast cancer. However, the complete lack of correlation between these cancers in Scotland and in Canada, and the strong, negative correlation found in England and Wales would indicate the contrary.

The division of the correlations for England and Wales into two subgroups composed generally of the northern and the southern regions might indicate a causal role for social class factors such as industrialization, urbanization or dietary habits, with smoking being only a confounding variable. The geographical distributions of lung cancer and breast cancer in Scotland, England and Wales, Italy and West Germany, show the highest frequencies of both diseases in areas where there was a preponderance of in- dustry and urbanization. For example, high rates were found in the Ruhr in West Germany and in Northern Italy.

In studies of time trends of diseases spanning several decades, a confident interpretation of the findings is se- verely hindered by the presence of confounding variables. Changes in the ICD coding, variations in diagnostic prac- tice, and reliability of death certification and of the cancer registries, can all affect the validity of a study, especially where differences between countries are being investigated.

Although the strong, positive correlation between male lung and female breast cancer (in Scotland) over time was also consistent with the findings of Horton,' it would be premature to argue that this relationship is causal. Firstly, epidemiologic studies have repeatedly failed to show a relationship between breast cancer and active smoking; if active smoking does not show an association with breast cancer, a link with passive smoking is highly implausible.

TABLE I . Mean Standard Errors and Correlations for Female Breast Cancer and Male Lung Cancer

Scotland, England, & Wales West Disease Scotland Canada Italy USSR Germany

parameters 1979- I983 1975- 1980 1985 1973- 1977 1975-1977 1969-1971 1976-1980

Lung cancer 87 (3) 78 (3) 101 (3) 49 (4) 91 (4) 33 (3) 47 (0.6) Breast cancer 101 (2) 59 ( 1 ) 99 ( I ) 64 (2) 93 (2) 15 ( 1 ) 20 (0.2) N 56 56 23 9 95 15 326 RHO 0.07 -0.15 -0.599 0.233 0.761 0.796 0.455 P value 0.608 0.264 0.002 0.546 0.000 0.000 0.000

Mean rates (SE)

No. 8 MALE LUNG AND BMALE BREAST CANCER - Williams and Lloyd 1767

FIG. 8. Scottish mortality rates from lung cancer, and tobacco consumption.

TOBACCO CONSUMPTION PER ANNUM W S )

T A

- MALE LUNG CANCER

1920- 1930- 1940- 1950- 1960- 1970- 1980- 22 32 42 52 62 72 82

YEAR

A second reason for skepticism comes from the differences in the long-term trends shown between lung cancer and breast cancer in countries such as Scotland, England and Wales, and the United States. For example, tobacco con- sumption in the United Kingdom became widespread during World War I , and although the marked increase in the male lung cancer rates during the 1940s was con- sistent with a 20-year latency between initiation and overt cancer (Fig. 8), proof of the causal relationship was con- sidered indisputable only after the large trial undertaken by Doll and Peto"; the leveling off of the rates after the

FIG. 9. Scottish mortality rates from breast cancer, and tobacco consumption.

TOBACCO CONSUMPTION

ILBS) PER t 2l 1

120

100

80

60

40

20

0

MORTALITY RATE PER 1 00,OOO

mid 1970s, and their subsequent decline in the 1980s, were consistent with the reduction of tobacco consump- tion in populations, and the increasing use of low-tar or filter-tipped cigarettes.

By contrast, extrapolation of the data in Figure 9 in- dicated that the rates for female breast cancer differed from those for male lung cancer in being already high at a time when tobacco consumption was negligible; this observation, and the evidence for only a gentle increase in the incidence of breast cancer over time, were not con- sistent with tobacco consumption (either passive or active)

0 4 I

MORTALITY RATE PER 100,000

. 20

- 10

I 0 1920- 1930- 1940- 1950- 1960- 1970- 1980-

22 32 42 52 62 72 82

YEAR

1768 CANCER October 15 1989 Vol. 64

forming a major component of the etiology of this cancer. The increase in breast cancer seen between 1930 to 1932 and 1987 could equally well be attributed to the noticeable change in diet which followed World War 11. Moreover, further evidence questioning the importance of tobacco consumption can be found in migrant studies: for in- stance, very low rates of breast cancer are found in native Japanese (despite their large tobacco consumption9), whereas higher rates are found in Japanese who have em- igrated to areas with high rates, where diet and other en- vironmental factors might be different."

The validity of using male lung cancer as a proxy for male smoking can be questioned, given the inconsistent epidemiologic association between lung cancer and bron- chitis. Both diseases are primarily attributable to smoking habit, and both show clear dose-response relationships between smoking habit and incidence. Yet, in Scotland between 1972 and 1985, male lung cancer remained static at around 120 per 100,000 and female lung cancer dou- bled to about 40 per 100,000, whereas male chronic bron- chitis declined from 75 to 22 per 100,000 and female chronic bronchitis declined from 22 to 10 per 100,000.12 Although the decline in rates was evident in all age groups and for both sexes, it could be argued that the decline in chronic bronchitis rates reflected improved medication, and that therefore the use of lung cancer as a proxy for smoking was valid.

In conclusion, the hypothesis linking exposure to pas- sive smoking with the etiology of breast cancer was not

supported consistentjy by the examination of the popu- lation data in this study. The inconsistencies of the cor- relation patterns might have resulted from the effects of other social differences such as diet, urbanization, and occupation. Carefully designed case-control studies are necessary to ascertain whether or not exposure to passive smoking increases the incidence of female breast cancer.

REFERENCES

1. Horton AW. Indoor tobacco smoke pollution: A major risk factor for both breast and lung cancer? Cancer 1988; 62:6-14.

2. Segi M. Atlas of Cancer Mortality in Japan by Cities and Counties, 1969-1971. Tokyo: Daiwa Health Foundation, 1977; 1-10,

3. Kemp I, Boyle P, Smans M, Muir C. Atlas of Cancer in Scotland 1975- 1980: Incidence and Epidemiological Perspective, IARC no. 72. Lyon: IARC, 1985; 113-120, 129-131.

4. Office of Population Censuses and Surveys. Mortality Statistics: 1978 and 1985. London: HMSO, 1980 and 1987; 2-1 1.4-13.

5. Waterhouse J, Muir C, Shanmugaratnam K, Powell J. Cancer In- cidence in Five Continents, vol. 4. Lyon: IARC, 1982; 2 18-257.

6. Cislaghi C, Decarli A, La Vecchia C, Laverda N, Mezzanotte G , Smans M. Data, Statistics and Maps on Cancer Mortality: Italia 1975/ 1977. Bologna, Italy: Pitagora Editrice, 1986; 374-385, 410-421.

7. Merabishvili VM, Parkin DM, Smans M, Muir CS. Cancer Inci- dence in the USSR, IARC scientific publication no. 48. Lyon: IARC,

8. Becker N, Frentzel-Beyme R, Wagner G. Atlas of Cancer Mortality in the Federal Republic ofGermany, ed 2. Berlin: Springer-Verlag, 1984.

9. Lee PN. Tobacco Consumption in Various Countries, research paper 6, ed. 4. London: Tobacco Research Council, 1975; 80-81.

10. Doll R, Peto R. Mortality in relation to smoking: Twenty years' observations on male British doctors. Br Med J 1976; 2: 1525-1536.

1 1 . Schottenfield D, Fraumeni JF Jr. Cancer Epidemiology and Pre- vention. Philadelphia: WB Saunders, 1982; 1-50.

12. Unpublished data, 1989.

1982; 88-93.