David A Leon, George Davey Smith

David A Leon
reader in epidemiology

George Davey Smith
professor of clinical epidemiology

Correspondence to: D A Leon d.leon{at}lshtm.ac.uk

Mortality from stomach cancer and stroke shows an international correlation, consistent inverse socioeconomic gradients, a particular dependence on socioeconomic circumstances in childhood, ⁽¹⁾ and parallel patterns of decline in most industrialised countries over the past 30-40 years. The plausibility of the hypothesis that salt intake underlies this similarity has been weakened over the past decade as evidence for Helicobacter pylori as the main factor in the aetiology of non-cardia stomach cancer has increased. ⁽²⁾H pylori is thought to be acquired in childhood, and risk of infection is closely related to living conditions, hygiene, and housing standards. Geographical, socioeconomic, and secular variations in the prevalence of H pylori fit well with the corresponding trends and differences in mortality from stomach cancer between and within countries. ⁽²⁾

Infant mortality in the early part of the 20th century indicated living conditions and, in particular, standards of hygiene. We have investigated how far international variations in infant mortality in the past predict adult deaths today from stomach cancer, stroke, and other causes.

Subjects, methods, and results

Death rates from stomach cancer and other causes were obtained from a database of the World Health Organization (www.who.int/whosis/mort/download.htm). We calculated sex specific mortality in 1991-3 for people aged 65-74 years (standardised to the European standard population), who were thus born around the year 1922 (range 1917-28). Infant death rates for 1921-3 (or 1920-24 when 1921-3 rates were not available) were obtained from various sources. ^{(3) (4)} Infant death rates for 1991-3 were obtained from the WHO Health for All database (www.who.dk/country/country.htm). ⁽³⁾ The 27 countries for which all variables were available and where death registration is believed to be complete were included in the analyses. Standards of certification of cause of death will, however, vary across these countries.

The figure shows close relation between infant mortality 1921-3 and stomach cancer rates 1991-3 for men aged 65-74. Data for women are similar (table). One of the striking features of the figure is that Japan and Russia are close to one another, both having historically very high rates of infant mortality and current high rates of stomach cancer. Their current socioeconomic circumstances differ dramatically as does overall mortality (in 1993 Japan had the highest life expectancy in the world, while Russia had the lowest of any industrialised country), which suggests that the association is not the result of confounding with current socioeconomic circumstances.
 

Figure 1

Infant mortality 1921-3 compared with mortality from stomach cancer 1991-3 for men aged 65-74 years in 27 countries
 

Table 1  Relation of adult mortality (age 65-74 years in 1991-3) with infant mortality at time of birth and at time of death for 27 countries*
 

	Infant mortality 1921-3						Infant mortality 1991-3
	Male			Female			Male			Female
	Coefficient	P value		Coefficient	P value		Coefficient	P value		Coefficient	P value
Pearson correlation coefficients:
All causes	0.52	0.005		0.51	0.007		0.58	0.002		0.63	<0.001
Respiratory tuberculosis	0.77	<0.001		0.73	<0.001		0.40	0.04		0.33	0.09
Stomach cancer	0.83	<0.001		0.82	<0.001		0.39	0.04		0.44	0.02
Lung cancer	0.10	0.61		0.48	0.01		0.02	0.91		0.23	0.24
Coronary heart disease	0.05	0.81		0.16	0.42		0.13	0.53		0.28	0.16
Stroke	0.66	<0.001		0.63	<0.001		0.61	<0.001		0.64	<0.001
Partial correlation coefficients:†
All causes	0.32	0.11		0.28	0.17		0.42	0.03		0.50	0.009
Respiratory tuberculosis	0.71	<0.001		0.69	<0.001		0.01	0.96		0.07	0.72
Stomach cancer	0.80	<0.001		0.77	<0.001		0.08	0.71		0.04	0.87
Lung cancer	0.10	0.60		0.43	0.03		0.04	0.86		0.02	0.92
Coronary heart disease	0.13	0.52		0.03	0.90		0.18	0.39		0.23	0.27
Stroke	0.51	0.008		0.45	0.02		0.42	0.03		0.48	0.01

*Australia, Austria, Belgium, Bulgaria, Canada, Chile, Czechoslovakia, Denmark, Finland, France, Greece, Hungary, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Russian Federation, Spain, Sweden, Switzerland, United Kingdom, United States.

†Sex and cause-specific correlations of adult mortality with infant mortality in one period adjusted for infant mortality in the other period.

The table shows strong correlations between infant mortality in the 1920s and current mortality from stomach cancer. To examine the possible confounding effect of current circumstances, the table also shows correlations between adult mortality 1991-3 and infant mortality in the same period. For stomach cancer these correlations are appreciable but considerably smaller than the correlations with mortality at the time of birth. Partial correlation coefficients are shown in the bottom half of the table, where the correlations of adult mortality with infant mortality in one period have been adjusted for infant mortality in the other. These partial coefficients indicate that the association is almost exclusively with infant mortality at the time of birth.

The table also shows correlation coefficients for other causes of death. Mortality from respiratory tuberculosis is more strongly related to infant mortality at the time of birth than currently, in agreement with the notion that many people dying of respiratory tuberculosis in old age were initially infected during their early years. Lung cancer shows an appreciable inverse correlation with infant mortality at birth for women only. This may be because historical levels of infant mortality may provide an indication of women’s position in society, which in turn is related to cohort differences in the uptake of tobacco smoking by women.

Coefficients for stroke and coronary heart disease are strikingly different, as previously reported. ⁽⁵⁾ Coronary heart disease shows the weakest correlations with historical levels of infant mortality of any of the causes in the table, and only weak correlations with current infant mortality, whereas stroke shows strong associations with both historical and current infant mortality.

Comment

Our analyses imply that that a poor environment during infancy and childhood, which is associated with high infant mortality, may explain some of the similarities in the descriptive epidemiology of stroke and stomach cancer, as has been previously suggested. ⁽¹⁾ Stroke risk may be influenced by undetermined infection(s) in childhood that may have epidemiological characteristics similar to H pylori. The results also reinforce the large differences in the aetiology of stroke and coronary heart disease, ⁽⁶⁾ with adverse circumstances during early life being considerably more important for stroke. ⁽¹⁾

The idea for this work was jointly developed; DL analysed the data and drafted the paper, which was revised by GDS. DL is guarantor for the paper.

Funding: None.

Competing interests: None declared.

1. Davey Smith G, Hart C, Blane D, Hole D. Adverse socioeconomic conditions in childhood and cause specific adult mortality: prospective observational study. BMJ 1998;316:1631-5.
2. Parsonnett J. Helicobacter pylori and gastric adenocarcinoma. In: Parsonnet J, ed. Microbes and malignancy. Infection as a cause of human cancers. Oxford: Oxford University Press, 1999:372-408.
3. Mitchell BR. International historical statistics: Europe 1750-1988. New York: Stockton Press, 1992.
4. UN demographic yearbook. New York: United Nations, 1948-70.
5. Elford J, Ben-Shlomo Y. Geography and migration. In: Kuh D, Ben-Shlomo Y, eds. Life course influences on adult disease. Oxford: Oxford University Press, 1997:220-41.
6. Gale CR, Martyn CN. The conundrum of time trends in stroke. J R Soc Med 1997;90:138-43.