[ Intro I Methods I Results and Discussion I Policy Implications ]

[ References I Exclusive Interview With Colin Campbell ]

'Division of Nutritional Sciences, Cornell University, Ithaca , Institute of Nutrition & Food Hygiene, Chinese Academy of Preventive Medicine, Beijing, 'Institute Of Mediterranean Agriculture, Montpellier, 'Department of Food and Nutritional Science, Kings College, University of London.

(Received 5th February, 1991; in final form November 5th, 1991)


One hundred and forty years ago John Hughes Bennett, Senior Professor of Clinical Medicine at the University of Edinburgh, when writing about the causes of cancer, said that nutritional conditions that favored the risk of tuberculosis were opposite to those favoring the risk of cancer (Bennett, 1849). Just three years before, Walshe had published cancer mortality data and proposed that cancer was a "disease of civilization" (Walshe, 1846) because death rates were higher amongst the upper income citizenry. On many occasions since then, it has been observed that disease prevalence changes radically after economic development (Pellett, 1989). The diseases that were responsible for a vast majority of deaths in Europe and North America in the nineteenth century are not the primary causes of death today.

Although the specific reasons for these changes in mortality rates are not well understood, a few of the more general characteristics of these trends are as follows. The decline in infectious and communicable diseases follows an increase in, and more equitable distribution of, economic resources. An extensive investigation of mortality rate trends in England and Wales in the eighteenth, nineteenth and twentieth centuries (McKeown and Record, 1955; McKeown and Record, 1962; McKeown et aL, 1975) indicates that the nineteenth century decline in mortality rates for the most prevalent infectious diseases was wholly attributable to environ- mental control, not to intervention with curative medicines and vaccines.

McKeown et al. (1975) suggest that a better diet, improved hygiene, and a more favorable relationship between microorganisms and humans were major contributors to these trends, with better nutrition accounting for about three-fourths of the effect. Deliberate improvements in sanitation, however, may have been somewhat more important than he supposed as suggested by Cairns (1985) who added that the wider use of cotton, which did not shrink like wool, allowed clothes to be washed and, in so doing, people also washed themselves. In many developing nations there has been a rapid increase in the incidence of cardiovascular disease, obesity, diabetes, lung cancer and a host of other health disorders concomitant with a rise in economic affluence (Pellett, 1989). In contrast with the communicable and infectious diseases affecting the rural poor, the more economically privileged urban sectors in these countries suffer from a rising preval- ence of chronic degenerative diseases appropriately referred to as 'diseases of misdevelopment' by Dumont (1989). Not only do these chronic diseases have a debilitating effect on a productive segment of the active elite but also the costs of treating these diseases tend to absorb a disproportionate share of the public health resources in favor of an already privileged social group. It is therefore of utmost importance to developing nations to avoid creating a new and costly pathology soon after emerging from the scourge of infectious and nutritional deficiency diseases (Pellett, 1989).

China, a vast and densely populated country in the process of rapid modern- ization offers a unique setting to study the relationship between dietary, environ- mental and economic changes and health indicators. Two recent surveys (National Cancer Control Office, 1979; Chen et aL, 1990) in the People's Republic of China have confirmed the emergence of a similar trend in certain areas of this country: that is, an increase in chronic diseases and a decrease in communicable diseases. This paper reports on an analysis of primary data collected in an ecologic survey of dietary and lifestyle factors (Chen et aL 1990) together with secondary data on disease-specific mortality rates for the survey sites (National Cancer Control office, 1979). Two broad research questions are addressed:

An analysis of these Chinese data, which is the focus of this communication, provides evidence to suggest that diseases such as cancers and coronary heart disease may share a common nutritional etiology based on the enrichment of diets with animal products. Policy implications of these findings for China and other countries also are briefly discussed.


Data from a 1976 retrospective mortality survey ft)r 1973-75 causes of death and a ecologic survey in 1983 84 were combined to study the relationship between various mortality rates and several dietary, lifestyle and environmental charac- teristics in 65 mostly rural counties in China (et al, 1990). Sampling procedures and analytical methodology have been described in detail by Chen et al ( 1990) and are only briefly noted below.

The mortality data were obtained from the same survey of causes of death for 1973-75 that resulted in the publication of the Atlas of Cancer Mortality in China (National Cancer Control Office, 1979). Cause of death was determined using medical records from clinics and hospitals and by questioning relatives and close friends of the deceased person in addition to area officials. Diagnosis and classi- fication of the cause of death were standardized by the Chinese Ministry of Public Health. The accuracy and reliability of these mortality rates is more fully discussed in the Foreward to the Atlas (National Cancer Control office, 1979) and further underscored by similar values obtained in neighboring counties juxtaposed on opposite sides of provincial borders and surveyed by different health survey teams (Le et aL, 1981). In addition the wide range of values across survey counties for most causes of death and the generally high correlation between values by sex, indicates a minimal impact of diagnostic errors. Deliberate precautions were taken to distinguish causes of death easily confused with one another such as liver cancer and hepatocirrhosis, stomach cancer and gastric ulcer, lung cancer and pulmonary TB, amongst others.

Disease-specific mortality rates for each county were calculated as cumulative mortality rates per 1,000 persons by sex for all ages up to age 64, and for all ages up to age 15. The cumulative mortality rate up to a particular age is "n" times the sum of the annual crude death rates in all age groups up to that age, where "n" is the number of single years in an age group (WHO, 1976). Thus, the cumulative mortality rate for a particular disease may be interpreted as approximately equivalent to the cumulative risk of death by age 64 (or age 15, as the case might be), in the absence of death from other causes. The question of competing risks is unlikely to be of relevance in the context of this analysis primarily because mortality data have been expressed as age-standardized rates. Furthermore, there is no reason to believe that risk of chronic degenerative disease during middle age is going to be different for individuals who otherwise would have died from non-degenerative diseases earlier in life.

The ecologic survey was conducted in 65 rural counties dispersed throughout the populated regions of China, selected to represent the full range of mortality rates for seven of the most prevalent cancers. A three-stage random cluster sampling procedure was used to select the smaller survey units of communes (2 per county), production brigades (2 per commune) and production teams (2 per production brigade). Within each of the 260 production teams, 25 households were randomly selected from an official registry of residences yielding 100 households per county. A total of approximately 6500 adults aged 35-64 were included in the survey, with 100 individuals per county, half of each sex. The study design and remaining experimental details are provided by Chen et al. (1990).

The information obtained from this survey included nutritional, metabolic, hormonal, environmental, reproductive, demographic and socioeconomic characteristics. The combined data from the two surveys of 65 counties (130 communes) consisted of 367 items of information including cumulative disease-specific county mortality rates (from the 1973-1975 retrospective study) and a comprehensive set of dietary, lifestyle and environmental characteristics. The analysis assumes the stability of biochemical indicators and dietary patterns in the population over time. The negligible migration in this population (an average of 94% of the survey subjects were born in their county of residence), food production and consumption patterns based on stable locally available crops, and the remarkably constant trend in the national food intake data (Piazza, 1986) underscore the validity of such an assumption.

Pearson product-moment correlation coefficients were used to relate every variable with all other variables in the compiled data set. These correlation co-efficients are cross-sectional measures of association between characteristics and were used as such with the recognition that they do not establish a causal link between the variables examined. Geographic clustering of disease mortality rates was investigated by visual inspection of a table of correlation coefficients for each disease with every other disease. Disease mortality rates that were directly correlated (p 0.05) with each other were placed in one group; inversely correlated (p 0.05) rates comprised the second group (Table 2). Any disease mortality rate that did not show a consistent correlation (when significant) with all diseases included in each of the two groups was omitted from the set.


Three features of these data make this epidemiologic investigation of mortality in China particularly appropriate and informative. First, the ranges of mortality rates across China as illustrated for a few diseases in Table 1, were far greater than those typically observed for countries that are more industrialized (and less economically diverse). Such broad ranges may facilitate the detection of otherwise unobserved relationships. Second, mortality rates were distributed around China in clusters that formed unique geographic patterns for each disease (National Cancer Control Office, 1979; Chen et aL, 1990). The uniqueness of these patterns indicates the presence of specific causes for each disease, perhaps related to routine consumption of locally grown foods. And third, among these several dozen disease categories

Average age-.standardized county mortality rates per 100,000 and ranges for selected diseases in The People's Republic of China, 1973 - 75*

Disease Males Females
CANCERS' Nasopharynx 8 (0-75)SS 4 (0-26)
Esophagus 120(1-435) 72(0 286) Stomach 9](6-386) 41 (2-141)
Liver 78(7-248) 26(3-67)
Colorectal 14(1-67) 10 (2-61)
Lung 23(3-59) 1O(O-26)
Leukemia 4 (0-9) 3 (0-7)

Pulmonary tuberculosis 113(31-270) 70(26 230)
Infectious diseases 31 (6 64) 23 (4-50)
Diabetes 4(0-9) 3(0-14)

Myocardial infarcation / coronary heart disease 12(0-52) IO(O-50)
Hypertensive heart disease 16(1-43) 15 (2-65)
Rheumatic heart disease 16(0-55) 27 (3-99)
Stroke 77 (20 197) 64 (21-251)
Pneumonia 14 (2-54) II(O-56)
Cirrhosis of the liver 51(6-155) 26(2-58)
Digestive disease other than ulcer 23(3-61) 18 (4 55)

Age truncated, 35-64 years; standardised to world population (WHO, 1976). 7Tht)se cancer @ites were used in the selection of the 1983 survey c,)unties. They represent the full range of mortality rates for all 2392 counties in China. The average annual age-standardized county mortality rates are based t)n age-sex-specific deaths in each of the 65 counties (49 counties for non- cancers) for the population aged 35-64 years.

Range of county mortality rate in the 65 (49 counties for non-cancers) counties, 35-64 years.

Some were significantly correlated with each other, thus suggesting that they might share a common underlying nutritional etiology. In addition to providing important information on the chief correlates and possible causes of geographic differences between these mortality rates, the data could also be useful in the investigation of mortality rate trends over time. At the turn of the century, for example, Reclus (1905) noted the similarities between chronological and geographic trends of economic development and concluded that ". . . geography is nothing but history into space and similarly history is geography on a time scale." If so, the present health profile of some economically developed counties in China might be used to predict changes in more traditional counties. Such projections could be useful in planning nutrition, health and agricultural strategies.

Diseases in group A (Table 11) are those generally associated with impoverished conditions while most of those in group B tend to be characteristic of more affluent societies. To investigate common causes underlying each disease group, cumulative mortality rates for each disease group were calculated for each survey county. The relationship between these group mortality rates and the various characteristics measured in the 1983-1984 ecologic survey were then analyzed (Chen et aL, 1990) and the correlations that were statistically significant at p<O.OL are listed in
Table 111.
As expected, diseases of poverty are associated more with agricultural than with industrial activity. Areas where these diseases are common are located further inland where mean elevation is higher and overall economic activity, literacy and population density are lower. In contrast, diseases of affluence are found in the more densely populated rural areas' nearer the seacoast where industrial activity

Self-clustered disease groups

Group A - Disease of Poverty Group B - Diseases of Affluence

Pneumonia (16) Stomach cancer (5)
intestinal obstructions (I 2) liver cancer (10)
Peptic ulcer (13) Col,in cancer (9)
Other digestive disorders ( 17) lung cancer (16)
Nephritis ( 12) Breast cancer ( I
Pulmonary tuberculosis ( 10) Leukemia (15)
Infectious diseases (other than tuberculosis) (17) Diabete@ (2)
Parasi(ic diseases (other than schisit)st)miasis) (I 0) Coronary heart disease( 1)
Eclampsia ( 1 3) Brain cancer (ages 0- 14) (13)
Rheumatic heart disease ( 13)
Metabolic and endocrine disease (other than diabetes) ( IO)
Diseases of pregnancy and birth (other than eclampsia ( 15)

*Each disease category, when significantly correlatted (p<0.05) with any other disease category is positive for disease categories in its own group and negative for disease categories in the second group.

'Numbers in parenthesis indicate the number of correlations which are statistically significant at p<0.05 (from a total of 20 comparisons).

The major municipalities of Shanghai, Beijing and Tianjin were not included in this survey.
and literacy rates are higher and more fish, eggs, soy sauce, beer and processed starch and sugar products are consumed. However, such univariate geographic correlations do not provide strong evidence of causal relationships between dietary factors and these diseases, particularly because relatively small amounts of several of these specific foods are consumed. For example, mean egg consumption in China nationwide is very low by Western practices, with mean daily intake being only about 5-10% that for the united States; egg consumption averages less than once per week in 78% of the 65 survey counties. Fish consumption averages only about 15% of the median intake in the United States, except for three counties in the Southeastern coastal provinces of Fujian, Guangdong and Zhejiang. Beer and processed starch and sugar products are also consumed in much lower quantities.
Therefore, consumption of these foods is probably more indicative of general economic conditions and other local circumstances than of biological relationships to disease. The correlation of diseases of poverty with higher levels of arsenic contamination of food remains unexplained.


Correlations of disease groups with various characteristics
Characteristics / Diseases of poverty / Diseases of affluence
Total cholesterol 0.48'
Urea nitrogen -0.47 0.40
Albumin 0.44
Antibody to core HBV 0.44 -0.32
Cotinine 0.37 0.50

Ribonavin excess -O..41 0.45
Chloride 0.48

Hemoglobin -0.39
Total n3 fatty acids (RBC) 0.44

Height -0.59 0.51
Weight -0.45 0.41
Beer (per day) -0.32 0.59
Fgg (per year) -0.54' 0.31
Total pregnancies 0.53' -0.38
Stillborn 0.51 -0.34
Infant mortality 0.69 -0.47

Fish (g/day) 0.56
Processed starch and sugar 0.5 1
Soysauce -0.41
Arsenic 0.41
Beer 0.59

Mean elevation 0.43 -0.39
Mean longitude 0.48 0.40
Ciross value of industrial and agricultural output, 0.44
Population density' 0.45 0.41
Literacy ratel 0.51 0.39
Agricultural employment' 0.64' -0,58
Industrial employment' 0.63' 0.58

Further characteristics of the areas where diseases of poverty are more common include greater infant mortality and a greater number of pregnancies per woman. The average age of the women surveyed in 1983 was 48 years and so these characteristics of reproduction and childhood disease describe childbearing practices of the 1950s and 1960s, a period that included some severe malnutrition and the great famine of 1959-61. In contrast, contemporary characteristics could be quite different.

In the China study, the analysis of blood samples shows that plasma cotinine (a nicotine metabolite from tobacco smoking) is inversely associated with diseases of affluence including lung cancer but directly associated with diseases of poverty. At first glance, this finding might appear anomalous, since the smoking of manufactured cigarettes in China has been centered mostly around major metropolitan areas such as Shanghai where diseases of affluence are more common. However, high plasma cotinine levels in the rural areas where diseases of poverty predominate, are strongly associated with the use of homemade cigarettes which contain high levels of nicotine (R. Peto, personal communication). Thus the inverse association between plasma cotinine levels and diseases of affluence may be explained by the high consumption of homemade cigarettes in less affluent areas and the lower consumption of homemade cigarettes in more industrialized and more literate regions of the country.

Nutritional status may be assessed by several indicators including body size, biochemical factors in blood and urine, and dietary intake. The mechanisms of action of nutritional factors on disease are complex and have not yet been fully elucidated but it has been estimated in Europe that diseases to which diet makes a significant contribution are responsible for approximately 50 percent of premature deaths in both men and women below the age of 65 (WHO, 1986). There is increasing evidence that in China the most affluent segments of the urban and rural populations are adopting unsatisfactory dietary practices that only can lead to disease and their costly consequences similar to those observed in industrial countries. It is reasonable to assume that the frequent occurrence of food restrictions and famines in China (Mellor and Gavian, 1987) has favored the survival of those with the ability to store fat in times of sufficient food to better resist starvation. If this assumption is true, then a longlasting period of food affluence may bring about a high prevalence of obesity and associated disorders such as hypertension, stroke, some cancers, and heart diseases. The physical stature of adults (both height and weight) is greater in areas where diseases of affluence are more common. Also, diseases of affluence are associated with higher levels of plasma cholesterol, plasma urea nitrogen and plasma albumin, each of which indicates a diet richer in fat and protein and lower in fiber and other plant food constituents. The proportion of a 5 mg experimental supplement of riboflavin excreted in urine
within the following 4 hours is significantly greater in areas where diseases of afflu-
ence predominate, indicating greater tissue repletion with this vitamin. Riboflavin intake is significantly positively correlated with the consumption frequency of meat and milk (Campbell, et aL, 1990).

Inclusion of liver and stomach cancers among "diseases of affluence" was some-what surprising because these cancers are much more common in economically underdeveloped countries. One interpretation could be that, for each of these cancers, there may be a predisposing associated factor which is:

First, it approaches being a prerequisite cause, as indicated by the extraordinary risk observed in case control studies (Beaseley, Huang and Lin, 1981; Blumberg and London, 1982); second, its mean prevalence is 13% in China but only 0.1-0.3% in the United States; and third, it is widely spread across all 65 survey counties (46 counties had 10-28% prevalence, while the remaining 19 counties had 1-9% prevalence). Moreover, liver cancer mortality was significantly positively correlated with plasma total cholesterol to indicate a promoting effect of nutritional enrichment (Campbell, et aL, 1990). In regards to stomach cancer, a predisposing cause could be something associated with the widespread use of non-refrigerated, non-canned foods. In Western countries, stomach cancer markedly declined with the introduction of refrigeration (Coggon and Acheson, 1984; Howson, et al., 1986; Coggon et al., 1989).

Presently in China, many foods are preserved either by salting or by fermentation, leaving opportunities for widespread exposure to endogenous nitro- samine formation and mold toxins, which could be a prerequisite cause upon which nutritional enrichment has an effect in the same manner as with the other diseases of affluence. For example, several investigators have reported a protective effect upon stomach cancer of vegetable consumption (Hirayama, 1971; Hirayama, 1981; Correa, et aL, 1982) and in this study, stomach cancer mortality was positively correlated with plasma albumin, plasma urea, and body weight and height, any or all of which indicate a promoting effect of nutritional enrichment. A model similar to that for liver and stomach cancers also appears to exist for tung cancer. Smoking is the prerequisite cause while nutritional enrichment of the same type as for stomach and liver cancers enhances risk of tumor (Peto, et aL, 1981; Shekelle, et aL, 1981; Byers, et aL, 1987).


China is presently at a stage of development where public health measures could be taken to avoid the mistakes of the West, a proposition similarly made by Keck (1991). Appropriate action now could prevent a large increase in the diseases of affluence. Lung can@er is one of the few cancers for which the most important cause is well known and which is therefore amenable to preventive action. There has been an alarming increase in cigarette smoking in China. According to Richard Peto, a leading British epidemiologist and one of the principal investigators of this study, the number of Chinese deaths from tobacco are expected to increase from about 100,000 per year at present to about 2 million per year by the year 2000 (R. Peto, personal communication). Strong measures involving health education, cigarette import restrictions and the curbing of domestic tobacco production are essential to reverse the smoking trend and prevent an otherwise inevitable leap in lung and other related cancer rates.

The results reported here have direct implications for public health policy. The ideal future policy, of course, would be to reduce mortality from the diseases of poverty without causing a large compensatory increase in the diseases of affluence. Among the characteristics listed in Table Ill, the number of characteristics that are correlated positively with one disease group and inversely with the other suggests that there could be a trade-off in disease mortality trends. There are, however, some characteristics that are significantly correlated only with one disease group. For example, relatively high levels of plasma cholesterol, plasma albumin and omega-3 erythrocyte phosphatidylcholine fatty acids (directly correlated with fish consumption) reflect increased mortality only from diseases of affluence without simultaneously being associated with decreased mortality from diseases of poverty.

Diseases of poverty are correlated with increased urinary chloride and decreased hemoglobin whereas diseases of affluence show no relationship to either. If these do at least partially reflect causal relationships then a diet low in protein, fat and salt, but sufficient to allow adequate synthesis of hemoglobin may be associated with reasonably low mortality rates for both disease groups.

China has developed a unique system of decentralized planning which has recently incorporated private initiative in agriculture, industry, and trade. From the early 1950s to the mid-1970s the Chinese government had strict control over agricultural production and trade. It gave priority to staple foods over preferred foods (legumes, meats, fruits) in order to ensure an adequate supply of essential grain for all provinces. Until recently government policy favored direct consump- tion of grain over consumption of animal products requiring feedgrains (Jamison and Piazza, 1987). However, policy has changed markedly in the last few years. With the consolidation of the new 'production responsibility system' the govern- ment expects a rapid growth in the livestock sector (World Bank, 1985). Cattle production will be limited by the carrying capacity of China's grasslands, which are already overgrazed. Poultry and pig production are more dependent on the avail- ability of feed concentrates. Such production has been increasing for the last ten years, and there is now a concern that it might be necessary to monitor the consumption of high-animal-fat food to prevent deleterious effects both economic and nutritional.

From an economic standpoint, increasing the share of consumption of animal products from 6 percent to 16 percent of total food intake (the official goal for the year 2000) presents technical difficulties and considerable risks of dependency on feedgrain imports (World Bank, 1985). China also might be tempted to take advan- tage of low grain prices on the world market to establish its poultry and pig industry. But imported feedgrains would present considerable social risks if, in the future, political and economic obstacles were to lead to cutbacks in meat availability or a rise in meat prices. Increasing food prices has often been the cause of social unrest in many developing countries. According to World Bank estimates (1985), even with a favorable feed conversion ratio, control of related factors, and use of local feeds only, this target would require the dedication of more than one third (35-40 percent) of China's cultivatable land to livestock production. Given severely limited arable land per capita and already intensive use of land resources this is neither desirable nor even possible.

From a food and nutritional standpoint, China has a choice to make between various consumption patterns. In terms of total per capital energy supply and the respective proportion of calories from carbc)hydrate, protein, and fat, China could follow either the Japanese or the European model of consumption. The Japanese dietary pattern seems more appropriate to China because it is characterized by lower intakes of energy and animal products, and a higher intake of fish products than the European pattern. Because Japan has the highest average life span in the world its food pattern would appear to be preferable. Riboflavin intake in China may not be as critically low as previously suspected, a finding discussed in depth in a separate manuscript (Campbell et al, 1990). An overestimated riboflavin allowance tends to foster the technological development of foods and food industries designed to alleviate the implied deficiency. For example, greater consumption of dairy products and other foods of animal origin have been justified, in part, because their assumed riboflavin richness would help alleviate the perceived problem. Not only in this recommendation of greater consumption of animal products highly questionable for China for very significant economic reasons (World Bank, 1985), but also the assumption that dairy products are a particularly useful riboflavin source is not entirely valid because it ignores the even richer sources of riboflavin provided by certain plant foods such as green vegetables.

The Chinese food system is unique in its ability to provide for more than a billion individuals a reasonably nutritious diet that may be closer to health promoting dietary guidelines than that of industrialized nations (Wittwer, et at, 1987): for example, the average Chinese diet is substantially lower in total fats, saturated fats, and sugar (soluble carbohydrate), and higher in dietary fiber, 0-carotene and vitamin C (Table IV). Chinese food policy planners are doing much better from a dietary point of view than their Western counterparts, while avoiding problems associated with increased obesity and higher incidence of cardiovascular disease (Sidet and Sidel, 1982), as well as a considerable waste of topsoil, water, fertilizers, pesticides, and other agricultural resources. Dietary iron intake and iron absorption in China are surprisingly high in a diet so low in animal products (Chen et al, 1990).

Comparison of American diet with rural Chinese diet (mean daily intakes)
U. S. China
Total fat (% of kcal) 38-40 15
Dietary fiber (g/day) 10-12 33
Solul)le carbohydrate (g/day) 240 470
Calcium (mg/day) 1140 540
Protein (g/day, 70 kg mate) 90-95 64
Animal protein (% of total protein) 70 7
Iron (mg/day) 18 34
Thiamin (mg/day) 1.4 2.3
Retinol (RE/day) 990 30
Total carotenoids (RE/day) 429 836
Vitamin C (mg/day) 73 140
Riboflavin (mg/day) 1.9 0.8
Encrgy intake (kcal/day) 2360 2640

'Chen et al., ( 1 990).

This study, which in the next phase will include survey data from Taiwan, provides an unprecedented opportunity for scientists and policy makers in China, the US and other developed and developing countries to explore the complex relationships between lifestyle and disease, thus helping to make policy recommendations to influence agroindustrial production and consumer behaviour. This also may help developed countries to find ways to curb their excesses while encouraging developing countries to avoid the mistakes of Western societies (Keck, 199 1). There is much to learn from China, a country that is successfully feeding a billion people primarily on plant derived food, and rapidly making the transition from an agrarian to an industrial society, but so far avoiding much of the major diseases affecting Western countries.


Beasley,R.P., L..Y. Huang, C.C. Lin, et aL (1981). Hepatocellular carcinoma and hepatitis B virus. A prospective study of 22,707 men in Taiwan. The Lancet. November 21, 1129 1132.

Bennett, J.1i. (1849). On (ancerous and Cancroid Growthv. Edinburgh: Sutherland and Knox.

Blumberg, B.S. and W.T. London. (1982). Primary hepatocellular carcinoma and hepatitis B virus. Current Problems in (ancer, 6, 3 23.

Byers, T.E., S. Graham, B.P. Haughey, etal (1987) Diet and lung cancer risk: Findings from the western New York diet study. Am. J. EpidemioL, 125, 351-363.

Cairns, J. (1985). The history of mortality and the conquest of cancer. Accompli.5hments in (,ancer Research, 1985, 90-105. oflavin recommen-

Campbell, T.C., T. Brun, J. Chen, F. Zulin and B. Parpia. (1990). Questioning rib dations on the basis of a survey in China. Am. J. ('Iin. Nutr., 5 1, 436 45.

Chen, J., I.C. Campbell, J. Li and R. Peto, (1990). Diet, li@style and mortality in China. A study of the characteristics of 65 (hinese countiev. Oxford, UK: Joint publication of Oxford University Press, Cornell University Press and China People's Publishing House.

Coggon, D. and E.D. Acheson. (1984). The geography of cancer of the stomach. Br. Med. BUIL, 40, 335-341.

Coggon, D., D.J.P. Barker, R.B. Cole and M. Nelson. (1989). Stomach cancer and food storage.

Dumont, R. (I 989). Paris, France: Pion (edit.). Ames combats.

Hirayama, T. ( 198 1). A large-scale cohort study on the relationship between diet and selected cancers of digestive organg, pp. 409-429. In: Bruce, R.W. Correa, P. Lipkin, M. Tannenbaum, S.R. Wilkins, T.C. (eds). Banbury Report 7. Gastrointestinal cancer: Endogenous priors. Cold Spring Harbor, NY: Cold Spring Harbor Laborattory.

T. (1971). Epidemiology of stomach cancer, Gann. Monogr. 11, 3-19,

Howson, C.P., T. ffirayama and E.L. Wynder. (1986). The decline in gastric cancer: Epidemiology of an tinplanned triumph. EpilemioL Rev. 9, 1-27.

Hamison, D.T. and A. Piazza. (1987). In: J.P. Gittinger, J. Leslie, and C. Hoisington (eds). Food policy: interpreting supply, distribution, andconsumption. EDI Series in Economic Development, Baltimore, MD: Johns Hopkins University Press.

Keck, C.W. ( 199 1). Health Promotion: ii nation@il health priority. International Sympovium on Health Promotion. Tapei, Taiwan. i,i, J., B. Liu, G.

Li, Z. Chen, X. Sun and S. Rong. (198 1). Atlas of Cancer mortality in the People's Repiibiic of China. An aid for cancer control and research. Int. J. Fpid. 10, 1 27-143.

McKeown, T. and R.G. Record. (1955). Medical evidence related to English population changes in the eighteenth century. Population Studies 9, 119-14 1.

McKeown, T. and R.G. Record. (1962). Reasons for the decline in mortality in England and Wales during the nineteenth century. Population Study 16, 94-122.

McKeown, 'F,, R.g. Record and R.D. lurner. (1975). An interpretation of the decline in mortality in England and Wales during the twentieth century. Population Studies 29, 39 1 422.

Mellor, J.W. ad S. Gavian. (1987). Famines: causes, preventic)n, and relief. Science 235, 539-545.

The National Cancer Control Office, Ministry of liealth and Nanjing Institute of Geography, Academics Sinica. (1979). Atlas of Cancer Mortality in the People'.5 Republic of China. Shanghai: China Map Press.

Pellett, P.L. (1989) Commentary: The IJ.S. Surgeon General report and the Third World. Ecot Food andNutr. 23, 237-242.

Petro, R., R. Doll and J.D. Buckley. (i981). Can dietary beta-carotene materially reduce human cancer rates? Nature 290, 201-208

Piazza, A. (1986). Food consumption and nuiritional status in the Peoples v Republic of China, Westview Special Studies on China. Boulder, CO: Westview Press.

Population Census Office of the State Council of the People's Republic of China and the Institute of 144 T. COLIN CAMPBELL ET AL. Geography of the Chinese Academy of Sciences. (1987) The Population Atlas oj China. New York, NY: Oxford University press, Inc.

Reclus, E. (1905). L'homme et la terre. Libraire Univervelle 6, 4.

Shekelle, R.B., W.J. Raynor Jr., M. Lepper, el at (1981). Dietary vitamin A and risk of cancer in the Western Electric Study. Lancet 2, 1185-1190.

Sidel, R. and V.W. Sidel. (1982). The health of China. Boston, MA: Beacon Press.

Walshe, W.H. (1846). The Nature and Treatment of Cancer. London: Taylor and Walton.

WHO. (1986). Healthy nutrition: preventing nutrition-related diseases in Europe. Final Draft ICP/NUT 114. Regional Office of Europe, Nutrition Unit, Copenhagen.

WHO. (1976). Cancer Incidence in Five Continenty. Volume IL (Eds., Waterhouse, J., Muir, C., Corryea, P., and Powell, J.) Lyon, France: Internatitonal Agency for Research on Cancer.

Wittwer, S., Y. Youtai, S. Han and W. Lianzheng. (1987). Feeding a billion:frontiers of Chinese agriculture. East Lansing, MI: Michigan State University Press.

World Bank. (1985). China: Long ferm development issuev and optiony. Baltimore, MD: Johns liopkins University Press.