nonassociation of aflatoxin with primary liver …...a comprehensive cross-sectional survey was...

[CANCER RESEARCH 50, 6882-6893, November I, 1990]

Nonassociation of Aflatoxin with Primary Liver Cancer in a Cross-SectionalEcological Survey in the People's Republic of China1

T. Colin Campbell,2 Junshi Chen, Chongbo Liu, Junyao Li, and Banoo Parpia

Division of Nutritional Sciences, Cornell University, Ithaca, New York 14853 [T. C. C., B. P.], Institute of Food Hygiene [J. C./ and Institute of I'irology/C. I..], Chinese

Academy of Preventive Medicine, and Department of Cancer Epidemiology, China Cancer Institute [J. LJ, Chinese Academy of Medical Sciences, Beijing, China

ABSTRACT

A comprehensive cross-sectional survey was undertaken in The People's Republic of China of possible risk factors for primary liver cancer

(PLC) to include 48 survey sites, an approximately 600-fold aflatoxinexposure range, a 39-fold range of PLC mortality rates, a 28-fold rangeof hepatitis B virus surface antigen (HBsAg+) carrier prevalence, andestimation of exposures for a large number of other nutritional, dietary,and life-style features.

PLC mortality was unrelated to aflatoxin intake (r = â€”0.17)but waspositively correlated with HBsAg+ prevalence (/' < 0.001), plasmacholesterol ( /' < 0.01), frequency of liquor consumption (/' < 0.01), and

mean daily intake of cadmium from foods of plant origin (P< 0.01).Multiple regression analyses for various combinations of risk factorsshowed that aflatoxin exposure consistently remained unassociated withPLC mortality regardless of variable adjustment. In contrast, associationsof PLC mortality with HBsAg-f, plasma cholesterol, and cadmium intakeremained, regardless of model specification, while the association withliquor consumption was markedly attenuated (was nonsignificant) withadjustment for plasma cholesterol.

The sharp contrast between the findings of no aflatoxin effect uponPLC prevalence in this survey and the positive correlation reported forprevious but more restricted surveys is discussed. Based on the resultsof this survey and the data of laboratory animal and in vitro studies, anexplanatory model for the etiology of PLC is proposed, taking intoconsideration the role of nutrition in the etiology of this disease.

INTRODUCTION

Aflatoxin is a group of difuranocoumarin metabolites ofAspergillus flavus that are exceedingly carcinogenic for selectedspecies of experimental animals, mostly producing hepatocel-lular carcinoma (1-3). The most carcinogenic member of thisgroup is aflatoxin B,, which also is the most commonly found.Accordingly, the association of aflatoxin with PLC' in humans

has been investigated in numerous populations, particularly inSoutheast Asia and Sub-Saharan Africa (4, 5).

Evidence of aflatoxin cancer risk for humans periodically hasbeen reviewed by expert committees of various institutionsincluding the IARC (4-6) and the World Health Organization(7), among other agencies. The accumulation and interpretationof this evidence has gradually produced a consensus, particularly in the marketplace (8), that aflatoxin is a significant humancarcinogen. A committee of the IARC (4) suggested in 1976that "studies of liver cancer incidence in relation to aflatoxinintake provide circumstantial evidence of a causal relationship"

Received 2/27/90; accepted 8/2/90.The costs of publication of this article were defrayed in part by the payment

of page charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

'Supported in part by NIH Grant 5RO1CA33638. Chinese Academy of

Preventive Medicine. United Kingdom Imperial Cancer Research Fund, UnitedStates Food and Drug Administration, American Institute for Cancer Research,and CPC International, Inc.

2To whom requests for reprints should be addressed, at Division of NutritionalSciences, Cornell University, Martha V. Rensselaer Hall, Ithaca. NY 14853-4401.

3The abbreviations used are: PLC, primary liver cancer; HBV, hepatitis Bvirus; HBsAg, hepatitis B virus surface antigen; HBsAg+, plasma positivity forHBsAg; IARC, International Agency for Research on Cancer (Lyon, France).

and a World Health Organization task force on mycotoxinsconcluded in 1979 that "aflatoxin ingestion may increase therisk of liver cancer" (7). But in 1987, using more recent infor

mation, another working group of the IARC made more certainthis finding by concluding that there was "sufficient evidencethat aflatoxin is a probable human carcinogen" (5), thus affirm

ing for many observers its human carcinogenicity. Accordingto IARC criteria (5) a conclusion of "sufficient evidence" meansthat "a positive relationship has been observed between expo

sure to the agent and cancer in studies in which chance, biasand confounding could be ruled out with reasonable confidence."

It is not clear, however, that there is reasonable agreementon this new certainty (9-13). Some of the lingering skepticismis due to the failure of the early aflatoxin studies to take accountof the almost certain and later discovered risk contributed bypersistent infection with hepatitis B virus (that is, for thoseindividuals who carry the viral surface antigen, HBsAg+) (14-16). Thus, when three more recent studies (17-19) reported anaflatoxin effect independent of HBsAg-l- prevalence, considerable justification undoubtedly was given for the 1987 IARCreport to change the strength of the evidence from "circumstantial" (4) to "sufficient" (5). However, the results of these three

studies hardly justified this important change. The first of thesestudies, a communication in 1984 by Sun and Chu ( 17), did notproduce comparative data on aflatoxin and HBsAg+ prevalenceas implied by the IARC report (5). Instead, these investigatorsonly speculated on preliminary methodological data from theirlaboratory and commented on other reports of two small localities in China (20, 21) where there was poor correspondencebetween HBsAg+ and PLC. No aflatoxin intake data weregiven in the Sun and Chu (17) report. The second communication (18), published in 1985, was limited to a cross-sectionalcomparison of two villages in the Guangxi Autonomous Regionin southern China, with one "high" and one "low" in aflatoxin

intake. The third cross-sectional study (19), reported in 1987,was a reinvestigation by the same investigators of four surveyregions in Swaziland studied earlier (22), although these fourregions were subdivided into 10 smaller areas and HBsAgH-prevalence also was measured. Aflatoxin intakes of 48-269 ng/kg/day were only modestly higher than intakes of 13-197 ng/kg/day estimated for southeastern regions of the United States(23) where no relationship with PLC was observed. The authorsof the 1987 Swaziland study also pointed out that aflatoxinintake estimates were crude because the consumption of peanuts, "... the most heavily contaminated food item... was notweighed at the time of collection" (intake was estimated indi

rectly by noting the frequency of peanut use in sauces). Inaddition, PLC mortality rates per 100,000 persons (5.8 formales and 1.8 for females in 1981) were surprisingly low,particularly in view of the very high prevalence (20-35% forthe 10 survey regions) of HBsAg-f- carriers (19). Further calcu

lation of these data indicate that only about 1.8% of the carriersmay be expected to die of PLC, compared with 40-50% in

6882

Association for Cancer Research. by guest on August 21, 2020. Copyright 1990 Americanhttps://bloodcancerdiscov.aacrjournals.orgDownloaded from

https://bloodcancerdiscov.aacrjournals.org

AFLATOXIN AND LIVER CANCER IN CHINA

Taiwan (16). These anomalies detract from the conclusion of companion interpretation of additional variables from this samean independent aflatoxin risk.

Two other studies were also cited by the 1987 IARC report(5) as providing evidence for an aflatoxin effect on PLC. Onestudy (24), which furnished "additional evidence for a causativeassociation between aflatoxin and human cancer" (5), was a

retrospective cohort study of 71 individuals who had beenexposed as workers to variable amounts of aflatoxin dust in anoilseed extraction plant (24). Although increases in mortalityrates for total and respiratory cancer were noted, these increaseswere not statistically significant and no cases of PLC werereported. A second case-control study in the Philippines (25)produced an aflatoxin effect but, again, no data were availableconcerning the prevalence of HBsAg+. Also, in this study, noconfirmatory analysis was indicated for the fluorescent materialsuspected of being aflatoxin BI; one-half of the total intake waspresent in cassava, a source likely to be only occasionallycontaminated, and confirmation of aflatoxin B, identity isrequired.

Since the 1987 IARC report (5) was released, two additionalcross-sectional studies (12, 26) have been reported. When fourcommunes in southern China were compared (26), a linearrelationship was observed for mean annual per capita consumption of aflatoxin during 1978-1984 and PLC mortality ratesfor 1982-1986. Aflatoxin intake (population-based rather thanpersonal assessment) ranged 170-fold [17-2896 ng/kg/day,based on an assumption of 49 kg average body weight measuredfor adult males in the same area by Chen et al. (27)], while therange of PLC mortality rates was 3.5-fold (175-614/100,000person years, 25- to 64-year-old men). In the cross-sectionalcomparison of these four communes, prevalence of HBsAg+was not associated with PLC, although, in a case-control studyof individuals within this same cohort, a relative risk of 38.6was observed for HBsAg-t- carriers when compared to noncar-riers. In the second report, a cross-sectional study of 8 districtsin Kenya (no aflatoxin data were available for one site), Autrupet al. (12) examined the association of PLC incidence withHBV prevalence (HBsAg+ and antibody to HBV core protein)and aflatoxin exposure, as measured by prevalence of individuals who excrete in their urine a major aflatoxin-DNA adduciexcision product. No association between HBV and PLC wasobserved. When all areas and ethnic groups were considered,no association between aflatoxin exposure and PLC incidencewas found, but when the analysis was limited to the Bantugroups a "reasonable correlation" was obtained, according to

the authors. However, no data were provided on the proportionof Bantu people in the total population. A comparison of thePLC prevalence rate between the study of Yen et al. (26) andAutrup et al. (12) reveals a substantial difference of perhaps 2orders of magnitude in prevalence rates between these twoareas, after adjustment for sex and age distributions and assuming incidence and mortality rates to be similar.

On the basis of these human population studies, particularlythose cited by the 1987 IARC report (5), it is reasonable toremain skeptical of the suggestion that there is "sufficient"

evidence for significant aflatoxin cancer risk in humans. Theuncertainty of the evidence that aflatoxin is a significant PLCrisk factor prompted interpretation of relevant portions of dataobtained from 48 counties in a comprehensive survey of dietary,life-style, and mortality characteristics in 65 rural counties inthe People's Republic of China undertaken by Chen et al. (27).

These data were uniquely comprehensive and broad rangingwhen compared with previous studies of aflatoxin and PLC. A

data set (27) is being prepared by Drs. Ann Hsing and WilliamBlot at the National Cancer Institute in the United States andwill be published elsewhere.

MATERIALS AND METHODS

The study design for the complete cross-sectional ecological surveydata used in this analysis was described by Chen et al. (27). Only thosedetails relevant to the data presented in this communication are presented below.

Study Design. A multistage sampling procedure, more fully describedby Chen at al. (27), was used to select survey counties. For the originalsurvey, 65 counties were selected from a total of 2392 in order to obtaina wide geographic scatter and to encompass the full nationwide rangeof mortality rates for 7 of the more common cancers (nasopharynx,esophagus, stomach, liver, colorectal, lung, and leukemia). Two communes were randomly selected within each county and similarly withineach commune, one production brigade, and within each productionbrigade two production teams were randomly chosen, yielding a totalof 130 communes and 260 production teams. However, the surveycommunes were restricted to those within 4 h travel time to the countylaboratory, necessitating replacement of 4-5 of the originally selectedcommunes. As described by Chen et al. (27), within each of the 260production teams, 25 households were randomly selected from anofficial registry of residences (100 residences/county, 50 households/commune). From each household either one male or one female aged35-64 years, randomly selected, was then invited to donate blood, withan approximately equal number of individuals for each sex and agegroup: 35-44, 45-54, and 55-64 years. "If the selected subject in each

household was absent or declined to participate (<1% of those initiallyidentified), then another subject in a neighboring household was selected. The total number of participating subjects aged 35-64 years was6500."

Because mortality data were available for 1973-1975 and biochemical data were collected in 1983, any analysis of the relationship betweenthese biochemical parameters and mortality data is valid only if thebiochemical indicators measure stable characteristics which have notchanged much over time. Evidence for the stability of these characteristics is indicated by the negligible migration in this population (anaverage of 94% of the survey subjects still lived in the county in whichthey were born) and the likelihood that dietary patterns at the time ofthe survey reflected past dietary patterns which have been dependenton stable, locally available crops. Time trend data of national foodintake show that dietary composition, for example, including grain(corn) intake (an indicator of aflatoxin exposure), has remained remarkably constant from 1950 until very recently (28). Also, a review ofaflatoxin contamination data for a county in the Guangxi AutonomousRegion where a good monitoring procedure has been in place since1972 revealed a mean annual contamination rate of 69.6 Â±15.3% (SD)positivity (range, 41-88%) for the years of 1972-1985 (data madeavailable to the Institute of Nutrition and Food Hygiene in Beijing).

Mortality Rates. PLC mortality rates were determined retrospectivelyfor the years 1973-1975 and were published originally as the Atlas ofCancer Mortality (29, 30). Details of data collection, methods, anddisease diagnosis and classification procedures were presented by Li etal. (30). The reliability of these data were affirmed on the basis ofseveral criteria. First, 96% of the PLC deaths, as recorded, wereattended by medical staff either at commune clinics (with one or morefully qualified physicians) or at county, prefecture, or province hospitals(including full medical services such as pathology, radiology, etc.). Moreimportant, 69% of liver cancer deaths had an antemortem diagnosisbased on the analysis of pathological or histological materials (9%) oron gross assessment of specimens, operation, X-ray, ultrasonography,isotope microassay, biochemical immunology, or the like (60%). Also,care was taken to distinguish causes of death easily confused with oneanother, such as liver cancer and hepatocirrhosis. Second, disease ratesfor counties juxtaposed on opposite sides of provincial borders and

6883




surveyed by different teams were similar. Third, diagnostic errors werevery small when compared with the nationwide range of mortality rates.Fourth, a repeat survey in the Guangxi Autonomous Region (18) in1987 showed a liver cancer rate of 94/100,000 persons compared witha rate of 96/100,000 persons obtained in the 1973-1975 survey for thesame county group.

In addition, it should be noted that any analysis of the relationshipbetween the mortality data of 1973-1975 and the dietary and ecologicaldata of 1983 is valid only if stable characteristics not changing muchover time were measured. Stability of the 1983 dietary characteristicswas suggested by negligible migration of the subjects (an average of94% of the survey subjects still lived in the county of their birth) andthe likelihood that the dietary patterns at the time of the survey reflectedpast dietary patterns which have been dependent on stable, locallyavailable crops. A current survey of causes of death for the years 1986-1988 coded according to the ninth revision of the International Classification of Diseases is presently being undertaken and these data willprove to be very' useful for comparison.

Blood Collection. Ten ml of blood was obtained from each of theadults between 6 a.m. and noon, using trace mineral-free Vacutainers(Becton-Dickinson) Inc.. Rutherford. NH). Samples were placed on icein light-free thermos containers and within 4 h were transported to thecounty laboratory where they were separated by cenlrifugation (2000rpm, 10 min) into RBCs and plasma. For each individual subject, one4 ml fraction of plasma, containing 20 mg of sodium ascorbate, wasthen saved at -15 C to -20 C. Blood samples were collected from

September to December 1983.Urine Collection. Only 51 of the 65 survey counties provided urine

samples (i.e. 4-h collections) considered to be adequate for this project(the remaining 14 counties only provided overnight collections whichwere not appropriate for this analysis). In each of these counties, about25 males aged 35-64 years and residing in the same commune of eachcounty gave 4-h urine samples between 6:00 am-12:00 noon after theyhad been given, as part of an accessory investigation, an oral 'load' of

riboflavin and ascorbic acid to measure the excretion of these vitamins.Samples were shipped from the survey sites to the nearby countylaboratories within 4 h of collection of the last sample. A 5% aliquotof each sample was then added to age-specific county pools (35-44,45-54, 55-64 years) which were stored at -15Â°Cto -20Â°C. Within 1-

2 months all samples were shipped on dry ice to the Institute ofNutrition and Food Hygiene laboratories in Beijing where they werestored at -30Â°Cuntil age-specific pools were prepared for other assays

not reported here. On February 21, 1984, the pools were shipped toCornell University in Ithaca, NY, and were stored at -80Â°C. During

June 1985, samples were thawed and divided into subsamples of varioussizes and then refrozen at â€”80Â°C.

Aflatoxin Analyses. Only 48 of the 51 counties providing urinesamples were analyzed for aflatoxin metabolites because samples from3 counties were lost at the time of the analyses. The possibility of theintroduction of an unintended selection bias was considered and thenecessary omission of 17 of 65 (26%) of the sampling units was judgedto have no serious statistical impact. First, the full range of the outcomevariable (PLC rates) was maintained (50% of the omitted counties wereabove the median PLC rate) and, second, the loss of these 17 datapoints appears not to have any unfavorable statistical effect when usingthe corn ration as a potential indicator of aflatoxin exposure (8 of 17counties being above the median corn ration for all 65 counties). Theseaflatoxin analyses were undertaken in the laboratories of Dr. JohnGroopman at Boston University according to modifications of previously published methods (31-33). Each age-specific pool was analyzedfor aflatoxin metabolites and then converted to weighted averages foreach county, depending on the number of individuals in each age group.Before analysis, each sample was cleaned by passing 25 ml through aCIS Sep-Pak and isolating the 80% methanol eluant which was thendried and redissolved in 4 ml of water. This procedure isolates thelipophilic compounds in the urine, including the aflatoxins. To obtainthe oxidative aflatoxin metabolites, the water fraction was extractedwith 3 volumes of ethyl acetate which was dried in vacuo anA redissolvedin 4 ml of water. This water sample was applied to an anti-aflatoxin

monoclonal antibody affinity column, was subsequently washed with 3volumes of phosphate-buffered saline, and then was eluted with 6 mlof 50% dimethyl sulfoxide to produce an eluant containing the purifiedaflatoxins. After concentration under reduced pressure to 500 p\ witha rotary evaporator, a lOO-^ilaliquot was analyzed by a competitive 'Fi-

base radioimmunoassay. For each assay, 400 ^1 total volume wasroutinely used [100 Â¿ilof which consisted of the |'H]aflatoxin B, tracer

(specific activity, 24 Ci/mmol)], representing a dilution in 1% normalmouse serum/0.1% bovine serum albumin in phosphate-buffered salineto a level of about 10,000 cpm/100 n\. The monoclonal antibody wasdiluted to a concentration which precipitated 40-60% of the aflatoxinB, tracer. The antibody was added to the reaction mixture in 100 //I of10% horse serum in phosphate-buffered saline. The test sample, eithercold aflatoxin B, (for standard curves) or antibody column eluantaflatoxin reactive material, was added to 200 n\ of a phosphate-bufferedsaline mixture. The reaction was incubated for 2 h at ambient temperature. An equal volume of ice-cold saturated ammonium sulfate wasadded to the reaction vessel, and the antibody was allowed to precipitatefor 15 min. The precipitate was centrifuged for 15 min at 12,000 x g,and 400 Â¡Aof the supernatant was counted with an LKB model 1211beta counter to quantify the amount of antibody inhibition.

Plasma Hepatitis B Virus Assays. Individual samples were used forthe determination of hepatitis B virus indicators and data are expressedas prevalence of positive samples. Only plasma HBsAg data are givenin this communication and these analyses were undertaken on individual samples (50 individuals in each of 48 counties) in the laboratory ofDr. Chongbo Liu using a SPRIA diagnostic kit (34). Plasma levels ofhepatitis B virus anti-core antibody also were determined on individualsamples by the enzyme-linked immunoabsorbent assay of Engvall andPerlman (35).

Biochemical Assays. Plasma total cholesterol was analyzed by themethod of Parekh and Jung (36) according to the modification of Fenget al. (37) This latter modification, which uses a ferric ammoniumsulfate color-developing reagent, enabled use of the ascorbate-preservedplasma samples. However, when this method was later compared withother methods, the estimates of plasma cholesterol are thought to beunderestimated by 10-15%.

Other Measurements. Food intakes were obtained either from aquestionnaire administered by a member of the provincial health surveyteam to all subjects who donated blood or from a 3-day weighing ofactual food intakes in the households of urine and blood donors (27).Cadmium intake was estimated from an analysis (38) of county-basedcomposites of foods of plant origin collected in each survey county[representative of both communes of the larger survey (27)]. Thesecomposites were comprised of proportions of foods actually consumed,as directly measured in the 3-day dietary survey. Further informationconcerning the questionnaire and the 3-day dietary survey is availablein the paper by Chen et al. (27).

Statistical Methodology. Cumulative PLC mortality rates/1000males aged 0-64 years for the 48 counties in which aflatoxin metaboliteswere analyzed were used as the outcome variable, both in the originalgeographic correlation study (39) and in the statistical analysis reportedhere. PLC mortality in persons <35 years of age was found to bemarginal and no differences were found in the results using the 35- to64-year age-truncated rates, thus confirming the validity of using ratesfor ages 0-64 years. This cumulative mortality rate is approximatelyequivalent to the cumulative risk of death by age 64 years from PLC inthe absence of death from other causes. Age-standardized mortalityrates for ages 35-64 years were also computed for meaningful comparison with the United States.

Pearson product-moment correlation coefficients were used to examine the relationship between PLC and selected variables includingaflatoxin, HBsAg+ prevalence, and dietary intakes of salient foods. Itshould be emphasized that these zero-order correlation coefficients arecross-sectional measures of the degree of linear association betweentwo factors and do not establish a causal relationship between thesefactors.

Multiple regression analysis was used to examine the combinedeffects of two or more factors on PLC and to determine the relationship

6884




Table 1 Age-truncated (35-64 years) primary liver cancer mortality rates in 48rural Chinese counties and the United States

Males Females

China (1973-1975)United States (1981-1985)

82.5"

2.626.4

1.2" Average annual rate/100.000 population, age-standardized to world popula

tion (41).

MEAN - 26.51S.D. - 21.21HANGE - 3.46-96.33

0 5 10 15

NUMBER OF DEATHS (AGES 0-64)/1000

Fig. 1. Cumulative liver cancer mortality rates for 0- to 64-year-old males in48 survey counties.

between selected variables such as aflatoxin and PLC, adjusting for theeffects of other relevant variables, such as HBsAg+ prevalence. Severalsuch theoretically plausible model specifications were examined usingthis multivariate framework. Natural logarithmic transformations ofPLC mortality rates were used in the analysis to obtain a normaldistribution of the outcome variable for reliable statistical significancetesting of the regression coefficients. For consistency, logarithms of themortality rates were used in the computation of correlation coefficientsas well. All probability levels of significance are two-sided.

RESULTS

The average age-truncated (35-64 years) PLC mortality ratesin these 48 Chinese counties, when compared to the mean ratefor the United States (40), with both rates being standardizedto the world population (41), is 32-fold greater for males and22-fold greater for females (Table 1). Mortality rates for malesvary 39-fold from the highest county to the lowest county (Fig.1), with the highest rates occurring along the eastern andsouthern coastal provinces (Fig. 2). Males have a higher prevalence than females, outranking females by 3.1-fold in Chinaand 2.1-fold in the United States. A male partiality also hasbeen observed in other societies (16, 42, 43), although the maleto female ratio varies considerably (10).

Means and ranges for a few characteristics selected for theirpossible associations with PLC mortality rates for males areshown in Table 2. Also shown are comparable United Statesdata4 (44-46) along with zero-order correlation coefficients

showing associations of these characteristics with logarithmically transformed PLC mortality rates in China. Aflatoxinexposure exhibited wide variation when compared with previousstudies. A much greater proportion of individuals were HBsAg+carriers in China than in the United States. This very highprevalence of HBsAg+ in China (Fig. 3) is consistent withprevious investigations showing carrier rates generally in excess

of 10% for Sub-Saharan Africa and other areas of SoutheastAsia (47, 48). The range of HbsAg+ prevalence in this study israther more than previous studies. For example, Yeh et al. (26)observed a range of only 21.6-24.8% in 4 communes in theGuangxi Autonomous Region in southern China, while Peerset al. (19) reported a range of 21-28% in Swaziland. The

prevalence of individuals positive for antibody to the HBV coreprotein (past infection) in these 48 counties was 18-81%, witha mean of 50% (data not shown). Plasma cholesterol levelswere substantially below those for the United States; Chinesehigh values were near American low values. Geographic distribution of average (males and females) plasma cholesterol valuesfor the 48 counties is shown in Fig. 4.

Intakes of all foods and about a dozen calculated nutrientswere recorded, with corn, moldy peanuts, liquor, and cadmiumselected for presentation in this paper. Information concerningother variables will be available in the report by Chen et al.(27). Liquor intake was considerably below that for the UnitedStates, with liquor contributing about two-thirds of total alcohol

consumption (the remainder came from wine and beer). Consumption of moldy peanuts was expressed as the proportion ofrespondents having eaten this food during the past year. Datafor moldy corn consumption also was sought, but this dietarypractice was so uncommon that these data are not presentedhere. Instead, total corn intake was estimated from the quantityof ration dispensed by central government authorities. Corn asa cereal staple is consumed mostly in northern and centralChina, with mean daily intakes of 200 g or more for communeslocated in 10 of the 24 survey provinces; thus, substantialamounts are consumed by a large number of people in China.Estimated intakes of these foods were sought because theyusually are considered the most common sources of mycotoxins,including aflatoxin.

Aflatoxin exposure was unrelated to PLC mortality (Table2), as were the intakes of the traditional sources of aflatoxincontamination, corn and moldy peanuts. In contrast, prevalenceof HBsAg carriers was highly significantly associated with PLCmortality (P < 0.001). The highly significant association ofPLC mortality rates with total alcohol consumption was mostlyattributed to liquor consumption. The intakes of several heavymetals (arsenic, cadmium, lead, mercury) were obtained from

mortality LIVER CANCER(cumulative rate. 0-64)/luOO

4 M. Nui, Centers for Disease Control. Atlanta, GA, personal communication,

1989.Fig. 2. Geographic distribution of liver cancer mortality rates in 48 survey

counties.

6885




Table 2 Means, ranges, and univariate correlation coefficients with male primarylirer cancer mortality rates for selected characteristics and putative risk factors in

48 rural Chinese counties and the L'nited States

plasma HEPATITIS B SURFACE ANTIGEN(% of individual samples that were positive)(non-pooled analysis}

ChinesedataCharacteristicUrinary

aflatoxin(ng/kg/4h)'HBsAg

positive'(%

prevalence)Liquorintake(g/person/day)''Alcohol

intake(g/person/day)''1'Proportion

consumingmoldypeanuts(%)Estimated

cornintake(g/day)'Cadmium

intakePlasma

cholesterol(mg/dl)'rÂ°-0.170.45s0.46'0.40;0.17-0.200.40J0.42Â«Mean156'15.529.421.45.184.518.6127.6Range*0-6

II71-28*0-1

110-750-870-6240-9094-162*(mean

= 127)United

States data(Means or

ranges)r-0.3(Footnote

4)9-168(45)1

5(46)212(44)

" Univariate correlation coefficients with logarithm of cumulative PLC mortality rates (ages 0-64 years).

* Range of means of communes or counties (two communes per county). See

footnotes d and f for further detail.' References shown in parentheses.d Data for China obtained from one commune/county.' Data for China obtained from two communes/county.fV it is assumed that (a) urinary aflatoxin is 5% of total aflatoxin intake. (A)

mean body weight is 55 kg (27), and (c) excretion in 4 h is one-sixth of dailyexcretion, then estimated aflatoxin intake would be 936 ng/kg/day.

*P< 0.001.* Range of county means for both sexes combined (100 individuals/county).1Approximately 80% comprised of liquor alcohol.

* Only the amount contained in plant origin foods which comprise about 90-

95% of total food intake (as indicated by protein intake).

analyses of food samples collected in each county and only theintake for cadmium showed a significant association with PLCmortality rates. Mean cadmium intake was similar to an estimate of mean intake in the United States (46). Plasma cholesterol also was highly significantly associated with PLC mortality rates. As is well known, levels of this constituent may berelated to dietary and other life-style conditions (49). In this

study, even though it was very low compared to the UnitedStates, it tended to be associated with the intakes of foods ofanimal origin (Table 3). Significant correlations with plasmaalbumin and body height are consistent with diets richer inanimal protein.

To determine whether the zero-order associations betweenPLC mortality and various characteristics (Table 2) might beaccounted for by the simultaneous presence of other possiblerisk factors, regression analyses for selected model specifications on logarithmically transformed PLC mortality rates werecarried out (Table 4). The simultaneous effects of all selectedrisk factors could not be validly assessed due to multicollinearityin the model. Thus, a series of models for which unbiasedestimates could be obtained with the maximum number ofindependent variables were fitted to the data. Plasma cholesterol, HBsAg carrier prevalence, liquor intake, and cadmiumintake were the most consistently associated PLC risk factors.Aflatoxin exposure consistently remained unassociated withPLC regardless of which characteristics were adjusted for inthe model. The model "explaining" the largest proportion of

PLC variation (model 4; 37%) did not contain aflatoxin. (Model6 had a higher R2 but the additional "variance explained" was

contributed by a theoretical inverse and biologically implausibleassociation between PLC mortality and aflatoxin intake.)

In contrast to aflatoxin intake, prevalence of HBsAg+ was

Fig. 3. Geographic distribution of HBsAg-l- carrier prevalence in 48 surveycounties.

plasma TOTAL CHOLESTEROL(mg/dl)

Fig. 4. Geographic distribution of plasma cholesterol levels in 48 surveycounties.

significantly associated with PLC mortality in all models. However, its contribution to the variation in PLC mortality wassomewhat attenuated when plasma cholesterol and liquor intake were simultaneously added (models 4 and 5).

The most significant association with PLC mortality waswith plasma cholesterol, regardless of model specification. Controlling for HBsAg+ prevalence, aflatoxin, liquor, cadmium, orany combination thereof still permitted a statistically significantplasma cholesterol association in each case. Daily liquor intakewas associated significantly with PLC mortality when controlling for HBsAg+ prevalence and aflatoxin exposure (ÃŸ= 0.30,P = 0.073), but this association disappeared when plasma

cholesterol was included in the model (models 4 and 5). Cadmium intake also was associated with PLC mortality regardlessof model specification. The proportion of the variance in PLCmortality explained by the three most significant risk factors,HBsAg+ prevalence, plasma cholesterol, and cadmium intake,was 36% (model 8). Excluding aflatoxin did not alter this result,except in the biologically implausible opposite direction.

6886




Table 3 Univariate correlation coefficients for various dietary and nutritionalfactors with average (males and females) plasma cholesterol in 65 rural Chinese

countiesVariableDietary

intakeLipid(%kcal)Dietary

fiber(g/day)Dietarycellulose(g/day)Total

protein(g/day)Animalprotein(g/day)Fish

protein(g/day)Dairy(g/day)Eggs(g/day)Meat(g/day)Legumes

(g/day)cBlood

constituentsPlasmaalbumin(g/dl)Other

characteristicsHeight(cm)r0.30Â°-0.27Â°-0.37*0.070.240.33*0.210.210.26Â°-0.35*0.33*0.28Â°

V<0.01.c Data based on amount rationed in each county; mostly soy beans.

Table 4 Results of multiple regression analysis for various model specificationson male primary liver cancer mortality0

ModelVariables1

AflatoxinHBsAg+Total

cholesterol2

AflatoxinHBsAg+Liquor3

AflatoxinHBsAg+Cadmium4

HBsAg+TotalcholesterolLiquorCadmium5

AflatoxinHBsAg-fTotal

cholesterolLiquor6

AflatoxinHBsAg+Total

cholesterolCadmium7

AflatoxinHBsAg+LiquorCadmium8

HBsAg+TotalcholesterolCadmium0-0.220.320.37(K2

= 0.34, P =0.0007)-0.110.280.30(R2

= 0.28, P =0.0036)-0.180.280.31(R2

= 0.30, P =0.0015)0.200.270.160.23(R2

= 0.37, P =0.0006)-0.180.250.310.15(R2

= 0.35. P = 0.0014)-0.220.250.340.27(R2

= 0.41, P=0.0002)-0.130.250.220.25(R2

= 0.33. />= 0.0018)-0.260.310.27(K'

= 0.36, P = 0.0003)Probability0.1000.0230.0100.4050.0860.0730.1780.0860.0250.1880.0650.3380.0920.1850.1060.0420.3750.0810.0620.0110.0370.3300.1100.1700.0800.0630.0220.046

Â°Logarithm of cumulative PLC mortality rates for males aged 0-64 years.

DISCUSSION

A comparison of the experimental features of this surveywith the most convincing cross-sectional surveys previouslyreported is shown in Table 5. The first four and, more recently,the first five studies (22, 50-53) listed in Table 5 have beenwidely cited (5, 7, 9, 10, 23, 52-56) and the data from thesestudies often have been combined (7, 52-56) to show an impressive correlation of PLC mortality rates with aflatoxin in

take; the experimental features of the combined data also areshown. Within these individual studies, all considered by the1987 IARC report (5), the range of aflatoxin intake was 3- to11-fold, the range of mortality rates was 3- to 4-fold, and thenumber of survey sites was 2-8. However, in none of thesestudies was HBsAg+ prevalence simultaneously surveyed, although a later study in Kenya (57) found a relatively constantHBsAg+ prevalence for the three survey regions in the studyby Peers and Linsell (51). The recent study in Swaziland (19),also considered by the 1987 IARC report, was the only one thatrecorded both aflatoxin exposure and HBsAg+ prevalence; noindependent HBsAg+ effect was observed. Also shown forcomparison are the more recent studies of Yeh et al. (26) andAutrup et al. (12), each of whom simultaneously measuredaflatoxin exposure and HBsAg+ carrier rates.

The experimental features of our China survey sharply contrast with the experimental features of these various cross-sectional studies. There is a much larger number of survey sites,a wider range of PLC mortality rates, a wider range of aflatoxinexposure, simultaneous measurement of a generous range ofHBsAg+ prevalence, and a large array of possible risk factors(more than 250 nutritional, dietary, and life-style variables). Animproved method of estimating aflatoxin exposure also wasused. Measurement of aflatoxin metabolites in urine directlymeasures exposure and integrates intake over a day or so, whilemeasurement of intake by aflatoxin analysis of foods is subjectto the highly erratic distribution of aflatoxin in different foods(58).

The lack of an association between aflatoxin exposure andPLC mortality in this China study is particularly remarkable inview of the findings of most previous investigations. The absence of an aflatoxin-PLC association is consistent with asimilar lack of association of PLC mortality with the consumption of the two foods most commonly contaminated with aflatoxin, corn and moldy peanuts; this consistency is particularlynotable because of the uniqueness of each method for estimatingaflatoxin exposure. A power calculation, undertaken in retrospect, shows that, with 48 counties and an a level of 0.05, acorrelation coefficient of +0.48 is required for significance; theobserved coefficient of â€”0.17is far below this reference point.

This nonassociation between aflatoxin intake and PLC mortality also persists after controlling for potential confoundingfactors such as plasma cholesterol, prevalence of HBsAg-t-, and

intakes of alcohol and cadmium.In contrast to the lack of an association with aflatoxin, PLC

mortality was highly correlated with HBsAg-f- prevalence. PLCmortality was not correlated with past HBV infection (r =â€”0.08),as assessed by the prevalence of antibody to the HBV

core protein, (data not shown). These relationships betweenPLC mortality and persistent and past HBV infection areconsistent with the hypotheses and data of Blumberg and London (15) and Beasley and Hwang (16).

The association of PLC mortality rates with liquor intake,making up about 80% of total alcohol intake, is in accord withsimilar observations of total alcohol intake (25, 59, 60). However, the liquor association was eliminated when controlling forplasma cholesterol, as in models 4 and 5 of Table 4. If some ofthe liquor effect was explained by higher plasma cholesterol,then this implies that a sustained intake of liquor may beassociated with elevated plasma cholesterol. Indeed, such arelationship has been reported (61).

The association of PLC mortality with plasma cholesterol isespecially interesting. Inspection of the partial regression coef-

6887




Table 5 Comparison of strengths of principal ecological studies to detect contributions ofaflatoxin intake and HBsAg+ prevalence lo PLC risk

ReportNo. of survey

districts CountryRange of PLC Range of aÃ±a- Range of HBsAg+mortality rate toxin intake prevalence Ref.

Pre-IARC (1987) 2 Thailand 3.03 Kenya 4.21 Mozambique4 Swaziland 4.28 Mozambique/Transkei 3.5

10 Combined data (first 4 10.8studies)

17C Combined data (first 5 14.8

studies)

9.03.0

8.511.1

25.6*

52.4

Shank el al.. 1972(50)Peersand Linseil, 1973(51)Van Rensburg et al., 1974 (52)Peers et al.. 1976(22)Van Rensburg et al.. 1985 (53)

Shank et ai., 1972(50)Peersand Linsell, 1973 (51)Van Rensburg et al.. 1974 (52)Peers et al.. 1976(22)Van Rensburg et al.. 1985 (53)

Post-lARC(1987)10"8'SJ48SwazilandKenyaGuangxi,

ChinaChina8.611.

2(M)'3.7

(ff3.539.013.9'3.5

(M)*4.6(Ff172.0600+'1.34.7

(M)17.2(F)*1.0728.0Peers

et al., 1987(19)Autrupetal..1987(12)\ehetal.,

1989(26)Chenet al.. 1990(27)

Â°A later study by Bagshawe et al. (57) found no differences in HBsAg+ prevalence between these 3 regions.* Range based on low value of 3.5 mg/kg body weight/day from Shank et al. (50) and high value of 89.5 mg/kg body weight/day in 1985 paper of Van Rensburg

et al. (53).f The 1985 Mozambique/Transkei study (53) includes the single survey site of the 1974 Mozambique/Transkei study (52), thus accounting for 17 sites rather than

18.d Reinvestigation of 1976 study of Peers et al. (22) and including 10 smaller areas within 4 original districts.'Total aflatoxins: 21% B,, 15% B2, 49% G,, 44% G2.^Ten areas were surveyed, but only 8 had complete data." Urinary excretion ofaflatoxin adducts.* Excludes one survey site with 0% HBsAg+ carrier rate.' Aflatoxin data not available for one site.

ficients in Table 4 shows that this association was even moreconsistent than the association of PLC mortality with HBsAg+prevalence. The chief determinants of elevated plasma cholesterol include dietary, environmental, and genetic factors.Among the dietary constituents, the effects of saturated fat arewell established. Less well known is the rather remarkable effectof animal protein (62, 63), along with somewhat more modestand variable effects of dietary fiber and, possibly, dietary cholesterol itself (49). The direct association of plasma albumin(64) and the inverse association of legume consumption (63)with plasma cholesterol have been previously observed. Theassociation of plasma cholesterol with PLC mortality also isconsistent in this study with similar associations of plasmacholesterol with mortalities from colon cancer (P < 0.01 ), rectalcancer (P < 0.05), lung cancer (P < 0.05), leukemia (P < 0.05),brain cancer (P < 0.05), and total aggregate cancer (P < 0.05)(27).

The association of adult height with plasma cholesterol indicates that young people who consume diets with greateramounts of protein and fat, such as with the inclusion of foodsof animal origin, grow faster and may become somewhat talleradults. The data from the parent study (27) showed that adultbody size was correlated highly significantly with several cancermortality rates (P < 0.01).

The link between plasma cholesterol and diets containingmore foods of animal origin was somewhat surprising in viewof the low plasma cholesterol and the low intake of such foodsin China. Plasma cholesterol in this cohort ranged up to about190 m/kg/dl (for pools of individuals and after correction formethodological underestimate) which is near the low end of therange for comparable Western subjects. And mean intake ofanimal products in China was quite low; for example, thesefoods contributed, on average, only about 11% of the totalprotein intake in the 65 counties of the parent survey (27).

The finding that elevated plasma cholesterol and increasedprevalence of HBsAg+ were significant risk factors for PLCmortality, while aflatoxin was not a risk factor, is consistent

with data from animal studies. When groups of animals all fedthe same level of dietary protein (20% casein) were givenincreasing doses of aflatoxin, the expected dose-response relationship was observed for the formation of presumptive preneo-plastic liver lesions (39). In contrast, when animals were fedeither lower levels of 5-10% animal protein (39, 65, 66) or thesame level (20%) of plant protein (67) after completion ofaflatoxin dosing, development of preneoplastic lesions andtumors5 was markedly inhibited. Both the promotion and progression stages of aflatoxin-induced tumor development maybe readily modulated by the level and type of dietary proteinintake, completely eliminating the dose-dependent response

reduced by the level ofaflatoxin intake (68). In these low proteinfed animals, plasma cholesterol also was reduced.5 The level of

dietary protein required for inhibition of hepatic foci and hepatic tumor development (<10% protein) is near the levelrequired for maximum growth rate (69) or, said another way,is the amount generally available in a diet mostly comprised offoods of plant origin.

The mean level of protein intake in the Chinese cohort, at0.99 g/kg body weight, is modestly above the United Statesrecommended dietary allowance of 0.8 g/kg (70) and significantly below the mean of 1.18 g/kg for comparable Americanadult males (71). The more significant difference in proteinintake between these populations, however, is the ratio of plantto animal protein (including fish). In China animal proteinrepresents only about 11% of the total (8% nonfish and 3%fish), while in the United States it is about 70% of the total orabout 0.13 and 0.82 g/kg body weight/day, respectively [seeChen et al.(21) for estimation details]. The intake of efficientlyutilized protein required for good growth both in rats and inpeople, when expressed per unit body weight, are very similar:about 9-10% of total dietary protein for both species, anamount approximately equivalent to the recommended dietaryallowance (70). Even though the mean intake in China is close

5L. D. Youngman and T. C. Campbell, unpublished data.

6888




to this recommended intake, about one-half of the populationconsume more. If this extra dietary protein were to be plantprotein consumed as part of a simple diet (that is, a diet withlittle food variety), its adverse effects on PLC mortality likelywould be limited; however, if the extra protein were to beobtained even from small additions of animal protein, measurable risk for PLC mortality may occur.

The finding of a nonassociation between aflatoxin intake andPLC mortality, which contrasts with the results of many previous studies, is provocative. An explanation of this differencemay be considered from several points of view. First, Chinesemight respond differently to aflatoxin, perhaps because ofunique genetic or environmental characteristics. While thisexplanation never can be fully discarded, very likely it is quiteuntenable, first, because some of the previous data interpretedas showing a positive association between aflatoxin intake andPLC mortality were based on investigations of Chinese subjects(18, 26) and, second, because major ethnic differences in riskfor other cancers are greatly reduced or eliminated after migration to new environments.

A second argument could be that the null effect in this studymay have occurred because measurement of aflatoxin exposureduring the survey period was not representative of past intakeswhen the cancers were forming. Although this possibility cannotbe discounted, it should be remembered that a similar limitationalso existed for all previous studies used to justify aflatoxin asa human carcinogen, especially those which relied on aflatoxinanalysis of marketplace samples and use of food disappearancerecords. In fact, in this study in China, aflatoxin exposure verylikely is more reliable because of the analysis of urinary aflatoxin metabolite excretion which directly represents and integrates over a day or so actual consumption. Moreover, aflatoxincontamination rates in a county in the Guangxi AutonomousRegion were relatively stable during the years 1972-1983 (seeabove). In a study documenting long-term aflatoxin exposurein the United States, no association with PLC was observed(23).

A third interpretation of these data suggests that aflatoxinmay not be a significant human carcinogen. Three types ofevidence support this view. First, the statistical power and morecomprehensive range, diversity, and inclusiveness of risk factorsavailable in this study far surpasses not only similar features ofprevious individual cross-sectional studies but also substantiallysurpasses their combined experimental features (Table 5). Second, the data from this study in China, which corresponds withlaboratory animal observations, may be explained by a biologically plausible hypothesis. And third, these data suggest humanresistance to this animal carcinogen, a finding which is supported by in vitro aflat ox in studies on species of varying resistance (72-74). Also, humans may be further refractory whenconsuming lower protein diets. [Incidentally, inhibition of af-latoxin-induced liver cancer with consumption of low proteindiets should not be confused with an enhancement of aflatoxin-induced acute toxicity (75). Under these conditions of lowprotein feeding, it has been hypothesized that there is an accumulation of unmetabolized parent compound which inhibitscell respiration to cause cell death (76), a hypothesis which isalso supported by the reports of acute toxicity in protein-malnourished children fed aflatoxin-contaminated foods (77).]

Although it might be argued that the China data are moreconvincing because of the supporting experimental animal andbiochemical evidence concerning the effects of nutrition andnatural species resistance and because of the considerable ad

vantage in experimental parameters, such a conclusion is notfully satisfying. Even if the previous studies were much morelimited in experimental design features, positive aflatoxin-PLCassociations nonetheless were reported and must be explained.

In an attempt to provide an explanation for these differingresults, we offer the following review of interpretations givento previous investigations and reviews, particularly the conclusions drawn by the 1987 IARC report (5). Unfortunately,several misinterpretations and misstatements of fact were foundin this 1987 IARC report (5) and some of these were notedearlier. These shortcomings, alone, cast doubt on the revisedconclusion that there was sufficient evidence to conclude aflatoxin human carcinogenicity.

The most compelling evidence in favor of aflatoxin humancarcinogenicity is the high correlation obtained [r = 0.88, (53)]for aflatoxin intake and PLC mortality when the results fromthe first five studies of Table 5 are combined. Such a correlationis impressive, especially when considering the rather severeexperimental limitations of the individual studies. One of themore severe limitations noted by others (10, 78) is the absenceof data concerning HBV infectivity; none of these studiesobtained this information, although one (51) provided complimentary data only in retrospect (57). If an independent aflatoxineffect is hypothesized, as suggested by the data, then it isnecessary to assume, among other assumptions, that HBsAg+prevalence is relatively constant among the 17 survey sites ofthese studies (that is, if this virus is to be considered animportant risk factor). This assumption seems unlikely. Forexample, later studies in these same regions showed thatHBsAg+ carrier prevalence rates were quite variable, being 5-23% in Kenya (12) and 5-35% in Swaziland (19) which aresimilar to the 1-28% range observed by us in China (27).Variation of this magnitude offers the very real possibility thatthe observed aflatoxin effect may be attributed to HBsAg+carrier prevalence, especially when this variation was reportedfor the same survey regions which supplied data points for thecombined regression. Furthermore, our calculation indeedshows that, in the study of Peers et al. (19), aflatoxin intakeand HBsAg+ prevalence are significantly correlated (r = 0.73,P < 0.05). The study of Yeh et al. (26) of four sites in Guangxi,China, is the only report showing an aflatoxin effect on PLCrisk in the presence of a reasonably constant HBsAg+ carrierprevalence (21.6-24.8%). However, these prevalence data contrast with those of Chen et al. (27) who included four countiesin the same Guangxi Autonomous Region and found variableprevalence rates of 8-23% (50 subjects/county, 50% of eachsex) (indeed, one county, Fusui, was included in both studies,with a 23% prevalence of HBsAg+ in both investigations). Insummary, considerable variation in HBsAg-t- prevalence rates

are likely to occur both in Africa and southeast China, andfurthermore, these rates may correlate with aflatoxin intake( 19), thus confounding observed relationships between aflatoxinand PLC.

In the earlier studies published prior to the 1987 IARC report(5), numerous other shortcomings (crude estimates of aflatoxinintake, disease diagnosis errors, incomparable population bases,etc.) have been noted by other commentators (10, 78) but errorsof this type should not negate the aflatoxin-PLC correlationsobserved; such shortcomings would only add "noise" to the

analysis. Another explanation of the observed aflatoxin effectis the possibility that aflatoxin exposure is a surrogate measurement for the presence of other unmeasured risk factors,including, for example, alcohol, cadmium, and plasma choles-

6889




terol or its dietary determinants, as illustrated in the China database. Surrogate representation, however, would require relatively constant exposure to these unknown factors among thesevarious studies.

An analysis of these data reported here would not be completewithout careful comparison with three recent studies (12, 19,26) each of which was cross-sectional in design and each ofwhich included simultaneous measurement of PLC mortality,HBsAg+ carrier rates, and aflatoxin exposure. One study (19)was cited by the 1987 IARC report (5) and one (26) is beingconsidered by current committees (United States Food andDrug Administration, California Proposition 65 Risk Assessment Committee, and industry groups) as a very importantadjunct to the 1987 IARC report (5). Aflatoxin exposure wasassociated with PLC mortality in the studies of Yeh et al. (26)and Peers et al. (19) but not in the study of Autrup et al. (12)(except a smaller indeterminant subset of the population). Comparisons of a few statistical parameters for these studies areshown in Table 6. HBsAg-l- carrier rates for males were comparable for Yeh et al. (19.5-24.8%) and Peers et al. (20-35%),both being somewhat higher than Autrup et al. (5%-23%). Thestudies of Yeh et al. (26) and Peers et al. (19) afford the mostappropriate comparison, both because of comparable meanHBsAg+ prevalence and because both studies observed a highlysignificant aflatoxin effect. However, in other ways, the resultsof these two studies were inconsistent as illustrated by comparison of the respective regression equations (Table 6). Althoughthe slope coefficients (0.15 and 0.12) were reasonably similar,the intercepts (mortality rates in absence of aflatoxin exposure)were markedly different. The most likely explanation might bethat the mean HBsAg+ carrier rates were much higher in thestudy by Yeh et al. (26), but this does not appear to be true; ifanything, the rates actually measured are lower than those ofPeers et al. (19). Thus, there appears to be a major discrepancyin PLC rates between southern China and east Africa (i.e.,different intercepts) that cannot be attributed either to a difference in the observed HBsAg+ prevalence or to a difference inaflatoxin exposure. Another discrepancy apparent from a comparison of these studies is the findings of Peers et al. (19) whoobserved a significant HBsAg+ effect on PLC mortality; in

contrast, Yeh et al. (26) saw no HBsAg-l- effect in the cross-

sectional analysis, although they did observe a highly significanteffect when comparing cases and controls within their cohort.And finally, the combined data of the previous cross-sectionalstudies (pre-1987 IARC report) exhibit a somewhat similarintercept to that of Peers et al. (19); in fact, these combineddata (17 points) include 4 survey sites representing the 10subregions included in the later Peers et al. ( 19) study. However,the slope constant of the combined studies is nearly triple theslope constant for the study by Peers et al. (19).

The third recent study was that by Autrup et al. (12) whofound no aflatoxin effect on PLC incidence rates when allethnic, social, and cultural groups were included in the analysis(our calculation of the correlation coefficients for males andfemales actually showed inverse associations, â€”0.40and â€”0.35,respectively), but a significant positive association was notedwhen the analysis was limited to the Bantu people (no datawere given on the apparently small number of these people).Not only was there no aflatoxin effect in this study (12) butalso there was no HBsAg-l- effect.

When all previous studies, before and after the 1987 IARCreport (5), are considered, there is substantial uncertainty concerning the role of aflatoxin in the etiology of human PLC. Ata minimum, we believe that it is definitely not possible toconclude that there is sufficient evidence for human carcinogen-icity (5), particularly because the criteria of "chance, bias andconfounding" cannot "be ruled out with reasonable confidence"

as required by the IARC guidelines (5). Also adding doubt onthe sufficiency of evidence are the observations, first, that thehuman species is highly resistant when compared in vitro toother species (72-74) and, second, that the dose-response relationship between aflatoxin exposure and PLC rates reportedfor humans is asymptotic to an upper PLC limit (concavedownward) which is opposite to that observed for the relationship between aflatoxin and putatively preneoplastic liver lesionsin experimental animals (concave upward) (68).

Comparison of these previous studies illustrates serious difficulties of interpretation which are not easily resolved and nonew information concerning causality is apparent. In contrast,the study by Chen et al. (27) indicates highly significant risk

Table 6 Statistical parameters" and results of three comparable studies on aflatoxin exposure, HBsAg+ carrier rate and PLC mortality

Correlationcoefficients*StudyPeers

et al. (I987)fAutrup et al. (1987)'

MF

Yeh Ã©tal.(1989)Â«Combined data (Table 5)*No.

ofsites108

8517AF

vs.PLC0.84J-0.40(NS)

-0.35 (NS)0.996''0.85'HBsAg+

vs.PLC0.73'0.17(NS)

0.49(NS)0.28(NS)AFvs.

HBsAg+0.73'0.23(NS)

-0.46 (NS)O.OO(NS)Regression

equationPLC= 4.1 +0.12AFPLC

= 3.7 - 0.05 AFPLC= 1.2-0.03 AFPLC= 159 + 0.15 AFPLC = 8.7 + 0.32 AF

Â°All parameters calculated from data in original papers.* AF, aflatoxin; NS. not significant (P > 0.05).c Original 5-year PLC incidence rates for mÃ¢les,annulized and truncated for Ã¢ges15-64 years. Aflatoxin exposure measured as /jg/person/day converted to ng/kg

body weight/day assuming 70 kg body weight.'/>' P < 0.05.' Original PLC incidence rates converted to truncated rates for ages 15+ years; rates in Table 5 of Autrup et al. (12) divided by 0.61 to account for 39% of total

population <1 5 years (based on demographic distribution for Senegal as a representative rural country (41) and assuming no PLC cases <1 5 years). Aflatoxin-guanineadduci in urine assumed to be 1% of aflatoxin intake by Autrup et al. (12).

' PLC incidence rates reported for males aged 25-64 years. Aflatoxin exposure of mg/person/day converted to ng/kg body weight/day assuming 49 kg body weight[recorded in same area by Chen et al. (27)]. Aflatoxin data not available for one site.

* No HBsAg+ data available. Originally published mortality rates made equivalent as follows: Shank et al. (50), original rates (all ages, both sexes) converted to

truncated rates for 15+ years, assuming 4; 1 male:female PLC incidence ratio, 1:1 sex distribution: and age distribution of Senegal (41) as representative of a ruraldeveloping country: Peers et al. (22), original rates (all ages, males only, ages 15+ years) used directly from Table 5: Peers et al. (51), original rates (all ages, malesonly, ages 15+ years), used directly from Table 4; Van Rensburg et al. (53), original rates (all ages, males only) converted to truncated rates, 15+ years, using Senegal(41) age distribution.

!P< 0.001.

6890




attribution for HBsAg+ prevalence and dietary/nutritional determinants of plasma cholesterol. To further advance this research, analytical cohort studies of individuals are needed thatpermit simultaneous analysis of as many potential risk factorsas possible. Presently, the IARC (79) is carrying out such astudy in Thailand among 3000 HBsAg+ carriers with the statedintention to measure several factors possibly associated withPLC risk. This study also offers the advantage that aflatoxinexposure will be estimated by measurement of circulatingplasma levels of aflatoxin-albumin adducts which integratesaflatoxin intake over a period of time even longer than thatused in the China study reported here. These results are eagerlyawaited.

In view of the comprehensive and reliable nature of the Chinadata, we propose the following model to explain the etiology ofPLC. For a given population, the vast majority of individualswho are susceptible to PLC are mostly limited to those whoremain persistently infected with HBV, a conclusion drawnfrom the extraordinary PLC risk observed for HBsAg+ carriers(16). Thus, the first approximation of population risk of PLCis the prevalence of HBsAg+ carriers. Then, within theHBsAg-l- carrier population, additional risk is contributed

chiefly by nutritional and dietary practices that enhance livercell proliferation, such as with diets containing significantamounts of animal protein. In contrast, aflatoxin contributeslittle or no PLC risk. Even though aflatoxin may act as acarcinogenic initiator, undoubtedly it contributes only a verysmall proportion of the initiating activity routinely exposingthe liver. Innumerable dietary constituents, upon metabolicactivation (mostly in the liver itself), may be genotoxic (80).Even modestly small amounts of these substances should beable to cause enough initiation for tumor occurrence providedthat, subsequently, there is adequate nutrient activity to causeliver cell proliferation. The HBV virion also may be capable ofcausing initiation through insertion of its DNA into liver cellDNA(81).

This model suggests that HBsAg+ positivity is a necessarybut insufficient cause of PLC, that aflatoxin is an unnecessaryand insufficient cause, and that sustained nourishment causingliver cell proliferation (and elevated plasma cholesterol) is anecessary and insufficient cause for HBsAg negative carriersbut a necessary and sufficient cause for HBsAg+ carriers. Thus,PLC may occur among HBsAg+ carriers consuming no aflatoxin but sufficient amounts of a nutrient-dense diet that stimulates liver cell proliferation. There is also the possibility thatsuch a diet first could stimulate replication of the incubatingvirus which then causes liver cell proliferation.6

An obvious question concerning the effect of "enriched"

nutrition upon PLC etiology is: why is this disease so muchmore common in undernourished and impoverished societies?According to this model, PLC is more common becauseHBsAg+ carriers are more common, not because undernourishment is more common. Then, among the HBsAg+ carrierpopulation, there is additional risk for those who consumesignificant amounts of dietary constituents, such as animalprotein, that encourage liver cell proliferation. Even though,within these societies, a large number of people consume nutritionally poor diets containing minimal food variety and littleor no animal protein, enough of the HBsAg+ carriers stillconsume enough animal protein (or total mixed protein andcompanion nutrients) to elevate plasma cholesterol, induce liver

6 S. Broder, personal suggestion.

cell proliferation, and promote tumor development. HBsAg+carriers are likely to be particularly vulnerable to PLC evenwith modest nutritional enrichment. For aflatoxin to contributesignificant risk, consumption must be high enough and prolonged enough to cause significant liver cell proliferation.

These hypothesized relationships suggest that, in addition tothe HBsAg+ carrier rate, the level of animal protein intake orsimilar nourishment would be the next most important determinant of PLC risk. This would be indicated by increasingprevalence of PLC among HBsAg+ carriers who consumeincreasing amounts of animal protein. Some evidence for thisrelationship is indicated by a crude comparison of PLC prevalence among HBsAg-l- carriers in Swaziland [estimated 1.8%

prevalence with negligible intake of animal protein, as indicatedin report of Peers et al. (19)], China [16% prevalence with lowbut significant animal protein intake (27)], and Taiwan [40-50% prevalence (16) with medium animal protein intake (82)].

This model emphasizes (a) the role of liver cell proliferationin tumor development, recently emphasized by others (83, 84),(Â¿>)the abundance of dietary constituents capable of causingliver cell initiation (80), (c) the great risk for PLC amongHBsAg-l- carriers (14), and (d) the ability of animal protein

(perhaps also other nutrients) to modulate tumor occurrence(even after relatively modest initiation may have occurred).

This model is not unlike that observed for alcoholics whoincur a higher risk for liver cancer after withdrawal of alcoholthan before. Lee (85), for example, concluded from his resultsthat after "liver disease is established (as with alcohol), and

with the substitution of a nutritious diet and a reduced alcoholintake, circumstances may be more favourable for the development of hepatoma." Accordingly, with HBV infection, if im

munity to HBV infection cannot be established early to eliminate persistent infection, then care must be used thereafter tominimize the consumption of a nutritionally rich (animal protein, fat) diet.

And finally, this explanation is in harmony, nutritionallyspeaking, with the preventive effect of nutrition upon the development of other cancers (86).

ACKNOWLEDGMENTS

The authors thank Carol Roertgen and Marilyn Vogel for manuscripttyping, Martin Root for technical assistance, and Dr. John Groopmanfor analyses of urine samples for aflatoxin adducts.

REFERENCES

1. Lancaster, M. C., Jenkins, F. P., and Philip, J. M. Toxicity associated withcertain samples of groundnuts. Part I. Nature (London), Â¡92:1095-1097.1961.

2. Butler, W. H., and Barnes, J. M. Toxic effects of groundnut meal containingaflatoxin to rats and guinea pigs. Br. J. Cancer, 17:699-710, 1964.

3. Wogan, G. N., and Newberne, P. M. Dose-response characteristics of aflatoxin B, carcinogenesis in the rat. Cancer Res., 27: 2370-2376, 1967.

4. International Agency for Research on Cancer. IARC Monographs on theEvaluation of the Carcinogenic Risk of Chemicals to Man: Some NaturallyOccurring Substances, Vol. 10, pp. 51-72. Lyon, France: InternationalAgency for Research on Cancer, 1976.

5. International Agency for Research on Cancer. IARC Monographs on theEvaluation of the Carcinogenic Risk of Chemicals to Humans: Aflatoxins.Suppl. 7, pp. 83-87. Lyon, France: International Agency for Research onCancer, 1987.

6. International Agency for Research on Cancer. IARC Monographs on theEvaluation of the Carcinogenic Risk of Chemicals to Man: Some InorganicSubstances, Vol. 1, pp. 145-156. Lyon. France: International Agency forResearch on Cancer, 1972.

7. World Health Organization Task Group on Environmental Health Criteriafor Mycotoxins. Mycotoxins. Environmental Health Criteria 11, pp. 1-127.

6891




Geneva, Switzerland: United Nations Environment Programme and theWorld Health Organization, 1979. 38.

8. Kilman, S. Spreading poisonâ€”fungus in corn crop, a potent carcinogen,invades food supplies, pp. Al, A8, A9. The Wall Street Journal (EasternEd.). White Oak, MD: February 23, 1989.

9. Linsell, A. Carcinogenicity of mycotoxins. In: Environmental Carcinogens. 39.Selected Methods of Analysis, IARC scientific publication 5, pp. 3-14. Lyon.France: International Agency for Research on Cancer, 1982.

10. Wagstaff, D. The use of epidemiology, scientific data, and regulatory author- 40.ity to determine risk factors in cancers of some organs of the digestive system.Liver cancer. Regul. Toxicol. Pharmacol., 5: 384-404, 1985.

11. Stoloff, L. A rationale for the control of aflatoxin in human foods. In: P. S. 41.Steyn et al. (eds.). Mycotoxins and Phytotoxins. pp. 457-471. Amsterdam,The Netherlands: Elsevier Science Publishers, 1986.

12. Autrup, H.. Seremet, T., Wakhisi, J.. and Wasunna, A. Aflatoxin exposure 42.measured by urinary excretion of aflatoxin Bl-guanine adduci and hepatitisB virus infection in areas with different liver cancer incidence in Kenya.Cancer Res., 47: 3430-3433, 1987. 43.

13. Stoloff, L. Aflatoxin is not a probable human carcinogen: the publishedevidence is sufficient. Regul. Toxicol. Pharmacol., JO: 272-283, 1989.

14. Beasley, R. P., Huang, L.-Y., and Lin. C.-C. Hepatocellular carcinoma and 44.hepatitis B virus. A prospective study of 22,707 men in Taiwan. Lancet, 2:1129-1132, 1981.

15. Blumberg, B. S.. and London. W. T. Primary hepatocellular carcinoma and 45.hepatitis B virus. Curr. Probi. Cancer, 6: 3-23, 1982.

16. Beasley, R. P.. and Hwang, L.-Y. Hepatocellular carcinoma and hepatitis Bvirus. Semin. Liver Dis., 4: 113-121, 1984.

17. Sun, T.-T., and Chu, Y.-Y. Carcinogenesis and prevention strategy of liver 46.cancer in areas of prevalence. J. Cell. Physiol., 3 (Suppl.): 39-44, 1984.

18. Yeh. F.-S., Mo, C.-C.. and Yen. R.-C. Risk factors for hepatocellular carcinoma in Guangxi, People's Republic of China. Nati. Cancer Inst. Monogr., 47.69:47-48, 1985.

19. Peers, F., Bosch, K., Kaldor, J., Linsell. A., and Pluijmen, M. Aflatoxin 48.exposure, hepatitis B virus infection and liver cancer in Swaziland. Int. J.Cancer, 39: 545-553, 1987. 49.

20. Li, W., Liu, Y., Hu, Y., Yeh, P., Chen. J., and Liu, C. Analysis of epidemiologie factors of liver cancer in Jiangsu, Zhejiang and Shanghai. Chin. J.Oncol., 5:48-51 (in Chinese), 1983. 50.

21. Wang, Y.. Lan, L., Ye. B.. Ku. Y., Liu, Y., and Liu, W. Correlation betweengeographical distribution of liver cancer and climate, aflatoxin B. in China.Sci. Sin. Ser. B.. 26: 1166-1175. 1983.

22. Peers. F. G., Gilman, G. A., and Linsell, C. A. Dietary aflatoxins and human 51.liver cancer. A study in Swaziland. Int. J. Cancer, 17: 167-176, 1976.

23. Stoloff, L. Aflatoxin as a cause of primary liver-cell cancer in the United 52.States: a probability study. Nutr. Cancer. 5: 165-185, 1983.

24. Hayes. R. B.. Van Nieuwenhuize. J. P.. Raatgever. T. W., and ten Kate. F.J. W. Aflatoxins exposures in the industrial setting: an epidemiological study 53.of mortality. Food Chem. Toxicol., 22: 39-43, 1984.

25. Bulatao-Jayme. J., Almero, E. M., C. A., C., Jardeleza. M. T. R.. andSalamat, L. A. A case-control dietary study of primary liver cancer risk fromaflatoxin exposure. Int. J. Epidemici., //: 112-118, 1982. 54.

26. Yeh, F.-S., Yu, M. C., Mo, C.-C., Luo, S., Tong, M. J.. and Henderson, B.E. Hepatitis B virus, aflatoxins, and hepatocellular carcinoma in SouthernGuangxi, China. Cancer Res., 49: 2506-2509. 1989. 55.

27. Chen, J.. Campbell, T. C.. Li, J., and Peto. R. Diet, lifestyle, and mortalityin China: A study of the characteristics of 65 Chinese counties. Joint 56.publication of Oxford University Press (Oxford. England). Cornell UniversityPress (Ithaca, NY), and China People's Medical Publishing House (Beijing. 57.

China), 1990.28. Piazza, A. Food Consumption and Nutritional Status in the PRC. Boulder 58.

CO: Westview Press, 1986.29. Editorial Committee. Atlas of Cancer Mortality in the People's Republic of 59.

China. Shanghai. China: China Map Press, 1979.30. Li, J., Liu. B., Li, G., Chen, Z., Sun, X., and Ring, S. Atlas of cancer

mortality in the People's Republic of China. An aid for cancer control andresearch. Int. J. Epidemiol., 10: 127-133. 1981. 60.

31. Groopman, J. D., Trudel, L. J.. Donahue. P. R., Marshak-Rothstein, A., andWogan. G. N. High-affinity monoclonal antibodies for aflatoxins and theirapplication to solid-phase immunoassays. Proc. Nati. Acad. Sci. USA, SI: 61.7728-7731, 1984.

32. Groopman, J. D.. Donahue, P. R.. Zhu, J.. Chen, J.. and Wogan, G. N.Aflatoxin metabolism in humans: detection of metabolites and nucleic acid 62.adducts in urine by affinity chromatography. Proc. Nati. Acad. Sci. USA. 82:6492-6496, 1985.

33. Groopman, J. D., Donahue. P. R., Zhu, J., Chen. J.. and Wogan. G. N.Temporal patterns of aflatoxin metabolites in urine of people living inGuangxi Province. PRC. Proc. Am. Assoc. Cancer Res., 28: 130, 1987. 63.

34. Ma, X.-K. Synthesis of HBcAg in and its application in diagnosis of viralhepatitis B. Chin. Peoples Liberation Army Med. J.. 9: 1-5. 1984.

35. Engvall, E., and Perlman, P. Enzyme-linked Â¡mmunoabsorbent assay(ELISA). Quantitative assay of immunoglobulin G. Immunochemistry, 8: 64.871-874, 1971.

36. Parekh, A. C., and Jung, D. H. Cholesterol determination with ferric acetate-uranium acetate and sulfuric acid-ferrous sulfate reagents. Anal. Chem.. 42: 65.1423-1427, 1970.

37. Feng, H., Pi, C., Wang. R.. and Chen. L. Use of ferric ammonium sulfate in

6892

serum cholesterol determination. Clin. Chem., 19: 121-122. 1973.Capar. S. G., Gajan. R. J.. Madzsar, E., Albert. R. H.. Sanders. M.. andZyren. J. Determination of lead and cadmium in foods by anodic strippingvoltammetry. II. Collaborative study. J. Assoc. Off. Anal. Chem., 65: 978-986. 1982.Dunaif, G., and Campbell. T. C. Dietary protein level and aflatoxin B,-induced preneoplastic hepatic lesions in the rat. J. Nutr.. 117: 1298-1302,1987.National Cancer Institute. Division of Cancer Prevention and C'ontrol, Annual Cancer Statistics Review. MH publication 88-2789. pp. N.29-IV.37.Washington, DC: Government Printing Office. 1987.Waterhouse, J., Muir, C.. Corryea. P., and Powell, J. Cancer incidence infive continents. Vol. II, IARC Scientific publication 15. pp. 338. Lyon,France: International Agency for Research on Cancer. 1976.Okuda. K., and Beasley, R. P. Epidemiology of hepatocellular carcinoma. In:K. Okuda et ai. (eds.), Hepatocellular Carcinoma, report 17, pp. 9-30.Geneva, Switzerland: World Health Organization. 1982.Young. J. L. J.. and Percy, C. L. Surveillance Epidemiology and End Results.Incidence and mortality data 1973-1977, DHHS publication NIH 81-2330,pp. 168. Bethesda, MD: National Cancer Institute. 1981.Lipid Research Clinics Program Epidemiology Committee. Plasma lipiddistributions in selected North American population. The lipid researchclinics program prevalence study. Circulation. 60: 427-439. 1979.Pao, E.. Fleming, K. H.. GÃ¼nther.P. M.. and Mickle, S. J. Foods commonlyeaten by individuals: amount per day and per eating occasion. Report No.44. pp. 431. Washington, D.C.: United States Department of AgricultureHuman Nutrition Information Service, 1982.Gunderson. E. L. FDA total diet study. April 1982-April 1984. dietaryintakes of pesticides, selected elements and other chemicals. J. Assoc. Off.Anal. Chem.. 71: 1200-1209. 1988.Szmuness. W. Hepatocellular carcinoma and the hepatitis B virus: evidencefor a causal association. Prog. Med. Virol.. 24: 40-69. 1978.Arthur. M. J. P., Hall. A. J.. and Wright. R. Hepatitis B. hepatocellularcarcinoma, and strategies for prevention. Lancet. /: 607-610, 1984.Zilversmit. D. B. Nutrition, lipids and coronary heart diseaseâ€”a global view.In: R. 1. Levy et al. (eds.), Dietary Fiber, pp. 149-174. New York: RavenPress, 1979.Shank. R.. Bhamarapravati. N.. Gordon, J. E., and Wogan. G. N. Dietaryaflatoxins and human liver cancer. IV. Incidence of primary liver cancer intwo municipal populations in Thailand. Food Cosmet. Toxicol., 10: 171-179. 1972.Peers. F. G., and Linsell, C. A. Dietary aflatoxins and liver cancerâ€”apopulation-based study in Kenya. Br. J. Cancer, 27:473-484, 1973.Van Rensburg, S. J., Van der Watt, J. J., Purchase, 1. F. H., Coutinho, L.P.. and Markham. R. Primary liver cancer rate and aflaloxin intake in a highcancer area. S. Afr. Med. J.. 48: 2508a-2508d. 1974.Van Rensburg, S. J.. Cook-Mozaffari. P.. Van Schalkwyk. D. J., Van DerWatt, J. J., Vincent. T. J., and Purchase, I. F. Hepatocellular carcinoma anddietary aflatoxin in Mozambique and Transkei. Br. J. Cancer, SI: 713-726.1985.Van Rensburg. S. J. Role of epidemiology in the elucidation of mycotoxinhealth risks. In: J. V. Rodericks et al. (eds.). Mycotoxins in Human andAnimal Health, pp. 699-711. Park Forest, IL: Pathotox Publishers, 1977.Linsell. C. A., and Peers. F. G. Aflatoxin and liver cell cancer. Trans. R. Soc.Trop. Med. Hyg., 71: 471-473, 1977.Peers. F. G.. and Linsell, C. A. Dietary aflatoxins and human liver cancer.Ann. Nutr. Aliment.. 31: 1005-1018, 1977.Bagshawe, A. F., Gacengi. D. M.. Cameron. C. H., DormÃ¡n. J.. and Dane.D. S. Hepatitis B antigen and liver cancer. Br. J. Cancer..?7:581-584, 1975.Whitaker, T. B., Dickens. J. W., and Monroe. R. J. Variability of aflatoxintest results. J. Am. Oil Chem. Soc., 51: 214-218, 1974.Oshima. A.. Tsukuma. H.. Hiyama. T.. Fujimoto, I., Yamano. H.. andTanaka. M. Follow-up study of HBsAg-positivc blood donors with specialreference to effect of drinking and smoking on development of liver cancer.Int. J. Cancer. 34: 775-779. 1984.Stemhagen. A.. Slade, J., Altman, R.. and Bill. J. Occupational risk factorsand liver cancer. A retrospective case-control study of primary liver cancerin New Jersey. Am. J. Epidemiol., 117: 443-454, 1983.Willen. W.. Hennekens. C. H.. Siegel. A. J.. Adner. M. M.. and Castelli, W.P. Alcohol consumption and high-density lipoprotein cholesterol in marathon runners. N. Engl. J. Med., 303: 1159-1161. 1980.Sirtori. C. R.. Noseda. G., and Descovich. G. C. Studies on the use of asoybean protein diet for the management of human hyperlipoproteinemias.In: M. J. Gibney et al. (eds.). Animal and Vegetable Proteins in LipidMetabolism and Atherosclerosis, pp. 135-148. New York: Alan R. Liss, Inc.,1983.Carroll. K. K. Dietary proteins and amino acidsâ€”their effects on cholesterolmetabolism. In: M. J. Gibney et al. (eds.). Animal and Vegetable Proteins inLipid Metabolism and Atherosclerosis, pp. 9-17. New York: Alan R. Liss.Inc., 1983.Silverman. L. M.. Cristenson. R. H.. and Grant. G. H. Amino acids andproteins. In: N. W. Tietz (ed.). Textbook of Clinical Chemistry, pp. 587.Philadelphia: W. B. Saundcrs. 1986.Appleton. B. S.. and Campbell, T. C. Effect of high and low dietary proteinon the dosing and postdosing periods of aflatoxin Bl-induced hepatic preneoplastic lesion development in the rat. Cancer Res., 43: 2150-2154. 1983.




66. Appleton, B. S., and Campbell, T. C. Dietary protein intervention during thepost-dosing phase of aflatoxin Bl-induced hepatic preneoplastic lesion development. J. Nati. Cancer Inst., 70: 547-549. 1983.

67. Schulsinger, D. A., Root. M. M.. and Campbell, T. C. Effect of dietaryprotein quality on development of aflatoxin B,-induced hepatic preneoplasticlesions. J. Nail. Cancer Inst.. 81: 1241-1245, 1989.

68. Dunaif. G.. and Campbell, T. C. Relative contribution of dietary protein leveland aflatoxin BI dose in generation of presumptive preneoplastic foci in ratliver. J. Nati. Cancer Inst.. 78: 365-369, 1987.

69. The National Research Council. Subcommittee on Laboratory Animal Nutrition. Nutrient Requirements of Laboratory Animals. Ed. 3, pp. 7-37.Washington, DC: National Academy Press. 1978.

70. Food and Nutrition Board. Recommended Dietary Allowances. Ed. 6, Chap.6, pp. 52-77.Washington, DC: National Research Council. National Academy of Sciences, 1989.

71. Dietary Intake Source Data: United States 1976-1980. DHHS publicationPHS 83-1681, series 11, no. 231. Hyattsville, MD: Department of Healthand Human Services. 1983.

72. Hsieh, D.. Wong. Z. A.. Wong. J. J.. Michas. C., and Reubner, B. H.Comparative metabolism of aflatoxin. In: J. V. Rodricksef al. (eds.), Myco-toxins in Human and Animal Health, pp. 37-50. Park Forest, IL: PathotoxPublishers, 1977.

73. Booth. S. C., Bosenberg. H.. Garner, R. C., Herzog, P. J.. and Norpoth. K.The activation of aflatoxin Bl in liver slices and in bacterial mutagenicityassays using livers from different species including man. Carcinogensis (London), 2: 1063-1068. 1981.

74. Norpoth, K., Grossmeier. R., Bosenberg. H., Themann. H.. and Fleischer,M. Mutagenicity of aflatoxin B, activated by S-9 fractions of human, rat,mouse, rabbit, and monkey liver towards S. typhimurium TA98. Int. Arch.Occup. Health. 42: 333-339. 1979.

75. Madhavan. T. V., and Gopalan. C. Effect of dietary protein on aflatoxin liverinjury in weanling rats. Arch. Pathol.. 80: 123-126, 1965.

76. Doherty, W. P.. and Campbell, T. C. Aflatoxin inhibition of rat livermitochondria. Chem. Biol. Interact.. 7:63-77. 1973.

77. Hendrickse, R. G.. Coulter. J. B. S., Lamplugh, S. M.. MacFarlane. S. B. J.,Williams, T. E., Omer. M. I. A., and Suliman. G. I. Aflatoxins and kwashior-kor: a study in Sudanese children. Br. Med. J., 285: 843-846, 1982.

78. Stoloff. L. Carcinogenicity of aflatoxins. Science (Washington DC), 237:1283, 1987.

79. Bosch. F. K., and MuÃ±oz.N. Prospects for epidemiological studies onhepatocellular cancer as a model for assessing viral and chemical interactions.In: H. Bartsch et ai. (eds.). Methods For Detecting DNA Damaging Agentsin Humans: Applications in Cancer Epidemiology and Prevention, pp. 427-438, IARC Scientific publication 89. Lyon, France: International Agency forResearch on Cancer, 1988.

80. Ames. B. Dietary Carcinogens and Anticarcinogens. Science (WashingtonDC). 221: 1256-1264. 1983.

81. Brechot. C., Nalpas, B., Courouce, A.-M., Duhamel, G., Callard, P., Camot.F.. Tiollais, P., and Bertholet. P. Evidence that hepatitis B virus has a rolein liver-cell carcinoma in alcoholic liver disease. N. Engl. J. Med., 306:1384-1387, 1982.

82. Hwang, P. C., Khaw, M. D., Lee, N. Y., Hong, C. L., You, S. L.. and Lee,S. M. The First Report: 1970-1971 Diet and Nutrition Survey in Taiwan,research report for diet and nutrition survey and anthropomÃ©trie measurement. Department of Health, Executive Yuan and National Taiwan University Medical School (in Chinese). Taipeh, 1972.

83. Farber. E. Cellular biochemistry of the stepwise development of cancer withchemicals: GHA Clowes Memorial Lecture. Cancer Res.. 44: 5463-5474.1984.

84. Ames, B. Ranking possible carcinogens. Science (Washington DC), 236:271-280, 1987.

85. Lee, F. 1. Cirrhosis and hepatoma in alcoholics. Gut, 7: 77-85. 1966.86. Committee on Diet Nutrition and Cancer. Diet. Nutrition and Cancer, pp.

478. Washington. DC: National Academy Press, 1982.

6893



nonassociation of aflatoxin with primary liver …...a comprehensive cross-sectional survey was...

Documents