Dramatic reduction of liver cancer incidence in young adults: 28 year follow-up of etiological interventions in an endemic area of China

Study populations and cancer registry

Qidong City has a relatively stable population of about 1.1 million; 11 000–13 000 children were born each year in the 1980s. About 90% of the population of Qidong City lives in rural areas. Corn and cotton were the main agricultural products in Qidong City before the early 1980s. We sometimes refer to Qidong City as Qidong.

Population-based age-specific liver cancer incidence data from 1980 to 2008 were collected by the Cancer Registry in the Qidong Liver Cancer Institute, Qidong City, which began collecting data in 1973. Data from this registry have been used by the International Agency for Research on Cancer (21).

Criteria for liver cancer diagnosis

We diagnosed liver cancer on the basis of progressive enlargement of a firm liver and the presence of one or more space-occupying lesions on ultrasonography, together with either serum alpha-fetoprotein (AFP) >300ng/ml or death within months or both. Since 1986, referring hospitals have also provided computerized axial tomography images. In >80% of these cases, confirmatory evidence was provided by elevated AFP >300ng/ml. Death usually occurred after about 8 months, providing additional confirmatory information. Practically, every case that went to surgery was found to have liver cancer pathologically diagnosed as HCC. The criteria of progressive enlargement of a firm liver and one or more space-occupying lesions on ultrasonography, together with elevated AFP or early death have not changed over the course of the study.

Identification of aflatoxin-producing Aspergillus flavus, average dietary AFB₁ intake and individual aflatoxin exposure measurements

We obtained samples of corn (n = 474), rice (n = 115) and wheat (n = 97) from rural families in Qidong in the 1970s. These samples were tested to detect the presence of A.flavus. We tested for AFB₁ from detected strains of A.flavus by using thin layer chromatography on extracts of rice media on which detected strains were cultured.

Corn was the main staple food in rural areas of Qidong rural before 1980. During 1973–80, a total of 2272 stored dietary corn samples (ranging from n = 83 to n = 522 samples each year) were collected from households in different rural towns in Qidong for detection of AFB₁. AFB₁ in samples was semiquantified by thin layer chromatography, using pure AFB₁ (Makor Chemicals, Jerusalem, Israel) as standard. The lower limit of detection was 5 p.p.b., equivalent to 5 µg/kg. Estimated average AFB₁ exposure per person per year was calculated as: [the annual amount (kg) of corn consumption per capita] × (the fraction of corn samples that were positive for AFB₁) × [the mean level of AFB₁ (µg/kg) in positive samples in the relevant year]. The AFB₁ in all corn samples from 1973 to 1980 was measured by the same group of scientists using the same AFB₁ standard purchased from Makor Chemicals, Jerusalem, Israel.

To measure an individual’s daily aflatoxin exposure, we had developed sensitive and reproducible techniques for detecting AFB₁ and AFM₁ (in urine). AFM₁ in a 24h urine sample was immunoconcentrated by monoclonal antibody coated small affinity columns and quantified by high-performance liquid chromatography. The AFB₁ exposure on the day tested was estimated by multiplying the measured urinary AFM₁ output (ng/l) by the conversion ratio 50 that was determined by our previous experiments (14–17).

Follow-up of a cohort of HBV surface antigen-positive men for AFB₁ exposure and incident HCC

A cohort of 145 men with CHB was recruited from two randomly selected towns using population-based sampling in 1987–88. Following informed consent and approval from the Review Board in the Qidong Liver Cancer Institute, the cohort was studied and followed for 13 years to 2000. Hepatitis B virus surface antigen (HBsAg) was measured at recruitment with a radioimmunoassay (RIA) kit from the Beijing Biological Products Institute, Beijing.

In 1988, mean AFM₁ was measured from 8 to 10 urine samples collected monthly from each individual and stored frozen at –20ºC. The pooled samples from each man were immunoconcentrated and AFM₁ was measured by high-performance liquid chromatography. In this way, we obtained average daily AFM₁ output and average daily aflatoxin exposure (by multiplying by 50) for each cohort member in 1988 (10,16,17). In 2000, after 13 years of follow-up, 36 of the 78 men with detectable AFM_1, in 1988, had died. Of the 78 – 36 = 42 survivors, 8 left Qidong for work elsewhere and were not accessible, whereas 21 of 34 accessible survivors provided urine samples in 2000 (2 urine samples spaced 4 days apart). Pooled samples from each man were retested for AFM_1. The 34 accessible survivors answered questions about their daily staple food consumption.

Data on consumption of corn, rice and wheat in Qidong

The Qidong Food Bureau kindly provided data on annual availability of corn, defined as Qidong production less exports, for the Qidong population from 1973 to 1981, and also provided data on computer media on annual consumption of commercial corn, rice and wheat in the Qidong population for the years 1988 through 2010. We converted these data to consumption per capita by dividing the availability by the yearly population size as shown in Table I for 1973–80 and in Figure 2 for 1973–81 and 1988 through 2010, respectively.

Additional information on diet came from a survey of staple food consumption in 2010 in six towns of the Wang Bao and Yun Yang districts of Qidong. This survey covered 36 420 households with 86 555 people in Wang Bao and 31 500 households with 77 226 people in Yun Yang. The total number of people surveyed was 163 781, which was about 15% of the total population of Qidong City.

HBV neonatal vaccination in Qidong and estimation of HBsAg prevalence

HBV neonatal vaccination began on 1 September 1983 as part of a large controlled clinical trial (18,19), promoted by World Health Organization to demonstrate the efficacy of neonatal HBV vaccination to control liver cancer. There were 1831, 9889 and 15 189 newborns vaccinated in 1983–84, 1985–86 and 1987–88, respectively. These numbers of vaccinees and the year vaccinated were useful data for this article. Note that all these vaccinees were <11 years old in 1990–93, but were 17–25 years old in 2005–08. In early November 1990 when the Chinese vaccine began to be commercially available in Qidong, our controlled neonatal HBV vaccination trial was ended. Under the Health Bureau’s administration, HBV vaccination continued without interruption to cover all newborns in Qidong clinics with the same standardized HBV vaccination regimen used in the trial until 2002. The National Expanded Program of Immunization of China assumed responsibility for free universal vaccination of neonates since February 2002 (19).

Following informed consent and Institutional Review Board approval, sera from 2390 men and 4284 women aged 30–69 years from two rural towns, Donghai and Dafeng, were measured to estimate age-specific HBsAg prevalence in rural Qidong adults in 1992. None of these adults had been vaccinated against HBV as neonates because neonatal vaccination first began in 1983. Sensitive RIA kits from Beijing Biological Products Institute were used. Earlier prevalence surveys in 1976 used a reverse passive hemagglutination assay, also from Beijing Biological Products Institute (4).

Evaluation of HCC prevalence in primary liver cancer in Qidong

Among 2606 cases of liver cancer clinically diagnosed in various Qidong hospitals from 1972 to 1976, 259 cases (9.9%) had available pathological diagnoses. Classification of subtypes of liver cancer was performed by senior pathologists using hematoxylin- and eosin-stained slides.

Statistical methods

Population-based age-specific incidence rates for liver cancer in successive time periods were calculated as incidence per 100 000 person-years based on data from the Cancer Registrar of Qidong City in the Qidong Liver Cancer Institute. Confidence intervals (CIs) for Figure 3 and Table II were calculated based on Poisson theory. In particular, the variance of the log relative risk was estimated as the sum of the reciprocals of numbers of cancers. Confidence limits on the log relative risk were based on normal theory and inverted to produce 95% confidence intervals on the relative risk. Poisson regression with allowance for overdispersion was used to compute rate ratios and to test hypotheses on age-specific rates and for comparing rate ratios for older ages with rate ratios for younger age groups (Supplementary Information, available at Carcinogenesis Online). Statistical tests were two sided with significance level 0.05.

Reduction in aflatoxin exposure from 1970s to 2010

Corn was the main staple food produced and consumed by residents in Qidong before 1982. Aflatoxin-producing Aspergillus flavus was detected in 29.1% (138 of 474) of corn samples, but only 2.6% (3 of 115) of rice samples and 2.1% (2 of 97) of wheat samples in Qidong in the 1970s. The prevalence of aflatoxin-producing A.flavus strains was 11 times higher in corn than in rice and 14.5 times higher than in wheat. Corn was the predominant source of aflatoxin exposure in Qidong residents, especially in rural poor families.

From 1973 to 1980, 2272 corn samples were tested. The percentage of samples with AFB₁ >5 p.p.b. rate varied from 26 to 99%, and the mean level of AFB₁ in positive samples varied from 11 to 55 p.p.b. in different years (Table I). The estimated AFB₁ intake from dietary corn in each year from 1973 to 1980 ranged from 0.4 to 3.3mg/person/year, with a mean exposure of 1.4mg/person/year (Table I). Ten years’ exposure at this average rate would be 14mg, which is comparable with the estimated cumulative exposure to the time of HCC diagnosis in seven men with chronic HBV hepatitis, which ranged from 7 to 29mg [Table 2 in ref. (6)].

By 1983–84, some data suggested that the intake of AFB₁ had decreased. Eleven of 40 (27%) rural residents of Qidong had detectable AFM₁ (>4ng/l) in urine samples. Among the 11 with detectable AFM₁, 2 had estimated intakes of 10–12.5 p.p.b. of AFB₁ per day (equivalent to 3.7–4.6mg/person/year), and the other 9 had estimated intakes of about 1.25 p.p.b. of AFB₁ per day (0.46mg/person/year). The mean AFB₁ intake in these 40 rural residents was 0.3mg/person/year, which was 4.7-fold less than the average AFB₁ intake of 1.4mg/person/year from 1973 to 1980 (Table I).

In 1987–88, a cohort of 145 HBsAg-positive men was recruited, and in 1988 they provided 8–10 repeated monthly urine samples. Seventy-eight (54%) of these 145 men had AFM₁ levels >4ng/l in pooled urine, with the elevated levels ranging from 5.7ng/l to 243ng/l and with a mean level of 48ng/l (Figure 1, panel A). In 2000, after 13 years of followed-up, 36 (46%) of these 78 HBsAg-positive men with detectable AFM_1, in 1988, had died, including 24 from HCC and 7 from end-stage cirrhosis. Eight men left Qidong to work elsewhere. The remaining 34 men were interviewed about diet in 2000, and all reported eating rice as their main staple food in recent years. Twenty-one (62%) of the 34 survivors in Qidong provided two 24h urine samples 4 days apart in 2000. All these 21 men had detectable (>4ng/l) urinary AFM₁ in 1988, but only one of them (5%) excreted detectable levels in 2000, and his level was only 9ng/l (Figure 1, panel B). In each of these 21 men, AFM₁ levels decreased between 1988 and 2000, and the decreases were statistically significant (P < 10^–6 by two-sided sign test).

Change in staple food consumption in Qidong

Annual per capita corn consumption ranged from 65 to 124kg/person/year in the 1970s and dropped to much lower levels from 1988 to 2010, when it was only 5kg/person/year (Figure 2). Rice became the dominant staple food by 1988, when data first became available, and rice consumption varied in a tight range from 225 to 235kg/person/year from 1988 to 2010 (Figure 2). Wheat consumption has grown since 1988, but remains small. Thus, corn was displaced by rice between 1982 and 1988, and by wheat to a much smaller extent.

This shift from corn to rice was consistent with the data from our survey of 165 781 people in the districts of Wang Bao and Yun Yang in 2010. All of these rural people ate rice in three meals per day as their standard staple food. Ten percent or fewer ate some wheat in the morning meal. The very limited corn consumption included food for domestic animals. The average annual per capita income in rural Qidong increased from 224–324 yuan in 1980–83 to 6069–8376 yuan in 2005–08, which was sufficient to allow people in Qidong to buy rice as their main staple food.

Reduction of age-specific liver cancer incidence in Qidong City

The age-specific incidence of liver cancer in Qidong and relative risks with 95% CIs are shown in Figure 3 (semi-log plot) and Table II. Age-specific liver cancer incidence was higher for each age group (except those aged ≥50 years) in 1980–83 than in 2005–08. Relative risks comparing 1980–83 with 2005–08 (grey hatched line and bold solid line) were 14.1, 9.0 and 4.1 for age groups 20–24, 25–29 and 30–34 years, respectively (Figure 3 and Table II). They diminished toward 1.0 for older age groups. The relative risk for those aged ≥50 years was 0.77 (0.72–0.83). The relative risks for each of the three age groups 20–24, 25–29 and 30–34 years were larger than that of the geometric mean of the relative risks for the age groups 35–39, 40–44 and 45–49 years, with respective P-values 0.003, <0.0001 and <0.0001 (Supplementary Information, available at Carcinogenesis Online). Thus, by 2005–08, after 28 years of preventive measures, there was a dramatic decrease in liver cancer incidence rates in young adults, especially in those in the age range 20–34 years. Although the relative reductions were smaller for those aged 35–49 years, the absolute reductions were comparable in the older and younger age groups because incident rates are higher in the older groups.

Relative risks (with 95% CIs) comparing 1980–83 with 1990–93 (grey hatched line and bold dashed line in Figure 3) were 3.4 (1.9–6.0) and 1.9 (1.4–2.5), respectively, for age groups 20–24 and 25–29 years (see also Table II). These decreases were highly statistically significant (P < 0.0001) after only 10 years of preventive effort. Rates of liver cancer incidence in 1990–93 were significantly higher than in 2005–08 for each age group >24 years, except for those aged 50 and over (Supplementary Information, available at Carcinogenesis Online).

Effect of HBV vaccination on liver cancer incidence

From September 1983 to June 1984, 1831 neonates were first vaccinated against HBV infection in the pilot phase of the controlled trial in Qidong. The oldest of these neonates would be 10 years old in 1993. Therefore, vaccination could have no effect on the 1990–93 incidence rates in those >19 years old. The 1831 neonates vaccinated in 1983–84 would be 24–25 years old in 2008. The number of these vaccinees at age 25 years in 2008 was 378, only 0.59% of the 64 001 people aged 25–29 years in 2005–08. And none of the vaccinees was older than age 25 years in 2005–08. Thus, vaccination could have had no effect on the incidence rates at ages >25 years in 2005–08, and negligible effect at ages >24 years. However, vaccination could have had an appreciable effect on those aged 20–24 in 2005–08 because {(1831×7/8 + 9889×7/8 + 15 186×3/8)/57 414) × 100} = 27.8% of the people aged 20–24 years had been vaccinated. The dramatic 14.1-fold reduction in liver cancer incidence at ages 20–24 years in 2005–08 cannot be entirely explained by vaccination of 27.8% of the population, which could at most reduce liver cancer incidence 1/(1 – 0.278)=1.38-fold. The reduction in liver cancer incidence probably also reflects the benefits of abatement of exposure to aflatoxin over two decades.

In summary, vaccination did not contribute to any of the decreases in liver cancer incidence observed in 1990–93 or to any of the decreases in 2005–08 among those aged >25 years. Although vaccination may have had an effect in 2005–08 among those aged 20–24 years, it can explain only a fraction of the dramatic reduction in liver cancer incidence in that age group.

Relative stability of HBsAg prevalence in non-vaccinated adults

In 1976, a general survey of HBsAg prevalence in adults aged 30–69 years in Qidong yielded prevalence estimates based on the reverse passive hemagglutination assay of 19.1% (1195 of 6270) in men and 16.2% (1365 of 8422) in women (4). In 1984, a survey of HBsAg prevalence in women aged 20–30 years who had just given birth in Qidong yielded a prevalence of 14.2% (167 of 1180) by RIA (18). In 1992, we measured serum HBsAg in adults aged 30–69 years in two rural Qidong towns (Donghai and Dafeng) using RIA. The estimated HBsAg prevalence was 16.6% (397 of 2390) in men and 11.4% (488 of 4284) in women. None of these adults had received neonatal vaccination, which began in 1983. These data indicate that the prevalence of HBV infection may have decreased slightly in non-vaccinated adults from 1976 to 1992, but still remained well >10%.

Prevalence of hepatocellular carcinoma in liver cancer cases

Pathological subclassifications were obtained for liver cancer specimens from 259 adults with clinically diagnosed liver cancer, which accounted for 9.9% of all such cases diagnosed from 1972 to 1976. The pathological diagnoses identified 249 hepatocellular carcinoma (96.1%), 7 mixed HCC and cholangiocarcinoma (2.7%) and 3 cholangiocarcinoma (1.2%). Thus, HCC and the mixed HCC–cholangiocarcinoma cancers constituted 98.8% of adult liver cancer in Qidong.