Hepatocellular carcinoma and liver tumors in South African children

A case for increased prevalence




The high regional incidence of hepatocellular carcinoma (HCC) in South Africa also may be present in children of the region, although the link to hepatitis B (HBV) appears less clear. The objective of this study was to assess the incidence and probable causes of HCC in South African children.


Data were obtained from seven participating pediatric oncology units and from the tumor registry to review hepatic tumors in children in South Africa.


One hundred ninety-four children (ages 0–14 years) presented with malignant primary hepatic tumors (1988–2003). One hundred twelve tumors (57%) were hepatoblastoma (HB), 68 tumors (35%) were hepatocellular carcinoma (HCC) (including 9 patients with the fibrolamellar variant, 6 of which occurred in black children), 10 tumors (5%) were sarcoma of the liver, and 4 tumors were lymphoma. The ratio of HB to HCC (1.67) was markedly lower compared with other reports, suggesting a greater prevalence of HCC. Correlation with population statistics indicated an incidence of 1.066 malignant liver tumors per year per 106 children age < 14 years (HB, 0.61 per 106 children; HCC, 0.39 per 106). Two-thirds of patients with HCC were positive for HBV surface antigen (HBsAg), and HCC occurred mostly in black African patients (93%). The mean age of onset was 1.47 years for HB and 10.48 years for HCC. A preponderance of males (3.5:1.0) was noted in the HBsAg-positive group that was not reflected elsewhere. Serum α-fetoprotein (AFP) levels were raised both in patients with HB (100%; most AFP levels were very high) and in patients with HCC (69%), although 15% of patients with HCC had low or normal AFP levels.


It appeared from the current results that HCC is more prevalent among children in South Africa compared with the children in more developed countries, although their rates were lower that the rates noted in adults. A collaborative approach will be required to improve their diagnosis and management. Cancer 2004. © 2004 American Cancer Society.

Primary malignant tumors of the liver, mainly hepatoblastomas (HB) and hepatocellular carcinomas (HCC), are uncommon in childhood, accounting for approximately 0.5–2.0% of childhood neoplasms.1 HCC occurs less frequently than HB and has a poorer prognosis,2 occurring mostly as a consequence of severe chronic liver disease and congenital metabolic errors in European and North American series.2

Although it is uncommon in Western countries, HCC remains 1 of the 10 most common adult malignancies encountered worldwide,3–5 largely due to the 2–3 times greater incidence in sub-Saharan Africa and Eastern Asia.3, 6–9 There also is a reported shift to younger age distribution in Africa,10, 11 which would include the pediatric age group, although to our knowledge little information exists. Considering the established relation between hepatitis B (HBV) carrier status and HCC in adults,3, 10, 12 with its assumed long latent period,13, 14 it is possible that, although carrier status in children remains high,15, 16 the pediatric population may not be affected as severely as adults in the same geographic setting.

Thus, the exact incidence of HCC in children age < 14 years in areas of high prevalence remains uncertain.1, 8, 13, 17–19 Most studies have included only small numbers of patients,9, 20–24 although there were two larger surveys.2, 25 The incidence of liver tumors in children has been estimated as 0.5–1.5 per 106 population.26, 27 Based on the Surveillance, Epidemiology, and End Results Program data, the age-standardized rates for males and females in the U.S. are reported as 1.9 for HB (Danish registry, 1.8) and 0.3 for HCC (Danish registry, 0.5).26 The ratio of HB to HCC varies between 4.0 and 6.5 patients with HB to each patient with HCC in the U.S.and in Western Europe (Fig. 1).2, 21, 23, 28–31 Preliminary data from South Africa suggest a much lower ratio in this region, indicating a probable increased prevalence of HCC,4, 32 provided that the incidence of HB is approximately the same. The objective of the current study was to calculate the prevalence of HCC in children in South Africa and to assess its clinical significance.

Figure 1.

Ratio of reported hepatoblastoma to hepatocellular carcinoma in children. RSA: Republic of South Africa. The numbers in parentheses indicate references.


Data were obtained from pediatric oncology units in all tertiary referral centers in South Africa (7 units representing 10 hospitals) with the purpose of reviewing the number and type of the diagnosed/treatable malignant hepatic tumors in children age < 14 years. Hemangioepitheliomas were excluded. The data base of the South African Children's Cancer Study Group (SACCSG) Tumor Registry as well as other source data were consulted to obtain as much information as possible. Information concerning age, gender, and ethnic group was obtained and correlated with population statistics when possible. Anatomic pathology of the tumor (as delineated by ultrasonography or computed tomography or as defined at surgery), histologic diagnosis, treatment, and outcomes also were recorded. When they were available, relevant biochemical investigations (especially α-fetoprotein [AFP] levels, HBV surface-antigen [HBsAg] status, and the presence or absence of cirrhosis) were documented.

Because of intercenter variations, the mean number of tumors diagnosed per annum was calculated by dividing the total number of tumors by the number of years of observation for each unit. The annual incidence was then calculated by determining the number of cases per million children in the estimated population. Because of the short mean survival time, prevalence was used to approximate incidence. Survival was defined as the time elapsed from the beginning of treatment until death or until the date of last communication. Histopathologic features were determined by institutional pathologic services, and the fibrolamellar (FML) variant was identified by established histologic features.33 Treatment was provided according to established chemotherapeutic protocols. Surgical resectability was assessed after four cycles of therapy.

Population data were obtained from the Census Statistics of the Central Statistical Services of South Africa (1991 census) and estimates of total population plus the Transkei, Boputhatswana, Venda, and Ciskei homelands from a report by Sitas.34 In addition, nongovernmental population estimates35 were included in the population calculations because of difficulties encountered in South Africa's recent history and the noninclusion of “homelands” in statistical reports.

Statistical Analysis

The statistical analyses were conducted using by means, medians, analyses of variance, and nonparametric tests for small numbers.


Overall, 194 children with malignant primary hepatic tumors were reported in children age < 14 years between 1988 and 2003 from 7 referral tertiary centers encompassing the combined experience of 10 pediatric cancer units in South Africa. Of these, 112 children (58%) had HB, 68 children (35%) had HCC, (including 9 children who had the FML variant; 13% of all patients with HCC), and 10 children (5%) had sarcomas of the liver (including 6 poorly differentiated sarcomas, 2 rhabdomyosarcomas, 1 angiosarcoma, and 1 mesenchymal carcinoma). The remaining 4 children (2%) had lymphomas that involved the liver. The HB:HCC ratio was 1.65:1.0 in this South African sample (Fig. 1). Study of the SACCSG registry data showed an HB incidence of 1.16 occurring in the second versus the first half of the study (the HCC incidence was 1.22).

Ninety-three percent of patients with HCC were black, compared with the estimated 74% black children in the general childhood population (1996 census). The male:female ratio was 1.95:1.0 for HB and 2.09:1.0 for HCC, with some variation in the particular histologic tumor types (Table 1). Geographically, nearly equal numbers of children with HCC came from coastal regions and from inland centers (35 children from coastal regions vs. 33 children from inland centers).

Table 1. Characteristics of Hepatic Tumours in Children
Histologic typeNo. of patientsM:F ratioMean age (yrs)
  • M: male; F: female; HCC: hepatocellular carcinoma; HBsAg: hepatitis B surface antigen; FML: fibrolamellar variant.

  • a

    Hepatitis B virus status was unknown in one patient with hepatocellular carcinoma.

 HBsAg positive3811.011.0
 HBsAg negative91.09.7

The mean age at onset was 2.16 years for children with HB and 10.48 years for children with HCC. Nine FML variant HCC tumors occurred at a median age of 9.7 years, including 6 tumors in black patients (Table 1). Age largely predicted histologic findings. Children age < 4 years mostly had features of HB, except for 6 exceptional younger patients (9%) ages 10 months, 2.0 years, 2.6 years, 5.0 years, 5.0 years, and 6.0 years). A recent patient age 10 months had trabecular histologic features. All children age > 8 years had HCC. One patient with HCC had retroviral disease and fared poorly.

Approximately two-thirds of the patients with HCC were HBsAg positive (Table 2). Hepatitis C status was negative when it was tested but mostly was not available because of the historic nature of the population that was studied. Virtually all liver function tests were within normal limits. AFP levels were elevated in 69% of patients, mostly to very high levels (> 500 000 IU), particularly in patients who had positive HBsAg status. Fewer than 10% of the patients had only moderately raised AFP levels at the time of diagnosis. Thirty-three percent of the 15% of patients who had normal AFP levels were negative for both HBsAg and AFP.

Table 2. Laboratory Tests in Patients with Hepatocellular Carcinoma
TestNo. of patientsNo. testedPositiveNegative
  1. HBsAg: hepatitis B surface antigen; sAFP: serum α-fetoprotein; HBsC = hepatitis C antigen.


A greater preponderance of males was observed in the HBsAg-associated HCC group (male:female ratio, 3.5:1.0 vs. 1.0:1.0 in HBsAg-negative patients). The M:F ratio for HBsAg-negative patients was similar to that encountered in the nine patients with the FML subtype. A reversal of this M:F ratio was noted in the sarcoma group, in which females predominated (Table 1).

Histologically, HCC tumors did not appear to arise from cirrhotic nodules, although the majority were multicentric (especially the HBsAg-positive tumors) and involved both lobes. The incidence of cirrhosis, however, was difficult to determine, because only a minority of tumors were amenable to resection because of the advanced stage of the disease at the time of diagnosis. The diagnosis of an FML variant of HCC was made on histologic grounds in nine patients, six of whom were black.

Treatment was comprised of chemotherapy with regimens that included vincristine, doxorubicin, cyclophosphamide, 5-fluorouracil, and a combination of the two most effective agents: cisplatin and doxorubicin.31 Palliative radiation therapy was given in selected patients, and a few selected patients underwent transarterial chemoembolization as part of a local protocol. Only 15% of patients were considered suitable for surgical resection.

The mean survival after diagnosis was 4 months. Only 11% of patients with HCC survived long term, with survival depending on the type of tumor, its multicentricity, its response to chemotherapy, and eventual surgical resectability. At least 4 known patients survived for > 5 years after undergoing surgical resection, and 1 patient who was treated with a cisplatin and doxorubicin regimen followed by complete surgical resection appeared to be incident-free at the time of last follow-up (> 120 months), despite the treatment of lung metastases.

Population Data

The population of South Africa varied from 28,087,439 in 1980 (1980 census) to 37,944,000 in 1991 (1991 census) and currently is estimated at 43,421,021(South African Central Statistical Services). The total population was estimated at 33,621,000 in 198635 (14,8% white, 74.1% black, 8.6% mixed race, and 2.6% Asian). For practical purposes, a median population of 35,782,500 for the period under review was used for calculating incidence. Because it has been estimated that the group ages birth–14 years represents 33.9% of the population (1996 census), the population of children age < 14 years would total 12,130,267 based on these figures.

The overall incidence for malignant liver tumors is on the order of 1.066 per 106 South African children (age < 14 years) per year. Of these, the incidence of HB is 0.61 per 106 South African children per year, and the incidence of HCC is 0.37 per 106 South African children per year.


The geographic clustering of HCC to sub-Saharan Africa and Asia has been reported widely in adult patients,3, 6, 7 with HCC the most common malignant tumor diagnosed in adult men in western, central, and southern Africa. Particularly high rates are reported from the entire southern and east African region36–40 (particularly Mozambique39, 40 and Zimbabwe38), and HCC was rated the fourth most common tumor in black males in South Africa in 1988.34 It appears to be most prevalent in adult males in the rural black population from the east coast,41 and the highest reported incidence in the world appears to be in the Shangaan people of Mozambique3, 39, 40 (age adjusted rate for males, 112.9 per 100,000 population per annum). In contrast, whites living in Africa appear to have the same low rate (2% of malignant diseases) as whites living in the U.S. or in the U.K.3, 42, 43

In the current study, which included all of the tertiary referral pediatric oncology units in South Africa, HCC appeared to be approximately 100 times less common in children than the highest regional incidence (Shangaan males) and 26 times less frequent than the overall regional incidence in adults.3 Despite obvious logistic difficulties encountered in a developing country, it can be concluded that the overall incidence of malignant liver tumors is at least 1.066 per 106 South African children (age < 14 years) per year (HB, 0.61; HCC, 0.37). In addition, the overall ratio of HB:HCC appears to be much lower in the South African population compared with other nonendemic populations, and no significant changes were noted between the first and second halves of the study. The lack of significant variation between the first and second halves of the study probably reflects population increase (among other factors, such as improved referral patterns). The incidence also appears to approximate that of other areas,8, 9, 44, 45 suggesting that the HB:HCC ratio of 1.65:1.0 for South Africa most likely indicates a presumed increased prevalence of HCC in children within the region.

Although a higher risk was identified previously for rural African adults from the east coast of South Africa,41 in the current study, the risk in children appeared throughout the region and involved patients from nearly all major centers. This may be related to progressive urbanization and population demographics (particularly as they apply to children).

The overall male:female ratio for patients with HCC was 2.2:1.0, although HBsAg-positive males had a 3.5:1.0 incidence (HBsAg-negative ratio, 1.0:1.0), reflecting the male preponderance reported in adult males from the region10 and similar to other reports.25, 29, 46 There appeared to be a higher incidence of females in the younger population as well as a higher percentage of males in the older age group.

In keeping with previous reports, the histologic forms of primary malignant childhood liver tumors appear to be related to age,2, 23, 25, 28, 46 with HB predominating in children age < 3 years (mean age, 2.16 years)21, 47 and 65% of HCC tumors developing after age 10 years.46 The median age of 10.48 years (range, 2–15 years) for children with HCC is somewhat younger than the median age of 12 years (range, 4–15 years) reported in recent international surveys.25, 48 The 35% of children with HCC that occurred at ages < 10 years were mostly in the 8–9 year age group, although 6.34% presented at age < 8 years (ages 10 months, 2.0 years, 2.6 years, 5.0 years, 5.0 years, and 6.0 years), in keeping with the suggested second peak of incidence and other reports of early HCC (21 months).2 The recent modern reclassification of histologic types of HB47, 49 may redefine these tumors, particularly those in very young patients with a trabecular pattern. Most atypical forms of HCC were HBsAg negative,20 although it is known that HBV-DNA positivity may exist in the tissue in the absence of detectable serologic markers for HBsAg.50

Only two-thirds of the children tested had elevated AFP levels (Table 2) that corresponded to positive HBsAg status. This is in keeping with the reported incidence in children from other countries27 as well as adults, with the markedly raised serum AFP levels51 in 75–92% of adults from these regions4, 52 linked to HBV infection.46, 53

Although chronic liver diseases (e.g., hepatitis C,54 cirrhosis, biliary atresia55, 56, hereditary tyrosinemia, Beckwith–Weidemann syndrome,2, 46 hemihypertrophy, membranous vena caval obstruction,57 and prolonged total parenteral feeding58) all have been implicated in the etiology of HCC, only one patient with associated chronic active hepatitis was reported in the current series. The risk-effect of longstanding hepatitis infection42 is the most likely pathogenic mechanism of HCC. Further indirect evidence comes from the significant decrease reported in HCC incidence after HBV vaccination in Taiwanese children.44, 45, 59 Sixty-seven percent of children with HCC were associated with positive HBsAg status in this series, compared with the 80% association in adults,.3, 12, 14, 16 in keeping with the high regional HBV total exposure rates.13, 16, 44, 60–62 A greater percentage of young black African patients were positive for HBsAg compared with whites or mixed ethnic groups.11, 16 However, not all black patients with HCC had positive HBsAg status, and patients with positive HBsAg status were not confined solely to the black population.

Although the exact mechanism of the development of HCC at the molecular level remains unknown, it is believed generally that the cycle of continuous liver necrosis and regeneration may result in cells that are susceptible to further carcinogenic influences. HBV infection usually results from the transfer of maternal HBsAg during infancy,60 and it has been shown that infants who acquire HBV in the perinatal period have an up to 90% risk of developing chronic HBV sequelae.62 However, it has been shown that perinatal transmission is lower in Sub-Saharan Africa than in the Far East16; and, despite a similar HBsAg exposure9 Chinese patients appear to develop HCC slightly later than Africans.8–10 This suggests a possible additional factor in the development of HCC in African children. The specific role of HBV infection9, 45, 63–65 in childhood HCC is difficult to understand in light of the generally held view that HBsAg-related tumors mostly develop after a long latent period (up to 20 years). This study supports the concept that there may be a population with a shorter HBV latent period.13, 14, 64–66 This is supported further by reports of secondary HCC in a leukemic patient age 10 years with a 3-year HBV exposure66 and an instance in which HCC occurred 2 years after a child contracted an HBV infection.13

Recent research has identified genotype B HBV as the major cause of severe liver disease,62 and Ni et al.63 have identified different mutations in the core region of the HBV virus in children with HCC compared with the mutations in children with chronic liver disease. This recent information may provide clues regarding the link between these conditions and may explain the discrepancies in the latent period between populations.

Apart from the known pathogenic mechanisms, it is possible that an increased local incidence may be linked genetically, and familial associations have been described recently in Scandinavia.67 Deletions of alleles on chromosome 17p and p53 mutations (codon 249) have been identified in 50% of adult patients,68, 69 indicating a possible link to specific etiologic carcinogens (e.g., aflatoxin B).68 In children, C-met mutations, decreased cyclin D levels, and a high frequency of loss of heterozygosity on 13q have been reported.70 The wnt/wingless pathway also has been implicated in HCC, with mutations of β-catenin resulting in its accumulation in the cytoplasm or nuclei as well as its overexpression (with the tcf-4 gene), which is correlated independently with c-myc overexpression.71 Although this opens new avenues for research in the etiology of HCC, histocompatability antigen studies in Southern African adult patients do not appear to support a genetic hypothesis for the increased susceptibility.68

The macroscopic and microscopic features of HCC are similar in children and adults, but it is becoming clear that differences in histology, presentation, and prognosis require different strategies of management and evaluation.25, 31 The FML carcinoma variant33 occurs in patients mostly age > 10 years (median age in young adults, 20 years33, 47, 48) but has been reported in childhood as early as age 5 years.1, 72, 73 In a recent North American survey, only 1 of the 10 FML tumors reported occurred in children age < 10 years.48 The FML subtype is considered to be more frequent among whites and is extremely rare in the Far East.43 It appears that the incidence reported previously in black African patients was regarded as low.10 Thus, it is interesting to note that 6 of 9 patients who had this variant in our series were black children, suggesting that the incidence of FML in this ethnic group is higher than expected. Although it is believed that these tumors are more localized, more frequently resectable, and imply a better prognosis in adults without cirrhosis, significant improvements in outcome have not been confirmed in a young cohort.48

Malignant primary liver tumors pose considerable management challenges, with successful management dependent on complete eradication of the tumor. Recognition of HBV infection as the most significant factor in the increased regional prevalence has led to widespread prophylaxis by immunization programs. A national HBV immunization program for children in Taiwan reportedly has contributed to a 50% decrease in the incidence of HCC.45 South Africa introduced a similar program in 1995, but it is too soon to evaluate its results, and the true benefits may be seen 10–20 years hence.16, 74

HCC appears to be less responsive than HB to adjuvant chemotherapy. A recent North American study in children failed to demonstrate any difference in the 5-year event-free survival rate for the FML subtype in patients age < 21 years.48 African patients with HCC usually have multifocal hepatic sites that involve both lobes and reportedly are associated frequently with liver damage and may or may not have associated cirrhosis.75 The 11% 5-year survival rate achieved in this series was not dissimilar to the 17% rate achieved in a recent North American study.48

It should be noted, however, that five children in the current series had tumors that were localized to one lobe, were HBsAg negative (serum, cell serum, and immunocytochemical stains), and responded well to chemotherapy, which made them amenable to surgical resection. At the time of last follow-up, two of these patients were alive and well after long-term follow-up, and one of these patients has fared well (> 120 months) despite having sustained lung metastases, which responded to chemotherapy and surgical resection.


HCC appear to be more prevalent in children in South Africa, in keeping with epidemiologic studies in adults. Although there remain considerable logistic difficulties in identifying and collecting data, the need for a collaborative effort is an important factor in approaching the management of patients in this difficult group.