Despite the fact that liver cancer is the main form of cancer in West Africa, there is little information available on its variations with age, time, gender and ethnicity. The Gambian National Cancer Registry has collected structured, nationwide information on liver cancer since 1986. One of the limitations of the data collected by the registry is the paucity of diagnostic information, with only about 40% of the cases assessed by liver ultrasonography. However, the diagnosis of liver cancer is unlikely to be strongly biased by other, space-filling liver lesions. A recent study of 323 clinically assessed cases from The Gambia showed that over 90% of the cases presented with very advanced liver cancer, 45% of them having tumors with a diameter over 10 cm. More than two-third had multiple lesions and 41% had lesions in both liver lobes. The main clinical features were hepatomegaly (92%), abdominal pain (94%) and weight loss (94%). This triad was present in 74% of the patients and the median duration of reported symptoms prior to HCC diagnosis was 8 weeks 
. Furthermore, only 67% of the cases showed evidence of cirrhosis at the time of HCC diagnosis. A morphological study of 35 liver biopsies obtained in the course of a case-control study confirmed HCC in 29 (83.3%) of patients having been diagnosed only on the basis of clinical criteria. The 6 excluded cases were 1 liver hemangioma and 5 liver metastases of unspecified primary tumors. The specificity increased to 95% using the combination of clinical assessment, a-foetoprotein levels and ultrasonography 
. Of note, cases of liver metastases of identified primary tumors were excluded from the present analysis.
Analysis of liver cancer incidence data over two time periods of registration (1988–1997 and 1998–2006) has allowed us to draw two main observations. First, between the two periods, there has been a decrease in male to female ratio (from 3.28 to 2.20, −33%), and this decrease is essentially attributable to an increase in females in all age groups. A modest decrease between the two registration periods is also noted in men aged over 30 years. However, in contrast with women, the decrease in men is no longer apparent in a regression analysis that takes into account each year's data over the two periods. Thus, it rates in men should be considered as relatively stable over the period covered by this study. In our analyses, there was no apparent decrease in liver cancer rates in young subjects despite the initiation of a vaccination program in 1986 (GHIS). It should be noted that the introduction of vaccination in The Gambia was performed in a stepwise manner during years 1986 to 1990, so that only half of the subjects born during that period actually received HB vaccine. Thus, the results presented here cannot be interpreted as revealing an absence of effect of vaccination in the younger age group (15–19 years). A recent review of the GHIS has shown that a significant effect of vaccination may not be measurable before 2017 
A significant decrease over time in male to female ratio for liver cancer has been observed in the Uganda (Kyadondo) cancer registry. In this area, the ASR of liver cancer for men is 8.7 per 105
person-years and 5.8 for women (ranking 4th
among men and women respectively) 
. Comparison between the periods 1960–1980 and 1991–2005 showed that liver cancer incidence (ASR) was quite stable or even slightly decreased among men, but increased by over 50% among women, with a change in male to female ratio from 2.32 to 1.30 
. Thus, our observations in The Gambia may correspond to a phenomenon which may also occur in other Sub-Saharan African countries.
Several explanations may be proposed. First, it is possible that female liver cancer has been systematically under-detected and underestimated over earlier cancer registration periods, and that the increase in the recent period (1998–2006) may primarily reflect improved registration in women. It should be noted, however, that no comparable change has been observed for other female cancers (breast and cervix cancer in particular 
), hence it is unclear why there should be a registration bias only for liver cancer. A second explanation may be changes in the prevalence of liver cancer risk factors and in their contribution to liver cancer burden in each gender. Three main factors may be considered: chronic infection by hepatitis B virus (Hepatitis B Virus), infection by hepatitis C virus (Hepatitis C Virus) and metabolic conditions associated with obesity and/or diabetes type 2.
Case control studies have shown that much of the imbalance in liver cancer incidence between men and women may be attributable to Hepatitis B Virus. In a study performed between 1997 and 2001, 
the prevalence of chronic Hepatitis B Virus carriage was of 18% in men and 10% in women (p
0.05) with a risk of liver cancer 2.6 (95% CI, [1.5–4.6]) fold higher in men than in women. In contrast, the prevalence of chronic Hepatitis C Virus infection was similar in both genders (2.6% in males and 3.6% in females, p
0.57). Among liver cancer patients, 32% of women were Hepatitis C Virus positive, as compared to 16% of men (p
0.03). Thus, the decrease in male to female ratio may be suggestive of an increasing impact of Hepatitis C Virus as cause of liver cancer. The possibility that Hepatitis C Virus may contribute to the increasing burden of female liver cancer deserves further investigation.
The other factor that may account for the observed increase in incidence of liver cancer in female is an increase in the prevalence of metabolic disorders associated with obesity, physical inactivity and/or diabetes type 2. Recent meta-analyses have found that in West Africa, women were more likely to be obese than men (odds ratio 3.16 95% CI, [2.51–3.98] and 4.79 (95% CI, [3.30–6.95]) in urban and rural areas, respectively 
. Time trend analyses indicated that the prevalence of obesity in urban West Africa had more than doubled (114%) over 15 years, accounted for almost entirely in women 
. Although data are limited, differences in overweight and obesity by gender and age have been observed in The Gambia. Among subjects aged 35–50 years (the age group with high rates of liver cancer), women were more likely to be overweight or obese as compared to men (34% versus 6%, and 50% versus 6%, respectively) 
. Cumulative time trends analyses suggested an increase in the prevalence of diabetes among adults in urban West Africa, from approximately 3% to 4% in the past 10 years, with similar prevalence and trends in males and females 
. A recent meta-analysis of 11 cohort studies including cohorts from Western Europe, US and Korea has shown a substantial association between excess bodyweight and increased risk of liver cancer. In this meta-analysis, compared with persons of normal weight, the relative risk of liver cancer was 1.89 (95% CI: 1.51–2.36) for those who were obese 
. Although there is currently no data on West African populations, it is plausible that overweight may also contribute to the increased burden of liver cancer in Gambian females. These associations need to be further tested in case-control and cohort studies.
Our results report a significant association between the risk of liver cancer and ethnicity. A recent genome-wide association analysis of malaria in The Gambia demonstrated that self-reported ethnicity correlates with genetically defined subpopulations 
. Our analyses show that Fula (18% of the population) and Wollof had a significantly higher risk of liver cancer than Mandinka (42% of the population). This association was previously reported in a case-control study in The Gambia 
. Two studies conducted on Gambian children reported that blood levels of aflatoxin-albumin adducts were slightly higher in Fula and Wollof than Mandinka 
; suggesting that these groups may have a higher sensitivity to aflatoxin due to genetic polymorphisms in enzymes involved in aflatoxin metabolism, and detoxification. In a case-control study, Fula were found to have a significantly higher prevalence than Mandinka in the Gly399 allele of XRCC1
, an enzyme involved in excision repair of aflatoxin-DNA adducts 
. These observations suggest that the higher risk observed in selected groups may be due to genetic susceptibility, although differences in lifestyle may also play an important role, in particular among rural populations.
In conclusion, our study based on almost two decades of population-based cancer registration in The Gambia identifies a significant trend in increase of hepatocellular carcinoma among females, with a significant reduction of the male to female ratio. This observation may be the consequence of changes in lifestyle or viral risk factors and suggest that recent increase in the prevalence of obesity among women deserves further attention in hepatocellular carcinoma prevention strategies.