Secular trends of salted fish consumption and nasopharyngeal carcinoma: a multi-jurisdiction ecological study in 8 regions from 3 continents

The distribution of nasopharyngeal carcinoma (NPC) has a distinctive geographic variation. Being a rare malignancy worldwide, it has particularly high rates in Southern China [1]. Residual risk remained high among migrants of Chinese descent [2] to non-endemic areas, with intergenerational decline [3], suggesting that the aetiology of NPC is a complex interplay between genetic influences and environmental exposures distinct in these populations.

Among known environmental exposures relevant to the aetiology of NPC; dietary consumption of salted fish is the most well recognised risk factor. First suggested by Ho and colleagues [4] to have a link with the high incidence of NPC in Southern China Guangdong Province in the early 1970s, subsequent studies quite consistently reported positive associations between salted fish consumption and risk of NPC in Chinese [5] and other ethnicities [6] such as Thai and Tunisian [7], and that early exposures during the weaning period in childhood may play a critical role [1, 8]. What makes salted fish an intriguing item to look at is that nitrosamine in Cantonese style salted fish is carcinogenic to humans and this food item is rated as a group 1 carcinogen by the International Agency for Research on Cancer (IARC)[9], whereas other salted fish are only classified as group 3 carcinogen [9]. This can be explained by the different levels of nitrosamines in various types of salted fish, depending on the method of preservation [10]. However, the low disease rates of NPC even in endemic regions render case-control studies to be the only feasible epidemiological method, of which limitations and deficiencies are well-recognised. The extent to which notorious effects of recall bias [11], lack of adjustment of caloric intake [12] and other confounders may affect the validity of the study findings remains unknown. Whether any publication bias may exist due to active reporting of salted fish consumptions as a classical risk factor of NPC is also uncertain. Although Cantonese-style salted fish has been classified by IARC as a group 1 carcinogen [9], the putative main biological active component N-nitrosodimethylamine, is only classified as a group 2A carcinogen [13]. Such incoherence makes the aetiological role of salted fish especially worthwhile for more thorough re-exploration.

In the past 30 to 40 years, the incidence and mortality rates of NPC have dropped noticeably in several Asia regions particularly among some Chinese subpopulations [14, 15]. Given the wide range of the relative risk of high salted fish consumption ranging from 2.5–17.2 [16],which translates into population attributable risk (PAR) of 39%–78%, adopting exposure prevalence of salted fish consumption in the population, it would be of interest to investigate whether the secular trend of NPC disease rates is explained by a corresponding change in salted fish consumption in relevant jurisdictions. To our best knowledge, reports of such studies were not available.

We, therefore, carried out this Hong Kong and multi-site ecological study to investigate the relationship between secular trends of salted fish consumption and standardised NPC incidence and mortality rates over 30 years. We also included an analysis in Hong Kong taking into account other key lifestyle related factors, including tobacco smoking [17] and vegetable [18] consumption, which were reported to be a risk and protective factor of NPC respectively.

This is the first multi-jurisdiction ecological study that examined the associations between secular trends of NPC risk and salted fish consumption per capita. We also adjusted for the potentially confounding effects of two most important lifestyle factors, namely tobacco and vegetable consumption, using per capita estimates. We included 7 regions in 3 continents to include both endemic and low-risk areas, to analyse the pattern with a more global perspective.

We observed a decline in both ASIR and ASMR of NPC in Hong Kong over the past three decades. Higher NPC ASMR was observed in males than females in most of the jurisdictions included, consistent with previous studies [11, 30, 31], which may be due to a combination of differential risk factor exposures such as greater tobacco consumption in males and genetic factors. The rather consistent; more than 3-fold risk in males compared with females needs further investigation. Our results showed markedly decreasing trends in salted fish intake and tobacco consumption, but an increasing trend in vegetable consumption in Hong Kong over the past several decades. All of these environmental risk factors were individually highly correlated with ASMR and ASIR. The decrease in salted fish consumption may represent a change in dietary culture with westernisation of dietary patterns among Hong Kong citizens from a traditional Cantonese diet and from increased awareness of the carcinogenicity of salted fish. Decrease in intake of salted fish should lower the risk of an individual getting NPC if the relationship is causal [5].

There were no substantial changes in the proportion of Cantonese residing in Hong Kong, thus any changes noted should not be confounded by migration effect. Additional file 1: Tables S1-S3 show the population structure in Hong Kong from 1961–2006. Nevertheless, in the multivariable model, the relations between salted fish consumption and NPC ASIR and ASMR were attenuated and became mostly insignificant. In contrast, the negative association between vegetable consumption and NPC ASIR remained statistically significant after adjustment and was robust across varying estimates of cumulative exposure windows prior to NPC outcomes. Consistent with findings from previous case-control studies [18, 32], dietary intake of fruits and vegetables may have a protective effect against NPC due to its antioxidant contents, potentially interfering with the actions of nitrosamines [33]. World Cancer Research Fund (WCRF) panel suggested fruit and vegetable consumption have limited to suggestive levels of evidence for protection against NPC [34], which is consistent with the result found in this study. Although positive association was observed for cigarette consumption and NPC ASIR in males, previous literature suggested that for lung cancer, of which smoking is known as one of the major risk factor, the lag year for the disease from exposure to cancer onset was 25–30 years [35], it is expected that the lag for NPC should not be shorter if not similar to lung cancer. As our study is ecological, further analytical studies, such as case control studies with genetic, viral and other biomarkers are warranted to examine the link between salted fish and NPC in Hong Kong, taking into account smoking, vegetable consumption and other potential confounders.

Despite the parallel secular trends of salted fish consumption and NPC rates in Hong Kong, the same pattern was not observed in all other jurisdictions. For instance, in China, salted fish consumption increased over the years and there was no sign in apparent increase of ASMR in both males and females. Similarly, there are jurisdictions where NPC ASIR/ASMR had markedly decreased, but with no accompanying drop in salted fish consumption. Caution needs to be taken in interpretation, since the salted fish consumed in other countries might not be the Cantonese-styled salted fish, which is categorised as a group I carcinogen by IARC. The effect of salted fish on NPC development might only have a notable impact among Hong Kong Chinese but not in other populations, due to the higher underlying disease rates of NPC in Hong Kong, or could be due to genetic differences e.g. nitrosamine metabolism [36]. Indeed, the vast majority of previous case-control studies on salted fish consumption and NPC were only conducted in Chinese [1, 4, 5, 37]. It remains uncertain whether the same carcinogenesis mechanism could apply to different Chinese sub-group or non-Chinese populations.

Our findings highlight the importance of examining the relations between key risk factor exposures and NPC on multiple jurisdictions rather than focusing on a single locality and a single exposure in an ecological study. Furthermore, since NPC includes a spectrum of squamous, and keratinising versus non-keratinising subtypes which are linked to differential risk factors, this may also be an important reason underlying the varying patterns of ecological associations found across different jurisdictions. Thus an ecological study investigates the links between risk and protective factors and different subtypes would be of interest, especially when the pathogenesis processes for various NPC subtypes appear to be distinct. Therefore it is important to first elucidate the secular trend of salted fish consumption and NPC incidence/mortality rates between multiple areas. The findings from this study have encouraged us to further examine the time trend of different NPC subtypes across regions, and so far our literature review has shown that a more thorough review and deeper understanding is needed of different NPC histological classification between China and the West, but such data are often scarce. Also, the differences between Chinese and WHO classifications need to be further examined and resolved. We believe this would warrant a separate paper for a detailed discussion.

There are several limitations in our study. First of all we could not be certain that the capture of NPC incidence and mortality was 100% complete and accurate in each jurisdiction, but most data were derived from cancer registries where completeness of data; especially that of mortality, is good with high accuracy. Variations of diagnostic criteria could exist but this effect is minimised by using International Classification of Disease (ICD 8–10) since NPC is a rare and uncommon cancer. Secondly, per capita consumption was calculated by taking the difference between export, import, production and re-export of salted fish. Consequently, there was a possibility of over or underestimating the actual amount of salted fish being consumed. Nevertheless, data obtained from Hong Kong Trade and Statistics and from the FAO were largely consistent. Even though we tried to obtain all the salted fish related data as precisely as possible, the data we used might include other dried fish products rather than just Cantonese style salted fish, details of which could not be obtained from other jurisdictions. Moreover, the uncertainty of latency period between the consumption patterns and ASIR/ASMR is well acknowledged, for that we included the sensitivity analysis by using year of cumulative consumption for regression and result were shown to be robust. Furthermore, vegetable consumption in Hong Kong was not available from Agriculture, Fisheries and Conservation Department HKSAR prior to 1991, thus multivariable analysis was done only between 1991–2008 but not from 1978–2008.

Finally, the limitations of ecological studies due to the ecological fallacy are well recognised, and unknown, unmeasured or immeasurable confounding factors may be present. We tried to take account of important and known factors by accounting for tobacco consumption, vegetable consumption and migration effects, which is the first example of its kind in NPC epidemiological studies. In future studies, age-specific rather than age-standardised mortality and incidence rates could be more informative to investigate disease risk across different ages and in different populations. Because NPC is endemic mainly in Southern China, ecological analysis focusing on the endemic regions of China would yield more important information than a national study. Data specifically in Southern China on salted fish, vegetable and cigarette consumptions, together with NPC incidence and mortality rates (with subtypes) would be highly valuable if such data are available in the near future and more in depth analysis is warranted.

In this study, we found markedly decreasing trends of NPC ASIR and ASMR in Hong Kong over the past 3 decades, which was correlated with corresponding secular changes in salted fish consumption per capita. However, this association no longer held after adjustment for tobacco and vegetable consumption per capita. Our result showed vegetable had a protective role in NPC ASIR. There was no clear or consistent pattern in the relations between NPC ASIR and ASMR with salted fish consumption across 7 regions in 3 continents. Whether the pathogenic role of salted fish consumption is specific to NPC among Hong Kong Chinese requires further studies.