Increase in mortality rates from intrahepatic cholangiocarcinoma in England and Wales 1968–1998
- S D Taylor-Robinsona,
- M B Toledanob,
- S Arorab,
- T J Keeganb,
- S Hargreavesc,
- A Beckd,
- S A Khana,
- P Elliottb,
- H C Thomasa
- aLiver Unit, Department of Medicine A, Imperial College School of Medicine, St Mary's Campus, South Wharf Street, London W2 1PG, UK, bDepartment of Epidemiology and Public Health, Imperial College School of Medicine, St Mary's Campus, South Wharf Street, London W2 1PG, UK, cKensington and Chelsea and Westminster Health Authority, 50, Eastbourne Terrace, London W2 6LX, UK, dDepartment of Biological Sciences, Wye College, University of London, Wye, Ashford, Kent TN25 5AH, UK
- Dr S D Taylor-Robinson, Liver Unit, Department of Medicine A, 10th Floor, QEQM Wing, Imperial College School of Medicine, St Mary's Hospital, South Wharf Street, London W2 1PG, UK.
- Accepted 19 December 2000
BACKGROUND The age standardised mortality rate per 100 000 population for all causes of liver tumours (International Classification of Disease 9 (ICD-9) 155) has almost doubled in England and Wales during the period 1979–1996. We further analysed the mortality statistics to determine which anatomical subcategories were involved.
METHODS Mortality statistics for liver tumours of ICD-9 155, 156, and subcategories, and for tumours of the pancreas (ICD-9 157), in England and Wales were investigated from the Office for National Statistics, London, from 1968 to 1996 inclusive. Data for 1997 and 1998 were also available on intrahepatic cholangiocarcinomas.
RESULTS There has been a marked rise in age standardised mortality rates for intrahepatic cholangiocarcinoma. Since 1993, it represents the commonest recorded cause of liver tumour related death in England and Wales. This is evident in age groups older than 45 years. In contrast, mortality trends from other primary liver tumours, including hepatocellular carcinoma, were unremarkable.
CONCLUSIONS The observed increase in mortality from intrahepatic cholangiocarcinoma may represent better case ascertainment and diagnosis due to improved diagnostic imaging, use of image guided biopsies, or increased use of ERCP. However, the trend started before ERCP was introduced nationally, mortality rates have continued to increase steadily thereafter, and there is no clear evidence that diagnostic transfers easily explains the findings. Alternatively, these observations may represent a true increase in intrahepatic bile duct tumours. Epidemiological studies are required to determine whether there is any geographical clustering of cases around the UK.
Studies from France, Italy, and the USA suggest that the incidence of hepatocellular carcinoma (HCC) is increasing,1 2 and this may be related to chronic hepatitis C virus (HCV) infection.1-4 Age standardised mortality rates (ASMR) for all causes of malignant liver tumours (International Classification of Disease 9 (ICD-9) 155) have increased from 1979 to 1994 in the UK,5 but unlike in Southern Europe, the ASMR for HCC has been reported to be relatively static over this time period while the ASMR for intrahepatic cholangiocarcinoma may have been increasing.5
We investigated ASMR and age specific mortality rates (ASpMR) for all liver tumours in England and Wales, starting in 1968, when comparable mortality records began, to determine when the rise in ASMR for intrahepatic cholangiocarcinoma first started and which age groups have primarily been affected.
Mortality data for 1968–1996 were obtained for the English and Welsh populations from the Office for National Statistics (ONS) in London. Information was requested on:
ICD-9 155 (all malignant liver tumours)
ICD-9 155.0 (primary liver tumours, mainly HCC)
ICD-9 155.1 (intrahepatic cholangiocarcinoma)
ICD-9 155.2 (histologically unspecified liver tumours)
ICD-9 156 (all extrahepatic biliary system tumours)
ICD-9 156.0 (gall bladder tumours)
ICD-9 156.1 (tumours of the extrahepatic bile ducts)
ICD-9 157 (pancreatic tumours).
The ICD-9 155.0 code contains information on all primary tumours of the hepatic parenchyma but is overwhelmingly composed of data on HCC. Mortality data, supplied in five year age bands, were considered to be an indicator of incidence because prognosis from liver cancer is poor. Additional mortality data for 1997–1998 were requested, once they had become available, for intrahepatic cholangiocarcinoma, the tumour of primary interest, as initial analysis indicated that the observed trends for this tumour warranted further up to date inspection.
NUMBER OF DEATHS
The total number of deaths attributed to each ICD category of cancers was analysed by year and sex. No deaths before 1968 are included because coding changes between ICD-7 (1958–1967) and ICD-8 (1968–1978) make it impossible to accurately trace deaths attributed to a particular category earlier than this. Coding changes between ICD-8 and ICD-9 for the tumours under study seemed to be minor, although it is recognised that any change in coding can have an impact on mortality statistics.6 In our analysis, we mapped the change in code for “unspecified primary or secondary tumours of the liver” across ICD revisions from 197.8 to 155.2.
Two other notable changes to the collection and coding of deaths in England and Wales occurred in 1993 which may affect interpretation of trends in mortality. These include the move back by the Office of Population Censuses and Surveys (OPCS) to the internationally accepted interpretation of World Health Organization (WHO) rule 3, regarding the selection of the underlying cause of death, which was in operation before 1984.7 In addition, until 1992 it was standard procedure for the OPCS to apply to the person certifying the death for further information regarding the underlying cause of death when the conditions mentioned on the death certificate were vague. The use of such “medical enquiries” to help assign a more definite code to the underlying cause of death was discontinued in 1993.7
AGE STANDARDISED MORTALITY RATES
ASMR per 100 000 population were calculated using direct standardisation with the European standard population as a reference population.8
AGE SPECIFIC MORTALITY RATES
Initially, ASpMR per 100 000 of the population were analysed by five year age bands but given that tumours are much less common in the younger age groups and that treatment and management may differ between middle and very old age, these were aggregated into more clinically relevant age specific groups. The following age bands were therefore chosen: 20–44, 45–64, 65–74, and 75+.
ALL LIVER TUMOURS
ASMR per 100 000 population for the combined causes of malignant liver tumours increased steadily in the period 1968–1996 from 1.29 to 1.93 in females and from 2.56 to 3.70 in males (fig 1A, B). Total numbers of deaths increased from 967 (396 females; 571 males) in 1968 to 1822 (770 females; 1052 males) in 1996 (table1).
Primary liver tumours including hepatocellular carcinomas
Data on primary liver tumours revealed fluctuations in ASMR about a stable mean between 1968 and 1978 for both sexes. However, in 1979 there was a noticeable increase in ASMR for both males and females until 1992 (fig 1A, B). In 1993 there was a sharp decline in ASMR to levels more in common with those before 1979.
ASMR per 100 000 population for intrahepatic cholangiocarcinoma increased markedly for both sexes over the period 1968–1996 (fig 1A, B). There was a 15-fold increase in ASpMR per 100 000 population in ages 45 and above (fig 2A, B). The total number of deaths increased from 38 (17 females, 21 males) in 1968 to 736 (387 females, 349 males) in 1996 (table 1). This cancer now represents the commonest primary liver tumour in England and Wales, having overtaken hepatocellular carcinoma in 1993 (table 1). Additional data for 1997–1998 showed a further marked increase in mortality rates from the 1996 statistics. ASMR increased from 0.92 and 1.22 in 1996 to 1.12 and 1.37 in 1998 for females and males, respectively, bringing the total number of deaths in that year to 864.
Unspecified tumours of the liver
ASMR for liver tumours in this ICD category showed a small decrease overall, although there was a marked fall in ASMR from 1978 to 1993 in both sexes (fig 1A, B).
TUMOURS OF THE GALL BLADDER AND EXTRAHEPATIC BILIARY TREE
There was a steady decrease in ASMR between 1968 and 1996 (fig 1A, B). Between 1968 and 1978, the total number of deaths for men and women was relatively stable at approximately 700 per year for women and approximately 400 per year for men. Subsequently, the total number of deaths fell from 1043 in 1979 to 585 in 1996 (table 1).
Tumours of the gall bladder
Between 1968 and 1996, both ASMR and total number of deaths attributed to malignant neoplasm of the gall bladder decreased steadily. ASMR fell from 1.21 and 0.84 in 1968 to 0.55 and 0.30 in 1996 for females and males, respectively.
Tumours of the extrahepatic bile ducts
ASMR for malignant tumours of the extrahepatic bile duct fluctuated between 1968 and 1979, after which they began to fall from 0.71 for females and 0.80 for males in 1980 to 0.23 in 1996 for both sexes.
Combined intrahepatic and extrahepatic cholangiocarcinoma (ICD-9 code 155.1 and 156.1)
The combined ASMR from all tumours of the biliary tract increased over the time period 1968–1996. The absolute number of deaths also increased from 466 (240 females, 226 males) in 1968 to 912 (493 females, 419 males) in 1996 (table 1).
ASMR showed a slight decline for both sexes but this was less marked for females (fig 1A, B). However, the absolute number of deaths rose between 1968 (total 4898; 2279 females; 2619 males) and 1996 (total 5873; 3044 females; 2829 males).
In England and Wales there has been a steady rise in mortality coded to intrahepatic cholangiocarcinoma. Although ASMR for extrahepatic cholangiocarcinoma have decreased, the increase in intrahepatic tumours outweighs this with a rise in ASMR and in absolute number of deaths for both these two ICD-9 categories combined. Cholangiocarcinoma, arising from the intrahepatic biliary tree, is much less common than HCC on a worldwide basis, but since 1993 this tumour is the commonest recorded cause of malignant liver tumour related death in England and Wales. In South East Asia and China, development of cholangiocarcinoma is mainly associated with liver flukes.9 In the West the aetiology is largely unknown although the obsolete radiological contrast agent thorotrast has been implicated in causing cholangiocarcinoma.10 Smoking and alcohol have also been implicated but the evidence is weak.11 12 Primary sclerosing cholangitis is the commonest predisposing factor in the UK but it is associated with only a minority of cases.9
The steadily increasing ASMR for intrahepatic cholangiocarcinoma may represent an artefactual trend, introduced into the mortality figures by improvement in diagnosis (case ascertainment) resulting from better imaging techniques, by changes in coding practices, or from misclassification of ICD coding (diagnostic transfer). It is also possible that this trend represents a real increase in mortality from this tumour.
Better ascertainment from improved diagnostic imaging, use of image guided biopsies, and from the national availability of techniques such as endoscopic retrograde cholangiopancreatography (ERCP) could account for these observations. ERCP was first available in the UK in the late 1970s after the rise in ASMR started, and became available in most UK hospitals in the mid 1980s.13 14 The increased mortality rates have continued after the initial endoscopic learning curve and the general availability of the technique should have led to an expected plateau in these values. It is therefore unlikely that our observations on intrahepatic cholangiocarcinoma are solely caused by an improvement in diagnosis from ERCP. Furthermore, it is the distally located extrahepatic bile duct tumours, total numbers of deaths from which have actually decreased, that are more easily visualised with ERCP, rather than the intrahepatic tumours which often involve much smaller bile ducts. The definitive diagnosis of intrahepatic bile duct tumours is histological, and in a large proportion of cases this is made from tissue obtained at surgery or at post mortem.15 However, this is dependent on the experience of the local pathologist and therefore there may be interobserver differences in interpretation. This factor in itself would not account for the year-on-year rise in intrahepatic cholangiocarcinoma that we have reported.
There have been a few changes to the collection and coding of mortality data from 1968 to 1998 which could potentially render our findings artefactual. However, no discernible “steps” in mortality trends for intrahepatic cholangiocarcinoma have been introduced by changes in coding practices. This is in stark contrast with the mortality trends for primary liver tumours (mainly HCC) and for unspecified liver tumours, which provide clear illustrations of the impact of coding changes. In 1979 when ICD-9 was first introduced, subtle changes in the number and type of diseases which fed into both these categories of tumours led to an increase in primary liver tumours and a simultaneous decline in unspecified liver tumours. In 1993, both the rule 3 coding change and the death certificate changes introduced further artefact, with the latter having the dominant effect on these tumour types. Prior to 1993, it was routine practice for OPCS to obtain follow up information from the certifying doctor about the cause of death, approximately six weeks after the initial death certification. In many instances, information that became available with post mortem confirmed or changed the cause of death and this was rectified with this process. With the absence of follow up medical enquiries on death certificates from 1993, there was an increase in the number of deaths assigned to categories of less specific causes.7 The sudden change in mortality statistics between 1979 and 1993 for primary and unspecified liver tumours are thus artefactual and any examination of time trends for these tumours is limited to cautious comparisons of data from before 1979 and after 1993.
Diagnostic transfer from tumours of the gall bladder and extrahepatic biliary tree to intrahepatic cholangiocarcinomas is an alternative explanation for our findings but the increase in ASMR and absolute number of deaths from intrahepatic tumours outweighed the decrease in the other malignancies. Simple diagnostic transfer is therefore unlikely. Diagnostic transfer may also occur between intrahepatic cholangiocarcinoma and undifferentiated metastatic adenocarcinoma in the liver but it is probable that histologically ill defined tumours tend to be reported in the unspecified category than reported specifically as intrahepatic cholangiocarcinoma.16Diagnostic transfer from pancreas to intrahepatic cholangiocarcinoma could account for the small decrease in ASMR for pancreatic tumours and the marked increase in ASMR for intrahepatic bile duct tumours but this is unlikely because it is unusual to confuse the two diagnoses from an anatomical standpoint.
Another explanation for the steady rise in ASMR for intrahepatic cholangiocarcinoma between 1968 and 1998 is a true increase in incidence of this tumour. The reason intrahepatic cholangiocarcinomas have increased in place of tumours of the rest of the biliary tract may be because of the larger surface area of the intrahepatic biliary tree17 compared with that of the gall bladder and extrahepatic biliary tree, allowing greater exposure to potential carcinogens, which may have been more prevalent over the past 30 years. Furthermore, intrahepatic bile ducts contain the majority of actively dividing cholangiocytes, which would make malignancies of this part of the biliary tree more likely.18 If improved imaging and the availability of ERCP has led to better case ascertainment, intrahepatic tumours may be detected at an earlier stage, before invasion of the extrahepatic biliary tree has occurred. However, the absolute rise in both ASMR and total number of deaths from the combined ICD-9 categories for intra- and extrahepatic bile duct tumours is also further evidence that the observed trend may, in part, be real rather than apparent.
The oral contraceptive pill has also been implicated in the development of cholangiocarcinoma19 20 although case control studies have failed to substantiate an increased risk.21 The time period when oral contraceptives became widely available in England and Wales and the age groups involved with increased mortality from intrahepatic cholangiocarcinoma do not entirely match.22Furthermore, this trend in mortality statistics is also seen in men, making the “pill” an unlikely candidate as the sole cause of this phenomenon, although there are many chemicals which could potentially have entered the food chain which have oestrogenic properties.
We have shown a dramatic increase in ASMR and ASpMR for intrahepatic cholangiocarcinoma from 1968 to 1998 in England and Wales. Epidemiological studies are required to determine whether there is any geographical clustering of cases around the UK and whether any clusters can be related to occupational exposure. Further studies are needed to compare the observed trend in mortality statistics with that for other countries. Given that current treatment outcomes for cholangiocarcinoma are poor, such studies may lead to a better understanding of the aetiology of these tumours and institution of appropriate preventative measures.
- Abbreviations used in this paper:
- age standardised mortality rates
- age specific mortality rates
- endoscopic retrograde cholangiopancreatography
- hepatocellular carcinoma
- hepatitis C virus
- International Classification of Disease 9
- Office for National Statistics
- Office of Population Censuses and Surveys
- World Health Organization