Discussion
Our study revealed contrasting hospital-related morbidity among individuals diagnosed with HBV and HCV infection, and that hospitalization rates were consistently higher in all co-infected cohorts, particularly those with HIV. The all cause and non alcoholic liver disease hospitalization rates and SHRs were two-fold higher in the HCV mono-infection cohort compared with the HBV mono-infection cohort, with all cause rates for HBV mono-infection even lower than in the NSW population. In contrast, the HBV mono-infected cohort had a hospitalization rate and SHR for primary liver cancer close to two-fold higher than that for the HCV mono-infected cohort. Perhaps the most notable contrast, however, is that between 2000 and 2006 rates for non alcoholic liver disease increased significantly in the HCV mono-infected cohort but decreased significantly in the HBV mono-infected cohort. Trends that were further accentuated in the HIV co-infection cohorts and are suggestive of an impact of improved treatment for advanced HBV-related liver disease [
8‐
10,
12,
14].
Contrasting all cause hospital-related morbidity between the HCV and HBV infected cohorts may relate to differing epidemiological patterns and health care utilization. In Australia, it is estimated that over 80% of HCV cases are in Australian-born injecting drug users, infected in their young adult years, with a smaller proportion among immigrants [
22]. In contrast, it is estimated that over 50% of HBV cases are immigrants from HBV endemic countries in South-East and North-East Asia [
23] with predominant perinatal or early childhood acquisition. A high proportion of the all cause hospitalizations amongst the HCV cohort in the less than 30 year age group has previously been reported to be lifestyle-related admissions such as drug and alcohol use [
24]. Such factors may also explain the high rates of all cause hospitalization in young adults in the HIV co-infected cohorts. The lower rates of hospitalization in the HBV mono-infection cohort compared with the NSW population across all age groups may be due to a 'healthy immigrant' effect or lower levels of health care utilization amongst immigrant populations [
25]. If it is the latter explanation, the comparative burden of morbidity associated with HBV may have been underestimated. Reasons for the downward trend in all cause hospitalization rates between 2000 and 2006 for both the HCV and HBV mono infected cohorts are unclear, as rates for the NSW population increased over the same period.
Rates of hospitalization for non alcoholic liver disease and primary liver cancer increased markedly with age in both the HBV and HCV infected cohorts, consistent with previous studies. In a systematic review, hepatic fibrosis progression was found to be non linear and significantly affected by duration of HCV infection [
26]. In addition, a large cross sectional study showed that, regardless of the duration of HCV infection, the risk of cirrhosis increased significantly after the age of 50 years [
27]. In HBV infection, risk of cirrhosis also increases with age, particularly after 50 years [
28].
As expected, co-infection with HIV increased the burden of morbidity for both the HCV and HBV cohorts. Compared with the mono-infected cohorts, there was at least a two-fold increase in SHRs for all three admission types examined, except for primary liver cancer in the HCV/HIV cohort. HIV co-infection has been shown to be associated with an increased HBV and HCV viral load, a higher rate of fibrosis progression and increased risk of cirrhosis [
29‐
35]. In addition, HIV co-infection has been associated with higher rates of HCC than HCV and HBV mono-infection [
36].
The burden of non alcoholic liver disease was expected have increased during 2000-2006. This is because the disease cohorts are aging, as well as continuing to expand, albeit at a slower rate than in previous decades [
37,
38]. However, hospitalization rates for non alcoholic liver disease in the HBV and HBV/HIV cohorts declined during the review period, in contrast to the (expected) upward trend in the HCV and HCV/HIV cohorts. This is unlikely to be due to differential changes in case finding, which remained relatively constant over the review period, but may suggest an impact of HBV treatment at a population level.
There are several factors that may help to explain why HBV treatment has impacted on liver-related morbidity at a population level. First, therapy for HBV improved during the study period. Although lamivudine (available since 1998) can be successfully used to treat advanced HBV-related liver disease [
12], the availability of adefovir (2004) and entacavir (2006), both of which have a higher genetic barrier to resistance than lamivudine, has generally lead to better maintenance of HBV DNA suppression, a further reduction in the risk of liver disease, and even reversal of decompensated liver disease [
11,
14]. Second, HBV treatment uptake improved over the study period [
38]. Therefore, even though uptake overall remains relatively low in Australia for both HBV (5%) and HCV (1.7%) [
38], patients presenting with HBV-related advanced liver disease would be expected to have high rates of treatment uptake and favourable outcomes [
12,
14]. This means that even low levels of HBV treatment uptake may reduce liver-related morbidity. In contrast, patients presenting with HCV-related advanced liver disease have either poor HCV treatment outcomes in the case of compensated cirrhosis [
39], or are not eligible for treatment if decompensated cirrhosis is present.
The greater decline in non alcoholic liver disease hospitalizations for the HIV/HBV cohort compared with the HBV mono-infected cohort adds to the evidence for an impact of HBV therapy at a population level. Tenofovir was approved for HIV treatment in Australia in 2002, prior to approval of adefovir (2004), entecavir (2006) or tenofovir (2008) for HBV mono-infection. Tenofovir provides high and sustained levels of HBV DNA suppression in HIV/HBV and HBV populations, with no described resistance [
10,
40], and has been shown to lead to resolution of decompensated liver disease [
13]. The earlier introduction and high uptake of tenofovir in HIV infected patients [
38] is consistent with a more rapid decline in non alcoholic liver disease hospitalizations amongst the HBV/HIV co-infected cohort and with an impact of HBV therapy at a population level.
The contrasting trends in non alcoholic liver disease morbidity are consistent with trends for liver transplantation and mortality data. U.S. liver transplant waiting list data showed a 27% decline in HBV-related end stage liver disease registrations between 1999 and 2006 [
41]. Similarly, Australian data for 1997-2006 show a decline in the number and proportion of transplant recipients with HBV-related cirrhosis, in contrast to the upward trend for HCV-related cirrhosis [
37]. In addition, rates of non-HCC liver deaths in notified cases of HBV in NSW declined by 5% while there was no change in the HCV cohort [
42]. The more pronounced trends for hospitalization rates seen in our study may be due to the relatively recent advances in antiviral treatment having a more immediate impact on morbidity than on mortality.
Rates of hospitalization for both HBV- and HCV-related primary liver cancer increased during 2000-2006, consistent with the increasing number of liver cancer diagnoses in Australia [
43,
44]. A further contribution to these trends may be increasing management options for primary liver cancer [
45]. The higher hospitalization rates for primary liver cancer in the HBV-mono-infected cohort compared with the HCV mono-infected cohort may be because a higher proportion were infected at an early age (in their country of birth) and therefore have a greater cumulative risk of disease progression [
46]. This may also explain the high SHRs in the less than 40 year age groups. However, only date of diagnosis (which may occur many years after the date of infection) is reported so we are unable to confirm what age cases were infected. Another reason for the difference may be that HBV-related liver cancer can occur in the absence of cirrhosis due to the direct oncogenic affects of the HBV virus [
47]. Therefore, even if the overall cirrhosis risk is similar there may be a greater risk of liver cancer in HBV cases. Our results are consistent with other populations-based linkage studies in Australia which show higher mortality from, and incidence of, liver cancer in people with HBV compared to those with HCV [
1,
42].
The major strength of our study is that it is a large population-based linkage study including HIV co-infection status. By linkage of the cohorts to their hospital records we were also able to account for the correlation between hospitalizations for the same patient in our analyses of changes over time. The temporal trends presented here may therefore differ from the crude rate changes reported in other studies that were unable to account for within patient clustering or increases in the HCV and HBV infected population [
48‐
50].
Population-based studies have some methodological limitations. In particular, caution needs to be exercised when comparing the different cohorts as some of the differences may be due to unmeasured confounding factors. Differences in the overall burden and trends over time may be due to variations in health care utilization rather than differences in actual morbidity. Second, while the aim of our study was to examine morbidity in cases of chronic HBV and HCV infection, more than 65% of HBV records and 77% of HCV records did not specify whether a case was acute or chronic and only 2% and 1% of cases, respectively, were recorded as newly acquired. For this reason all records were included, which would lead to an underestimate chronic HBV and HCV-related morbidity as some of the notified cases may have cleared their infection. However, it is estimated that 74% HCV infections are chronic [
51] and most notified cases of HBV are also likely to be chronic given that a high proportion are thought to have acquired their infection in endemic countries during childhood [
23]. Third, we were unable to exclude members of the disease cohorts from the reference (NSW) population. Although the cohorts only accounted for 1.5% of all cause hospitalizations in the NSW population, they accounted for one quarter of the liver cancer and 19% of the non alcoholic liver disease admissions. A comparison with the non infected population would therefore have yielded higher SHRs. Fourth, identifiers for matching HIV cases were not complete, so some of the HBV and HCV mono-infected cohort may include HIV positive cases, but again this would only lead to an underestimate of morbidity associated with HIV co-infection. Fifth, hospital data with identifiers for linkage were only available for a six year period from June 2000, so the trends identified here need to be monitored for a longer period of time. Sixth, there is likely to be some misclassification of alcoholic and non alcoholic liver disease diagnoses. However, this misclassification is likely to be non differential and there were no changes to the coding procedure over the period of the review. Therefore coding errors are unlikely to explain any divergent trends. Finally, for some of the analyses there were too few cases of HIV co-infected patients to examine trends.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
The study was designed by HG and GD. Data extraction was undertaken by HG. The analysis was conducted by HG with assistance from ML and JA. The manuscript was drafted by HG with modifications by GD and ML. All authors read and approved the final manuscript.