Accepted Manuscript Trends in mortality from chronic liver disease Ugo Fedeli , MD Francesco Avossa , Stefano Guzzinati , Emanuela Bovo , Mario Saugo PII:
S1047-2797(14)00156-2
DOI:
10.1016/j.annepidem.2014.05.004
Reference:
AEP 7654
To appear in:
Annals of Epidemiology
Received Date: 14 February 2014 Revised Date:
17 April 2014
Accepted Date: 6 May 2014
Please cite this article as: Fedeli U, Avossa F, Guzzinati S, Bovo E, Saugo M, Trends in mortality from chronic liver disease, Annals of Epidemiology (2014), doi: 10.1016/j.annepidem.2014.05.004. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Trends in mortality from chronic liver disease
SER – Epidemiological Department, Veneto Region, Italy
2
Veneto Cancer Registry, Italy
M AN U
SC
1
RI PT
Ugo Fedeli1, Francesco Avossa1, Stefano Guzzinati2, Emanuela Bovo2, Mario Saugo1
Corresponding author: Ugo Fedeli
Passaggio Gaudenzio, 1 - 35131 Padova (PD), Italy Tel:
+39 049 8778252
Fax +39 049 8778235
AC C
Word count: 2353
EP
TE D
e-mail: [email protected]
Abstract word count: 200 Tables 2 Figures: 3
1
ACCEPTED MANUSCRIPT ABSTRACT Background: Mortality from liver cirrhosis has sharply declined in most countries in the last few decades, but includes only a fraction of deaths due to chronic liver disease.
RI PT
Methods: Mortality records for liver cirrhosis, liver cancer, viral hepatitis, and the three disease categories combined were extracted for the period 1995–2010 in the Veneto Region (North-Eastern Italy). The presence of a birth cohort effect was assessed by fitting Poisson regression models. The alcoholic or viral etiology was retrieved through multiple causes of death analysis for the years 2008-2010.
SC
Results: Liver cirrhosis mortality represented only one-third of all liver disease deaths; this proportion was even smaller for deaths with a viral etiology and among elderly people. Through the study period age-standardized rates declined by 50% for liver cirrhosis, and by 30% (world standard) or by 25% (regional standard) for all liver diseases; the reduction in proportional mortality was limited to 11%. Mortality rates started to decline in subjects born after the early 1920s.
M AN U
Conclusion: The decline in mortality was smaller than that reported by standard statistics for chronic liver disease; these statistics should routinely include viral hepatitis and liver cancer deaths. When available, multiple causes of death data should be analyzed.
TE D
Keywords
AC C
EP
Mortality; liver cancer; liver cirrhosis; viral hepatitis
List of abbreviations and acronyms HCV HBV
hepatitis C virus hepatitis B virus
ICD
International Classification for Diseases
HCC
hepatocellular carcinoma
2
ACCEPTED MANUSCRIPT Introduction In the last few decades mortality from cirrhosis has been decreasing in most countries around the world. With the exception of the increasing mortality recorded in the UK and Eastern Europe, steeply declining rates have been reported in Southern and Western Europe, North America, Oceania and some Asiatic countries [1]. A reduction in the prevalence of alcoholic liver disease, a
RI PT
fall in the transmission of hepatitis C virus (HCV) infection, and vaccination campaigns against hepatitis B virus (HBV) have all contributed to the observed decrease in mortality in Southern Europe [2]. In Italy, a greater than 5% yearly decline in cirrhosis mortality rates has been estimated for the period 1984–2002 among males, with only a slightly less favorable trend observed in
SC
females [1]. According to the Global Burden of Disease profile for Italy, cirrhosis has dropped from the 5th to the 8th most common cause of premature death, due to a 50% reduction in years of life lost
M AN U
from 1990 to 2010 [3].
However, liver cirrhosis mortality captures only a limited fraction of deaths due to chronic liver disease. In the US, by adopting the usual definition for chronic liver disease mortality, a 38% decrease in rates was observed from 1979 to 2008; if other liver-related causes of death including viral hepatitis and liver cancer were also considered, mortality rates remained essentially unchanged throughout the study period [4]. Chronic viral hepatitis, liver cirrhosis, and hepatocellular
TE D
carcinoma constitute a pathogenic continuum and could therefore be usefully merged in a broader definition of liver-related deaths. However, further intricacies in interpreting mortality trends could arise from coding practices and rules for liver cancer deaths [5], and from the strong birth cohort
EP
effects observed for liver cancer mortality in several countries [6-8]. In Italy healthcare, including disease surveillance and preventive services, is administered by
AC C
Regions based on general rules set at the national level. The aim of the study was to examine time trends of liver-related deaths in the Veneto Region (North-Eastern Italy) by addressing different methodological issues that impact on liver mortality statistics: restrictive or broader inclusion of codes across subsequent versions of the International Classification for Diseases (ICD), analysis of age-specific or age-standardized rates, birth-cohort effects, and the multiple causes of death approach.
3
ACCEPTED MANUSCRIPT Methods The Veneto region has about 4,900,000 inhabitants and is subdivided into 21 Local Health Units (LHUs) and 581 municipalities. Death certificates are sent by each municipality to the LHU; until 2005 causes of death were coded in each LHU according to the ICD, 9th Edition (ICD-9). Since 2006, LHUs have transmitted a copy of the certificate to the Regional Epidemiological Department
RI PT
for coding of the underlying cause (from 2007 in ICD-10) and entry into the electronic regional archive of mortality data. The underlying cause is the single condition usually reported in mortality statistics, identified from all diseases mentioned in the death certificate according to internationally adopted selection and modification rules. Since 2008 the regional archive has included ICD codes
SC
not only for the underlying cause, but also for all diseases reported in the certificate.
From the mortality records of residents in the Veneto Region in the period 1995–2010, the
M AN U
following underlying causes of death were extracted: liver cirrhosis (ICD-9 571; ICD-10 K70, K73, K74), liver cancer (ICD-9 155; ICD-10 C22), and viral hepatitis (ICD-9 070; ICD-10 B15-B19). The above three disease categories were thereafter merged as “liver disease deaths”; a broader definition reported in a recent US study [4] was adopted including deaths from liver failure, other and unspecified liver disease, and esophageal varices (“all liver disease extended”, the above codes plus ICD-9 570, 572, 573, 4560, 4561; ICD-10 K71, K72, K75, K76, I85). Limited to the period
TE D
2008–2010, liver disease deaths were classified by etiology, retrieving ICD-10 codes for alcoholic disease (codes for alcoholic liver disease and for mental and behavioral disorders due to the use of alcohol, K70, F10), and for acute or chronic HBV or HCV infection (B16, B18.0, B18.1 and B17.1,
EP
B18.2, respectively) from all conditions reported in the certificate. Crude and age-specific mortality rates were computed for the above disease categories across
AC C
calendar periods (1995–1998, 1999–2002, 2003–2006, 2007–2010) as the number of deaths per 100,000 population, with population data derived from the National Institute for Statistics (http://demo.istat.it/). Age-standardized rates were obtained by the direct standardization method, taking as reference either the World standard population, or the regional population in 2007. The proportional mortality in each age class was defined as the percentage of liver disease deaths out of all registered deaths. Lastly, to assess the presence of a birth cohort effect, liver disease death rates among subjects aged 45–94 were classified in 5-year age and calendar period classes, and the corresponding birth cohort was identified by the mid-cohort year of birth. Death rates were the dependent variable in different Poisson regression models including as explanatory variables age alone, age and period, age and cohort, or age, period and cohort. Such models were compared in both genders using the likelihood ratio test and the Akaike information criterion [9]. 4
ACCEPTED MANUSCRIPT
Results The overall number of registered deaths in 1995–2010 was 12,741 for liver cirrhosis, 1,226 for viral hepatitis, 12,883 for liver cancer, 26,850 for all liver diseases (the previous categories combined),
RI PT
and 27,684 when adopting the broader definition (all liver diseases extended). Figure 1 shows time trends in crude mortality rates for the above five disease categories by gender: mortality due to liver cirrhosis steeply declined, liver cancer rates were relatively stable, and viral hepatitis became an important cause of death during the study period. As a consequence, all liver diseases mortality
SC
showed a limited decrease; the inclusion of additional causes of death did not change this finding, and was not considered in further analyses. Time trends were similar between genders, but among
more pronounced than among males.
M AN U
females the relative reduction in liver cirrhosis and the increase in viral hepatitis mortality were
Liver cirrhosis death rates dropped by about 50% from 1995–1998 to 2007–2010 across all age classes (Figure 2); in contrast liver cancer rates displayed different patterns by age, with a limited decrease in younger subjects, and almost no change among the elderly. Table 1 shows mortality figures for all liver diseases combined: there was a 40% decline in age-
TE D
specific mortality rates during the study period in subjects younger than 55 years, but only a 15% decline in those aged 75–84 years, while rates remained almost unchanged among the very elderly. Given the diverging time patterns, the reference adopted for age-standardization influenced the
EP
overall estimated decrease, which was thus larger with a reference giving more weight to younger classes (World standard) than with an internal reference reflecting the ageing regional population. Liver diseases accounted for a substantial proportion of deaths among subjects aged 45–74 years;
AC C
the overall proportional mortality declined by about 10% (from 4.2% in 1995–1998 to 3.7% in 2007–2010).
In contrast, if analyses were limited to liver cirrhosis deaths, age-standardized rates halved during the study period irrespective of the reference population adopted, and the proportional mortality decreased by 40% (from 2.3% to 1.4% of overall deaths, data not shown). Among the Poisson regression models performed, the most informative was the age-cohort model (the one with the minimum Akaike information criterion value): if the period effect was added to the explanatory variables, the model fit did not improve. Figure 3 summarizes the results of the agecohort analysis: liver disease mortality remained high among people born in the first two decades of 5
ACCEPTED MANUSCRIPT the last century, and started to decline in both genders only for birth cohorts born after the early 1920s. This pattern was quite different from that observed for overall mortality, which had already begun to decline in older birth cohorts. Among females, the relative reduction in liver mortality among cohorts born after 1920 seemed to be steeper than that estimated for overall mortality. Table 2 shows liver disease mortality rates by gender and age class for the period 2008–2010, when
RI PT
multiple causes of death were also available, allowing the etiology of the disease to be retrieved; the ratio to mortality rates estimated by analyses limited to liver cirrhosis deaths is also provided. Death rates rose with age in both genders, but the male to female ratio, which was about four in younger subjects, declined among the elderly. The etiology of liver disease was reported in about 40% of
SC
deaths, being mainly HCV among females and alcohol among males. Mortality rates were about three times those estimated by liver cirrhosis deaths alone, with a higher rate ratio among older age
M AN U
classes; furthermore, whereas alcoholic liver disease is included among ICD-10 codes for cirrhosis (K70), viral hepatitis is in another ICD-10 chapter, and standard mortality statistics greatly underestimated deaths with a viral etiology.
Discussion
TE D
The present findings support the inclusion of viral hepatitis and liver cancer deaths among routine statistics of chronic liver disease mortality [4]. With respect to liver cirrhosis mortality, rates were about three times higher, depending on etiology. Mortality due to liver disease decreased in 1995–
EP
2010 in North-Eastern Italy, but the extent of this reduction differed according to the adoption of a more or less restrictive selection of ICD codes: age-standardized rates (World standard) declined during the study period by 50% for liver cirrhosis mortality, and by 30% when examining all liver
AC C
disease mortality. Moreover, since time trends were less evident in older age classes, the reduction declined to 25% when rates were standardized using the regional population as a reference. When the proportional mortality (share of total mortality) was analyzed, the overall reduction was limited to 11%, and liver disease deaths still accounted for a substantial fraction of overall mortality among subjects aged 45–74 years. The fall in liver cirrhosis mortality analyzed alone was large and homogeneous across age classes. This finding only in part reflects a real epidemiological trend, possibly being amplified by the following mechanisms: strict surveillance and increased diagnostic rate of hepatocellular carcinoma (HCC) among cirrhotic patients; better control of complications of cirrhosis with improved survival and risk of HCC development at older ages [10]; adoption of the ICD-10 with related coding 6
ACCEPTED MANUSCRIPT instructions and specification [11], with an increasing number of liver disease deaths classified as chronic viral hepatitis. According to data from an Italian multicentric study, about 96% of HCC cases were associated with cirrhosis [12]. Therefore, the inclusion of liver cancer mortality could take into account the above-described diagnostic shift and allow for a better evaluation of time trends. Moreover, the overall analysis of all liver diseases classified as chronic, viral, or neoplastic
RI PT
is likely to be more robust to coding discontinuities in mortality statistics. It should be noted that according to the ICD-9, which was adopted in the earlier study periods, liver cancer deaths were coded as primary (1550), of intrahepatic bile ducts (1551), and unknown if primary or metastatic (1552). As a consequence, earlier mortality studies focusing on primary liver
SC
cancer adopted different strategies: the inclusion of all 155 codes [6], of all liver cancers other than intrahepatic bile ducts [13], or more often only of those specified as primary [7,10]. The latter
M AN U
option however could be affected by time shifts in the adoption of the less specific 1552 code instead of the 1550 code [5]; the same holds for the inclusion/exclusion of intrahepatic cholangiocarcinoma, mortality from which is reported to be steeply increasing in some European countries [14]. Although entailing a possible misclassification of some tumors metastatic to the liver, the most feasible choice in evaluating time trends across different ICD versions appears to be the inclusion of all 155 codes; this strategy is also consistent with reports examining the
TE D
comparability between ICD-9 and ICD-10 in different countries [15,16]. Mortality from all liver diseases displays wide variation in time trends by age, supporting the presence of a strong birth cohort effect. The steep decline in mortality in cohorts born after the early
EP
1920s suggests that the downward time trend, although less favorable than that reported using standard liver cirrhosis mortality statistics, could be a long-term tendency that will continue into subsequent decades. Such a birth cohort effect has already been reported for primary liver cancer,
AC C
and is usually attributed to the HCV epidemic in Italy in the 1940s that mainly arose due to contaminated medical instruments [7,12]. The birth cohort effect in our data extended to mortality for all liver diseases, and was similar in females, in whom according to the multiple causes of death analysis the viral etiology largely prevailed – possibly due to a higher exposures to medical procedures –, and males, in whom a major role was also played by alcohol abuse. Multiple causes of death data can be analyzed to investigate liver disease mortality by etiology in different population subgroups [17]; where available for long time periods, they have been usefully employed to examine trends in mortality related to viral hepatitis and alcoholic liver disease [1820]. Such analyses are limited by the certifying physician’s knowledge of the etiology of the disease, and by problems of underreporting. However, the present results are consistent with 7
ACCEPTED MANUSCRIPT seroepidemiologic data showing a higher prevalence of anti-HCV antibodies in subjects aged ≥75 years [21], and with a report from multiple data sources in Northern Italy showing that HCV was the main cause of chronic liver disease in both genders, closely followed by alcohol abuse among males [22]. Furthermore, multiple causes of death analyses can contribute to understanding comorbidities in subjects with liver disease.
RI PT
This improving epidemiological scenario could be disrupted in the near future by several potential threats: HBV infection among an increasing immigrant population from high-risk countries; a bimodal distribution of anti-HCV prevalence, with the highest prevalence observed among the elderly and a second smaller peak reported in younger subjects [21]; the diffusion of binge-drinking
SC
among adolescents and young adults [23]; and the diabetes epidemic, with the associated risk of liver disease mortality [24]. In view of the above, continuous surveillance of mortality from chronic
M AN U
liver disease is warranted, including all the main nosologic categories (viral hepatitis, liver cancer,
AC C
EP
TE D
cirrhosis), and, when available, multiple causes of death data.
8
ACCEPTED MANUSCRIPT References 1. Bosetti C, Levi F, Lucchini F, Zatonski WA, Negri E, La Vecchia C. Worldwide mortality from cirrhosis: an update to 2002 J Hepatol 2007;46:827-839. 2. Blachier M, Leleu H, Peck-Radosavljevic M, Valla DC, Roudot-Thoraval F. The burden of liver disease in Europe: a review of available epidemiological data. J Hepatol 2013;58:593-608.
http://www.healthmetricsandevaluation.org/gbd/country-profiles
RI PT
3. Institute for Health Metrics and Evaluation. GBD profile: Italy. Accessed at
4. Asrani SK, Larson JJ, Yawn B, Therneau TM, Kim WR. Underestimation of liver-related mortality in the United States. Gastroenterology 2013;145:375-382.
SC
5. Capocaccia L, Capocaccia R. The rise and fall in primary liver cancer mortality in Italy. Dig Liver Dis 2002;34:606-607.
M AN U
6. Lee LT, Huang HY, Huang KC, Chen CY, Lee WC. Age-period-cohort analysis of hepatocellular carcinoma mortality in Taiwan, 1976-2005. Ann Epidemiol 2009;19:323-328. 7. Coviello E, Caputi G, Martinelli D, Germinario CA, Prato R. Mortality trends for primary liver cancer in Puglia, Italy. Eur J Cancer Prev 2010;19:417-423.
8. Matsuda T, Saika K. Trends in liver cancer mortality rates in Japan, USA, UK, France and Korea based on the WHO mortality database. Jpn J Clin Oncol 2012;42:360-361.
TE D
9. Clayton D, Schifflers E. Models for temporal variation in cancer rates. II: age-period-cohort models. Stat Med 1987;6:469-481.
10. La Vecchia C, Negri E, Pelucchi C. The rise and fall in primary liver cancer mortality in Italy. Dig Liver Dis 2002;34:169-171.
EP
11. Istituto nazionale di statistica. Istruzioni integrative per l’applicazione dell'Icd-10 nella codifica delle cause di morte. Metodi e Norme n. 35 - 2007. Accessed at
AC C
http://www3.istat.it/dati/catalogo/20080108_00/ 12. Santi V, Buccione D, Di Micoli A, Fatti G, Frigerio M, Farinati F, et al. The changing scenario of hepatocellular carcinoma over the last two decades in Italy. J Hepatol 2012;56:397-405. 13. Altekruse SF, McGlynn KA, Reichman ME. Hepatocellular carcinoma incidence, mortality, and survival trends in the United States from 1975 to 2005. J Clin Oncol 2009;27:1485-1491. 14. von Hahn T, Ciesek S, Wegener G, Plentz RR, Weismüller TJ, Wedemeyer H, et al. Epidemiological trends in incidence and mortality of hepatobiliary cancers in Germany. Scand J Gastroenterol 2011;46:1092-1098. 15. Anderson RN, Miniño AM, Hoyert DL, Rosenberg HM. Comparability of cause of death between ICD-9 and ICD-10: preliminary estimates. Natl Vital Stat Rep 2001;49:1-32. 9
ACCEPTED MANUSCRIPT 16. Istituto nazionale di statistica. Analisi del bridge coding Icd-9 - Icd-10 per le statistiche di mortalità per causa in Italia. Metodi e Norme n. 50 – 2011. Accessed at http://www3.istat.it/dati/catalogo/20111020_01/ 17. Fedeli U, Schievano E, Lisiero M, Avossa F, Mastrangelo G, Saugo M. Descriptive epidemiology of chronic liver disease in northeastern Italy: an analysis of multiple causes of
RI PT
death. Popul Health Metr 2013;11:20. 18. Wise M, Bialek S, Finelli L, Bell BP, Sorvillo F. Changing trends in hepatitis C-related mortality in the United States, 1995-2004. Hepatology 2008;47:1128-1135.
19. Paula H, Asrani SK, Boetticher NC, Pedersen R, Shah VH, Kim WR. Alcoholic liver disease-
SC
related mortality in the United States: 1980-2003. Am J Gastroenterol 2010;105:1782-1787. 20. Ly KN, Xing J, Klevens RM, Jiles RB, Ward JW, Holmberg SD. The increasing burden of
2012;156:271-278.
M AN U
mortality from viral hepatitis in the United States between 1999 and 2007. Ann Intern Med
21. Fabris P, Baldo V, Baldovin T, Bellotto E, Rassu M, Trivello R, et al. Changing epidemiology of HCV and HBV infections in Northern Italy: a survey in the general population. J Clin Gastroenterol 2008;42:527-532.
22. Zani C, Pasquale L, Bressanelli M, Puoti M, Paris B, Coccaglio R, et al. The epidemiological
TE D
pattern of chronic liver diseases in a community undergoing voluntary screening for hepatitis B and C. Dig Liver Dis 2011;43:653-658.
23. D'Alessio M, Baiocco R, Laghi F. The problem of binge drinking among Italian university students: a preliminary investigation. Addict Behav 2006;31:2328-2333.
EP
24. De Marco R, Locatelli F, Zoppini G, Verlato G, Bonora E, Muggeo M. Cause-specific mortality
AC C
in type 2 diabetes. The Verona Diabetes Study. Diabetes Care 1999;22:756-761.
10
ACCEPTED MANUSCRIPT Figure legends
Figure 1. Crude mortality rates from liver cirrhosis (ICD-9 571 / ICD-10 K70, K73, K74), liver
RI PT
cancer (ICD-9 155 / ICD-10 C22), viral hepatitis (ICD-9 070 / ICD-10 B15-B19), all liver diseases (cirrhosis + cancer + viral hepatitis), and all liver diseases extended (previous codes plus ICD-9 570, 572, 573, 4560, 4561 / ICD-10 K71, K72, K75, K76, I85): males and females of the Veneto
M AN U
SC
Region (Italy), 1995–2010.
Figure 2. Time trends in liver cirrhosis and liver cancer mortality by age class (rates x 100,000
TE D
population, logarithmic scale on the y axis): Veneto Region (Italy), 1995-2010.
Figure 3. Age-cohort analysis of liver disease mortality in males and females. Rate ratio (reference = mean year of birth equal to 1920) with bars corresponding to 95% confidence intervals; dotted
AC C
EP
line = all causes mortality.
11
ACCEPTED MANUSCRIPT Table 1. Age-specific and age-standardized rates, and proportional mortality from all liver diseases: Veneto Region (Italy), 1995–2010. 1999–2002
2003–2006
2007–2010
6.6 25.9 66.4 137.4 191.6 183.4
6.2 21.4 59.3 122.4 182.7 160.5
5.1 21.6 51.2 107.0 171.1 170.6
4.0 15.2 43.5 96.4 162.9 174.2
39.5 19.8 44.8
37.6 17.6 40.4
36.1 15.9 37.2
5.6% 8.7% 8.5% 6.6% 3.5% 1.2% 4.2%
AC C
EP
Proportional mortality (%) 35–44 yrs 45–54 yrs 55–64 yrs 65–74 yrs 75–84 yrs 85+ yrs All ages
SC
M AN U
Crude rate World standard Veneto population 2007
6.0% 8.1% 8.6% 6.7% 3.8% 1.1% 4.0%
TE D
Rates (x105) 35–44 yrs 45–54 yrs 55–64 yrs 65–74 yrs 75–84 yrs 85+ yrs
5.9% 9.2% 8.6% 6.9% 3.8% 1.1% 4.0%
∆ 07–10 vs.95– 98
RI PT
1995–1998
-39% -41% -34% -30% -15% -5%
33.9 13.8 33.5
-14% -30% -25%
5.4% 7.6% 8.1% 7.1% 3.8% 1.2% 3.7%
-3% -13% -5% +7% +11% +1% -11%
ACCEPTED MANUSCRIPT Table 2. Mortality from all liver diseases, Veneto Region (Italy), 2008–2010: number of deaths, mortality rates, ratio to liver cirrhosis mortality (RR) by gender and age class, and by gender and etiology retrieved from multiple causes of death analysis.
rate (x10 )
RR vs. cirr
rate (x105)
RR vs. cirr
7 27 64 161 375 762 382 1778
0.3 2.2 6.2 18.0 47.0 123.1 143.7 24.0
2.3 1.9 1.3 1.7 2.7 3.9 4.5 3.1
2.2 7.7 0.5
1.1 20.3 17.0
AC C
EP
TE D
M AN U
Age
S1047-2797(14)00156-2
DOI:
10.1016/j.annepidem.2014.05.004
Reference:
AEP 7654
To appear in:
Annals of Epidemiology
Received Date: 14 February 2014 Revised Date:
17 April 2014
Accepted Date: 6 May 2014
Please cite this article as: Fedeli U, Avossa F, Guzzinati S, Bovo E, Saugo M, Trends in mortality from chronic liver disease, Annals of Epidemiology (2014), doi: 10.1016/j.annepidem.2014.05.004. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Trends in mortality from chronic liver disease
SER – Epidemiological Department, Veneto Region, Italy
2
Veneto Cancer Registry, Italy
M AN U
SC
1
RI PT
Ugo Fedeli1, Francesco Avossa1, Stefano Guzzinati2, Emanuela Bovo2, Mario Saugo1
Corresponding author: Ugo Fedeli
Passaggio Gaudenzio, 1 - 35131 Padova (PD), Italy Tel:
+39 049 8778252
Fax +39 049 8778235
AC C
Word count: 2353
EP
TE D
e-mail: [email protected]
Abstract word count: 200 Tables 2 Figures: 3
1
ACCEPTED MANUSCRIPT ABSTRACT Background: Mortality from liver cirrhosis has sharply declined in most countries in the last few decades, but includes only a fraction of deaths due to chronic liver disease.
RI PT
Methods: Mortality records for liver cirrhosis, liver cancer, viral hepatitis, and the three disease categories combined were extracted for the period 1995–2010 in the Veneto Region (North-Eastern Italy). The presence of a birth cohort effect was assessed by fitting Poisson regression models. The alcoholic or viral etiology was retrieved through multiple causes of death analysis for the years 2008-2010.
SC
Results: Liver cirrhosis mortality represented only one-third of all liver disease deaths; this proportion was even smaller for deaths with a viral etiology and among elderly people. Through the study period age-standardized rates declined by 50% for liver cirrhosis, and by 30% (world standard) or by 25% (regional standard) for all liver diseases; the reduction in proportional mortality was limited to 11%. Mortality rates started to decline in subjects born after the early 1920s.
M AN U
Conclusion: The decline in mortality was smaller than that reported by standard statistics for chronic liver disease; these statistics should routinely include viral hepatitis and liver cancer deaths. When available, multiple causes of death data should be analyzed.
TE D
Keywords
AC C
EP
Mortality; liver cancer; liver cirrhosis; viral hepatitis
List of abbreviations and acronyms HCV HBV
hepatitis C virus hepatitis B virus
ICD
International Classification for Diseases
HCC
hepatocellular carcinoma
2
ACCEPTED MANUSCRIPT Introduction In the last few decades mortality from cirrhosis has been decreasing in most countries around the world. With the exception of the increasing mortality recorded in the UK and Eastern Europe, steeply declining rates have been reported in Southern and Western Europe, North America, Oceania and some Asiatic countries [1]. A reduction in the prevalence of alcoholic liver disease, a
RI PT
fall in the transmission of hepatitis C virus (HCV) infection, and vaccination campaigns against hepatitis B virus (HBV) have all contributed to the observed decrease in mortality in Southern Europe [2]. In Italy, a greater than 5% yearly decline in cirrhosis mortality rates has been estimated for the period 1984–2002 among males, with only a slightly less favorable trend observed in
SC
females [1]. According to the Global Burden of Disease profile for Italy, cirrhosis has dropped from the 5th to the 8th most common cause of premature death, due to a 50% reduction in years of life lost
M AN U
from 1990 to 2010 [3].
However, liver cirrhosis mortality captures only a limited fraction of deaths due to chronic liver disease. In the US, by adopting the usual definition for chronic liver disease mortality, a 38% decrease in rates was observed from 1979 to 2008; if other liver-related causes of death including viral hepatitis and liver cancer were also considered, mortality rates remained essentially unchanged throughout the study period [4]. Chronic viral hepatitis, liver cirrhosis, and hepatocellular
TE D
carcinoma constitute a pathogenic continuum and could therefore be usefully merged in a broader definition of liver-related deaths. However, further intricacies in interpreting mortality trends could arise from coding practices and rules for liver cancer deaths [5], and from the strong birth cohort
EP
effects observed for liver cancer mortality in several countries [6-8]. In Italy healthcare, including disease surveillance and preventive services, is administered by
AC C
Regions based on general rules set at the national level. The aim of the study was to examine time trends of liver-related deaths in the Veneto Region (North-Eastern Italy) by addressing different methodological issues that impact on liver mortality statistics: restrictive or broader inclusion of codes across subsequent versions of the International Classification for Diseases (ICD), analysis of age-specific or age-standardized rates, birth-cohort effects, and the multiple causes of death approach.
3
ACCEPTED MANUSCRIPT Methods The Veneto region has about 4,900,000 inhabitants and is subdivided into 21 Local Health Units (LHUs) and 581 municipalities. Death certificates are sent by each municipality to the LHU; until 2005 causes of death were coded in each LHU according to the ICD, 9th Edition (ICD-9). Since 2006, LHUs have transmitted a copy of the certificate to the Regional Epidemiological Department
RI PT
for coding of the underlying cause (from 2007 in ICD-10) and entry into the electronic regional archive of mortality data. The underlying cause is the single condition usually reported in mortality statistics, identified from all diseases mentioned in the death certificate according to internationally adopted selection and modification rules. Since 2008 the regional archive has included ICD codes
SC
not only for the underlying cause, but also for all diseases reported in the certificate.
From the mortality records of residents in the Veneto Region in the period 1995–2010, the
M AN U
following underlying causes of death were extracted: liver cirrhosis (ICD-9 571; ICD-10 K70, K73, K74), liver cancer (ICD-9 155; ICD-10 C22), and viral hepatitis (ICD-9 070; ICD-10 B15-B19). The above three disease categories were thereafter merged as “liver disease deaths”; a broader definition reported in a recent US study [4] was adopted including deaths from liver failure, other and unspecified liver disease, and esophageal varices (“all liver disease extended”, the above codes plus ICD-9 570, 572, 573, 4560, 4561; ICD-10 K71, K72, K75, K76, I85). Limited to the period
TE D
2008–2010, liver disease deaths were classified by etiology, retrieving ICD-10 codes for alcoholic disease (codes for alcoholic liver disease and for mental and behavioral disorders due to the use of alcohol, K70, F10), and for acute or chronic HBV or HCV infection (B16, B18.0, B18.1 and B17.1,
EP
B18.2, respectively) from all conditions reported in the certificate. Crude and age-specific mortality rates were computed for the above disease categories across
AC C
calendar periods (1995–1998, 1999–2002, 2003–2006, 2007–2010) as the number of deaths per 100,000 population, with population data derived from the National Institute for Statistics (http://demo.istat.it/). Age-standardized rates were obtained by the direct standardization method, taking as reference either the World standard population, or the regional population in 2007. The proportional mortality in each age class was defined as the percentage of liver disease deaths out of all registered deaths. Lastly, to assess the presence of a birth cohort effect, liver disease death rates among subjects aged 45–94 were classified in 5-year age and calendar period classes, and the corresponding birth cohort was identified by the mid-cohort year of birth. Death rates were the dependent variable in different Poisson regression models including as explanatory variables age alone, age and period, age and cohort, or age, period and cohort. Such models were compared in both genders using the likelihood ratio test and the Akaike information criterion [9]. 4
ACCEPTED MANUSCRIPT
Results The overall number of registered deaths in 1995–2010 was 12,741 for liver cirrhosis, 1,226 for viral hepatitis, 12,883 for liver cancer, 26,850 for all liver diseases (the previous categories combined),
RI PT
and 27,684 when adopting the broader definition (all liver diseases extended). Figure 1 shows time trends in crude mortality rates for the above five disease categories by gender: mortality due to liver cirrhosis steeply declined, liver cancer rates were relatively stable, and viral hepatitis became an important cause of death during the study period. As a consequence, all liver diseases mortality
SC
showed a limited decrease; the inclusion of additional causes of death did not change this finding, and was not considered in further analyses. Time trends were similar between genders, but among
more pronounced than among males.
M AN U
females the relative reduction in liver cirrhosis and the increase in viral hepatitis mortality were
Liver cirrhosis death rates dropped by about 50% from 1995–1998 to 2007–2010 across all age classes (Figure 2); in contrast liver cancer rates displayed different patterns by age, with a limited decrease in younger subjects, and almost no change among the elderly. Table 1 shows mortality figures for all liver diseases combined: there was a 40% decline in age-
TE D
specific mortality rates during the study period in subjects younger than 55 years, but only a 15% decline in those aged 75–84 years, while rates remained almost unchanged among the very elderly. Given the diverging time patterns, the reference adopted for age-standardization influenced the
EP
overall estimated decrease, which was thus larger with a reference giving more weight to younger classes (World standard) than with an internal reference reflecting the ageing regional population. Liver diseases accounted for a substantial proportion of deaths among subjects aged 45–74 years;
AC C
the overall proportional mortality declined by about 10% (from 4.2% in 1995–1998 to 3.7% in 2007–2010).
In contrast, if analyses were limited to liver cirrhosis deaths, age-standardized rates halved during the study period irrespective of the reference population adopted, and the proportional mortality decreased by 40% (from 2.3% to 1.4% of overall deaths, data not shown). Among the Poisson regression models performed, the most informative was the age-cohort model (the one with the minimum Akaike information criterion value): if the period effect was added to the explanatory variables, the model fit did not improve. Figure 3 summarizes the results of the agecohort analysis: liver disease mortality remained high among people born in the first two decades of 5
ACCEPTED MANUSCRIPT the last century, and started to decline in both genders only for birth cohorts born after the early 1920s. This pattern was quite different from that observed for overall mortality, which had already begun to decline in older birth cohorts. Among females, the relative reduction in liver mortality among cohorts born after 1920 seemed to be steeper than that estimated for overall mortality. Table 2 shows liver disease mortality rates by gender and age class for the period 2008–2010, when
RI PT
multiple causes of death were also available, allowing the etiology of the disease to be retrieved; the ratio to mortality rates estimated by analyses limited to liver cirrhosis deaths is also provided. Death rates rose with age in both genders, but the male to female ratio, which was about four in younger subjects, declined among the elderly. The etiology of liver disease was reported in about 40% of
SC
deaths, being mainly HCV among females and alcohol among males. Mortality rates were about three times those estimated by liver cirrhosis deaths alone, with a higher rate ratio among older age
M AN U
classes; furthermore, whereas alcoholic liver disease is included among ICD-10 codes for cirrhosis (K70), viral hepatitis is in another ICD-10 chapter, and standard mortality statistics greatly underestimated deaths with a viral etiology.
Discussion
TE D
The present findings support the inclusion of viral hepatitis and liver cancer deaths among routine statistics of chronic liver disease mortality [4]. With respect to liver cirrhosis mortality, rates were about three times higher, depending on etiology. Mortality due to liver disease decreased in 1995–
EP
2010 in North-Eastern Italy, but the extent of this reduction differed according to the adoption of a more or less restrictive selection of ICD codes: age-standardized rates (World standard) declined during the study period by 50% for liver cirrhosis mortality, and by 30% when examining all liver
AC C
disease mortality. Moreover, since time trends were less evident in older age classes, the reduction declined to 25% when rates were standardized using the regional population as a reference. When the proportional mortality (share of total mortality) was analyzed, the overall reduction was limited to 11%, and liver disease deaths still accounted for a substantial fraction of overall mortality among subjects aged 45–74 years. The fall in liver cirrhosis mortality analyzed alone was large and homogeneous across age classes. This finding only in part reflects a real epidemiological trend, possibly being amplified by the following mechanisms: strict surveillance and increased diagnostic rate of hepatocellular carcinoma (HCC) among cirrhotic patients; better control of complications of cirrhosis with improved survival and risk of HCC development at older ages [10]; adoption of the ICD-10 with related coding 6
ACCEPTED MANUSCRIPT instructions and specification [11], with an increasing number of liver disease deaths classified as chronic viral hepatitis. According to data from an Italian multicentric study, about 96% of HCC cases were associated with cirrhosis [12]. Therefore, the inclusion of liver cancer mortality could take into account the above-described diagnostic shift and allow for a better evaluation of time trends. Moreover, the overall analysis of all liver diseases classified as chronic, viral, or neoplastic
RI PT
is likely to be more robust to coding discontinuities in mortality statistics. It should be noted that according to the ICD-9, which was adopted in the earlier study periods, liver cancer deaths were coded as primary (1550), of intrahepatic bile ducts (1551), and unknown if primary or metastatic (1552). As a consequence, earlier mortality studies focusing on primary liver
SC
cancer adopted different strategies: the inclusion of all 155 codes [6], of all liver cancers other than intrahepatic bile ducts [13], or more often only of those specified as primary [7,10]. The latter
M AN U
option however could be affected by time shifts in the adoption of the less specific 1552 code instead of the 1550 code [5]; the same holds for the inclusion/exclusion of intrahepatic cholangiocarcinoma, mortality from which is reported to be steeply increasing in some European countries [14]. Although entailing a possible misclassification of some tumors metastatic to the liver, the most feasible choice in evaluating time trends across different ICD versions appears to be the inclusion of all 155 codes; this strategy is also consistent with reports examining the
TE D
comparability between ICD-9 and ICD-10 in different countries [15,16]. Mortality from all liver diseases displays wide variation in time trends by age, supporting the presence of a strong birth cohort effect. The steep decline in mortality in cohorts born after the early
EP
1920s suggests that the downward time trend, although less favorable than that reported using standard liver cirrhosis mortality statistics, could be a long-term tendency that will continue into subsequent decades. Such a birth cohort effect has already been reported for primary liver cancer,
AC C
and is usually attributed to the HCV epidemic in Italy in the 1940s that mainly arose due to contaminated medical instruments [7,12]. The birth cohort effect in our data extended to mortality for all liver diseases, and was similar in females, in whom according to the multiple causes of death analysis the viral etiology largely prevailed – possibly due to a higher exposures to medical procedures –, and males, in whom a major role was also played by alcohol abuse. Multiple causes of death data can be analyzed to investigate liver disease mortality by etiology in different population subgroups [17]; where available for long time periods, they have been usefully employed to examine trends in mortality related to viral hepatitis and alcoholic liver disease [1820]. Such analyses are limited by the certifying physician’s knowledge of the etiology of the disease, and by problems of underreporting. However, the present results are consistent with 7
ACCEPTED MANUSCRIPT seroepidemiologic data showing a higher prevalence of anti-HCV antibodies in subjects aged ≥75 years [21], and with a report from multiple data sources in Northern Italy showing that HCV was the main cause of chronic liver disease in both genders, closely followed by alcohol abuse among males [22]. Furthermore, multiple causes of death analyses can contribute to understanding comorbidities in subjects with liver disease.
RI PT
This improving epidemiological scenario could be disrupted in the near future by several potential threats: HBV infection among an increasing immigrant population from high-risk countries; a bimodal distribution of anti-HCV prevalence, with the highest prevalence observed among the elderly and a second smaller peak reported in younger subjects [21]; the diffusion of binge-drinking
SC
among adolescents and young adults [23]; and the diabetes epidemic, with the associated risk of liver disease mortality [24]. In view of the above, continuous surveillance of mortality from chronic
M AN U
liver disease is warranted, including all the main nosologic categories (viral hepatitis, liver cancer,
AC C
EP
TE D
cirrhosis), and, when available, multiple causes of death data.
8
ACCEPTED MANUSCRIPT References 1. Bosetti C, Levi F, Lucchini F, Zatonski WA, Negri E, La Vecchia C. Worldwide mortality from cirrhosis: an update to 2002 J Hepatol 2007;46:827-839. 2. Blachier M, Leleu H, Peck-Radosavljevic M, Valla DC, Roudot-Thoraval F. The burden of liver disease in Europe: a review of available epidemiological data. J Hepatol 2013;58:593-608.
http://www.healthmetricsandevaluation.org/gbd/country-profiles
RI PT
3. Institute for Health Metrics and Evaluation. GBD profile: Italy. Accessed at
4. Asrani SK, Larson JJ, Yawn B, Therneau TM, Kim WR. Underestimation of liver-related mortality in the United States. Gastroenterology 2013;145:375-382.
SC
5. Capocaccia L, Capocaccia R. The rise and fall in primary liver cancer mortality in Italy. Dig Liver Dis 2002;34:606-607.
M AN U
6. Lee LT, Huang HY, Huang KC, Chen CY, Lee WC. Age-period-cohort analysis of hepatocellular carcinoma mortality in Taiwan, 1976-2005. Ann Epidemiol 2009;19:323-328. 7. Coviello E, Caputi G, Martinelli D, Germinario CA, Prato R. Mortality trends for primary liver cancer in Puglia, Italy. Eur J Cancer Prev 2010;19:417-423.
8. Matsuda T, Saika K. Trends in liver cancer mortality rates in Japan, USA, UK, France and Korea based on the WHO mortality database. Jpn J Clin Oncol 2012;42:360-361.
TE D
9. Clayton D, Schifflers E. Models for temporal variation in cancer rates. II: age-period-cohort models. Stat Med 1987;6:469-481.
10. La Vecchia C, Negri E, Pelucchi C. The rise and fall in primary liver cancer mortality in Italy. Dig Liver Dis 2002;34:169-171.
EP
11. Istituto nazionale di statistica. Istruzioni integrative per l’applicazione dell'Icd-10 nella codifica delle cause di morte. Metodi e Norme n. 35 - 2007. Accessed at
AC C
http://www3.istat.it/dati/catalogo/20080108_00/ 12. Santi V, Buccione D, Di Micoli A, Fatti G, Frigerio M, Farinati F, et al. The changing scenario of hepatocellular carcinoma over the last two decades in Italy. J Hepatol 2012;56:397-405. 13. Altekruse SF, McGlynn KA, Reichman ME. Hepatocellular carcinoma incidence, mortality, and survival trends in the United States from 1975 to 2005. J Clin Oncol 2009;27:1485-1491. 14. von Hahn T, Ciesek S, Wegener G, Plentz RR, Weismüller TJ, Wedemeyer H, et al. Epidemiological trends in incidence and mortality of hepatobiliary cancers in Germany. Scand J Gastroenterol 2011;46:1092-1098. 15. Anderson RN, Miniño AM, Hoyert DL, Rosenberg HM. Comparability of cause of death between ICD-9 and ICD-10: preliminary estimates. Natl Vital Stat Rep 2001;49:1-32. 9
ACCEPTED MANUSCRIPT 16. Istituto nazionale di statistica. Analisi del bridge coding Icd-9 - Icd-10 per le statistiche di mortalità per causa in Italia. Metodi e Norme n. 50 – 2011. Accessed at http://www3.istat.it/dati/catalogo/20111020_01/ 17. Fedeli U, Schievano E, Lisiero M, Avossa F, Mastrangelo G, Saugo M. Descriptive epidemiology of chronic liver disease in northeastern Italy: an analysis of multiple causes of
RI PT
death. Popul Health Metr 2013;11:20. 18. Wise M, Bialek S, Finelli L, Bell BP, Sorvillo F. Changing trends in hepatitis C-related mortality in the United States, 1995-2004. Hepatology 2008;47:1128-1135.
19. Paula H, Asrani SK, Boetticher NC, Pedersen R, Shah VH, Kim WR. Alcoholic liver disease-
SC
related mortality in the United States: 1980-2003. Am J Gastroenterol 2010;105:1782-1787. 20. Ly KN, Xing J, Klevens RM, Jiles RB, Ward JW, Holmberg SD. The increasing burden of
2012;156:271-278.
M AN U
mortality from viral hepatitis in the United States between 1999 and 2007. Ann Intern Med
21. Fabris P, Baldo V, Baldovin T, Bellotto E, Rassu M, Trivello R, et al. Changing epidemiology of HCV and HBV infections in Northern Italy: a survey in the general population. J Clin Gastroenterol 2008;42:527-532.
22. Zani C, Pasquale L, Bressanelli M, Puoti M, Paris B, Coccaglio R, et al. The epidemiological
TE D
pattern of chronic liver diseases in a community undergoing voluntary screening for hepatitis B and C. Dig Liver Dis 2011;43:653-658.
23. D'Alessio M, Baiocco R, Laghi F. The problem of binge drinking among Italian university students: a preliminary investigation. Addict Behav 2006;31:2328-2333.
EP
24. De Marco R, Locatelli F, Zoppini G, Verlato G, Bonora E, Muggeo M. Cause-specific mortality
AC C
in type 2 diabetes. The Verona Diabetes Study. Diabetes Care 1999;22:756-761.
10
ACCEPTED MANUSCRIPT Figure legends
Figure 1. Crude mortality rates from liver cirrhosis (ICD-9 571 / ICD-10 K70, K73, K74), liver
RI PT
cancer (ICD-9 155 / ICD-10 C22), viral hepatitis (ICD-9 070 / ICD-10 B15-B19), all liver diseases (cirrhosis + cancer + viral hepatitis), and all liver diseases extended (previous codes plus ICD-9 570, 572, 573, 4560, 4561 / ICD-10 K71, K72, K75, K76, I85): males and females of the Veneto
M AN U
SC
Region (Italy), 1995–2010.
Figure 2. Time trends in liver cirrhosis and liver cancer mortality by age class (rates x 100,000
TE D
population, logarithmic scale on the y axis): Veneto Region (Italy), 1995-2010.
Figure 3. Age-cohort analysis of liver disease mortality in males and females. Rate ratio (reference = mean year of birth equal to 1920) with bars corresponding to 95% confidence intervals; dotted
AC C
EP
line = all causes mortality.
11
ACCEPTED MANUSCRIPT Table 1. Age-specific and age-standardized rates, and proportional mortality from all liver diseases: Veneto Region (Italy), 1995–2010. 1999–2002
2003–2006
2007–2010
6.6 25.9 66.4 137.4 191.6 183.4
6.2 21.4 59.3 122.4 182.7 160.5
5.1 21.6 51.2 107.0 171.1 170.6
4.0 15.2 43.5 96.4 162.9 174.2
39.5 19.8 44.8
37.6 17.6 40.4
36.1 15.9 37.2
5.6% 8.7% 8.5% 6.6% 3.5% 1.2% 4.2%
AC C
EP
Proportional mortality (%) 35–44 yrs 45–54 yrs 55–64 yrs 65–74 yrs 75–84 yrs 85+ yrs All ages
SC
M AN U
Crude rate World standard Veneto population 2007
6.0% 8.1% 8.6% 6.7% 3.8% 1.1% 4.0%
TE D
Rates (x105) 35–44 yrs 45–54 yrs 55–64 yrs 65–74 yrs 75–84 yrs 85+ yrs
5.9% 9.2% 8.6% 6.9% 3.8% 1.1% 4.0%
∆ 07–10 vs.95– 98
RI PT
1995–1998
-39% -41% -34% -30% -15% -5%
33.9 13.8 33.5
-14% -30% -25%
5.4% 7.6% 8.1% 7.1% 3.8% 1.2% 3.7%
-3% -13% -5% +7% +11% +1% -11%
ACCEPTED MANUSCRIPT Table 2. Mortality from all liver diseases, Veneto Region (Italy), 2008–2010: number of deaths, mortality rates, ratio to liver cirrhosis mortality (RR) by gender and age class, and by gender and etiology retrieved from multiple causes of death analysis.
rate (x10 )
RR vs. cirr
rate (x105)
RR vs. cirr
7 27 64 161 375 762 382 1778
0.3 2.2 6.2 18.0 47.0 123.1 143.7 24.0
2.3 1.9 1.3 1.7 2.7 3.9 4.5 3.1
2.2 7.7 0.5
1.1 20.3 17.0
AC C
EP
TE D
M AN U
Age
