Burden of Liver Cancer Attributable to Five Risk Factors by Specic Etiologies in 204 Countries and Territories

Background Liver cancer has caused a serious disease burden worldwide wherein prognostic risk factors may have had a signicant role. However, previous studies have focused on liver cancer distributions and trend variations without analyzing the role of risk factors. Methods We assessed liver cancer burden attributable to ve recognized risk factors as well as temporal trends from 1990 to 2019 based on the 2019 Global Burden of Disease Study. The risk exposure levels and their contributed burden were estimated by cause, sex, age, and location. Furthermore, the relationship between the risks and social-demographic index (SDI) was investigated. Results

There are ve types of LC by different etiologies, namely liver cancer due to alcohol use (LCAU), liver cancer due to NASH (LCNA), liver cancer due to hepatitis B (LCHB), liver cancer due to hepatitis C (LCHC), and liver cancer due to other causes (LCOC). Five recognized risk factors, including alcohol use, drug use, high body mass index (BMI), high fasting plasma glucose, and smoking, are closely related to LC burden through a comparative risk assessment framework. Summary exposure value (SEV) was applied to estimate the population level exposure, which is a measure of a population's risk-weighted exposure or the risk-weighted prevalence of a separate risk factor [7]. The age-standardized SEV rate was used in this research, which ranges from 0% (no risk of exposure) to 100% (maximum risk possibility for the entire population).

Estimation Of Lc Burden Attributable To Risk Factors
The indicators used to estimate LC burden included the number of deaths, years of life lost, years lived with disability, and DALYs [8]. The ve risk factors and ve types of LC constituted several risk-outcome pairs, and for a given risk-outcome pair, the risk exposure and the related cancer burden were estimated by age, gender, region, and social development.
Social development was evaluated using the social-demographic index (SDI), which was calculated based on fertility rate, average schooling years (for people aged ≥ 15 years old), and per capita after-tax income distribution. Each GBD location has an annual SDI score, and countries are divided into ve SDI quintiles (high, high-medium, medium, medium-low, and low levels). The relationship between a location's risk attributed LC burden and the corresponding SDI value was investigated.

Statistical Analyses
All indicator estimates were presented with 95% uncertainty intervals (UIs) to ensure that all the sources of uncertainty have been captured. A 95% UI excluding 0 was considered statistically signi cant. All statistical analyses were performed using the R program (Version 4.0.2, R core team).

Risk factors attributable to LC burden
In 2019, there were 484,577 (95% UI: 444,091 to 525,798) LC deaths and 12,528,422 (11,400,671 to 13,687,675) LC DALYs globally, with nearly half of these being attributed to the ve risk factors. Among the 254.7 thousand risk attributed LC deaths, alcohol use accounted for 30.4%, smoking accounted for 27.2%, drug use accounted for 20.6%, high BMI accounted for 20.6%, and high fasting plasma glucose accounted for the remaining 1.3% (Table 1). In 2019, the ASR of the total risk attributed LC deaths and DALYs were 3.1 (2.7 to 3.6) per 100,000 persons and 75.0 (63.0 to 88.2) per 100,000 persons, respectively.      (Fig. 1).

Global Lc Burden By Sex And Age Group
The LC burden in males (120.9) was more than three times that in females (32.5) in 2019, in terms of age-standardized DALY rates. From 1990 to 2019, the LC burden in females and males demonstrated a trend of rst increasing and then decreasing.
The general ranking of LC burden due to risk factors was similar in both sexes. However, drug use attributed LC burden has been increasing in females from 1990, and then decreased in the past decade, and the alcohol use attributed burden has gradually decreased from 1990 to 2019 (Fig. 2).
The LC burden was mainly concentrated in the 45 to 69 years old age group with the heaviest in the 60 to 64 years old for males and the 65 to 69 years old for females (Fig. 3). Overall, LC burden increased with age from 20 to 60 years old and decreased with age after 60 years old. Among patients over 50 years old, LC burden attributable to drug use increased and burden attributable to high BMI decreased in males.

Regional And National Lc Burden
In 1990, the risk attributed LC burden was the highest in East Asia, followed by high-income Asia Paci c, and the lowest in Central Sub-Saharan Africa and Southern Latin America, in terms of DALYs. With respect to speci c etiologies, LCHB burden was the highest in East Asia and Oceania; LCHC burden was the highest in high-income Asia Paci c, North Africa and the Middle East; and LCAU burden was the highest in the 17 remaining GBD regions. The highest risk attributed LC burden was observed in Mongolia, Sierra Leone, and Burkina Faso.
By contrast, Central Asia and High-income Asia Paci c had the highest risk attributable LC burden in 2019, and the lowest burden was in Central Sub-Saharan Africa and Eastern Sub-Saharan Africa. LCHB burden was still the highest in East Asia; LCHC burden was the highest in high-income Asia Paci c, North Africa and the Middle East and Oceania; and LCAU burden was the highest in the 17 remaining GBD regions (Fig. 4). The three countries with the heaviest risk-related LC burden were Mongolia, Gambia, and Tonga in 2019.

Trends Of Risk-related Lc Burden
Globally, total risk factors attributed LC burden increased from 1990 to 2000, then decreased from 2001 to 2005, and remained stable. In low SDI regions, LC burden remained stable, while it rst increased and then gradually stabilized in high SDI regions, and other regions remained consistent with global trend (Fig. 5).
LC burden due to high BMI, and high fasting plasma glucose continued to grow from 2005 to 2019. Burden due to alcohol use and drug use fell after 2000, and rose slightly from 2005 to 2019, while burden due to smoking continued to reduce from 2005, in terms of ASDR. For alcohol use, drug use, and high BMI, the related LC burden was the highest in the high SDI region, and was the lowest in the low SDI region. For high fasting plasma glucose and smoking, middle SDI region had the highest burden.

Relationship Between Risk-related Lc Burden And Social Development
In general, the SEVs of drug use and high BMI increased with SDI, whereas the alcohol use and smoking exposure levels rst increased with increasing SDI and decreased when the SDI became > 0.8. In particular, the high fasting plasma glucose SEV decreased with an increase in the SDI value from 0.5 to 0.8 (Fig. 6).
In terms of the DALY rates, the relationship between the overall LC burden with SDI was roughly similar to that of LC by different causes. The risk-related burden rst increased as SDI increased from 0.2 to 0.5, then decreased as SDI ranged from 0.5 to 0.7, and increased again after the SDI value became > 0.7 (Supplementary Figure S3).

Discussion
The worldwide LC burden remains serious, and several prognostic risk factors have signi cantly contributed to this over the years. In the present study, we evaluated exposure levels of the ve recognized risk factors, the risk-related LC burden, and temporal trends from 1990 to 2019, as well as the speci c contribution to each LC subtype. Generally, the ve risk factors contributed to nearly half of the total LC burden, with alcohol use as the greatest risk factor, followed by smoking. The increase in risk attributed LC burden showed signi cant differences among different sexes, age groups, and SDI regions. In addition, exposure level, attributable LC burden, and SDI were signi cantly correlated. Therefore, recognizing risk factors that increase LC burden is important in improving the disease prognosis.
Previous studies have focused on LC burden by different etiologies, whereas our study shifted the emphasis to the risk factors that increased LC burden. Consistent with previous literature, the distribution of risk attributed LC burden is similar to the total LC burden, with East Asia exhibiting the highest burden among the regions worldwide where HBV is an epidemic [9]. These ndings demonstrated that the measures taken to reduce HCV and HBV infection have achieved good results. Among the risk factors, drug use has a certain in uence [10]. Nevertheless, a lag effect may have been existing after the implementation of the controlling policies. High fasting plasma glucose was the risk factor with the least contribution, exhibiting a difference only in LCNA and LCOC [11]. Although the change is minimal, burden due to high fasting plasma glucose has been increasing over the years, especially in the high SDI regions [12]. From here on, we will discuss the three risk factors with the greatest contribution to LC burden.
Smoking is a risk factor especially in LCHB and LCNA. Large cohort studies have shown that smoking was signi cantly associated with impaired quality of life in HCC patients [13] and smoking history was an independent prognostic factor in HBV/HCV-infected HCC patients [14]. Moreover, other studies indicated that smoking was associated with increased non-graft related mortality after liver transplantation [15]. Possible explanations to date included the effect of smoking on increased brosis, age, tness, and compliance difference between the two groups [16]. The smoking-related LC burden has declined in the last decade owing to aggressive smoking cessation and tobacco control policies in various countries worldwide [17].
However, the contribution of smoking to LC burden is still non-negligible, especially among middle-aged and elderly. Taken together, these ndings strongly suggest that smoking cessation should be considered as an important element of healthy living habits that must be fully implemented.
Large-scale cohort studies have demonstrated the association between high BMI and the prognosis of LC [18]. A signi cant increase in high BMI attributed LC mortality was observed in our study, which was consistent with the rising exposure level of high BMI. In this case, the risks of LC morbidity and mortality due to high BMI would continue to increase without effective interventions [19]. Moreover, the reduction in LC burden due to the national prevention and control of HBV and HCV infections might be offset by high BMI, especially in developed countries, where LC exhibited a prominent upward trend and high BMI exposure may have worsened the problem [19]. Therefore, it is imperative to adopt policies to prevent and control obesity including governments emphasizing the seriousness of the obesity problem, increasing related research funding, and raising public awareness [20].
Alcohol use and obesity have a synergistic role in the development of liver diseases [15]. Alcohol consumption is not only an etiology but also a vital prognostic factor for LC, accounting for nearly 15% of LC cases and 25% of risk-related DALYs worldwide. Our study indicated that alcohol use attributed LC deaths doubled from 1990 to 2019, whereas the related ASDR slightly decreased during this period. This discrepancy might be the result of the acceleration of population aging worldwide [21,22]. Globally, alcohol consumption is roughly consistent with the distribution of alcohol use-related LC burden [24]. However, in some other countries like China, a converse association between alcohol consumption and related LC burden was observed. This mismatch might be explained by the following reasons: (1) these countries are dominated by LCHB, where alcohol use has little effect; (2) LC development is a long process, and alcohol consumption has increased in these countries only in recent years, with its effect not being fully demonstrated yet.
Several limitations need to be addressed in this study. First, there may be some prognostic risk factors that have a role in LC but are not yet recognized and thus were not included in this research. Second, the interactions between these risk factors are not taken into account in the estimation of LC burden, leading to the possibility of overestimation. Third, there was no pathology-related analysis in our study owing to the lack of pathological information in the database. Regardless of these de ciencies, the current study has provided the most comprehensive and in-depth analysis of global LC burden and its related prognostic risk factors, lling in the gaps in previous studies.

Conclusion
In summary, the recognized risk factors signi cantly contributed to the global burden of LC. On the one hand, the total LC burden attributable to the risk factors has gone through a process of increasing and decreasing, and has reached a stable state in the last decade. And the LC burden attributable to smoking have particularly shown a downward trend in the recent years. On the other hand, the risk-related burden in males was more than three times as that in females, and the regions of Central Asia, high-income Asia Paci c and East Asia exhibited the highest burden worldwide. In addition, both the exposure level and the LC burden attributable to the risk factors were associated with the SDI. The ndings of this study could help in understanding the role of prognostic risk factors in LC burden, promoting the development of relevant studies, and proposing better prevention and control policies tailored to local conditions.

Declarations
Ethics approval and consent to participate This paper is exempt from ethical committee approval, because it is based on the GBD 2019 database, which does not contain identi able personal information, and a waiver of informed consent was reviewed and approved by the Institutional Review Board of the University of Washington.

Not applicable
Availability of data and materials The datasets supporting the conclusions of this article are included within the article.

Competing interests
The authors declared that they have no con ict of interests.   The variation trends of risk attributable liver cancer burden by causes from 1990 to 2019. (A) total liver cancer; (B) liver cancer due to alcohol use; (C) liver cancer due to hepatitis B; (D) liver cancer due to hepatitis C; (E) liver cancer due to NASH; (F) liver cancer due to other causes. Abbreviation: ASDR, age-standardized death rate; NASH, non-alcoholic steatohepatitis; SDI, social demographic index.