Diabetes Poses a Higher Risk of Hepatocellular Carcinoma and Mortality in Patients With Chronic Hepatitis B

A Population-based Cohort Study

Yu-Chiau Shyu; Ting-Shuo Huang; Cheng-Hung Chien; Chau-Ting Yeh; Chih-Lang Lin; Rong-Nan Chien


J Viral Hepat. 2019;26(6):718-726. 

In This Article


To the best of our knowledge, this cohort study has the longest follow-up period exploring the impact of DM on the development and prognosis of HCC in CHB patients and is also the first to investigate the causal relationship between DM and HCC development in patients with HBV infection. Furthermore, it is the first study to estimate temporal transition probability of HCC and mortality using a multi-state model approach. In this study, we found that DM significantly increases the risk for HCC development, HCC-related mortality and all-cause mortality in CHB patients. After adjusting for possible confounders, DM remained a significantly independent risk factor associated with all the aforementioned outcomes. Since HCC was a transition state and HCC-related mortality was greater, the long-term cumulative probability of being alive with HCC was lower in the DM cohort than the non-DM cohort. These findings suggest that DM may function as an important aggravating factor for CHB patients' long-term outcome.

DM is a worldwide health problem that may be increasing the risk of HCC, as previously reported in two meta-analyses.[8,9] However, some controversy remains with regard to whether DM is a risk factor for HCC development after adjusting for alcoholic liver disease, viral hepatitis and other risk factors. Little is known about the causal relationship between DM and HCC development. Furthermore, viral hepatitis, a well-known aetiological factor of HCC, induced more than 90% of cases in Taiwan.[28] When a CHB patient is diagnosed with newly onset DM, understanding whether the risk of HCC and HCC-related complications is increasing or decreasing is a vital issue. As a result, we carried out this nationwide cohort study, which revealed that DM was an independent risk factor for increasing development of HCC, HCC-related mortality and all-cause mortality in CHB patients.

Previous studies have yet to address the temporal relationship between DM and HCC in CHB patients. The significant correlation between DM and HCC, as well as its mortality rate, was corroborated by the current study. Our data revealed that DM substantially increased the cumulative hazards of HCC development in the DM cohort when compared to the non-DM cohort. Furthermore, the mortality rate following the development of HCC increased more in the DM cohort than the non-DM cohort over time after 2 years of follow-up. However, living patients with HCC remained within an intermediate state throughout our model. After further delineation by transition probabilities, the cumulative probability of being alive with HCC was lower in the DM cohort than the non-DM cohort within 2 years of follow-up and was similar from 2 years of follow-up through the end of the follow-up period. Clinically, CHB patients with new-onset DM should be carefully monitored for any HCC complications, particularly within the first 2 years. Furthermore, we adopted the Cox proportional hazard model to examine the association between DM complications or DM medication and HCC development (Table S3). Univariate analysis revealed that only sulfonylurea was associated with a lower rate of HCC development (P = 0.018).

Our study discovered that liver cirrhosis has time-varying effects on HCC-related mortality, but this finding must be interpreted with caution. In fact, liver cirrhosis in CHB patients may be composed of heterogeneous groups. In the early follow-up period, the decreasing effect can be attributed to the high mortality rate of HCC in decompensated cirrhotic CHB patients. According to previous studies,[29,30] a higher mortality rate is not uncommon in the first 2 years of treatment when HBV-related decompensated cirrhosis is combined with HCC. Afterwards, HCC-related mortality in the remaining patients with compensated CHB cirrhosis decreased dramatically, which may thus imply that the majority of the causes of death after 2 years are related to DM rather than CHB. Therefore, CHB patients should be more carefully screened for 2 years after being diagnosed with liver cirrhosis. However, it should be noted that the diagnosis of liver cirrhosis is difficult to verify in this insurance database because the Taiwan NHIRD does not contain laboratory tests or examination results. Clinically, if liver cirrhosis was suspected, ultrasonography, transient elastography scan, computerized tomography scan, magnetic resonance imaging scan or liver biopsy may be necessary, and without these confirmatory tests, the incidence of liver cirrhosis may be underestimated based on claims data.

Our study has several strengths. First, the current database enrolled all patients that were newly diagnosed with DM in Taiwan during the study period. With propensity score matching, a large sample size (2966 patients in the DM cohort and 2966 in the non-DM cohort), and a long follow-up period (up to 11 years), we are able to strongly suggest that DM increases the risk of HCC development, HCC-related mortality and mortality without HCC in CHB patients. Secondly , controlling for relevant confounders by adopting multivariable regression analysis minimized the chance of selection bias. Finally, applying the nationwide, population-based Taiwanese NHIRD with reliable information enabled us to conduct such a large cohort with favourable qualifications by using a multi-state model approach.

However, our study also has a number of limitations caused by using insurance data. First, the Taiwan NHIRD does not contain several risk factors that may contribute to the development of HCC, such as genetic parameters, body mass index, diet, cigarette smoking, aflatoxin exposure, lifestyle and family history. Our inability to account for these variables may prevent the analysis of potential confounders. Secondly, laboratory test results were unavailable in Taiwan NHIRD, so we were prohibited from evaluating the impact of the individual degree of blood sugar control because the haemoglobin A1c level was not available. Relevant metabolic factors, such as hypertension and dyslipidemia, were also not included in our analysis. Thirdly, we were unable to investigate the effect of CHB serological changes in this population-based study. Since the severity of HBV infection ranges from inactive carriers to active hepatitis, the incidence of HCC may be varied, and our failure to address the factors associated with HBV replication remains a major limitation of this study. In addition, a liver cirrhosis diagnosis was based on ICD-9 codes and may have been underestimated. Furthermore, this observational study does not explore the mechanism through which DM may be related to HCC. Finally, in this study, we could not exclude the effects of antiviral therapy or DM drug compliance.

In conclusion, in this nationwide, population-based cohort study, we discovered that DM is an independent risk factor associated with increased HCC, HCC-related mortality and all-cause mortality in CHB patients. These findings provide new support for adopting aggressive therapy for HCC and strict DM control for such patients. In the future, additional prospective studies may be required to investigate the underlying mechanisms for this result.