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

Disclosures

J Viral Hepat. 2019;26(6):718-726. 

In This Article

Results

Patient Characteristics

The enrollment flow chart is shown in Figure 1. To achieve a 1:1 comparison between the DM and non-DM cohorts, we used propensity score matching with the following key characteristics: age, gender, alcohol-related liver disease and baseline liver cirrhosis, all within the same follow-up period. In total, 2966 CHB patients were identified in each cohort. Their baseline demographics and clinical features are summarized in Table 1. After comparing the comorbidities in the Deyo-Charlson Comorbidity Index, the DM cohort had more myocardial infarction (P = 0.044), congestive heart failure (P = 0.002) and cardiovascular disease (P = 0.036). No other clinical comorbidities were observed to have significant differences. In the present study, the median follow-up time of the DM and non-DM cohorts was 11.4 years (interquartile range [IQR], 11.1-11.7) and 11.4 years (IQR 11.2-11.7), respectively.

Cumulative Incidence of HCC, All-cause Mortality and HCC-related Mortality

To differentiate between the incidence of HCC, all-cause mortality and HCC-related mortality, we used multi-state models to represent the transitions of "start-to-HCC," "start-to-death" and "HCC-to-death." The events of multi-state models in the total cohort, DM cohort and non-DM cohort are displayed in Figure S2. During the 11-year follow-up period, 701 CHB patients (11.8%) developed HCC. A total of 731 patients (12.3%) suffered from mortality of all causes, and 363 patients (6.1%) had HCC-related mortality (Table S1). Compared to the non-DM cohort, the incidence of HCC in the DM cohort was significantly higher (13.3% vs 10%; P < 0.001). The DM cohort also demonstrated significantly higher all-cause mortality and HCC-related mortality (16.9% vs 8.2%; P < 0.001, 7.5% vs 4.7%; P < 0.001, respectively).

Supplemental Figure 2.

A multi-state model, including the "start-to-HCC" transition, "start-to-death" transition, and "HCC-to-death" transition in (A) the total cohort, (B) the DM cohort, and (C) the non-DM cohort.

Nonparametric Estimates of Cumulative Hazards and Transition Probabilities of the Multi-state Model

To further characterize these events based on temporal relationships, time-dependent, nonparametric estimates of the cumulative hazard were compared between the DM and non-DM groups within the same follow-up period (Figure 2). The cumulative hazards of the incidence of HCC were greater in the DM cohort than in the non-DM cohort, with this effect increasing with time (Figure 2A). Furthermore, the cumulative hazards of all-cause mortality were greater in the DM cohort than in the non-DM cohort, and this effect was more prominent than the development of HCC (Figure 2B). The cumulative hazards of HCC-related mortality were also greater in the DM cohort than in the non-DM cohort, with this effect becoming more prominent after 2 years (Figure 2C).

Figure 2.

Nonparametric estimates of cumulative hazards in the DM and non-DM cohorts. Cumulative hazards of (A) "start-to-HCC" transition, (B) "start-to-death" transition and (C) "HCC-to-death" transition, stratified by DM status. DM, diabetes mellitus; HCC, hepatocellular carcinoma

The transition probabilities of the multi-state model were utilized to show relative outcome probabilities (Figure 3 and Table S2). The cumulative probability of transitioning to HCC was higher in the DM cohort than in the non-DM cohort, while all-cause mortality and HCC-related mortality were significantly higher in the DM cohort than in the non-DM cohort. The cumulative transition probabilities of living patients with HCC were greater within 2 years of follow-up and similar after 2 years of follow-up between the DM and non-DM cohorts (2.2%, 4.0%, 5.1%, 6.6% and 8.5% vs 1.4%, 3.1%, 4.8%, 6.7% and 8.3%, respectively). However, the cumulative transition probabilities of mortality after HCC were higher in the DM cohort than the non-DM cohort (0.5%, 1.0%, 2.0%, 2.7% and 3.2% vs 0.2%, 0.5%, 0.8%, 0.8% and 1.0%, respectively). Therefore, the cumulative transition probabilities of dying without HCC were once again higher in the DM cohort than the non-DM cohort up to 10 years (1.4%, 2.5%, 3.9%, 5.4% and 7.2% vs 0.1%, 0.5%, 1.4%, 2.2% and 3.1%, respectively). When combining the living patients with HCC and patients who died after HCC, the cumulative transition probability of CHB patients with HCC development in the DM cohort was 1% higher than the non-DM cohort at 2 years and increased to 2.4% at 10 years.

Figure 3.

Stacked transition probabilities among four states in the DM and non-DM cohorts. (A) DM cohort and (B) non-DM cohort. DM, diabetes mellitus; HCC, hepatocellular carcinoma

Risk Factors Through cox Proportional Hazard Model Analysis

The multi-variate, transition-specific Cox proportional hazard model was adopted to examine the association between the development of HCC and potential risk factors. As shown in Table 2, DM was an independent risk factor that was significantly associated with HCC development (hazard ratio [HR] 1.35; 95% CI 1.16-1.57; P < 0.001) after adjusting for such other confounders as age, gender, liver cirrhosis and alcohol-related liver disease.

Similarly, the association between HCC-related mortality and all-cause mortality was analysed with potential risk factors using a Cox proportional hazard model. DM was also an independent risk factor that was significantly associated with HCC-related mortality (HR 1.31; 95% CI 1.06-1.62; P = 0.014) after adjusting for such other confounders as age, gender, liver cirrhosis, alcohol-related liver disease and renal disease. Furthermore, DM was an independent risk factor that was significantly associated with all-cause mortality (HR 2.32; 95% CI 1.84-2.92; P < 0.001) after adjusting for such other confounders as age, gender, liver cirrhosis, alcohol-related liver disease and the comorbidities in the Deyo-Charlson Comorbidity Index.

To investigate the degree to which liver cirrhosis contributed to HCC-related mortality, we used post-estimation simulation distribution techniques to determine the time-varying impact (Figure 4). The results demonstrated that the effect of liver cirrhosis on HCC-related mortality declined with time, particularly within 2 years.

Figure 4.

Post-estimation simulation results of time-varying effects. The graph of the effect of liver cirrhosis on HCC-related mortality demonstrates decreasing relative hazards over time. Blue lines represent the median values of simulation distributions. The darker blue bands represent the 50% confidence bands, while the lighter blue bands represent the 95% confidence bands of the simulated distributions. HCC, hepatocellular carcinoma

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....