Type 2 diabetes mellitus increases the risk of hepatocellular carcinoma in subjects with chronic hepatitis B virus infection: a meta-analysis and systematic review.

BACKGROUND: Type 2 diabetes mellitus has been proved to be a risk factor of hepatocellular carcinoma, but how diabetes affects incidence of hepatocellular carcinoma among patients with chronic hepatitis B virus infection remains controversial.
METHODS: A comprehensive search of Medline and Embase was performed. Incidence of hepatocellular carcinoma in chronic hepatitis B patients was the primary outcome. Pooled HRs and 95% CIs were calculated to assess the correlation between diabetes and incidence of hepatocellular carcinoma.
RESULTS: Five cohort studies and two case-control studies were identified, with a total of 21,842 chronic hepatitis B patients. The diabetes mellitus cohort was found to have increased incidence of hepatocellular carcinoma (pooled HR 1.77, 95% CI 1.28-2.47; fixed effect) and worse overall mortality (pooled RR 1.93, 95% CI 1.64-2.27; fixed effect) in comparison with those without diabetes. In case-control studies, hepatocellular carcinoma cases were found to have an insignificantly elevated diabetes mellitus rate in comparison with the control group.
CONCLUSION: Type 2 diabetes mellitus is significantly associated with increased risk of hepatocellular carcinoma among patients with chronic hepatitis B virus infection, and aggressive management of diabetes mellitus is strongly suggested.

Introduction

Hepatocellular carcinoma (HCC) is the fifth-most common cancer worldwide, leads to nearly 1 million deaths every year,1 and is the third-most frequent cause of cancer-related death. The incidence of HCC is particularly high in Asia (over 20 in 100,000 men and over ten in 100,000 women) and in Africa, intermediate in southern Europe, and much lower in most developed countries.2 Hepatitis virus infection, mainly hepatitis B virus (HBV) and hepatitis C virus (HCV), has been widely accepted as the major recognized risk factor of HCC globally, accounting for over three-quarters of primary HCC cases.2,3 However, when HBV or HCV is not involved, the etiologic factor of HCC varies, of which diabetes mellitus (DM),4 heavy alcohol drinking,5 smoking,6 obesity,7 and aflatoxin8 are relatively important.

DM, which has been proved to be a risk factor of various kinds of malignancies, is strongly associated with nonalcoholic fatty-liver disease and many other metabolic processes.9 Insulin resistance10 was believed to play an important role in hepatocarcinogenesis in HBV patients with type 2 DM or even prediabetes.11 The association between DM and HCC risk was indicated to be independent of cirrhosis, though most HCC cases presented with cirrhosis.12 A recent systematic review demonstrated that concurrent DM is strongly associated with increased HCC risk among chronic HCV patients,13 but scanty evidence is available about the correlation between DM and HCC in chronic HBV (CHB) patients. The clinical landscape of HCV is currently facing a great change, such that its cure would be universal for patients for whomever has access to effective therapy, which will definitely result in a decrease in HCC developments. Therefore, HBV infection, alcohol consumption, and metabolic disorders, such as DM and obesity, are supposed to be the leading etiologic factors of HCC in the coming future. There are mixed results14–17 of the few studies on the association between DM and the risk of HCC in patients with CHB. As such, we performed this meta-analysis and systematic review of the literature to achieve further understanding of the impact of DM on the risk of developing HCC in patients with CHB.

Methods

Literature-search strategy

A comprehensive search of Medline and Embase was performed to retrieve studies published in English (cutoff date February 5, 2018) using the keywords “diabetes” or “diabetes mellitus, type 2” or “DM”, “hepatitis B” or “HBV”, and “hepatocellular carcinoma” or “HCC” or “liver cancer”. We also examined the reference lists of eligible studies to identify additional articles, in order to guarantee a systemic search. Figure 1 depicts the search strategy in detail.

Figure 1

Flow diagram showing the search strategy along with the selection and screening processes for the eligible studies.

Abbreviations: DM, diabetes mellitus; HCC, hepatocellular carcinoma; HBV, hepatitis C virus.

Inclusion and exclusion criteria

Inclusion criteria from the literature search were: studies that focused on the relationship between DM and the risk of HCC: HCC incidence and/or related mortality as outcomes; results of HR/RR/OR and their corresponding 95% CIs for DM and incidence of HCC; and if two or more studies were reported on the same cohort and objectives, either the higher-quality publication or more recent publication was included in the analysis. Studies were excluded if they had not presented data on the relationship between DM and incidence of HCC in patients with HBV infection or specific results were unable to extract. Studies reporting on the effect of DM on the prognosis of HCC or where HCC was not the only outcome (eg, including cholangiocarcinoma) or including patients with type 1 DM were not considered. Reviews, case reports, letters, animal or in vitro studies, conference abstracts, and non-peer-reviewed articles were also excluded from the meta-analysis.

Assessment of quality of articles according to the New-castle–Ottawa Scale, and data extraction were independently performed by YT and SW, as well as the literature search. Quality assessment of both cohort studies and case–control studies included three main characteristics: selection of groups, comparability of cohorts/cases and controls, and outcome/exposure. Studies with seven or more scores were defined as highly qualified.

Data extraction

Incidence of HCC was the main outcome assessed in this meta-analysis, while overall mortality or HCC-related mortality was the secondary main outcome. HR, RR, and OR values with their corresponding 95% CIs were extracted from each study. Other information included (but was not limited to) study design, population characteristics, source of cases and controls, ascertainment of type 2 DM, treatment for HBV infection, anti-HCV status, variables adjusted for in the multivariate regression models, and the measure of association between DM and HCC incidence.

Statistical analysis and exploration of heterogeneity

We used pooled HRs for primary outcomes to assess the relationship between type 2 DM and incidence of HCC in HBV patients. RRs and weighted mean differences, both with corresponding 95% CIs, were computed for binary data and continuous data, respectively. Funnel plots were visually inspected to identify publication bias. Sensitivity analysis was conducted by removing a single study each time and recalculating pooled results of the remaining studies. The χ2 test was used to explore potential heterogeneity with I2 and P-values. I2>50% or P<0.10 was defined as increased heterogeneity and a random-effect model18 used, while a fixed model was used when P>0.10. Results are presented as P-values and 95% CIs, where appropriate, and two-sided P<0.05 was considered to indicate statistical significance. Meta-analyses were performed using RevMan software (version 5.3.5; Cochrane Collaboration, Copenhagen). Quality of evidence was evaluated using the software GradePro, comprising four levels: high quality, moderate quality, low quality, and very low quality.

Results

The online search initially found 1,682 studies, and 1,624 were excluded after screening of titles and abstracts. A total of 58 full-text articles were reviewed. Finally, two case–control studies17,19 and five cohort studies14–16,20,21 were identified, of which six were conducted in Asia14–17,19,20 and one in New Zealand.21 In all cohort studies, patients diagnosed with HCC before the inception point were excluded. The number of eligible CHB patients ranged from 223 to 6,545, with a total of 21,842. Diagnosis of type 2 DM was obtained from patient self-report, abnormal fasting/random glucose, positive oral glucose tolerance test, and DM management (oral hyperglycemic agent or insulin injection). HCC cases were confirmed with the combination of increased AFP and imaging findings (ultrasound, enhanced computed tomography, or angiography) in five studies,14,15,17,19,21 and positive histology or cytology was also used to define HCC in the two case– control studies.17,19 Two cohort studies16,20 in Taiwan defined HCC cases according to the national cancer registry alone (Table 1). Only Hsiang et al21 reported details of dropouts and withdrawals. In their study, seven patients were lost to follow-up or moved overseas, resulting in a loss to or unavailability for follow-up rate of 3%. Other studies, however, did not provided data about details of follow up.

Table 1

Baseline characteristics of the seven studies included

Study	Period, location	Study design	Group	n	Participants	DM ascertainment	HCC diagnosis
Lai et al14	1999–2002, Taiwan	Cohort	DMNDM		A total of 54,916 subjects aged ≥30 years from a community-based program; HBV-positive patients(n=6,545)	FBG >126 mg/dL, self-reported	AFP >400 mg/nL, positive imaging finding (enhanced CT, angiography), and histopathological confirmation
Chen et al15	1991–1992, Taiwan	Cohort	DMNDM	623,862	23,567 residents from a multiple township-based cancer-screening program; seropositive for HBsAg only (n=3,924)	Self-reported	1) National cancer registry; 2) histopathological confirmation; 3) two imaging findings (ultrasonography, angiography, or CT), one imaging diagnosis and AFP >400 ng/mL
Wang et al16	1997–2004, Taiwan	Cohort	DMNDM	47649	5,929 residents >35 years old with hepatitis from a population-based screening program; seropositive for HBV only (n=696)	FBG ≥126 mg/dL, nonfasting glucose ≥200 mg/dL, or using hypoglycemic drugs	National cancer registry
Ko et al17	2004–2005, Taiwan	Case– control	DMNDM		Hospitalized inpatients first diagnosed with HCC (n=359) and non-HCC controls (n=1,536) randomly selected from a health-checkup program	FBG ≥126 mg/dL, current use of oral hyperglycemic agent or insulin injection at the time of recruitment	1) Histopathological verification; 2) AFP >400 ng/mL and at least one imaging study (ultrasonography, enhanced CT, or angiography)
Li et al19	2004–2008, Chinese mainland	Case– control	DMNDM	4215,854	Patients aged ≥30 years hospitalized for HCC (n=1,105) or CHB (n=5,170) without HBV treatment	FBG ≥7 mmol/L, self-reported	1) Histopathological confirmation; 2) two imaging findings (ultrasonography, enhanced CT, or MRI); 3) AFP >400 ng/mL and one positive image finding
Fu et al20	1997–2009, Taiwan	Cohort	DMNDM	2,0992,080	Chronic HBV patients with(n=2,099) or without DM (n=2,080) randomly selected from the national health-research database (1:1 ratio)	NA	National insurance-program registry (for catastrophic illness patient database)
Hsiang et al21	2000–2012, New Zealand	Cohort	DMNDM	50173	223 HBV patients with transient elastography or radiological features of established cirrhosis with (n=50) or without DM (n=173)	FBG ≥7.0 mmol/L, random glucose level ≥11.1 mmol/L, or positive OGTT and/or HbA1c ≥6.5%, and active DM follow-up*	Combination of elevated AFP and positive imaging finding

Note:

Under primary care with diabetes disease-management program or under hospital diabetes-service follow-up.

Abbreviations: CT, computed tomography; HbA1c, glycosylated hemoglobin; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HBsAg, HBV surface antigen; DM, diabetes mellitus; NDM, non-DM; MRI, magnetic resonance imaging; FBG, fasting blood glucose; OGTT, oral glucose-tolerance test; NA, not applicable.

No statistical differences were found between subjects with and without type 2 DM for average age (mean difference 2.81, 95% CI −2.91 to 8.52), male sex (RR 0.99, 95% CI 0.91–1.08), years of follow-up (mean difference −0.35, 95% CI −1.02 to 0.32), or HBV-treatment rate (RR 1.07, 95% CI 0.73–1.55).

Type 2 DM and risk of HCC in CHB subjects

Of the seven studies included in this meta-analysis, three cohort studies15,20,21 and one case–control study17 demonstrated a positive association between type 2 DM and risk of HCC in CHB patients, while other two cohort studies14,16 and one case–control study19 failed to find a statistical difference between the two groups. Four in seven studies reported correlations between DM and HCC with HRs,14,16,20,21 two with ORs,17,19 and another with RRs,15 all with corresponding 95% CIs (Table 2). All studies had different variables adjusted in multivariate regression analysis (Table 3).

Table 2

DM and incidence of HCC risk

Study	Group	n	HCC cases	HCC, %	OR/HR/RR		95% CI
Lai et al14	DM	NA	NA	NA	HR	1.04	0.37–2.93
	NDM	NA	NA	NA
Chen et al15	DM	62	8	12.9	RR	2.41	1.17–4.95
	NDM	3,862	179	4.6
Wang et al16	DM	47	NA	NA	HR	1.3	0.3–5.6
	NDM	649	NA	NA
Ko et al17	DM	NA	NA	NA	OR	4.32	1.92–9.70
	NDM	NA	NA	NA
Li et al19	DM	421	93	NA	OR	0.9	0.7–1.2
	NDM	5,854	1,012	NA
Fu et al20	DM	2,099	69	3.29	HR	1.798	1.194–2.707
	NDM	2,080	42	2.02
Hsiang et al21	DM	50	13	26	HR	2.36	1.14–4.85
	NDM	173	23	13.29

Abbreviations: DM, diabetes mellitus; HCC, hepatocellular carcinoma; NDM, non-DM; NA, not applicable.

Table 3

Factors adjusted in the multivariate regression analysis

Study	Adjusted confounders
Lai et al14	Age, gender, HCV status, smoking and cumulative consumption of alcohol
Chen et al15	Age, sex, cigarette smoking, habitual alcohol consumption, and education levels
Wang et al16	Age, sex, smoking habit, alcohol consumption, BMI, and diabetes status before the study
Ko et al17	Age, sex, and other viral hepatitis infection
Li et al19	Age, sex, city of residence, family history of liver cancer, HBeAg status and cirrhosis
Fu et al20	Age, sex, hyperlipidemia, HBV treatment, statin therapy, cirrhosis, comorbidity index, and obesity
Hsiang et al21	Age, sex, antiviral therapy, sustained viral suppression, and MELD score

Abbreviations: HCV, hepatitis C Virus; BMI, body-mass index, MELD, model for end-stage liver disease.

The incidence of HCC in DM cohorts varied from 3.29% to 26.0% in cohort studies, while the rate was 2.02%–13.3% in the non-DM group. The four cohort studies reporting on HR14,16,20,21 found increased incidence of HCC among patients with DM over those without DM among CHB patients (Figure 2), resulting in a pooled HR of 1.77 (95%CI 1.28–2.47, heterogeneity I2=0; fixed effect). In case–control studies,17,19 an elevated DM rate was indicated in HCC cases in comparison with control groups (12.35% vs 6.53%), but not statistically, with an RR of 2.10 (95% CI 0.84–5.25; random-effect). Among the three studies reporting RRs or ORs, Chen et al15 and Ko et al17 also found significant associations between type 2 DM and risk of HCC in HBV subjects, with an RR of 2.41 (95% CI 1.17–4.95) and an OR of 4.32 (95% CI 1.92–9.70), respectively, but Li et al19 found no significant relationships between DM and risk of HCC when comparing all HCC cases with cross-sectional controls, with an OR of 0.9 (95% CI 0.7–1.2). The publication bias of studies assessing the relation between DM and risk of HCC is shown in Figure 3.

Figure 2

Forest plot of meta-analysis results comparing the incidence of HCC between patients with DM and those without DM.

Abbreviations: HCC, hepatocellular carcinoma; DM, diabetes mellitus.

Figure 3

Funnel plot of studies assessing the relationship between DM and risk of HCC.

Abbreviations: DM, diabetes mellitus; HCC, hepatocellular carcinoma.

Analysis of sensibility was carried out by excluding one article at a time to guarantee stability of the meta-analysis. As a result, pooled results remained statistically positive when any of the cohort studies was excluded from the meta-analysis. For example, when Wang et al was excluded, the pooled HR of remaining studies remained at 1.80 (95% CI 1.29–2.53). When it came to heterogeneity, I2 remained at 0 when any study was excluded.

DM, overall mortality, and HCC-related mortality

Two cohort studies20,21 reported overall mortality between two groups, revealing that subjects with type 2 DM suffered significantly higher overall mortality in comparison with those without DM (Figure 4), with a pooled RR of 1.93 (95% CI 1.64–2.27, I2=18%; fixed effect). Only Hsiang et al21 reported on HCC-related mortality in patients with CHB, and patients assigned to the DM group had significantly higher HCC-related mortality than those in the non-DM group (27.9 vs 8.8 per 1,000 patient-years, P=0.02). They also demonstrated increased liver-related mortality or orthotopic liver-transplantation rate in the DM group, at 23.4%, compared to the non-DM group, at 9.4% (P=0.009).

Figure 4

Forest plot of results comparing overall mortality between DM and non-DM subjects.

Abbreviations: DM, diabetes mellitus; NDM, non-DM; M–H, Mantel–Haenszel.

Risk of bias and quality evaluation of evidence

We conducted a quality evaluation on all five cohort studies and two case–control studies included, based on the coding manual for cohort-studies and for case–control studies. All studies scored at least 6, and five studies scored 7 or more. The main outcomes and some other results are summarized in Table 4. The evidence was considered to be of low quality.

Table 4

Quality evaluation of evidence

	Studies, n	Design	Patients, n	Effect	95% CI	Quality	Level
Incidence	4	Observational	11,643	1.88	1.28–2.47	ÅÅOO	Low
Overall mortality	2	Observational	4,402	1.93	1.64–2.27	ÅÅOO	Low
Age	2	Observational	4,402	2.81	−2.91 to 8.92	ÅÅOO	Low
Sex	2	Observational	4,402	0.99	0.91–1.08	ÅÅOO	Low
HBV treatment	2	Observational	4,402	1.07	0.73–1.55	ÅÅOO	Low
Follow-up, years	2	Observational	4,402	−0.35	−102 to 0.32	ÅÅOO	Low
% Diabetes mellitus	2	Observational	8,180	2.18	0.82–5.83	ÅÅOO	Low

Discussion

Chronic hepatitis virus infection, including HBV and HCV,22 has always been widely acknowledged as a major risk factor of primary HCC. DM, as well as other metabolic abnormalities, has been proved to be associated with quite a few kinds of malignancies,23,24 and this association was suggested not to be mediated by body-mass index.25 A recent meta-analysis26 and a systemic review13 demonstrated a strong associations between concurrent DM and risk of HCC among chronic HCV patients. While well studied in anti-HCV-positive subjects, the potential relationship of type 2 DM and risk of HCC in the HBV-infected population remains unclear. In addition, with effective treatment for HCV infection, HCV-related HCC cases are expected to decrease worldwide.27 Therefore, HBV infection and metabolic factors, including obesity, DM, and hyperlipidemia, are predicted to account for most of the increase in HCC.

Five cohort studies and two case–control studies were included in the current meta-analysis, with >20,000 individuals enrolled. A recent meta-analysis by Chen et al26 summarized three studies (two reporting with HRs and the other with RRs), revealing a similar risk of HCC in diabetics without DM among CHB patients. However, in the current meta-analysis, we found a significant association between type 2 DM and increasing incidence of HCC among CHB patients, with a pooled HR of 1.77 (95% CI 1.28–2.47) and no heterogeneity detected (I2=0). Previous meta-analyses have failed to find statistical differences between the two groups, possibly due to the small number of studies included and statistical issues, as not all results were presented with HRs. The biological mechanism for how type 2 DM influences HCC development is not well understood and remains controversial. Most researchers have suggested that type 2 DM contributes to HCC development independently or synergetically with other risk factors, such as HBV28 or HCV29,30 infection and alcohol consumption.31

Though relatively uncommon in the setting of HCC, DM is of growing importance, because of its rapidly increasing incidence among adults, as well as nonalcoholic fatty-liver disease, especially in developed countries.9 It has been reported that type 2 DM and/or obesity had the greatest population-attributable fractions (36.6%) and proportion of cases that could be attributed to specific risk factors: to HCC, higher than alcohol-related disorders (23.5%) or HCV (22.4%).32 Insulin resistance9,33 and hyperinsulinemia34,35 are believed to be key factors of HCC oncogenesis in diabetics, mostly through the process of inflammation and cellular injury.36–38 While playing an important role in glucose and lipid metabolism, insulin also has pleiotropic effects on regulation of inflammation and cell proliferation. IGF1, the most powerful activator of cellular proliferation, and IRS1 are crucial downstream targets of insulin.36 HCC cells have been found to overexpress39 IGF1 and IRS1, revealing that they are of great importance in the process of HCC development. On the other hand, the inflammatory milieu caused by insulin resistance and nonalcoholic steatohepatitis leads to multiple pathways of inflammatory processes, which in return activated oncogenic signaling pathways, such as PI3K–PTEN–Akt and JAK–STAT.9 Insulin resistance also causes hepatic inflammation and fibrosis by the accumulation of fat within hepatocytes, which produces oxidative stress and results in hepatosteatosis.40 All these activations of pathways are associated with cellular proliferation promotion, increased angiogenesis, and decreased apoptosis, which finally lead to HCC development.

DM was also found to be a poor prognostic factor for overall mortality in the cohort of patients with CHB infection (RR 2.33, 95% CI 1.64–3.31) in the current meta-analysis, which echoed previous studies of worse overall mortality and HCC recurrence after potential curative therapy in patients with DM.41,42 DM has been attributed to increases in both HCC related mortality and mortality from other causes, such as cardiovascular factors.21 Since insulin resistance and hyperinsulinemia are key factors in the process of hepatocarcinogenesis, any other underlying disease or medication that affects the insulin level is a potential risk factor for the progression of HCC. Several studies21,43,44 have demonstrated higher incidence of HCC among HBV subjects who were treated with insulin or antidiabetic drugs that increase circulating insulin, while metformin, which is supposed to improve insulin sensibility, has been reported to be related to declined HCC risk and mortality.45,46 Therefore, we recommend better management of DM and cautious selection of therapy for DM in CHB patients.

In our meta-analysis, five studies were conducted in Tai-wan. Although they were conducted using different databases in different periods (Table 1), it is still difficult to confirm whether the cohort pool used in each study was completely different. Besides, methods used to identify HCC and DM were different in the studies included. Therefore, we think this may be a limitation of our study, and our results should be interpreted with caution.

The major limitation involves the variability of the adjustments within the studies. As is well known, HCC is related to quite a few independent risk factors, of which HBV DNA levels, antiviral therapy, and liver cirrhosis are of great importance. However, specific data about these effecting factors were not available in most of the studies included in this meta-analysis. Three studies19–21 clearly adjusted for cirrhosis in the multivariate analysis, two20,21 adjusted for antiviral therapy, and one19 just excluded patients who accepted antiviral therapy, while only one21 adjusted sustained viral suppression (HBV viral load <3log IU/mL on at least two occasions 6 months apart). Also, the evaluation of evidence turned out to be of low quality. All seven studies included were observational: five cohort studies and two case–control studies. Many of the researches did not report on the details of follow-up, including precise follow-up years of both groups and loss rate. More well-designed research and randomized controlled trials are needed to confirm further the association between type 2 DM and the risk of HCC. Though the evidence was evaluated as low quality, most studies were designed to minimize potential bias by adjusting for age, sex, and other covariates that were possibly influential (Table 3).

Consequently, the findings of type 2 DM significantly related to HCC risk may shed some light on the prevention of HCC. Better management of DM and correlated metabolic factors is strongly recommended, and HBV patients with DM should be tested more frequently, in order to improve early detection of HCC or other malignancies.