This is the first meta-analysis to specifically examine the association between HCV infection and risk of diabetes (DM) in the general population as well as in sub-groups at particularly increased risk of chronic liver disease (CLD) including those with hepatitis B (HBV) or HIV infection, or with other causes of liver disease (OLD) like alcohol-related liver disease. Among 34 eligible studies identified for this review, eighteen (15 retrospective and 3 prospective) evaluated DM risk in HCV-infected cases in comparison to general controls without HCV infection. Our meta-analysis which combined the adjusted odds ratios from these retrospective studies demonstrated an approximately 1.7-fold significant increase in DM risk with HCV infection. Similarly, the overall unadjusted pooled estimator demonstrated a significant 2-fold excess risk. Although there was evidence of potential small study or publication bias among these retrospective studies, this effect appears to be largely explained by the single largest study(18
) removal of which did not change the overall trend. Further, none of the other potential sources of between-study heterogeneity examined including geographic region, year of publication, or type of controls were significant (p>0.15).
Three prospective studies also evaluated whether HCV infection increases risk of developing type II diabetes(16
). All had serological confirmation of HCV and exclusion of DM at baseline. Results from our meta-analysis pooling adjusted HRs suggested HCV infection conveys an approximately 1.7-fold excess DM risk. Interestingly, essentially the same significant excess risk was observed by pooling the unadjusted HRs. In contrast to retrospective studies which have well-established limitations, long-term longitudinal studies with prospectively collected data such as these are particularly valuable as they establish a temporal relationship between HCV infection and subsequent occurrence of diabetes and help support an argument of a causal association. The significant excess DM risk observed in our meta-analysis of prospective studies (HRadjusted
=1.67) is also highly consistent with the significant excess risk observed in our meta-analysis of retrospective studies (ORadjusted
=1.70) and adds further support of those retrospective results. Taken together, the findings of our combined meta-analyses clearly indicate that chronic HCV infection is associated with a modest but significantly increased risk of developing type 2 diabetes in comparison to uninfected controls.
The reasons why chronic HCV infection would induce type 2 diabetes could be manifold. Several experimental studies have suggested a direct role of the virus in promoting DM risk. Within HCV core-transgenic mice, hepatocyte-associated degradation of the HCV core protein leads to negative interaction with insulin signaling by reducing IRS-1 phosphorylation and downstream signaling by Akt(48
) and by promoting IRS-1 and IRS-2 degradation(49
). In one study, the virus has also been localized in 39% of pancreatic islets in HCV-infected humans and occurs in approximately 54% of all cells within affected islets. Although there is no evidence of increased apoptosis, these HCV+ islet cells exhibit morphologic changes as well as derangement in glucose-stimulated insulin release (β-cell dysregulation)(50
). Other experimental studies have suggested a more indirect role of the virus, or rather that it is host response to the virus that promotes DM risk. For example, hepatic levels of pro-inflammatory cytokine TNF-α are doubled in HCV core-transgenic mice with blockade of TNF-α leading to restored hepatic sensitivity to insulin.(51
) However, it has also been suggested that HCV infection promotes DM risk as a tertiary consequence of HCV-induced liver damage. Indeed, it is well-established that advanced cirrhosis induces dysregulation of glycemic control which may result in overt diabetes(52
). Some support for such a tertiary mechanism comes from a clinical study demonstrating severe fibrosis is the only independent predictor of insulin resistance (IR) as measured by the surrogate marker the HOMA index in HCV-infected patients(53
). However, other studies have shown higher IR in HCV-infected patients irrespective of degree of liver injury(4
) with increases in IR evident even at early fibrosis stages(55
If it is characteristics specific to HCV infection itself rather than just the tertiary liver damage it generates that induces insulin resistance and increases diabetes (DM) risk in human populations, then it would be expected that the prevalence of diabetes should be higher with chronic HCV, than, for instance, with other causes of chronic liver disease. Nine retrospective studies evaluated this hypothesis with respect to HBV infection(5
). Suggestive evidence in support of this hypothesis came from the 1.7-fold significant excess DM risk conveyed by the unadjusted pooled OR. Only three studies also provided an adjusted estimate, with all three including adjustment for degree of liver pathology(17
). Our pooled adjusted OR demonstrated that HCV infection conveys a significant 1.8-fold excess risk of DM beyond that conveyed by relative degree of liver pathology. Four studies also provided unadjusted risk estimates stratified according to the presence of chronic hepatitis or cirrhosis(10
). The unadjusted pooled estimators in the context of cirrhosis and in the context of chronic hepatitis both demonstrated only modest non-significant excess risk of DM. However, given low study power as well as lack of adjustment for other possible confounders, these findings are difficult to interpret. The single study comparing DM risk with HCV-infection to that with chronic liver disease (CLD) attributable to a mixture of other causes including alcohol-related disease(27
) demonstrated strong and significant excess risk with HCV infection only in the context of chronic hepatitis. Unfortunately, without individual patient data it is not possible to further clarify the impact of liver injury on the relative risk of diabetes associated with HCV infection, both for fibrosis and necro-inflammatory activity.
An estimated 25–30% of HIV cases in the U.S. and Western Europe are co-infected with HCV(56
). Highly active anti-retroviral therapy (HAART) used to treat HIV infection is well-known to increase risk of CLD, with development of hepatotoxicity an important reason why HAART is discontinued(57
). However, HAART is still recommended for HCV co-infected treatment candidates, with some data suggesting its use may also lessen HCV-related liver disease progression(57
). Five studies included in this review evaluated whether HIV cases co-infected with HCV have increased DM risk in comparison to HIV mono-infected cases(20
). A small to modest excess risk of DM with HCV co-infection was generally observed though significance of findings was variable. Only three studies provided comparable unadjusted risk estimators (ORs) that were able to be combined in a pooled estimator (25
). Two studies included cases and controls prior to initiation of HAART(25
) while the third study adjusted for use of HAART(31
). This unadjusted meta-analysis demonstrated co-infection with HCV conveys a 1.8-fold significant excess risk of DM in comparison to that observed in HIV mono-infected cases. However, as only two studies provided adjusted estimates, we did not obtain an adjusted pooled estimator and our unadjusted meta-analysis finding of excess DM risk among HIV patients co-infected with HCV must be considered as suggestive only.
In evaluating findings from our meta-analyses, it is important to consider the potential impact of confounders of the relationship between HCV infection and occurrence of DM, particularly from such well-established risk factors as BMI. The three available prospective studies that evaluated diabetes risk in comparison to uninfected controls reached different conclusions as to what categories of HCV-infected individuals are at increased risk. Specifically, the two larger prospective cohort studies both demonstrated HCV infection conveys additional DM risk beyond that conveyed by age or BMI(16
). This finding is similar to that observed among all 6 retrospective studies which also included adjustment for these factors(8
). In contrast, the much smaller American case-cohort study, which included only 15 HCV-infected cases, showed only individuals who were already at increased diabetes risk (mainly overweight individuals older than 50) have an additional DM risk due to HCV infection(16
). Although the preponderance of current evidence suggests hepatitis C infection may convey additional DM risk beyond that conveyed by BMI, additional prospective studies are therefore needed to sort out the important aspect of the interaction between HCV infection and other risk factors for diabetes including current and historical obesity.
The present study has several strengths as well as some limitations. We used exhaustive search methods to identify all eligible studies and attempted to increase comparability and quality of included studies by using pre-specified eligibility criteria including publication in a peer-reviewed journal, a minimum total sample size and presence of appropriate as well as adequately identified case and control groups. To help assess the validity and reliability of our findings, we also performed additional quality control analyses including meta-regression and sensitivity analyses in order to identify possible sources of between-study heterogeneity. Further, we systematically investigated the potential for small study or publication bias and the impact of removal of individual studies on the pooled estimator was also assessed.
Our application of rigorous eligibility criteria to assure the internal validity of our findings has also imposed some potential limitations, particularly with respect to the generalizability of our results. First, our restriction to studies performed in adults means we are unable to extrapolate these findings to HCV-infected children. Second, our restriction to articles published in the English language means it is possible that complex interactions between ethnicity, HCV infection and the occurrence of diabetes could have been missed in particular ethnic groups if these findings were published exclusively in non-English language journals. However, only a small minority (6%) of studies reviewed were excluded specifically because they were not published in English. Further, 62% of all included studies were performed in countries where English is not the primary language. Third, we included only studies with a minimum total sample size of 200, with at least 100 exposed or cases and 100 controls or unexposed. We employed this criterion to help mitigate the potential for small study bias given the greater likelihood that small studies in particular will be published if they report significant or interesting results(58
). Although less than 12% of studies were specifically excluded due to sample size, it is therefore possible that our reported effect sizes might actually be increased if we had included these smaller studies. Finally, we selected only studies reporting on prevalence or incidence of overt diabetes. This has the advantage of robust and reproducible clinical definitions of the outcome of interest across studies. However, it can also underestimate the magnitude of the relationship between HCV infection and impaired glucose metabolism.
Other limitations are due to insufficient information provided by the eligible studies themselves. Dose-dependent effects could not be demonstrated because viral load and duration of HCV infection were not typically recorded. Other potentially important viral- (e.g., genotype) or host-related factors (e.g., family history of diabetes and visceral adiposity) could also not be examined. Additional prospective studies are therefore needed to determine what specific combination of viral- and host-related factors explain the observed excess risk of type II DM conveyed by HCV infection.
The findings of this meta-analysis could have important clinical implications. Given the demonstrated increased risk of diabetes conveyed by HCV infection, a strong case can be built for screening for glucose abnormalities in all HCV-infected individuals. Second, these data might provide a better insight into the overall burden of disease in chronic hepatitis C. Indeed, if risk of diabetes increases with the duration of exposure to HCV, then diabetes might become a prominent HCV-induced health problem in some patients like those with a low risk of fibrosis progression (e.g., women contaminated with HCV at a young age). Finally, since some reports have shown that HCV eradication improves insulin sensitivity(3
) and reduces the incidence of diabetes(59
), a reasonable inference would be that some HCV-infected patients at high risk of diabetes occurrence might benefit from antiviral therapy beyond hepatological reasons. Future work is needed in order to determine if diabetes could be prevented or reversed with successful HCV eradication.