Potential conflict of interest: Yao-Chun Hsu reports having received lecture fees from the Merck Sharp & Dohme (Taiwan) and the Roche company (Taiwan). There is no other conflict of interest to declare.
Hepatitis C virus (HCV) infection is causally associated with insulin resistance and diabetes mellitus. This population-based cohort study aimed to investigate whether antiviral therapy for HCV infection was associated with improved clinical outcomes related to diabetes. From the Taiwan National Health Insurance Research Database, 2,267,270 Taiwanese residents diagnosed with diabetes mellitus were screened for eligibility. HCV infection was defined by a specific diagnosis code and measurement of serum antibody. After excluding patients with serious comorbidity, we enrolled a total of 1,411 eligible patients who received pegylated interferon plus ribavirin (treated cohort), and matched them 1:1 with 1,411 untreated controls by propensity scores (untreated cohort). We also matched the treated cohort 1:4 with 5,644 diabetic patients without HCV infection (uninfected cohort). Participants were followed up for the occurrence of endstage renal disease (ESRD), ischemic stroke, and acute coronary syndrome (ACS) after receiving antiviral treatment or the corresponding calendar date. From 2003 to 2011, the 8-year cumulative incidences of ESRD in the treated, untreated, and uninfected cohorts were 1.1% (95% confidence interval [CI], 0.3-2.0%), 9.3% (95% CI, 5.9-12.7%), and 3.3% (95% CI, 2.3-4.3%), respectively (P < 0.001); those of stroke were 3.1% (95% CI, 1.1-5.0%), 5.3% (95% CI, 3.0-7.5%), and 6.1% (95% CI, 4.8-7.4%), respectively (P = 0.01); and those for ACS were 4.1% (95% CI, 2.1-6.1%), 6.6% (95% CI, 3.7-9.5%), and 7.4% (95% CI, 5.9-9.0%), respectively (P = 0.05). As compared with the untreated cohort, antiviral treatment was associated with multivariate-adjusted hazard ratios of 0.16 (95% CI, 0.07-0.33%) for ESRD, 0.53 (95% CI, 0.30-0.93) for ischemic stroke, and 0.64 (95% CI, 0.39-1.06) for ACS. Conclusion: Antiviral treatment for HCV infection is associated with improved renal and cardiovascular outcomes in diabetic patients. (Hepatology 2014;59:1293-1302)
Both diabetes mellitus (DM) and chronic infection with hepatitis C virus (HCV) are serious public health problems around the world, globally affecting 347 million and 170 million people, respectively.[1, 2] A large and growing body of evidence has unraveled a complex association between these two diseases. On the one hand, patients with HCV infection as compared with the general population or those with another viral hepatitis are more likely to develop insulin resistance and DM.[3-6] On the other hand, insulin resistance with or without overt manifestation of DM adversely impacts the clinical outcomes in HCV-infected patients, in terms of poor response to antiviral therapy, accelerated progression of liver fibrosis, and increased risk of hepatocellular carcinoma.
It has been demonstrated consistently in laboratory experiments and clinical observations that infection with HCV can induce insulin resistance and hence susceptibility to DM.[10, 11] The molecular mechanism has not been fully elucidated, but appears to involve intracellular oxidative stress, dysregulation of cytokines, inhibition of insulin downstream signaling, and reduced expression of glucose transporters.[12-14] Along with these lines of evidence, viral eradication has been shown to effectively ameliorate insulin resistance, and may attenuate the risk of new-onset DM.[16, 17] It remains unknown, however, how treatment of concomitant HCV infection will influence outcomes in patients with established DM.
In light of its efficacy in resolving insulin resistance,[15-17] we hypothesized that antiviral therapy for HCV might improve clinical outcomes related to DM. We analyzed a national healthcare database to investigate whether anti-HCV treatment was associated with improvement of renal and cardiovascular outcomes among diabetic patients.
Materials and Methods
Study Design and Data Source
This population-based cohort study used the Taiwan National Health Insurance Research Database (NHIRD), which has been prospectively collecting nationwide healthcare data since the implementation of Taiwan National Health Insurance (NHI) in 1995. Because NHI is a single-player compulsory program that covers all forms of healthcare for residents in Taiwan, the NHIRD comprehensively includes claim data on both outpatient and inpatient services for nearly the entire 23.7 million population of the country.[19-21] Data acquisition and the study protocol was approved by the National Health Research Institute of Taiwan (NHIRD-102-020) and the Institutional Review Board at the E-Da Hospital (EMRP29101N), respectively.
Identification and Definition of Study Cohorts
This study enrolled diabetic patients ages 20 to 70 years who had been continuously using diabetes medication for more than 90 days. Therefore, eligibility required not only a specific diagnosis of DM (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code 250), but also ongoing use of diabetes medication (Supporting Table S1). In order to optimize comparability among the study cohorts, patients were not enrolled if they had the following serious comorbidity or condition that might bias the use of antiviral drugs: malignancy, chronic kidney disease, acute coronary syndrome (ACS), stroke, hepatitis B virus infection, esophageal or gastric varices, complications of hepatic decompensation including hepatic encephalopathy, hepatorenal syndrome, and ascites, severe psychosis or mood disorder, or any serious illness certified in the Registry for Catastrophic Illness Patient Database (RCIPD), a subpart of the NHIRD.[19-21] The ICD-9-CM codes used for disease definitions and conditions are detailed in the Supporting material (Tables S2, S3). Eligible enrollees were grouped into three cohorts according to HCV infection and antiviral therapy.
The definition of HCV infection was based on measurement of serum antibody against HCV and a diagnosis with the specific ICD-9-CM code (070.41, 070.44, 070.51, 070.54, 070.70, 070.71, V02.62). Among HCV-infected patients, those who received antiviral treatment with pegylated interferon (Peg-IFN) plus ribavirin between October 1, 2003, and December 31, 2010, were grouped into the treated cohort. This combination regimen has been reimbursed for HCV-seropositive patients with liver fibrosis or active viral hepatitis since October 1, 2003. Therapeutic duration generally ranged from 16 weeks to 48 weeks, according to the administration date, viral genotype, serum viral load, on-treatment virological response, and patient tolerability.[22, 23]
The treated cohort was matched with two other distinct diabetic cohorts in the propensity score that was calculated to adjust for the baseline differences between patients with and those without antiviral therapy. The propensity score was estimated by the logistic regression built on demographic factors, comorbidity, and diabetes medication. The untreated cohort consisted of diabetic patients with HCV infection who never received any prescription of interferon or ribavirin throughout the study period. Each untreated patient was matched with one treated patient. The uninfected cohort, which was matched 4:1 with their treated counterpart, comprised diabetic patients never coded for HCV infection. The baseline for matching was set at the day when antiviral treatment commenced in the treated cohort and the corresponding calendar date in the untreated and uninfected cohorts.
Definition and Ascertainment of Outcomes
The treated cohort was followed up after administration of an antiviral regimen, whereas the untreated and uninfected cohorts after the matched calendar date. All participants were observed for the occurrence of outcomes, until death or December 31, 2011, whichever came first. Enrolled subjects were followed up for renal and cardiovascular outcomes that included endstage renal disease (ESRD), ACS, and ischemic stroke. ESRD was defined as irreversible renal failure that necessitated maintenance dialysis and should be ascertained by the certification in the RCIPD (HV type 4). The occurrences of ACS and ischemic stroke were defined by hospitalizations with their disease codes as the primary diagnosis (ICD-9-CM codes: 410-414 for ACS and 433-434 for ischemic stroke, respectively). Because antiviral therapy has been shown to decrease mortality in HCV-infected patients, censoring resulting from death was regarded as informative and was adjusted by using competing risk analyses.
Adjustment for Confounding Factors
In addition to excluding severe illness that might have affected the prescription of antiviral regimen, we adjusted for several comorbidities including hypertension, dyslipidemia, chronic obstructive lung disease, and peripheral arterial occlusive disease as potential confounders. The definition of hyperlipidemia and hypertension required both the specific ICD-9-CM codes and the use of disease-defining medications for a minimum of 90 days. We also took into consideration the impact of pharmacotherapies. Antidiabetes agents were classified according to type (metformin, oral agent other than metformin, or insulin) and number (single, two, or more than two oral drugs) of the medication. Patients were classified as insulin-dependent if they had been using insulin every day for more than 6 months. In contrast, metformin-only users exclusively took metformin. Potential confounding by aspirin, nonsteroidal antiinflammatory drug (NSAID), angiotensin converting enzyme inhibitor (ACEI), angiotensin receptor blocker (ARB), and statin were also accounted for. Exposure to these drugs was defined as an average frequency of at least one tablet per month.
Study outcomes were adjusted for death as the competing risk event. The modified Kaplan-Meier method and Gray's method were used to calculate and to compare the cumulative incidences of outcomes. After ensuring the assumption of proportional hazards, we applied the modified Cox proportional hazard model to examine the independent association of antiviral therapy with outcomes. Models were developed by assessing goodness-of-fit. The effect of antiviral treatment among patients with HCV infection was further explored in stratified analyses. Data were managed with SAS software (v. 9.2, SAS Institute, Cary, NC). The cumulative incidence and hazard ratio (HR) in the competing risk analysis were calculated by the R software with the “cmprsk_2.1-4” package (http://biowww.dfci.harvard.edu/∼gray/). Continuous variables were summarized with mean ± SD, and categorical variables with number and proportion. Estimated results were expressed along with 95% confidence intervals (CIs). All analyses were two-sided with significance set at P < 0.05.
Baseline Characteristics of the Study Population
From a total of 2,267,270 individuals with a diagnosis of DM between January 1, 1997, and December 31, 2011, we identified 746,280 diabetic patients without HBV infection or serious comorbid conditions. There were 3,957 patients who had HCV infection and ever received Peg-IFN plus ribavirin, 20,239 patients whose HCV infection was never treated, and 720,302 patients without HCV infection. Among them, 1,411 treated patients were eligible as the treated cohort (therapeutic duration ≧16 weeks in 1,321 patients or 93.6%). Therefore, the untreated and uninfected cohorts consisted of 1,411 and 5,644 patients, respectively (Fig. 1). These three diabetic cohorts, which were matched in the propensity scores, did not differ in demographic factors, comorbidities, or diabetes medication, although exposure to aspirin, NSAID, and statin was not similar (Table 1).
Table 1. Baseline Characteristics and Follow-Up Status of the Three Study Cohorts
Incidences of Death, ESRD, Ischemic Stroke, and ACS in Three Study Cohorts
The treated, untreated, and uninfected cohorts were followed up until death for a mean ± SD duration of 3.8 ± 2.2 years, 3.7 ± 2.2 years, and 3.8 ± 2.2 years, respectively, with the longest observation of 8 years (from October, 2003 to December, 2011). The cumulative incidence of death at 8 years was significantly highest in the untreated cohort (23.6%; 95% CI, 19.1-28.1%), as compared with the treated (13.0%; 95% CI, 8.3-17.7%) and uninfected cohorts (11.4%; 95% CI, 9.68-13.17%), respectively (P < 0.001).
The incidences of ESRD, ischemic stroke, and ACS were all lowest in the treated cohort. ESRD occurred cumulatively at 8 years in 1.1% (95% CI, 0.3-2.0%), 9.3% (95% CI, 5.9-12.7%), and 3.3% (95% CI, 2.3-4.3%) of the treated, untreated, and uninfected cohorts, respectively (P < 0.001; Fig. 2). The 8-year cumulative incidence of ischemic stroke was 3.1% (95% CI, 1.1-5.0%), 5.3% (95% CI, 3.0-7.5%), and 6.1% (95% CI, 4.8-7.4%) in the treated, untreated, and uninfected cohorts, respectively (P = 0.01; Fig. 3). That of ACS was 4.1% (95% CI, 2.1-6.1%), 6.6% (95% CI, 3.7-9.5%), and 7.4% (95% CI, 5.9-9.0%) in the treated, untreated, and uninfected cohorts, respectively (P = 0.05; Fig. 4).
Antiviral Treatment in Association With Renal and Cardiovascular Outcomes After Multivariate Adjustment
The association of treatment for HCV infection with lower risks of ESRD, ischemic stroke, and ACS was ascertained in the multivariate-adjusted analyses. As compared with the untreated control, the treated cohort was associated with a multivariate-adjusted HR of 0.16 (95% CI, 0.07-0.33) for ESRD (Fig. 5), 0.53 (95% CI, 0.30-0.93) for ischemic stroke (Fig. 6), and 0.64 (95% CI, 0.39-1.06) for ACS (Fig. 7), respectively. Antiviral treatment was consistently associated with an attenuated risk of ESRD across all subgroups (Fig. 5). On the other hand, the association with a lower risk of ischemic stroke was observed in most strata, except for patients with peripheral arterial occlusive disease and metformin-only users (Fig. 6). Similarly, the association with ACS was compatible in most subgroups but not the one with peripheral artery disease (Fig. 7).
In the multivariate-adjusted Cox proportional hazard models, which were built on all three cohorts, the risks of ESRD, ischemic stroke, and ACS were all lowest in the treated cohort (Supporting Tables S4-S6). Taking the treated cohort as the reference, the Cox analysis demonstrated that the adjusted HRs for ESRD were 6.58 (95% CI, 3.11-13.95) and 1.92 (95% CI, 0.92-4.03) in the untreated and uninfected cohorts, respectively. The adjusted HR for ischemic stroke in the untreated cohort was 1.97 (95% CI, 1.12-3.48) and that in the uninfected cohort was 2.07 (95% CI, 1.26-3.40). Finally, the adjusted HRs for ACS were 1.59 (95% CI, 0.97-2.63) and 1.65 (95% CI, 1.08-2.51) in the untreated and uninfected cohorts, respectively.
This nationwide population-based research reveals that antiviral therapy for concomitant HCV infection is associated with improved renal and cardiovascular outcomes in patients with DM. The incidences of ESRD, ischemic stroke, and ACS were all lower in HCV-infected patients treated with Peg-IFN and ribavirin, as compared with infected individuals without antiviral treatment or diabetic patients without HCV infection. By performing multivariate-adjusted analyses that accounted for various potential confounders including death as a competing cause of risk, we were able to estimate that antiviral treatment was associated with an 84% reduction in the risk of ESRD (adjusted HR, 0.16; 95% CI, 0.07-0.33), 47% in that of ischemic stroke (adjusted HR, 0.53; 95% CI, 0.30-0.93), and 36% in ACS (adjusted HR, 0.64; 95% CI, 0.39-1.06) over an 8-year study period. These findings suggest that HCV infection may have a role in the pathogenesis of renal and cardiovascular complications among diabetic patients, and also implicate that treatment of concomitant HCV infection may improve clinical outcomes related to DM.
The efficacy of anti-HCV therapy in ameliorating insulin resistance and restoring glucose homeostasis, which has been convincingly demonstrated in previous studies,[15-17] may underlie the associations uncovered in our research. The mechanism through which antiviral therapy alleviates insulin resistance has not been fully elucidated but is most likely mediated via viral clearance, instead of direct pharmacological effects of interferons or ribavirin. Conjeevaram et al. reported that successful viral eradication was essential to sustain the beneficial effects in insulin resistance. Weight loss is frequent during the interferon-based therapy, but is usually transient and unlikely the reason for clinical benefits. We believe that our observation should result from viral elimination in the treated patients, although this study regrettably could not determine the status of virological response. Direct measurement of therapeutic response was not possible in that linking from the NHIRD to individual patients' laboratory results was forbidden for privacy protection. Nonetheless, we are confident of the antiviral efficacy in the treated cohort because Peg-IFN plus ribavirin generally achieves an eradication rate exceeding 70% in Taiwan,[26, 27] where a favorable genetic variation in IL28B is prevalent.[28, 29] Further research is warranted to clarify how the therapeutic response correlates with outcomes.
In addition to amelioration of insulin resistance, antiviral therapy may also improve renal and cardiovascular outcomes through other mechanisms. It has long been noted that HCV infection may cause mixed cryoglobulinemia and immune-mediated glomerulonephritis,[30, 31] which can progress to renal failure. Moreover, a growing body of data supports that the interferon-based regimen is able to confer renal protection in patients with HCV-associated glomerulonephritis. This may also contribute to the striking reduction of ESRD in our treated cohort. In fact, the significantly higher incidence of ESRD in the untreated HCV-infected cohort as compared with the uninfected counterpart implied that diabetic complication was not the only reason for renal failure.
Whereas the association between antiviral treatment and reduced cardiovascular events was consistent in most subgroups, it is interesting to note that both the risk of ischemic stroke (adjusted HR, 0.94; 95% CI, 0.14-6.63) and that of ACS (adjusted HR, 2.90; 95% CI, 0.30-27.58) were not attenuated in treated patients with peripheral arterial disease. Since the presence of atherosclerosis in the peripheral artery may indicate a more diffuse process, this finding suggests that the pathogenic role of HCV is limited at the early phase of atherosclerosis and that antiviral treatment cannot reduce cardiovascular morbidity at an advanced stage. The other plausible explanation for this discrepant result is insufficient statistical power in this subgroup, with only 95 and 91 subjects in the treated and untreated cohorts, respectively. Currently, little has been unraveled regarding the impact of HCV infection on the development, progression, and manifestation of atherosclerotic disorders.
This study undertook several methods to minimize bias and potential confounding. First, enrolled participants were systemically identified from a nationwide population with a diagnosis of DM. We utilized restrict eligibility criteria to exclude patients with severe physical or psychological conditions that might contraindicate interferons or ribavirin. Besides, enrolled subjects were matched in the propensity scores. Furthermore, the time when patients received antiviral treatment was deliberately chosen as the baseline of matching and entry of observation. All these measures optimized comparability among study cohorts and precluded the concern of immortal time bias. Moreover, we adopted an “intent to treat” approach to distinguish cohorts, in order to mitigate possible selection bias related to completion of antiviral therapy. Finally, competing risk analysis that recognized the informative censoring by death prevented overestimation of nonfatal outcomes in the untreated patients, in whom the mortality was significantly higher.
Several limitations are recognized and discussed. First, our findings should not be extrapolated to patients with significant comorbidity, who were explicitly excluded to ensure internal validity. Recruitment of these patients would have resulted in selection bias, for whom interferon or ribavirin might be contraindicated.[22, 23] Second, given that the Taiwan NHIRD did not contain direct laboratory results, we were unable to determine how viral genotype and viral load might influence the outcomes. We also could not clarify the correlation of virological response with outcomes. Third, the actual adherence to prescribed medication was unknown. Nonetheless, excessive prescription is improbable by reason of the strict regulations for antivirals in Taiwan. Fourth, this study lacked information of several important covariates such as the length of diabetic history, body weight, cigarette smoking, alcohol consumption, physical inactivity, and presence of steatohepatitis. Finally, antiviral efficacy of Peg-IFN plus ribavirin is considerably higher in Taiwan than in most Western countries, largely as a result of IL28B genotypic polymorphism.[28, 29] Caution is thus recommended before directly applying our results to the West.
In conclusion, antiviral treatment with Peg-IFN plus ribavirin is associated with improved renal and cardiovascular outcomes in diabetic patients with concomitant HCV infection. Risks of ESRD, ischemic stroke, and ACS are significantly reduced in HCV-infected patients who received antiviral therapy, as compared with untreated controls. These findings imply that HCV infection may have a pathogenic role in the development of clinical complications related to DM. The causal relationship and pathophysiological mechanisms underlying this association warrant further research.
The authors thank Tzu-Ting Chen for assistance in data analysis. This study is based on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health, and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health or National Health Research Institutes. Preliminary results presented at the 20th United European Gastroenterology Week (UEGW 2012) on October 23, 2012, Amsterdam, the Netherlands (OP242A).