Add-on neurological benefits of antiviral therapy in HCV patients with chronic kidney disease — a nationwide cohort study

Data source

The data source for this population-based longitudinal cohort study is the Taiwan National Health Insurance Research Database (NHIRD), which has been prospectively recording claim data of all forms of healthcare service in Taiwan since 1995 [16]. Because the NHI program is single-player and compulsory for citizens and legal residents, it covers almost all 23 million residents in Taiwan. By the end of 2014, > 99.6% of the population of Taiwan was enrolled in the NHI, and the contract rate was >93% [17]. In the NHIRD, disease is coded according to the International Classification of Diseases, Revision 9, Clinical Modification (ICD-9-CM). Patient information in the NHIRD is de-identified. The study protocol was approved by the Chang-Gung Memorial Hospital Research Ethics Committee (Institutional Review Board number: 103-6439B).

Patients with specific diagnosis codes (one inpatient or two outpatient codes) for HCV (070.41, 070.44, 070.51, 070.54, 070.70, 070.71, V02.62) recorded between 1 January 2003 and 31 December 2013 were identified.

Eligibility criteria and identification of study cohorts

First, we excluded HCV-infected patients without a claims-based diagnosis of CKD (Additional file 1) including dialysis (ICD-9-CM code 585) and registered in the Catastrophic Illness Patient Database (CIPD). Second, we excluded patients <20 years old and >80 years, co-infected with hepatitis B virus (HBV) or comorbidities possibly associated with interferon intolerance: stroke, myocardial infarction, heart failure, hepatic decompensation (esophageal or gastric varices, hepatic encephalopathy, hepatorenal syndrome, portal hypertension, and ascites), hepatic surgery or liver transplantation, all malignancies, catastrophic illness for autoimmune disease, and severe psychosis/mental disorder (Additional file 1). Additionally, to reduce survival bias, all patients who had died, had clinical outcomes, or had been lost to follow-up within 6 months after the index date were excluded. We finally enrolled 9872 HCV-infected patients with CKD in the study.

Patients undergoing anti-viral pegylated interferon (Peg-IFN) treatment (n = 1684) were assigned to the IBT-treated cohort and patients, who had never undergone interferon or ribavirin treatment throughout the study period (n = 8188) were assigned to the IBT-untreated cohort. Each patient in the treated cohort was matched with one patient in the untreated cohort using the probability calculated by logistic regression based on the following confounding variables: gender, age, hospital levels, comorbidities, dialysis status, and medications (listed in Table 1). Figure 1 shows a detailed flowchart of the study. The index date was set as the date when HCV was first diagnosed in the untreated cohort and as the date when interferon was first prescribed in the treated cohort.

Fig. 1

Flowchart of using stepwise exclusion to identify and enroll patients

Definition of Comorbidities and drug exposure

The comorbidities were diagnosed and then recorded in NHIRD using ICD-9-CM codes. All of the comorbidities listed in this study were confirmed on at least two consecutive clinic visits or whenever the patient had been hospitalized before the index date. In addition, some comorbidities were also reconfirmed using long-term associated medications. The definitions of the comorbidities are listed in the Additional file 1. Personal socio-economic status was divided to five levels according to monthly income (<10,000, 10,000 ~ 19,999, 20,000 ~ 29,999, 30,000 ~ 39,999, ≥ 40,000 New Taiwan Dollar [NTD]). Based on Tzeng’s classification [18], we modified the score and categorized all Taiwan areas into four levels (most to least urbanization). Drug information—drug names, date of prescription, and duration of use—was obtained from the outpatient pharmacy prescription database. Long-term drug use was defined as a prescription for at least 6 months after the index date.

The therapeutic duration of IBT generally ranged between 16 and 48 weeks [19] based on viral information (genotype and virological response) and the patient’s ability to tolerate the treatment. According to the response-guided trial [20] and Taiwan Health insurance regulations, we will check early viral response (EVR) at week 12 and premature termination (before 16 weeks) usually results from either intolerance to side effects or failure of early virological response or death. Treated patients were consequently subgrouped based on whether or not their antiviral treatment was ≥16 weeks or <16 weeks. “Incompletely treated” cohort means “treated <16 weeks”.

Definition and ascertainment of outcomes

Enrollees were observed until occurrence of the clinical outcomes, death, or 31 December 2013, whichever came first. The clinical outcomes were all-cause mortality, acute myocardial infarction, ischemic stroke, and hemorrhagic stroke, all of which were confirmed based on the primary diagnosis of the index hospitalization. These outcomes have been validated before and confirmed as highly accuracy [8, 9]. The other outcome was new-onset dialysis, which was defined as first-time and necessary maintenance dialysis by certification in the CIPD registry.

Statistical analysis

We performed propensity score matching (PSM) before comparing the treated cohort with the untreated cohort because there was substantial difference between the two cohorts that might confound the result [21]. Covariates to be calculated in the propensity score included index year (as the year of IBT) and variables listed in Table 1 (except for the follow up year). Propensity score indicates the predicted probability of the logistic regression and each patient in the treated cohort was matched with a corresponding patient in the untreated cohort. To warranty the quality of matching, the tolerance level on the maximum propensity score distance (caliper) was set as 0.2 × standard deviation of the propensity score under the greedy nearest neighbor matching algorithm.

We compared the patient’s characteristics between groups (i.e. treated cohort vs. untreated cohort) using independent sample t-test for continuous variable or using chi-square test for categorical variable. The incidence of outcome (mortality, acute myocardial infarction, ischemic stroke, hemorrhagic stroke, and new onset dialysis) was estimated using incidence density as number of event per 1000 person-years since entry of index date. The time-to-event between the two cohorts was compared using a Cox proportional hazard analysis adjusted for the propensity score. In an additional subgroup analysis, we stratified the treated cohort into two groups (incompletely treated <16 weeks and treated ≥16 weeks) and compared the time-to-event between the groups using a log-rank test. Data analysis was performed with SAS Version 9.3 (SAS Institute, Cary, NC).

Baseline characteristics of the study population

We identified 17,792 HCV-infected patients with CKD among the 242,568 patients diagnosed with HCV between January 1, 2003, and December 31, 2013. After patients who did not meet our inclusion criteria had been excluded, 1684 patients (mean age: 57.7 ± 10.2 years; men: 65%) were in the pre-PSM treated group, and 8188 (mean age: 60.5 ± 11.8 years; men: 53%) were in the pre-PSM untreated group (Table 1). Most patients also accepted antiviral therapy at medical center and regional hospital. Despite higher income status in pre-PSM treated group, the residential urbanization level was insignificantly different among both cohorts. The pre-PSM treated HCV patients had a higher prevalence of diabetes and hypertension and they tended to receive oral hypoglycemia agents and ACEi/ARB other than CCB. Only 15% of the patients in the pre-PSM treated cohort were on dialysis, but 38% in the pre-PSM untreated cohort were. The prevalence of liver cirrhosis without hepatic decompensation was higher in the pre-PSM treated than in untreated cohorts (11.1% vs. 6.9%, P < 0.001). Finally, after 1:1 PSM, the distribution of demographic factors, hospital levels, comorbidities, and associated medications were not significantly different between the two cohorts (Table 1).

Outcomes during the 10-year longterm follow up

Outcomes were assessed after 1:1 PSM, which yielded 1531 treated patients and 1531 untreated patients. The annual incidence of all-cause mortality was apparently higher in the untreated cohort (incidence density [ID], 42.89; 95% confidence interval [CI], 37.43–48.35) than in the treated cohort (ID: 19.00; 95% CI: 15.45–22.55), which was significantly different (adjusted hazard ratio [aHR], 0.45; 95% CI: 0.36–0.56, P < 0.001). The risk of secondary outcomes, including acute myocardial infarction and ischemic stroke, were comparable between the two cohorts. Noticeably, the ID of hemorrhagic stroke was 1.21 (95% CI: 0.31–2.11) in the treated cohort and 4.19 (95% CI: 2.48–5.91) in the untreated cohort, a significantly different risk reduction in the treated cohort (aHR: 0.31; 95% CI, 0.13–0.71, P = 0.006). The cumulative incidence between the two cohorts for all-cause mortality is plotted in Fig. 2a, and for hemorrhagic stroke in Fig. 2b.

Fig. 2

a All-cause mortality survival curves in the treated and untreated cohorts; b incidence density of hemorrhagic stroke between the treated and untreated cohorts

An evaluation of the effect of dialysis on IBT showed that all-cause mortality reduction was less efficacy for HCV patients on dialysis than for HCV patients not on dialysis (aHR: 0.80 vs. 0.39; P for interaction = 0.018). The effect of IBT on the other three outcomes was comparable between the two cohorts (Additional file 2).

Effect of incomplete IBT on clinical outcomes

We stratified the treated cohort into two groups (incompletely treated <16 weeks and treated 16 ≥ weeks). The mortality rate was higher in the incompletely treated (<16 weeks) cohort than in the treated (≥16 weeks) cohort (P < 0.001), but it was comparable with the untreated cohort (P = 0.118) (Fig. 3a). The risk of hemorrhagic stroke in the incompletely treated cohort was comparable with the risks in the other two groups (Fig. 3b).

Fig. 3

a All-cause mortality survival curves in the “treated ≥16 weeks” and incompletely treated (<16 weeks) and untreated cohorts; b incidence density of hemorrhagic stroke in the three cohorts based on the duration of therapy

Patients in the incompletely treated cohort tended to be older, have a higher prevalence of liver cirrhosis and dialysis, took more beta blockers, number of anti-HTN agents and diuretics, including spironolactone, a potassium-sparing diuretic (Additional file 3).

Renoprotective effect in non-dialytic patients

After patients on dialysis before the index date had been excluded, antiviral therapy was associated with a lower risk of new-onset dialysis (aHR: 0.31; 95% CI: 0.20–0.48, P < 0.001) (Table 2 and Fig. 4a). Additionally, the incidence of new-onset dialysis in the incompletely treated (<16 weeks) cohort was greater than that in the treated (≥16 weeks) cohort (P < 0.001), but it was comparable with that of the untreated cohort (P = 0.357) (Fig. 4b).

Fig. 4

a Incidence density of new-onset dialysis between the treated and untreated cohorts in CKD patients without dialysis; b the treated cohort was subgrouped by therapeutic duration

Survival benefits of IBT in HCV-infected patients with comorbid CKD

High mortality of chronic HCV infection is attributable not only to hepatic decompensation, liver cirrhosis, and hepatocellular carcinoma (HCC), but also to multiple extrahepatic complications, which can be ameliorated by using complete IBT in HCV-infected patients without CKD [8, 9]. Because of the genotype distribution of 50% for HCV-1 and 45% for HCV-2 in Taiwan [19], sustained viral response (SVR) for HCV with comorbid CKD or dialysis was 50–64% [10, 11]. Hsu et al. [8] demonstrated a small reduction of cardiovascular events (<20%) in a treated cohort, which was comparable with the findings of a study on diabetes [9]. However, our study showed neural cardiovascular effects of antiviral therapy in HCV-infected patients with CKD, which might be related to drug intolerability, IBT inefficacy, and severe atherosclerosis in CKD patients, but not in patients without CKD.

IBT might reduce all-cause mortality because it ameliorates the hepatic complications of decompensation, liver cirrhosis, and HCC, or non-hepatic problems like a tendency to hemorrhage, immunocompromised sepsis, advanced renal impairment, hemodialysis, etc.

Cerebrovascular protection of IBT in patients with chronic HCV and CKD

Our study showed a significant reduction of hemorrhagic stroke (10-year hemorrhagic stroke incidence rate: untreated = 1.5% vs. treated = 0.46%), which is comparable to the findings of Arase et al. [22] on HCV clearance (10-year incidence rate in HCV non-clearance = 1.1% vs. clearance = 0.4%; P = 0.02). However, their study lacked renal function data before IBT and background matching between cohorts. Despite conventional factors like age, being male, smoking, hypertension, diabetes, and drinking alcohol [23], HCV contributed to hemorrhagic stroke [24, 25], liver cirrhosis [26–28], and unstable hemostasis [29]. Moreover, HCV was critically involved in systemic atherosclerosis, vascular remodeling, ectasia change, and microaneurysms, which manifested as cerebral microbleeds (CMBs) in the MRIs of patients with advanced hepatic fibrosis [30]. In addition to viral replication and HCV receptor expression in microvascular endothelium [31], neurological vasculitis was highly prevalent with mixed cryoglobulinemia in 25–60% of HCV-infected patients [32, 33]. Both consequences increased the risk of microvascular destruction and disturbance of the brain–blood-barrier related to spontaneous hemorrhagic events. Indeed, the protection of complete IBT against hemorrhagic stroke comes not only because it ameliorates advanced liver cirrhosis, but also because it attenuates extrahepatic immunological complications.

Both the brain and the kidneys are the end-organs for high blood pressure and autoregulation in hypertensive patients [34]. Thus, CKD might be another risk factor of hemorrhagic stroke because of an altered response to antiplatelet agents, platelet dysfunction, endothelial hyper-permeability [35], microvascular dysregulation [36], and occult CMBs [37], Our subgroup analysis showed that IBT promised a greater reduction of new-onset dialysis and constant protection against hemorrhagic stroke, regardless of whether the patient was on dialysis, and that antiviral therapy provided many benefits for the liver, kidney, and brain triangle.

Effect of incomplete IBT on clinical outcomes

According to a meta-analysis [11], the dropout rate in patients on dialysis was 18%. Our finding was 14.4% for incompletely treated patients. Both rates were higher than that of Hsu et al. [8, 9]: 10.4% in a general population without CKD. Hsu et al. showed that patients who underwent incomplete IBT were older, as were our patients, who also had more liver cirrhosis and more dialysis. Additionally, many incompletely treated patients had been prescribed diuretics because they had more fluid overloading or ascites due to hepatic decompensation. Despite the loss of benefits because of incomplete IBT, our analysis and other studies also found slightly higher mortality in patients treated for <16 weeks, possibly because they had more critical comorbidities, marked anemia, immunocompromised status, or malnutrition. In addition, interferon might cause adverse effects or complications, including heart failure and infection [38]. Because of high mortality in the incompletely treated group, identifying the precipitating factors before starting IBT is critical.

Limitations

This study has some limitations. First, details of the HCV genotype, IL-28B polymorphism, virus load, viral response, serum creatinine or eGFR, duration of renal impairment, duration of dialysis, platelet and coagulation function, echogenic study, degree of hepatic fibrosis, body weight, and personal lifestyles are not included in the Taiwan NHIRD. However, by carefully excluding confounding comorbidities and estimating associated variables, including medications, we obtained information that contributed to outcomes. Second, the misdiagnoses of CKD or cirrhosis might occur because of the ICD-9 codes used in several NHIRD-based nationwide cohort studies, but we used the standard methodology (1 inpatient or 2 outpatient diagnosis codes) to identify CKD patients in the NHIRD. Underdiagnosis of cirrhosis might also occur in CKD or dialysis population due to relatively low aminotransferase [39]. Moreover, our study tried to the gap of longterm outcomes in HCV-infected patients with CKD, which is difficult to do in clinical researched because of individual decision making unpredictable dose adjustments at various stages. Third, alcohol consumption might be involved in hemorrhagic stroke because of alcohol-induced hypertension and coagulation disorders, whether or not the patient has cirrhosis [23]. Moreover, we had carefully excluded alcohol-related liver cirrhosis the number of confounding factors as much as possible. Fourth, variable doses of ribavirin according to renal function, body weight or drug tolerance might influence effect of viral eradication, but all treated patients would receive guided interferon therapy. Fifth, the effects of events that occurred within the therapeutic course (<6 months) and their association with clinical benefits, adverse side effects of drugs, and underlying comorbidities were not clear. Therefore, we excluded patients who had died, or had a myocardial infarction, ischemic stroke, hemorrhagic stroke, new-onset dialysis, or who were lost to follow-up within 6 months, which is different from other studies [8, 9].