Interferon for Hepatitis C Virus in Hemodialysis—an Individual Patient Meta-analysis of Factors Associated with Sustained Virological Response

Abstract

Background and objectives: Hepatitis C virus (HCV) infection is prevalent in hemodialysis patients and causes excess mortality. Interferon (IFN) treatment of chronic HCV infection in hemodialysis patients results in high sustained virological response (SVR) rates 6 mo after treatment. The authors aimed to identify factors associated with SVR in hemodialysis patients through analysis of individual patient data obtained from systematic review of published literature.

Design, setting, participants & measurements: Medline was searched from 1966 through February 2009, and prospective studies describing IFN treatment of hemodialysis patients with chronic HCV infection with published individual patient data were included. To identify factors associated with SVR, logistic regression was applied with adjustment for study.

Results: Twenty studies of IFN treatment provided data on 428 patients. Overall SVR was 45% and in univariate analyses was higher with: 1) three million units or higher three times weekly of IFN; 2) treatment for at least 6 mo; 3) treatment completion; 4) lower baseline HCV RNA; 5) female gender; and 6) early virological negativity. Although limited by missing data, these relationships persisted in multivariate regression.

Conclusions: SVR is more likely with larger IFN dose, longer treatment duration, treatment completion, female gender, lower HCV RNA and early virological negativity. For appropriate treatment candidates, regimens should consist of three million units of IFN three times weekly for at least 6 mo, with patients encouraged to complete the full course.

Hepatitis C virus (HCV) infects an estimated 170 million people worldwide (1). The prevalence of HCV in hemodialysis (HD) patients ranges from 3 to 23% in developed countries (2) and exceeds 50% in some developing countries (3). HCV-infected HD patients have higher mortality rates than noninfected HD patients, with reported relative risks from 1.25 to 1.57 (4,5). Untreated, spontaneous viral clearance occurs in only 0.5% of chronic HCV-infected patients per year (6). The standard measure of treatment success, sustained virological response (SVR), is defined as achieving HCV RNA negativity six months after treatment completion. In non-HD patients, interferon (IFN) monotherapy achieves SVR in 9 to 22% of patients (7–9) but combination pegylated IFN and ribavirin achieves SVR in 50 to 60% (10,11). However, IFN and ribavirin are associated with significant toxicity including influenza-like symptoms, anemia and depression with IFN (7–9), and hemolytic anemia with ribavirin (7,8).

Most studies of HCV-infected HD patients have investigated IFN monotherapy. Only recently, studies have explored pegylated IFN or ribavirin (12–18). IFN treatment after kidney transplantation is associated with increased rates of allograft rejection (19,20), so treatment before transplantation is advised (21,22). Our recent meta-analysis of summary data in HCV-infected HD patients demonstrated an overall SVR rate of 41% with IFN (22), higher than rates in IFN-treated non-HD patients (7,8), but rates of treatment discontinuation due to adverse events were also higher (22).

Identifying factors associated with a higher likelihood of SVR among HD patients has important implications for selecting treatment candidates and the optimal treatment regimen. In non-HD patients, higher SVR rates are associated with younger age, female gender, lower patient weight, HCV genotypes other than 1, lower baseline HCV RNA, and absence of cirrhosis on liver biopsy (7–11). Early virological response (EVR), defined as a 2-log₁₀ or larger decrease in HCV RNA by the 12^th week of treatment, is a powerful predictor of SVR (23,24).

We investigated whether these previously identified factors associated with SVR in non-HD patients could be validated in the HD population. The majority of studies of IFN in HD patients reported individual patient data, allowing us to extend our prior subgroup analysis and meta-regression of summary data (22) to identify factors associated with SVR.

Materials and Methods

Study Selection

The literature search strategy and study selection process for our systematic review have been described previously (22). We included prospective studies of IFN-based treatment of HD patients with chronic HCV infection and with reported SVR at 5 mo or more after treatment completion. Studies involving kidney transplant recipients, CKD stages 1 to 5 not on dialysis, or not reporting SVR were excluded as were studies with sample sizes less than 10. Studies of pegylated IFN were excluded because pharmacokinetic differences may result in different SVR rates than with standard IFN. The literature search was updated on February 1, 2009.

Individual Patient Data Analysis

All studies specified treatment protocols such as dose and duration, but only 65% contained individual patient data. A single investigator (Craig E. Gordon) extracted published individual patient data and contacted the authors of all included studies directly to request additional unpublished data. If contact was unsuccessful, we inferred data from study inclusion and exclusion criteria (e.g., transaminase elevation or presence of cirrhosis) when able to do so with certainty. When data were unclear or required adjudication, additional authors were consulted until consensus was achieved. Based on these criteria, 20 studies of IFN treatment of HCV-infected HD patients met inclusion criteria (Figure 1) and one was excluded (16).

Figure 1.

Systematic review of the literature. GN, glomerulonephritis; IPD, individual patient data; PEG-IFN, pegylated IFN.

Statistical Analyses

Using individual patient data, we investigated the relationship between SVR and patient, viral, and treatment characteristics based on factors previously identified in non-HD patients (7–9), HD patients (25,26), and our previous meta-analysis of summary data (22). Patient factors included age, gender, duration of HD, and previous kidney transplantation. Viral factors included HCV genotype, baseline quantitative HCV RNA, cirrhosis or bridging fibrosis on liver biopsy, duration of HCV infection, and elevation of transaminases. Treatment characteristics included intended dose and duration of IFN and measures of treatment completion, including achieved treatment duration and the proportion of intended treatment completed. Because studies did not uniformly report 2-log₁₀ decrease in quantitative HCV RNA at 3 mo, the standard EVR definition, we instead defined “early virological negativity” (EVN) as HCV RNA clearance following 1 to 3 mo of treatment. We analyzed HCV RNA as a continuous (log₁₀ HCV RNA) and a categorical variable (cutoff of < or > = 400,000 IU/ml) based on exploratory analysis of the data. As in the original studies, we defined elevated transaminases as greater than 40 IU/L. Selecting lower cutoff values did not affect the relationship between this variable and SVR.

We applied univariate logistic regression using SAS version 9.1 (Cary, NC) to investigate the relationship between predictors and SVR. To account for potential confounding by study, we reanalyzed these relationships using study as a categorical set of dummy variables. We performed multivariate logistic regression using factors associated with SVR in univariate regression with p-values <0.05 and with clinically important factors. Multivariate analysis was limited because individual variables were often missing for entire studies. We also found unanticipated clustering of data; e.g., all patients identifiable by age, gender, HCV RNA and genotype were treated with 3 MU of IFN or higher and for 6 mo or longer. This clustering precluded multiple imputation and necessitated multiple sets of multivariate analyses using alternate subsets of studies.

To analyze the relationship of the above covariates and treatment discontinuation due to adverse events, we applied identical methods. We focused on treatment discontinuation as opposed to adverse events alone because it is more consistently reported and a marker of serious adverse events.

Results

Study Characteristics

Twenty studies met inclusion criteria (Figure 1) with individual patient data published in 13 studies (27–39) and obtained for three additional studies after contacting authors (40–42). Limited data based on inclusion criteria were imputed for the remaining four studies (43–46), but no further data were provided despite multiple attempts to contact authors.

The 20 studies of IFN treatment comprised 428 patients with varying completeness of data for specific covariates (Table 1). Patients had a mean age of 46 yr and 60% were male. The mean duration of HD was 72 mo and of HCV infection was 73 mo. Mean pretreatment HCV RNA was 363,307 IU/ml. The prevalence of HCV genotype 1 was 65%; 4% had cirrhosis, and 17% had bridging fibrosis on liver biopsy.

Table 1.

Characteristics of all patients and those subsets with data on HCV RNA, gender, or early virological negativity

Characteristic	Complete Dataset (n = 428)		HCV RNA Data (n = 112)		Gender Data (n = 275)		Early Virological Negativity Data (n = 196)
	Number with Data on Characteristica	Mean or Proportion	Number with Data on Characteristica	Mean or Proportion	Number with Data on Characteristica	Mean or Proportion	Number with Data on Characteristica	Mean or Proportion
Patient Characteristics
Age (yr)	275	46	86	41	275	46	148	46
Female gender (%)	275	40	86	33	275	40	148	41
Duration of hemodialysis (mo)	210	72	80	71	210	72	107	89
Prior kidney transplantation (%)	126	31	60	32	126	31	75	23
HCV Characteristics
HCV RNA (IU/ml)	112	363,307b	112	363,307	86	395,505	59	219,891
HCV genotype (% genotype 1)	220	65	110	51	135	56	113	67
Duration of HCV infection (mo)	175	73	78	62	153	74	129	77
LFTs, pre-treatment (% elevated)	219	78	63	62	160	76	87	67
Liver biopsy
Cirrhosis (%)	242c	4	70	4	172	4	141	4
Cirrhosis or bridging fibrosis (%)	188c	21	59	7	146	25	113	19
Treatment Characteristics
Interferon dose
1 to 1.5 MU (%)	24	6	0	0	0	0	0	0
3 MU (%)	342	80	112	100	213	77	149	76
5 to 6 MU (%)	62	14	0	0	62	23	47	24
Duration
4 mo (%)	37	9	0	0	37	14	37	19
6 mo (%)	129	31	41	37	74	28	45	23
12 mo (%)	253	60	71	63	155	58	114	58
Treatment complete (%)	419	75	112	85	266	76	196	80
Treatment length (mo)	419	8	103	9	267	8	196	8
Treatment Outcomes
Early virological negativity (%)	182	60	59	81	138	56	182	60
Sustained virological response (%)	411	45	98	32	271	49	182	50

^aNumber of patients with data on presence or absence of characteristic (e.g., number with data on any HCV genotype or results of liver biopsy).

^bReported value is mean HCV RNA. Median value was 100,952 IU/ml.

^cNumber of patients with data on cirrhosis greater because cirrhosis variable included studies where presence of cirrhosis was inclusion or exclusion criterion. Relaxing this assumption did not affect results of univariate or multivariate analyses.

LFT, liver function tests; MU, million units.

Because we identified a significant proportion of missing individual patient data, in particular for important factors such as HCV RNA, gender, and EVN, we examined baseline characteristics in the cohorts with data for these covariates and in the overall cohort. Most characteristics were similar both in the subpopulations and the overall cohort, and not surprisingly, the largest discrepancies occurred in the smaller subsets (those with data for HCV RNA and EVN).

Treatment Response

Overall SVR was 45% (95% CI, 40 to 50%). SVR was similar in the subpopulations with data on gender and EVN but was somewhat lower in patients with data on HCV RNA (Table 1). EVN was achieved in 60% (95% CI, 55 to 69%).

Univariate Logistic Regression of SVR

Univariate logistic regression (Table 2) demonstrated that 3 MU or higher of IFN (OR 3.3, 95% CI 1.2 to 9.1) and intended treatment duration of 6 mo or longer (OR 2.0, 1.1 to 3.9) both were associated with higher SVR. Treatment length, whether defined by months of actual treatment (OR 1.13, 1.07 to 1.19 per month), completion of the intended treatment (OR 4.1, 2.4 to 6.8), or proportion completed of intended treatment (OR 1.3, 1.2 to 1.4 per 10% of intended treatment), was associated with higher SVR. Lower baseline HCV RNA was associated with higher likelihood of SVR, whether analyzed continuously (OR 3.6, 1.9 to 6.7 per log₁₀ decrease in HCV RNA) or categorically (OR 11.1, 1.4 to 100.0 for HCV RNA < = 400,000 IU/ml). Female gender (OR 2.1, 1.3 to 3.5) and EVN (OR 5.1, 2.6 to 10.0) were also associated with increased odds of SVR. Patient age, duration of HCV or HD, HCV genotype, cirrhosis/bridging fibrosis, and other covariates were not statistically significantly associated with SVR. Limiting the analysis to the 16 studies with individual patient data had minimal or no effect on the results (results not shown).

Table 2.

Odds ratio of sustained virological response after interferon (univariate and bivariate logistic regression)

	Unadjusted		Adjusted for Study
	Number with Covariate	Odds Ratio (95% CI)	Odds Ratio (95% CI)
Patient Characteristics
Age (per additional 10 yr)	271	1.1 (0.9 to 1.3)	1.0 (0.8 to 1.3)
Female gender	271	2.1 (1.3 to 3.5)	2.4 (1.4 to 4.2)
Duration of hemodialysis (per month)	210	1.0 (0.9 to 1.1)	1.0 (0.9 to 1.1)
Prior kidney transplantation (versus none)	126	0.9 (0.4 to 2.0)	0.7 (0.3 to 1.3)
HCV Characteristics
HCV RNA (per log10 decrease)	98	3.6 (1.9 to 6.7)	3.1 (1.6 to 6.3)
HCV RNA (<= versus >400,000 IU/ml)	98	11.1 (1.4 to 100.0)	11.1 (1.3 to 100.0)
HCV genotype 1 (versus others)	207	1.1 (0.6 to 2.1)	0.9 (0.4 to 1.8)
Duration of HCV infection (per month)	162	1.0 (0.99 to 1.01)	1.0 (0.99 to 1.01)
Elevated transaminases, pretreatment	218	0.9 (0.5 to 1.7)	0.7 (0.3 to 1.6)
Cirrhosis	230	1.9 (0.5 to 7.4)	1.7 (0.4 to 7.4)
Cirrhosis or bridging fibrosis	177	0.9 (0.5 to 1.9)	0.6 (0.3 to 1.3)
Treatment Characteristics
Interferon dose: 3 MU or higher	411	3.3 (1.2 to 9.1)	3.2 (0.8 to 12.5)
Duration: 6 mo or longer	411	2.0 (1.1 to 3.9)	5.3 (0.8 to 36.3)
Treatment completed	402	4.1 (2.4 to 6.8)	6.4 (3.4 to 12.0)
Proportion completing Treatment (per 10%)	399	1.3 (1.2 to 1.4)	1.4 (1.3 to 1.6)
Treatment duration (per month)	403	1.13 (1.07 to 1.19)	1.3 (1.2 to 1.5)
Early Virological Negativity
Early virological negativity	172	5.1 (2.6 to 10.0)	13.1 (5.3 to 32.3)

Logistic Regression of SVR with Adjustment for Study

To account for possible unmeasured differences between studies that might confound the relationship between candidate variables and SVR, we performed logistic regression of SVR with adjustment for study (Table 2) and did not identify any additional factors. As might be expected, because of less power after adjusting for study, the effects of treatment dose and duration, while remaining large, were no longer statistically significant. Although the strong relationships with SVR for treatment completion and EVN remained statistically significant, estimates became less precise. The relationship between patient factors (gender, HCV RNA) and SVR remained virtually unchanged in magnitude and statistical significance.

Multivariate Logistic Regression of SVR

We performed multivariate logistic regression to determine whether factors identified in univariate analyses remained independently associated with SVR after adjustment for possible confounding variables (Table 3). Due to missing data, selection of a single model was not possible, so in separate analyses we demonstrated the independent association of increased SVR with 3 MU or higher of IFN and intended or achieved treatment duration (Models 1 to 3), as well as with female gender and intended or achieved treatment duration (Models 4 to 6).

Table 3.

Multivariate logistic regression of sustained virological response after interferon

		Odds Ratio (95% Confidence Interval)
		Number of Patients	Dose 3 MU or Highera	Female Gendera	Intended Treatment Duration (≥6 mo)	Completed Intended Treatment	Actual Treatment Duration (per month)	Log10 HCV RNA (per log10 decrease)b
Models using interferon dose 3 MU or highera
Model 1:	Interferon dose	411	3.7	a	2.2
	Intended duration		(1.3 to 10.0)		(1.1 to 4.3)
Model 2:	Interferon dose	402	3.9	a		4.2
	Completed treatment		(1.4 to 10.8)			(2.5 to 7.1)
Model 3:	Interferon dose	403	3.2	a			1.1
	Actual duration		(1.2 to 8.9)				(1.1 to 1.2)
Models using female gendera
Model 4:	Female gender	271	a	2.2	2.7
	Intended duration			(1.3 to 3.6)	(1.4 to 5.4)
Model 5:	Female gender	262	a	2.2		3.3
	Completed treatment			(1.3 to 3.7)		(1.8 to 6.1)
Model 6:	Female gender	263	a	2.3			1.1
	Actual duration			(1.4 to 3.9)			(1.1 to 1.2)
Models using log10 HCV RNA
Model 7:	HCV RNA	98	c		c	d	1.2	3.9
	Actual duration						(1.1 to 1.4)	(2.0 to 7.7)
Model 8:	HCV RNA	83	c	2.0	c	d		3.6
	Female gender			(0.7 to 6.3)				(1.8 to 7.1)
Model 9:	HCV RNA	83	c	1.9	c	d	1.3	4.0
	Actual duration			(0.6 to 6.3)			(1.1 to .5)	(2.0 to 8.3)
	Female gender

^aAll patients with data on dose or gender received 3 MU or higher of interferon. Multivariate models could not be developed with these variables.

^bSimilar relationships existed when HCV RNA analyzed using 400,000 IU/ml as a cutoff.

^cAll patients with data on HCV RNA were treated with 3 MU or higher of interferon and for an intended treatment course of 6 mo or longer.

^dUnable to evaluate models with HCV RNA and other measures of treatment completion due to nonconvergence of data.

Multivariate analysis of HCV RNA was limited because all patients with HCV RNA data received 3 MU or higher of IFN for 6 mo or longer, and because there were strong associations between treatment completion and reporting of HCV RNA. Actual treatment duration and HCV RNA were independently associated with SVR (Model 7), even after adjustment for gender (Model 9). Female gender maintained an OR of 2 when adjusted for HCV RNA (Model 8) and treatment duration (Model 9), but lost statistical significance. We did not investigate EVN in multivariate models because it is an intermediate outcome.

The relationship between SVR and clinically relevant but nonstatistically significant factors in univariate analyses remained nonsignificant in multivariate models. Because of their importance in non-HD patients, we forced HCV genotype and cirrhosis/bridging fibrosis into multivariate models, but they remained nonsignificant.

Treatment Discontinuation

All 20 studies provided data on treatment discontinuation for 428 patients. The rate of treatment discontinuation ranged from 0 to 50% with an overall rate of 26% (95% CI, 20 to 34%). We examined the same variables used in logistic regression of SVR with treatment discontinuation as the outcome (Table 4). Of these, longer duration of HD was associated with a reduced likelihood of treatment discontinuation (OR 0.88, 0.80 to 0.96) and previous kidney transplantation (OR 2.55, 1.03 to 6.31) with an increased likelihood. No other factors were associated with treatment discontinuation. Multivariate logistic regression demonstrated that the effect of previous kidney transplantation was of borderline significance (OR 2.5, 1.0 to 6.3), and duration of HD (OR 0.9, 0.8 to 1.0) was nonsignificant in two variable models.

Table 4.

Odds ratio of treatment discontinuation due to adverse events after interferon (univariate regression)^a

	Number with Covariate	Odds Ratio of Treatment Discontinuation	95% Confidence Interval
Patient Characteristics
Age (per year)	275	1.0	0.9 to 1.1
Female gender	275	1.2	0.7 to 2.1
Duration of hemodialysis (per month)	215	0.88	0.80 to 0.96b
Prior kidney transplantation (versus none)	131	2.55	1.03 to 6.31b
HCV Characteristics
Log10 HCV RNA (IU/ml(per log10 decrease)	112	0.9	0.5 to 1.4
HCV RNA <= 400,000 IU/ml	112	0.7	0.2 to 2.3
HCV genotype 1 (versus others)	225	0.9	0.5 to 1.6
Duration of HCV infection (per month)	175	1.0	0.99 to 1.01
Elevated transaminases, pretreatment	219	0.8	0.4 to 1.7
Cirrhosis	242	2.3	0.6 to 9.1
Cirrhosis + bridging fibrosis	188	1.9	0.8 to 4.5
Treatment Characteristics
Dose 3 MU or highera	428	1.9	0.6 to 5.6
Duration: 6 mo or longera	428	0.6	0.28 to 1.1

^aSensitivity analysis using adverse event as outcome did not reveal any statistically significant relationships except when compared with 3 MU and 1 to 1.5 MU dosing, 5 to 6 MU of interferon had an OR of an adverse event of 1.9 (95% confidence interval, 1.1 to 3.4) and 5.0 (1.5 to 16.3), respectively. The OR of an adverse event with 3 MU of interferon was 2.6 (0.9 to 7.7) when compared with 1 to 1.5 MU of interferon. Interestingly, the OR of 6 mo of intended treatment duration was 1.7 (1.1 to 2.6) when compared with 12 mo of treatment. We did not investigate treatment completion with adverse events because of a high likelihood of confounding.

^bBivariate models showed prior kidney transplantation remained marginally statistically significantly associated (2.55, 1.03 to 6.31) with increased likelihood of treatment discontinuation after adjustment for duration of hemodialysis, but the latter relationship was nonsignificant.

We performed sensitivity analysis using the likelihood of an adverse event as the outcome without regard to treatment discontinuation. No patient or viral characteristic was associated with adverse events. However, IFN dose of 5 to 6 MU was associated with an increased likelihood of adverse events when compared with 3 MU (OR 1.9, 1.1 to 3.4) or 1 to 1.5 MU (OR 5.0, 1.5 to 16.3). The odds of an adverse event were higher with 3 MU of IFN (OR 2.6, 0.9 to 7.7), when compared with 1 to 1.5 MU. Surprisingly, intended treatment duration of 6 versus 12 mo was associated with a higher likelihood (OR 1.7, 1.1 to 2.6) of adverse events, but this effect did not persist after adjustment for dose.

Discussion

The decision to treat a HD patient for HCV infection is complex and requires weighing the benefits and risks of treatment, durability of treatment response, risks of untreated HCV infection, co-morbid conditions, and transplantation candidacy. IFN treatment of chronic HCV infection in HD patients results in a 45% SVR, significantly higher than the 10 to 20% rate reported for IFN monotherapy in non-HD patients (7–9,22). This difference may be related to lower IFN clearance in HD patients with fixed dosing, but differences in patient, viral, or treatment characteristics may also contribute. This meta-analysis of individual patient data extends the present understanding of factors associated with SVR in HD patients and provides information that clinicians should consider in deciding which patients to treat and how to treat them.

IFN dose of 3 MU or higher, intended treatment duration of 6 mo or longer, actual treatment duration, female gender, lower baseline HCV RNA levels, and EVN were associated with statistically significant increases in the likelihood of SVR. These relationships persisted after we adjusted for study. Multivariate analysis demonstrated independence of these relationships although, due to missing data, we were unable to establish independence of all these predictors in a single model. However, we could show that actual treatment duration remained independently associated with SVR in models adjusted for dose, gender, or HCV RNA.

In non-HD patients, HCV genotypes 1, 4 and cirrhosis/bridging fibrosis are associated with lower SVR (7–10) but, surprisingly, neither of these had a statistically significant association with SVR in HD patients. The wide confidence intervals surrounding estimates for these variables, however, highlight the uncertainty about their true effect size. These nonsignificant relationships may represent important clinical differences in HD patients or insufficient statistical power because of missing data for both variables.

Few of the studied patient, viral, or treatment factors were associated with treatment discontinuation except increasing duration of HD and previous kidney transplantation, but these did not persist in bivariate models. Although there was no relationship between IFN dose and treatment discontinuation, we observed a statistically significant association between higher dose and the odds of developing an adverse event. Our previous analysis found that specific adverse events had different rates of treatment discontinuation (22), so increasing doses of IFN may lead to adverse events (e.g., influenza-like symptoms or anemia) that do not result in treatment discontinuation.

EVN was a strong predictor of SVR in univariate analysis. Achieving viral negativity anytime in the first 3 mo of treatment was associated with an SVR of 67% (positive predictive value), whereas 25% of HD patients who did not achieve EVN ultimately had an SVR yielding a negative predictive value of 75% (Figure 2).

Figure 2.

Impact of early virological negativity on sustained virological response. Data on early virological negativity and SVR were available for 182 patients, of whom 109 (60%) achieved early virological negativity. Seventy-three of the 109 patients who achieved EVN had SVR, equal to a positive predictive value of 67%. Of 73 patients who did not achieve EVN, 18 achieved SVR and the remainder were HCV RNA positive 6 mo after treatment. Thus, the negative predictive value of not achieving SVR after not achieving early virological negativity was 75%. If treatment was discontinued in patients not achieving EVN, 18 patients (20% of the total achieving SVR) would have missed the opportunity to become HCV RNA negative. *Value different from overall SVR because EVN data only available in 182 of 428 patients (see Table 1). ^†If treatment was discontinued in patients who did not achieve EVN, 18 (20%) of 91 total SVR would have been missed. ^‡Proportion of patients who would stop treatment early because they did not achieve EVN.

Because of limitations in reporting within the original studies, our definition of EVN differs from that used in non-HD patients, where EVR is defined as a 2 log₁₀ or larger decrease in HCV RNA or viral negativity after 3 mo of treatment. Nonetheless, it is similar to the recent complete EVR definition (HCV RNA negative at 12 wk) (47) in non-HD patients, and our results suggest that EVN in HD patients is also associated with a higher likelihood of SVR. In non-HD patients, however, the absence of an EVR implies a less than 3% likelihood of SVR with continued treatment (negative predictive value of 97 to 100%) (23,24), so treatment may be stopped for patients with HCV genotype 1 to avoid morbidity associated with further treatment (23). In contrast, the negative predictive value of EVN in HD patients is only 75%, so continued treatment could still achieve SVR in 25%. The 1- to 3-mo virological negativity definition used in studies of HD patients should not be used to stop treatment early. Instead, the more conservative 2-log₁₀ decrease in quantitative HCV RNA after 12 wk of treatment definition used in non-HD patients (23) could be considered for HD patients, but further research is needed. Determining the optimal definition for EVR in HD patients has important implications for kidney transplant candidates because transplantation is delayed during IFN treatment due to the risk of allograft rejection in IFN-treated patients who undergo deceased donor kidney transplantation (21).

Publication bias is always a concern in meta-analysis. We addressed this by searching the proceedings of recent nephrology and hepatology meetings, but did not identify additional studies of IFN for HCV-infected HD patients. Second, the generalizability of our results may be limited by selection bias because studies typically enrolled kidney transplantation candidates and excluded patients with psychiatric or hematologic diseases. Missing data were a particular problem for this analysis and limited our ability to investigate whether statistically significant univariate relationships persisted in complete multivariate models. Additionally, missing data may have decreased the statistical power to detect univariate relationships. Other important factors associated with SVR in non-HD patients, such as race and insulin resistance (47), were also unreported. Most of the included studies involved IFN monotherapy; but recent studies suggest that combined pegylated IFN and low-dose ribavirin, with careful monitoring of ribavirin plasma levels, may result in higher SVR rates (48). Finally, although SVR rates appear higher in IFN-treated HD patients, SVR is only a surrogate outcome. Future research should investigate whether achieving SVR in HD patients translates into improved clinical outcomes.

Despite these limitations, this study is the first to provide important information on predictors of SVR in IFN-treated HD patients using individual patient data. There appears to be a threshold dose of 3 MU three times weekly, below which SVR is less likely to be achieved. When compared with 3 MU of IFN, higher doses do not appear to increase the likelihood of SVR but result in higher adverse event rates. We did not identify a significant difference between 6 and 12 mo of treatment, but less than 6 mo of IFN resulted in lower SVR. Our analysis did not have the statistical power to study interactions between HCV genotype and treatment duration. Current guidelines in non-HD patients recommend that treatment lengths be tailored by genotype (47). In the absence of specific data, this may be extrapolated to HD patients.

In HD patients, there appears to be a strong effect of treatment completion. Patients should be encouraged to complete the intended treatment course, regardless of the selected treatment length. Achieving viral negativity in the first 3 mo of treatment appears to be a strong predictor of ultimately achieving SVR. However, EVN, as reported in included studies, had insufficient negative predictive value to guide decisions to terminate treatment early. Clinicians may consider adopting the 2-log₁₀ definition of EVR from the non-HD literature until further data become available in HD patients.

The combined treatment data suggest that HD patients with chronic HCV infection, who are appropriate treatment candidates, should be treated with 3 MU of IFN three times weekly for at least 6 mo. Treated patients, particularly those who achieve EVN, should be encouraged to complete the planned treatment. Female gender and low pretreatment HCV RNA level are two factors that predict higher likelihood of SVR. Future research is needed to determine the persistence of viral negativity in patients who achieve SVR, a particularly important issue for viral negative patients awaiting kidney transplantation.

Disclosures

None.