Correspondence to: Dr K. A. Chan, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA. E-mail: firstname.lastname@example.org
Aim : To use meta-analysis to study the risk of anaemiarelated to ribavirin therapy for chronic hepatitisC.
Methods : The MEDLINE database up to January 2001 was searched for randomized controlled trials of ribavirin (monotherapy or combined with interferon) for chronic hepatitis C. The outcomes evaluated were withdrawal from the study due to anaemia, ribavirin dosage reduction due to a decrease in haemoglobin and haemoglobin levels below 10 g/dL.
Results : Based on 17 studies, the overall risk difference (ribavirin vs. no ribavirin) for anaemia was 0.09 [95% confidence interval (CI), 0.04–0.13]. Two Asian studies reported risk differences of 0.29 and 0.22, greater than the pooled risk difference of 0.07 (95% CI, 0.03–0.12) for 15 non-Asian studies. The risk associated with 1 g or more of ribavirin per day was higher (risk difference, 0.09; 95% CI, 0.04–0.14) than that for 0.8 g of ribavirin per day (risk difference, 0.01; 95% CI, − 0.04–0.06).
Conclusions : Chronic hepatitis C patients treated with 1 g or more of ribavirin per day were at a higher risk of developing anaemia. Reported risks were higher among Asian studies, which may be due to differences in study entrance criteria, dosage titration strategy or ethnic vulnerability.
Chronic hepatitis C is a major risk factor for liver cirrhosis and hepatocellular carcinoma world-wide.1,2 The combination therapy of interferon and ribavirin for 24–48 weeks has been shown to be efficacious in treating chronic hepatitis C in clinical trials.3,4 A sustained response, defined as persistent normal serum alanine aminotransferase concentrations and undetectable serum hepatitis C virus RNA levels for at least 6 months after treatment, is achieved in approximately 40% of patients, almost twice as many as compared with interferon monotherapy.5 In most industrialized countries, this combination therapy has been approved for the treatment of both naive chronic hepatitis C patients, who have not previously received interferon treatment, and patients who do not respond to or relapse after previous interferon therapy. Although the efficacy of interferon and ribavirin combination therapy is promising, adverse reactions during the treatment period have been reported, one being haemolytic anaemia. As more chronic hepatitis C patients have received ribavirin treatment, more cases of severeanaemia following combination therapy have been reported.6,7 Pre-treatment platelet level, interferon-α dosage and the haptoglobin phenotype can influence ribavirin-induced haemolysis.8 This adverse effect is a major concern for patients with underlying cardiovascular disease and in populations with a high prevalence of haemoglobinopathy. The aim of this quantitative study was to evaluate the haematological safety of ribavirin therapy. We evaluated the range of cumulative incidence of haemolytic anaemia during ribavirin monotherapy or ribavirin plus interferon combination therapy among chronic hepatitis C patients (interferon naive, relapsers or non-responders) reported in the literature. We explored the source of heterogeneity of the cumulative incidence among patients with different baseline profiles, assessed the dose–response relationship and tried to identify high-risk subgroups who were most vulnerable to the development of haemolytic anaemia during ribavirin therapy.
Patients and methods
This study was a meta-analysis of clinical trial results published in widely available English journals. We searched MEDLINE and the references of identified reports published before January 2001 to identify clinical trials that had ribavirin in a treatment arm. Medical subjects heading terms used for key and text word searching included hepatitis C, hepatitis C chronic, ribavirin, randomized controlled trials (as both publication types and free text words), clinical trials, random allocation, double-blind method, single-blind method, meta-analysis and safety. The references of three recently published review articles on treatment for chronic hepatitis C9–11 were reviewed to identify additional studies that had not been found in the computerized literature databases. Further trials were identified from the reference lists of the identified studies. Only reports of randomized controlled trials written in English were included. Observational studies and quasi-randomized trials were not included.
Participants in the trials
Chronic hepatitis C was defined by a persistently elevated serum aminotransferase level for at least 6 months, positive anti-hepatitis C virus antibody, hepatitis C virus RNA detected by polymerase chain reaction, and pathology findings compatible with chronic hepatitis with or without cirrhosis on liver biopsy. Eligible patients, aged 18 years or over, who had never received any anti-viral therapy (naive), had shown an initial response to interferon-α therapy, but with subsequent elevation of serum aminotransferase levels (relapsers), or had shown resistance to previous interferon therapy (non-responders), were included and analysed separately. Studies that had enrolled patients co-infected with hepatitis B virus or human immunodeficiency virus, patients who had undergone organ transplantation or patients who had received any immunosuppressive therapy were excluded.
Types of intervention
Randomized controlled trials comparing ribavirin plus interferon vs. interferon only, vs. placebo or vs. no intervention were included, as were trials comparing ribavirin monotherapy vs. placebo or no intervention. Trials were included regardless of the dosage, length of therapy or treatment regimen of ribavirin and interferon.
The outcome of interest was the proportion of patients who met one of the following three criteria: withdrawal from the study due to anaemia; reduction of ribavirin dosage due to decrease in haemoglobin; and decrease in haemoglobin level to less than 10 g/dL.
Data analysis plan
Two physicians independently evaluated each identified study. Once the reviews had been completed, disagreement on specific studies was resolved through discussion between the physicians. Abstracted information included the study characteristics (author, location of research institute, time of the study, regimen, dose and duration), patient characteristics (naive, relapser or non-responder, proportion of patients who were male, mean age and percentage of patients with cirrhosis), sample size and efficacy outcomes. The proportions of patients who withdrew or required ribavirin dosage reduction due to anaemia or whose haemoglobin levels decreased to below 10 g/dL during treatment were recorded for both the intervention groups and the comparison groups. Studies that did not provide the above safety outcome information were excluded from the final analysis.
We also evaluated the quality of the trials according to the descriptions of the study population, interventionregimens and follow-up.12 Questions regarding the protocol design, especially the subjects' entry criteria and strategies for ribavirin dosage titration due to decreasing haemoglobin level, were added for completeness of quality evaluation. For each of the questions, studies were assigned a score of 0 (inappropriate), 1 (fair) or 2 (appropriate); the highest possible score was 14 (see Appendix).
We used the random effect model proposed by DerSimonian and Laird13 to estimate the summary risk difference and 95% confidence interval (CI) of the combined haematological safety end-points between treatment and control arms from eligible studies. We applied this model separately on stratified subsets of trials enrolling naive vs. relapsers or non-responders, a treatment duration of 24 weeks vs. 48 weeks, and a ribavirin daily dosage of 0.8 g or less vs. 1.0 g or more. We also calculated the summary estimate for the subset of reports from North American and European countries.
We identified 69 studies involving ribavirin from a MEDLINE search (Figure 1); 40 were reported in English and were retrievable for review. Two physicians independently reviewed all reports and 23 studies met the pre-defined eligibility criteria. Only 17 studies clearly presented the haematological outcome information of interest and were included in our analysis.3,4,14–28 These 17 studies, published between 1993 and 1999, with the number of study subjects ranging from 20 to 912, and a total of 3520 chronic hepatitis C patients, are summarized in Table 1. Their quality scores were in the range 4–12. Six of the studies were placebo-controlled trials. The mean age of the patients ranged from 33 to 49 years, and the proportion of male patients ranged from 40% to 100%. The proportion of patients with liver cirrhosis enrolled in each study varied: 45% and 25% in the studies of Brillanti et al.18 and Sostegni et al.,28 respectively, compared with 0% in those of Lai et al.25 and Chemello et al.19 None of these 17 studies provided results of the haematological outcome of interest separately for males vs. females, or for patients with cirrhosis vs. those without cirrhosis.
Table 1. Characteristics of the 17 studies included in the meta-analysis
N, naive; NR, non-responder; R, relapser; NA, not available; INF, interferon; R, ribavirin; tiw, three times a week.
The outcome of interest was the proportion of patients who met one of the following three criteria: withdrawal from the study due to anaemia; reduction of ribavirin dosage due to decrease in haemoglobin (Hb); and decrease of Hb level to less than 10 g/dL.
INFαn-3 3 MU tiw + R 1 g/day R 1 g/day followed by INFαn-3 3 MU tiw INFαn-3 3 MU tiw
27.5 24 —
The treatment regimens employed in these studies were quite diverse with different types of interferon being used in combination with ribavirin simultaneously or sequentially. Most of the studies used interferon-α, either a human leucocyte-derived product or a recombinant product containing single interferon. We assumed that all types of interferon were comparable with respect to the haematological adverse effects that were specific to ribavirin. Except for one arm in the studies of Barbaro et al.,15,16 where interferon monotherapy (6 MU) was administered three times a week, the dosage of interferon was quite uniform, i.e. 3 MU three times a week; the dosage of ribavirin ranged from 600 mg/day to 1200 mg/day. The study population inall studies received at least 24 weeks of therapy. Whilst McHutchison et al.3 and Poynard et al.4 also included treatment arms with a duration of 48 weeks, Di Bisceglieet al.21 only compared the results between ribavirin monotherapy and placebo for 48 weeks.
There was significant heterogeneity among the 17 studies with regard to the proportion of patients who developed the combined haematological end-point (P < 0.001). The observed risk difference for the haematological end-point ranged from 0 to 0.30, with an overall effect estimate of 0.09 (95% CI, 0.04–0.13) based on the random effect model (Figure 2). There was a trend of decreasing incidence of anaemia during ribavirin treatment reported in recent studies compared with earlier studies. There was no significant difference in risk reported between studies enrolling more than 100 patients (risk difference, 0.05; 95% CI, − 0.01–0.10) compared with studies enrolling less than 100 patients (risk difference, 0.12; 95% CI, 0.04–0.20). The two Asian studies (Lai et al.25 and Kakumuet al.23) reported risk differences between patients treated with a ribavirin-containing regimen and a non-ribavirin-containing regimen of 0.29 and 0.22, respectively, fargreater than that reported by the 15 non-Asian studies with a pooled risk difference of 0.07 (95% CI, 0.03–0.12). In addition, those who were treated with a ribavirin dosage of less than or equal to 0.8 g/day did not show a significant risk of developing anaemia (risk difference, 0.01; 95% CI, − 0.04–0.06), whereas those who were treated with a ribavirin dosage equal to or more than 1 g/day were at significantly higher risk (risk difference, 0.09; 95% CI, 0.04–0.14) (Figure 3). There was no significant difference in risk between naive patients (risk difference, 0.07; 95% CI, 0.05–0.08) vs. non-responders/relapsers (risk difference, 0.08; 95% CI, 0.02–0.14); those treated for 24 weeks (risk difference, 0.07; 95% CI, 0.02–0.11) vs. 48 weeks (risk difference, 0.08; 95% CI, 0.05–0.10); and those who received ribavirin plus interferon combination therapy (risk difference, 0.06; 95% CI, 0.02–0.10) vs. ribavirin monotherapy (risk difference, 0.12; 95% CI, − 0.02–0.26).
One previous meta-analysis of individual patient data from placebo-controlled trials or open studies with ribavirin as part of the treatment regimen, conducted in European centres before 1995, reported that less than 5% of patients had haemoglobin levels below 6 mmol/L (9.6 g/L) during the treatment period.29 As more data were collected by randomized controlled trials, the cumulative incidence of patients with haemoglobin levels decreasing to less than 75% of the lower limit of normal (e.g. 9.75 g/dL if the lower limit of normal is 13 g/dL) increased to 5% for patients with or without cirrhosis at 12 weeks, 7% for patients without cirrhosis at 24 weeks and 10% for patients with cirrhosis at 24 weeks.30 Our study suggested that, during ribavirin treatment, the cumulative incidence of patients with a haemoglobin level less than 10 g/dL or anaemia requiring withdrawal from the trial or a reduction in ribavirin dosage was 9%.
Variation in reporting the safety outcomes did not allow us to utilize all information from the clinical trial reports. Some studies reported the mean haemoglobin change, whilst others only reported the end of treatment haemoglobin levels; yet, these studies provided sufficient information for this meta-analysis. In addition, different operational definitions of anaemia were used in different studies, and the proportions of patients who were diagnosed with anaemia were based on haemoglobin concentrations below different levels. This inconsistency led to a limited number of studies with comparable safety end-points. We chose a combined end-point because it had important clinical implications and it was reported in 17 of the 23 eligible articles. The limited reported information did not allow us to evaluate the risk of anaemia in males vs. females, or in patients with liver cirrhosis vs. those without liver cirrhosis.
There was a wide range of anaemia risk between studies, substantiated by the statistically significant heterogeneity measure. In some studies, minimal decreases in haemoglobin levels were reported, whilst other studies found that nearly 30% of patients suffered from anaemia leading to dosage reduction. Because the objective of a meta-analysis is not synthesis but comparison, and the identification of heterogeneity,31 we explored factors that might contribute to the difference in the incidence of anaemia. The 17 studies used in the meta-analysis differed in their study populations, treatment strategy and the quality of safety end-points (Table 2). We used a quality score to quantitatively describe these differences, although it may not be possible to differentiate between poor quality resulting from deficiencies in the study design or conduct of the clinical trials vs. deficiencies in the reporting of the studies. Only five of the 17 studies had clearly defined, pre-specified criteria and a strategy for ribavirin discontinuation or dosage reduction for patients with decreasing haemoglobin during treatment. In these studies, patients' doses were reduced whenever their haemoglobin level dropped below a certain value (e.g. 10 g/dL). In other studies, the reasons for ribavirin dosage reduction were not specifically described. As more data have become available in recent years, investigators are more likely to monitor patients with ribavirin-related haemolytic anaemia, which might explain the decreasing incidence of anaemia reported in more recently published studies. Another possible source of heterogeneity is the enrolment criteria for the study subjects. In seven studies, only patients with normal haemoglobin levels (> 13 g/dL for men and > 12 g/dL for women) were included, and those with a history of haemolytic anaemia and haemoglobinopathy were excluded. In other studies, patients with lower haemoglobin levels due to subclinical haemoglobinopathy, such as thalassaemia, may have been enrolled. The interaction between the underlying defect of the red cells, oxidative damage to the membrane and increased binding of autoantibodies, leading to the premature removal of erythrocytes exposed to ribavirin from the circulation, will present as a more marked decrease in haemoglobin level.32,33
Table 2. Comparison of subject enrolment criteria and management strategy in the 17 studies
Exclude patients with history of haemolytic anaemia or haemoglobinopathy
We found the risk of anaemia during ribavirin treatment to be much higher in the two Asian studies, which were published before 1997, than in the non-Asian studies. Although the prevalence of certain subclinical haemoglobinopathy is higher in Asia than in other regions of the world, different strategies for ribavirin dosage reduction might be a contributing factor to this higher risk of anaemia. In the study of Lai et al.,25 29% of the patients receiving a ribavirin-containing regimen had haemoglobin levels lower than 10 g/dL. However, the ribavirin dosage was reduced in only two patients because of dizziness and palpitations. The ethnic vulnerability for the development of haemolytic anaemia is also a reasonable concern in Asians, and whether or not Asian populations are at a higher risk for the development of this adverse effect can only be examined in studies enrolling both Caucasians and Asians and comparing the risks by subgroup analysis.
The risk of haemolytic anaemia may be dose dependent. Vlierbergheet al.8 reported that a ribavirin dosage of 1.0–1.2 g/day did not influence patients' haemoglobin levels. We found a threshold effect for ribavirin and the risk of anaemia: those treated with ribavirin at 0.8 g/day were not at risk of developing anaemia, whereas those treated with 1 g or more per day were at a higher risk. This dose-dependent risk of anaemia has major implications in the risk–benefit evaluation of ribavirin for chronic hepatitis C, and the risk needs to be interpreted in the context of the higher treatment success rates associated with a higher ribavirin dosage (1.0–1.2 g/day).11 The outcomes for patients requiring ribavirin dosage reduction due to anaemia are inconsistent. McHutchison et al.3 reported that dose reduction did not affect the rate of sustained response, whereas Lai et al.25 reported that dose reduction was associated with the relapse of persistently abnormal serum aminotransferase activities in one of two patients. Nevertheless, we wish to emphasize the importance of close monitoring of the haemoglobin levels and judicious adjustment of the ribavirin dosage in order to avoid the premature discontinuation of therapy and to attain optimal treatment results.
In conclusion, our study suggested a 9% excess risk of developing clinically significant anaemia during ribavirin therapy among patients with chronic hepatitis C, but there was a significant difference in this risk between the reported studies. Chronic hepatitis C patients receiving combination therapy with a ribavirin dosage of more than 1.0 g/day were at a higher risk of developing haemolytic anaemia. Close monitoring of haemoglobin levels and judicious adjustment of the ribavirin dosage will markedly reduce the number of patients experiencing clinically significant anaemia during ribavirin therapy.
This study was sponsored in part by the Harvard Pharmacoepidemiology Research and Teaching Fund.
Scoring system for assessment of quality of report
Consecutive eligible consenting patients or random series
Attempt made to enrol as such, with failure due to reasons outlined explicitly
Selected patients (not consecutive or random) or not described
Clearly specify that patients without pre-existing anaemia or a minimal level of Hb to be enrolled in the trial
Only one of the above criteria
None of the above
Computerized generated allocation, random number table