A mixed treatment comparison demonstrated that FCR significantly prolonged PFS compared with all other treatments and significantly increased ORR compared with all other treatments.
There has been a lack of meta-analyses and systematic reviews conducted in chronic lymphocytic leukemia (CLL). Structured searches were conducted in publication and conference databases to identify randomized controlled trials (RCTs) that are reporting efficacy data for treatment-naïve CLL and both RCTs and non-RCTs for relapsed/refractory CLL. Eight RCTs met prespecified inclusion criteria to permit a meta-analysis of fludarabine/cyclophosphamide/rituximab (FCR) with alemtuzumab, bendamustine, chlorambucil, fludarabine, and fludarabine/cyclophosphamide (FC). A mixed treatment comparison demonstrated that FCR significantly prolonged progression-free survival (PFS) compared with all other treatments (hazard ratios, 0.24-0.56) and significantly increased overall response rate (ORR) compared with all other treatments. FCR significantly increased the complete response (CR) rate compared with chlorambucil, fludarabine, and FC (odds ratios of 30.3, 10.1, and 2.7, respectively). For relapsed and refractory CLL, 9 RCTs and 86 non-RCTs were identified that reported efficacy data. A meta-analysis was deemed inappropriate due to trial heterogeneity. Overall, overall survival (OS) and PFS ranged from 24 to 33.8 months and from 6 to 30.6 months, respectively, depending on the therapy. Across all trials, FCR showed the greatest improvement in PFS, with robust ORR and CR rates. The identified studies indicated that FCR is highly efficacious for the treatment of both untreated and relapsed/refractory patients with CLL. These studies and the benefit of FCR are largely applicable to young and fit patients with CLL, and the best treatment for older patients could not be extracted from this analysis. Going forward, this evidence-based review should be updated with longer follow-up from trials to better determine OS benefit.
Chronic lymphocytic leukemia (CLL) is the most common adult leukemia in the Western world, with an incidence rate of approximately 3.9 per 100,000 persons per year.1 Approximately 14,570 patients will be diagnosed with CLL in the United States in 2011.2 Primarily a malignancy of the elderly, the median age at diagnosis is 72 years in the United States, and approximately 70% of patients are aged >65 years.3 The disease is largely characterized by a slow, often asymptomatic progression over several years as malignant cells gradually accumulate in the blood, bone marrow, and lymph nodes.4 Anemia, thrombocytopenia, and immunosuppression can accompany painless lymph node enlargement and organomegaly. Systemic symptoms, such as fever, sweating, and weight loss, may be rare. Patients with early disease are usually left untreated; however, most patients will eventually require treatment.5
The treatment of CLL has advanced rapidly over the past 2 decades due to the use of chlorambucil as standard first-line treatment and subsequent clinical determination that the purine-based analog fludarabine showed superior clinical benefit over alkylator- and anthracycline- based regimens.6-8 Initial treatment with fludarabine is associated with response rates of 63% to 80%, and when combined with cyclophosphamide (FC) showed increased response rates of 74% to 94%.9,10 Adding the anti-CD20 monoclonal antibody rituximab to FC (FCR) chemotherapy resulted in further improvements in the overall response rates (ORRs) and complete response (CR) rates.11,12 Additional biologic and chemical agents, including alemtuzumab, ofatumumab, and bendamustine, have been added to the CLL treatment armamentarium, and a number of investigational agents are undergoing study in the clinic. Treatment options for patients with relapsed or refractory CLL remain limited, and the prognosis for these patients is still poor.13 The management of CLL depends on a number of factors, and in the first-line setting it has been proposed that 3 overriding considerations dictate treatment: tumor load with or without symptomatic or progressive disease (ie, determined by Rai or Binet stage or by lymphocyte doubling time); the physical condition of the patient (ie, fitness and comorbidity); and the prognostic risk of the leukemia (ie, due to genetic and other molecular factors).14 For relapsed disease, these factors, in addition to the choice of prior therapy and the length of remission duration from first treatment, will influence treatment decisions. Despite treatment, the vast majority of patients relapse and CLL remains an incurable disease.15 Randomized controlled trials (RCTs) are needed to determine optimal treatment for appropriate patient types, depending on stage of disease, cytogenetics, performance status, and treatment status.
Published data from RCTs and systematic reviews have provided more definitive evidence for the role of treatment. A meta-analysis performed by the CLL Trialists’ Collaborative Group and a systematic review reported by the Swedish Council on Health Technology Assessment have provided substantial evidence that early treatment with chlorambucil with or without corticosteroids is not associated with a better outcome compared with deferred treatment.16,17 A systematic 2006 Cochrane Review in patients with previously untreated CLL18 and an analysis by Zhu and colleagues19 on 5 randomized trials involving 1300 patients confirmed the greater response rates achievable by using purine antagonists, but provided inconclusive evidence for improvement in survival.
Here we report on an updated systematic review and meta-analysis of first-line RCTs and a systematic review of trials and observational studies conducted in the relapsed and refractory settings.20,21 Where appropriate, results were combined using a Bayesian mixed treatment comparison (MTC). MTC is an extension of traditional meta-analyses; trials of direct (eg, A vs B and B vs C) comparison are used to yield indirect (eg, A vs C) estimates across multiple different pairwise comparisons and across a number of different interventions in the absence of head-to-head trials.22-24
Click image to enlarge
CAP indicates cyclophosphamide/adriamycin/prednisone; CHOP, cyclophosphamide/hydroxydaunorubicin/vincristine/ prednisone; CLL, chronic lymphocytic leukemia; CVP, cyclophosphamide/vincristine/prednisone; FC, fludarabine/cyclophosphamide; FCR, fludarabine/cyclophosphamide/rituximab; y, years.
Systematic literature reviews of untreated and previously treated patients were undertaken separately, according to prespecified protocols. Search terms used both free text and medical subject headings (MeSH) and were conducted in MEDLINE, EMBASE, BIOSIS, the Cochrane Library, and conference abstract databases (ie, ASCO, ASH, EHA, and ESMO). Conference abstracts were excluded for first-line treatments because the reports rarely provided sufficient information or were not of acceptable quality to be included in meta-analyses. Articles were assessed on the basis of title and abstract according to prespecified patient population, intervention, comparison, and outcome (PICO) criteria; the full text was reviewed, where necessary. Inclusion criteria specified open or blinded RCTs of treatment-naïve adult patients aged >18 years with good performance status (ECOG 0 to 2) (Table 1). Treatment- naïve patients were administered alemtuzumab, bendamustine, chlorambucil, fludarabine, fludarabine/cyclophosphamide (FC), and fludarabine/cyclo-phosphamide/rituximab (FCR). For treatment- ex-perienced patients, a larger number of treat- ments and treatment combinations were included (Table 1). Studies that enrolled mixed populations (eg, both treatment-naïve and relapsed patients) were included where data for the relevant population were reported separately. Observational (non-RCT) studies in relapsed or refractory patients were also included but were not eligible for meta-analyses and were reviewed separately.
The primary outcome was progression-free survival (PFS), defined as the time between randomization and disease progression, relapse, or death by any cause. Other outcomes of interest included ORR, partial response (PR), CR, overall survival (OS), disease-free survival, duration of remission, and mortality. Study quality was assessed using either the Oxford scale or Cochrane Handbook assessment tool.25,26
Meta-analyses were prespecified where studies evaluated the same treatment in a comparable patient population with few other indicators of clinical or methodological heterogeneity. Where permissible, results were combined using a Bayesian MTC.22-24 The Bayesian framework also permitted the derivation of relative treatment rankings and best treatment likelihoods. Information from both continuous (time-to-event) and binary outcomes were of interest for meta-analysis. Analysis of binary outcomes (such as response rates) was based on the log odds ratio. A Cox regression model was assumed for time-to-event outcomes, and the log hazards were summarized across RCTs. Analyses were performed using WinBUGS 1.4 statistical software. A noninformative prior distribution was used (ie, the mean [μ] was set to 0 and the variance [δ] was set to 1000). Credibility intervals were calculated in order to indicate uncertainty around the point estimate. The appropriateness of a fixed or random effects model was investigated using overlap of credibility intervals for residual deviance (ie, the goodness of fit).
Click image to enlarge
aThe PFS results reported by Knauf and colleagues are from a later analysis of this trial and were not included in the MTC analysis reported in Table 3. They are reported here for completeness.
A indicates alemtuzumab; B, bendamustine; C, cyclophosphamide; Chl, chlorambucil; CR, complete response; FC, fludarabine/ cyclophosphamide; FCR, fludarabine/cyclophosphamide/rituximab; mo, months; NA, not available; NR, not yet reached; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; y, years.
The systematic review identified 683 abstracts, of which 7 trials met the inclusion criteria (see Appendix A) (Table 2).12,27-33 The CLL8 trial from the FCR clinical development program was also available for inclusion.12 Neither the age (approximately 60 y) nor gender (approximately 70% male) varied significantly across trials. The LRF CLL4 trial included a notably greater proportion of patients with Binet stage A disease and higher chlorambucil dose.27 All 8 RCTs were considered sufficiently similar in design, patient-inclusion criteria and characteristics, and outcomes to conduct an MTC. Data were available to compare PFS, CR, and ORR for FCR with alemtuzumab, bendamustine, chlorambucil, fludarabine, and FC, although only small patient numbers were reported for some treatments and outcomes.
The MTC showed that FCR significantly prolonged PFS compared with each of the other treatments (Table 3). Hazard ratios (HRs) ranged between 0.24 (0.17 to 0.34) compared with chlorambucil and 0.56 (0.43 to 0.72) compared with FC. The median PFS was prolonged with FCR by 20 and 8 months when compared with chlorambucil and FC, respectively. FCR had a 100% probability of being the best treatment option when PFS is used to assess efficacy.
The MTC suggested that FCR would be the best therapy for treatment-naïve CLL for achieving response (Table 3). The ORR was highest for FCR and superior to all comparators. The odds ratio for FCR with respect to the ORR endpoint ranged between 2.3 (1.6 to 3.4) with fludarabine to 14.2 (8.3 to 24.7) with chlorambucil. The probability of FCR being the best treatment option was greater than 99% when comparing ORR as the clinical endpoint.
FCR increased the probability of a CR in comparison with chlorambucil, fludarabine, and FC, and was shown to be the best treatment in 58% of cases (Table 3). When assessing CR, the lower probability of being the best treatment option when compared with PFS or ORR is due to the relative lack of data, with fewer than 5 patients achieving a CR in either treatment arm.
Click image to enlarge
Data presented as median values.
CI indicates confidence interval; FC, fludarabine/cyclophosphamide; FCR, fludarabine/cyclophosphamide/rituximab; NR, not recorded.
The LRF CLL4 Trial reported a greater proportion of patients with Binet stage A disease and used a higher chlorambucil dose; however, sensitivity analyses demonstrated no significant effect of excluding this study (PFS HR = 0.54; 0.42-0.68).27 All outcomes were analyzed using a fixed effects model, but results from a random effects model were also considered; the credibility intervals overlapped for all outcomes and the overall results remained the same.
The systematic review in relapsed or refractory patients identified 2318 abstracts, of which 9 articles relating to 8 RCTs (see eAppendix B and Table 4) and 1 publication reporting additional follow-up data, were eligible for inclusion.8,34-41 The most frequent reasons for exclusion were inappropriate disease (N = 708) and insufficient reporting of either relevant outcomes (N = 385) or interventions (N = 314) (see eAppendix B). Assessment of trial methodology, treatment comparisons, and reported outcomes revealed significant heterogeneity across trials. As such, a meta-analysis was not deemed appropriate due to the risk of introducing bias and generating misleading results. Where 2 similar trials were available, further analyses were constrained by poor study quality, limited reporting quality, patient-inclusion criteria and length of follow-up, treatment duration, outcomes, and frequency of outcomes assessment. Given the heterogeneity across RCTs in relapsed and refractory patients with CLL, the systematic review was extended to non-RCTs to provide a comprehensive assessment of the wider evidence base. This search identified 90 publications reporting data from 86 non-RCTs.
Click image to enlarge
aRandomized controlled trial that reported data only for patients who were treated with FC GM-CSF.
C indicates cyclophosphamide; CAP, cyclophosphamide/adriamycin/prednisone; Chl, chlorambucil; CLL, chronic lymphocytic leukemia; COP/CVP, cyclophosphamide/vincristine/prednisone; CLR P, chlorambucil and prednisone; ECOG, Eastern Cooperative Oncology Group; FC, fludarabine/cyclophosphamide; FCR, fludarabine/cyclophosphamide/rituximab; FCM, fludarabine/cyclophosphamide/mitoxantrone; FCM-R, fludarabine/cyclophosphamide/mitoxantrone/rituximab; FCO, fludarabine/cyclophosphamide/oblimersen; FCR, fludarabine/cyclophosphamide/rituximab; GM-CSF, granulocyte macrophage colony-stimulating factor; mo, months; NA, not available; NCI, National Cancer Institute; NHL, non-Hodgkin lymphoma; NR, not yet reached; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; P, prednisone; R, rituximab; WHO, World Health Organization; y, years.
Overall, fludarabine with or without cyclophosphamide was the most frequently used intervention in clinical studies. Of the monoclonal antibody therapies, the largest body of evidence was identified for rituximab with 2 RCTs and 25 publications relating to 22 non-RCTs.34-40,42-66 To date, the phase III REACH RCT is the largest available study in this setting.40 Only non-RCTs were identified for alemtuzumab or ofatumumab.45,50,67-86,87,88 Although not the focus of this study, it was noted that adverse event data were infrequently reported and were often incomplete, selective, or combined across treatment groups and/or study populations.
All RCTs were parallel-group design and size, and the duration of study follow-up ranged from 16 to 552 patients and from 6 to 60 months, respectively (Table 4); 6 out of 9 were full publications and 3 were abstracts.8,34-41 No 2 studies compared similar treatments except for 2 related publications comparing FC with FC/oblimersen (FCO).37,38 Six studies reported on OS (Table 4).8,36-38,40,41 In REACH, the largest randomized trial conducted to date in this setting, the median OS was 52 months with FC; but the median OS has not yet been reached for the FCR comparator.40 One study did not present evaluable data (to be considered “evaluable,” data were required to be reported separately for all treatment arms).39 This trial compared FC plus granulocyte macrophage colony-stimulating factor with FC plus supportive care but reported only composite results for both treatment arms. OS data were unavailable or not reported for 2 studies.34,39
Three evaluable studies reported PFS.8,40,41 Results from REACH showed a substantial and significant improvement; intent-to-treat analyses showed a median PFS of 31 months with FCR compared with 21 months with FC (P = .0002).40 Johnson and colleagues8 showed a benefit of fludarabine (11 mo) over CAP (6 mo). Sawitsky and colleagues41 reported an improvement in PFS with monthly chlorambucil and prednisone (16 mo) over daily chlorambucil and prednisone (7 mo) and prednisone alone (7 mo). Three studies reported CR and PR; 2 studies each reported on stable disease and progressive disease.8,34,36,37,39,41 The most robust data were available for ORR, with 7 studies reporting for a number of different treatment regimens.8,34-36,38,40,41 Across all the treatments, the ORR ranged from 11% with prednisone treatment to 70% with FCR.40,41
The non-RCTs ranged similarly in size (N = 5-724) and duration of follow-up (3-60 mo) to RCTs in this setting but reported on a wider range of treatment comparisons. Ninety publications relating to 86 studies were included, with 66 studies published in full and 24 available only as abstracts. Fludarabine was the most commonly evaluated treatment as monotherapy or in combination with other agents (N = 44).
Click image to enlarge
Disease response was the most frequently reported outcome. CR was reported by 77 publications, PR by 81 publications, stable disease by 33 publications, and progressive disease by 23 publications. ORR was reported by 59 publications (Figure).
OS was reported by 26 publications.44,52,53,61,63,68,72-74,76,77,85,88-101 Twenty-one studies reported median survival, and across all treatments the median OS ranged from 5.9 months with alemtuzumab to 48 months with fludarabine and epirubicin.85,100 Direct comparisons of median OS across non-RCTs were complicated by a lack of consistency around the timepoint at which median OS was reported, ranging from 12.6 to 48 months.89,91,94-96,100,101 Fludarabine was the most frequently reported intervention, either administered alone (median OS values from 3 studies of 13, 18, and 20 mo) or as combination FC (24.5 mo).91,92,94,101 Across all treatments the reported survival rates ranged from 29% to 100% (years 2002-2004).76,85
PFS was reported by 10 publications.44,52,53,65,66,91,92,94,98,101 Two publications were associated through long- term follow-up of patients after the initial report of results.65,66 Four studies reported the proportion of patients with PFS, and 8 of the 10 studies reported median PFS. The median PFS ranged from 6.5 months with rituximab and etanercept to 24.5 months with FC.66,92 The proportion of patients with progression-free disease ranged from 29.4% with rituximab to 75% with FCR (years 2005-2009).52,53,65
The systematic reviews and meta-analysis presented here outline the evidence base of RCT or non-RCT studies in patients with treatment-naïve CLL or with relapsed or refractory CLL. Good systematic reviews are important in that they can identify clinically relevant studies, classify them by quality, combine the data, and, where appropriate, draw reasonable conclusions. Their conclusions may have implications for clinical practice but also, importantly, implications for clinical research itself. This systematic review showed that clinical trials in the relapsed setting were heterogeneous in design (and in many cases lacked a comparator arm) and were not sufficiently powered to draw definitive conclusions about treatment. These results highlight the need to improve the quality of individual trials with patients with relapsed or refractory CLL.
The results of the MTC, including data from the 8 identified RCTs in treatment-naïve patients with CLL, confirmed the superior clinical benefits of FCR. Based on the calculated HRs, FCR significantly prolonged PFS compared with all other treatments, with the risk of progression reduced by between 44% and 76%, compared with FC and chlorambucil, respectively. ORRs were shown to significantly increase compared with all other treatments, with the likelihood of achieving response when treated with FCR ranging from more than double compared with FC, and up to 14 times compared with chlorambucil. When comparing CR rates, FCR showed significant benefits compared with chlorambucil, fludarabine, and FC, with the likelihood of achieving CR more than doubling compared with FC and increasing to >30 times compared with chlorambucil. Although FCR did show a benefit in CR compared with both bendamustine and alemtuzumab, these findings were nonsignificant as the credibility intervals overlapped. The direction of the effect is similar to that observed when comparing CR rates for FCR to FC, fludarabine, or chloram-bucil; the lack of statistical significance may be attributed to the small number of patients achieving CR in the trials.
The findings of the MTC should be treated with some caution due to the relatively small number of trials available on which to conduct the analysis, and the possibility that heterogeneity may have confounded the analysis. Efforts were made to assess the heterogeneity across comparisons prior to conducting the MTC. However, in order to assess the potential impact of heterogeneity due to unknown effect modifiers, additional comparable trials would be needed. Finally, it should be noted that there are likely systematic differences among the patients enrolled in the first-line trials because physicians are often less likely to treat an elderly, debilitated patient with FCR, or a young healthy one with chlorambucil. These unmeasured effects may contribute to the heterogeneity.
The MTC was conducted using the first-reported FCR data from the CLL8 trial. Subsequent to this analysis, updated results from a longer follow-up (median, 37.7 mo) of the patients in CLL8 have been reported; it was demonstrated that patients treated with FCR benefited from a significantly more prolonged PFS benefit than first observed.102 The median PFS for FCR from this longer follow-up was reported to be 51.8 months compared with 32.8 months with FC alone (HR = 0.56; P <.0001). Most importantly, for the first time in the treatment of CLL disease, a significant survival advantage was demonstrated with FCR therapy in the first-line setting (3-y OS rate: 87.2% vs 82.5% for FC; HR = 0.664; P = .012). The median survival has not yet been reached in either treatment arm. These improved PFS and OS benefits with FCR were observed across the majority of patient subgroups analyzed, with the exception of patients with Binet stage C disease. These data confirm FCR as the standard of care for young and fit patients with untreated CLL. Health-related quality-of-life analyses from CLL8 showed that the addition of rituximab to chemotherapy had little or no impact.103 The clinical benefits are provided with a reasonable incremental cost per quality-adjusted life year (QALY) gained.104 Of note, the median ages of patients in all trials in the meta-analysis was <65 years, with the exception of the study by Eichhorst and colleagues (median age, 70 y) (Table 2), which is below the median age of 72 at diagnosis. The median age of patients in CLL8 was 61 years; exploratory analysis in CLL8 patients aged >70 years has suggested that the benefit to patients is questionable when treated with FCR.105
Of the 8 RCTs reporting treatment of relapsed or refractory CLL, all were parallel-group RCTs. The quality indicators essential to the assessment of clinical and methodological biases were infrequently reported or too variable in these trials to justify meta-analysis. Similarly, 90 non-RCT publications met the inclusion criteria relating to 86 studies for this review and reported approximately 30 different treatment regimens. Where similarities were found, the results were typically variable, making overall conclusions difficult. OS was the most frequently reported survival outcome.
RCTs represent the methodological gold standard for demonstrating the efficacy and safety of treatments. Further, based on the results of CALGB-9011, longer follow-up in such studies will better determine clinically significant survival benefits. The follow-up time of trials conducted in relapsed and refractory patients ranged from 5 to 60 months, and ORR was the most frequently reported endpoint. This systematic review demonstrates the lack of robust, high-quality comparator RCT data, despite the numerous non-RCTs that have reported on a wide variety of treatment regimens. Only 2 RCTs (CLL8 and REACH) were identified; both evaluated rituximab. REACH was the largest RCT retrieved, with almost twice the number of enrolled patients of any other trial included in the systematic review. Regardless of the disease setting, it is imperative as the development of treatments for CLL evolves that both investigational and approved therapies be compared in RCTs to optimize survival outcomes for patients. Such studies continue for rituximab.
The significant clinical benefits achieved with FCR highlight the need for more tolerable and efficacious treatments for patients with CLL who are elderly and/or have comorbidities and who cannot tolerate such immunochemotherapeutic combinations. Clinical trials in elderly patients with therapies such as chlorambucil, lenalidomide, and bendamustine in combination with monoclonal antibodies are underway. Until the results of such large studies become available and heterogeneity is reassessed, this analysis provides the most current evidence-based review and serves to facilitate the development of consensus on the use of rituximab alongside other therapies in the treatment of CLL.
Ann Janssens, MD, is a hematologist at the University Hospitals Leuven in Belgium; Robin Foà, MD, is a professor of Hematology at Sapienza University of Rome in Italy; Michael J. Keating, MB, BS, is a professor of Medicine in Hematology at the University of Texas MD Anderson Cancer Center in Houston; Iain Tatt, PhD, and Emma S. C. Carr, BSc, MSC, PhD, are employees of F. Hoffman-La Roche in Basel, Switzerland.
Support for third-party writing assistance for this manuscript was provided by F. Hoffman-La Roche. Dr Foa reports having served on advisory boards for Roche and lecturing upon the company’s invitation. Drs Tatt and Carr are employees of F. Hoffman-La Roche. The other authors report no financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Address correspondence to:
Ann Janssens, MD, Department of Hematology, University Hospitals Leuven, Leuven 3000, Belgium. E-mail: email@example.com.