Welcome international healthcare professionals

This site is no longer supported and will not be updated with new content. You are welcome to browse and download all content already included in the site. Please note you will have to register your email address to access the site.

You are here

Comparative efficacy of first-line therapies for advanced-stage chronic lymphocytic leukemia: A multiple-treatment meta-analysis

Cancer Treatment Reviews, 4, 39, pages 340 - 349

Abstract

Since the introduction of chlorambucil as a treatment for chronic lymphocytic leukemia (CLL) in the 1960s, several alternative treatment regimens have been explored. We performed a multiple-treatment meta-analysis using direct and indirect data based on all available head-to-head randomized controlled trials (RCTs) to compare the benefits and harms of first-line treatments for untreated advanced-stage CLL. Two reviewers independently identified RCTs comparing overall survival and progression-free survival between two or more first-line treatments. Twenty-five trials involving 7926 patients were included. Of the 25 eligible RCTs, 30 (n = 7741 patients) and 12 (n = 3910 patients) treatment pairs were included in the multiple-treatment meta-analysis of overall and progression-free survival, respectively. Trials generally enrolled younger and less complicated patients than actual clinical practice. There was no evidence for inconsistency between direct and indirect data. Based on combined direct and indirect data, no single treatment showed significantly better overall survival than any other, and credible intervals were wide. Among six newer treatments with longer progression-free survival compared with chlorambucil, fludarabine-rituximab-based chemoimmunotherapy (HR = 0.24, 95% CrI: 0.13–0.51) and bendamustine (HR = 0.23, 95% CrI: 0.13–0.42) had the largest PFS benefit. Limited data on treatment-related mortality precluded multiple-treatment meta-analysis. In conclusion, published randomized evidence on overall survival is insufficient to recommend any particular first-line treatments. Any progression-free survival differences may be applicable to relatively young uncomplicated patients.

Keywords: CLL, Treatment, Meta-analysis.

Introduction

Chronic lymphocytic leukemia (CLL) is the most commonly diagnosed lymphoid malignancy in Western countries.1 and 2 Its incidence increases with age with a peak among patients older than 70 years.2 and 3 Although the clinical behavior of untreated CLL varies from a long-term indolent to a rapidly progressive disease, it remains as an incurable condition. In advanced stages of CLL, timely therapeutic interventions are employed to control disease progression and its complications, such as cytopenias and opportunistic infections.4 and 5

Since the introduction of chlorambucil as a treatment for CLL in the 1960s, alternative treatment strategies for untreated advanced-stage CLL have continued to multiply. Nearly 40 years later, clinicians can choose from any one of a handful of regimens. However, the comparative effectiveness of these treatment strategies has not been systematically evaluated. Previous meta-analyses6, 7, and 8 assessed direct head-to-head comparisons from randomized controlled trials (RCTs) of earlier treatments (i.e., chlorambucil monotherapy versus combination chemotherapies such as cyclophosphamide, doxorubicin, vincristine, prednisone [CHOP], and single-agent purine analogues, or other combinations) and found no evidence of better overall survival with a particular regimen. Subsequent trials have tested more aggressive regimens such as combinations of fludarabine and cyclophosphamide (FC), or chemoimmunotherapy such as fludarabine, cyclophosphamide, and rituximab (FCR) against various comparator treatments. Current practice guidelines, an expert consensus on the basis of the limited head-to-head comparative trials, recommend chemoimmunotherapy such as FCR for non-frail younger and chlorambucil for older patients.9 and 10 Expert recommendations are also in line with these guidelines.11 and 12

To date, the comparative effectiveness and safety of all clinically relevant treatments for CLL has not been assessed in a systematic and quantitative manner. To address this question, we performed a systematic review and multiple treatment meta-analysis (MTM) of all clinically relevant CLL treatments. The advantage of MTM is that it can estimate the comparative effectiveness and safety of a network of treatments, even when not all of them have been compared head to head. MTM is based on the premise that one can obtain indirect information on the comparison of two treatments by examining their effects versus a third reference treatment while preserving randomization.13 and 14 For example, treatment A can be indirectly compared to treatment C by using a set of trials that compare treatment A to treatment B, and another set of trials that compare treatment B with treatment C.

Methods

We developed and followed a standard protocol for all steps of this research. The Appendix provides comprehensive descriptions of the methods.

Search strategy

We searched PubMed from inception to June 30, 2011 for RCTs. We used a highly sensitive strategy for identifying RCT reports 15 crossed with terms for CLL. 16 We also searched SCOPUS (from inception to June, 2011) and the Cochrane Central Register of Controlled Trials (through the 2nd quarter, 2011), and examined the reference lists of eligible trials and relevant systematic reviews and meta-analyses.6, 7, 16, 17, and 18 Finally, we hand-searched conference abstracts between 2004 and 2010 for RCT reports that have not been published in full. We set no language restrictions. The Appendix provides the exact strategies.

Grouping of treatment regimens

We grouped treatment regimens a priori in the 11 operational categories as described in Table 1 , based on our clinical expertise and the historical developments of treatment strategies for CLL.

Table 1 Grouping of treatment regimens in main analyses and sensitivity analyses a .

# Main analysis Example Sensitivity analysis b
1 Single-agent chlorambucil c Chlorambucil Conventional chemotherapies
2 Conventional combination regimens Cyclophosphamide, doxorubicin, vincristine, prednisone (CHOP)
3 Single-agent fludarabine Fludarabine Purine analogue monotherapies
4 Single-agent cladribine, Cladribine
5 Fludarabine-based combination regimens Fludarabine, cyclophosphamide (FC) Purine analogue-based combination regimens
6 Cladribine-based combination regimens Cladribine, cyclophosphamide
7 Fludarabine–rituximab-based chemoimmunotherapies Fludarabine, cyclophosphamide, rituximab (FCR) Purine analogue-based chemoimmunotherapies
8 Pentostatin–rituximab-based chemoimmunotherapies Pentostatin, cyclophosphamide, rituximab
9 Single-agent alemtuzumab Alemtuzumab Single-agent alemtuzumab
10 Single-agent bendamustine Bendamustine Single-agent bendamustine
11 High-dose chemotherapy with autologous stem cell transplantation d NA NA

a These treatment categories may or may not include steroids in addition to the chemotherapies (e.g., chlorambucil and prednisone, a steroid, was still categorized as single-agent chlorambucil).

b Some of the treatment categories in the main analysis are aggregated: conventional chemotherapies (combining single-agent chlorambucil and conventional combination regimens); purine analogue monotherapies (single-agent fludarabine and single-agent cladribine); purine analogue-based combinations (fludarabine-based combination regimens and cladribine-based combination regimens); purine analogue-based chemoimmunotherapies (fludarabine–rituximab-based chemoimmunotherapies and pentostatin–rituximab-based chemoimmunotherapies).

c Any chlorambucil monotherapies regardless of dose intensity.

d No eligible studies used this treatment, and it is not analyzed further.

Inclusion of eligible trials

One investigator (TTe) performed searches and screened abstracts. Two investigators (TTe, NAT) independently assessed full-text papers for eligibility. We included all first-line therapy RCTs comparing at least two chemo- or chemoimmunotherapy regimens belonging to the 11 categories in adults with intermediate- to high-risk (Rai classification) or Stage B to C (Binet staging) B-cell CLL, who needed treatment. 19 We a priori defined as first-line therapy trials those where at least 80% of enrollees were previously untreated. In sensitivity analysis performed post-hoc, we selected only trials that exclusively enrolled previously untreated patients. We included only comparisons across the specified treatment categories. For example, if a 3-arm trial compared fludarabine versus CHOP versus cyclophosphamide, doxorubicin, prednisone (CAP), we only evaluated the comparisons between fludarabine and CHOP, and fludarabine and CAP, but excluded that between the two combination chemotherapies (CHOP and CAP) as trials comparing regimens belonging to the same treatment category (i.e., CHOP versus CAP) cannot inform the estimates between different treatment categories. We included trials irrespective of sample size, or whether they were stopped early for any reason.

Data extraction

One reviewer (TTe) extracted general information, which was verified by another extractor (NAT). This included bibliographic information (e.g., first author name, journal or conference and year of publication), information on trial design (e.g. description of comparison arms, patient eligibility criteria, definitions of primary and secondary outcomes, methodological quality items as described later), patient characteristics (e.g., demographics, comorbidities, diagnostic criteria, stage) and intervention characteristics. Details on the operational definitions of extracted items are provided in the Appendix .

Our primary outcome of interest was overall survival (OS). We also examined progression-free survival (PFS) and treatment-related mortality (TRM) as secondary outcomes. Two reviewers (TTe, NAT) independently extracted hazard ratios (HRs) for OS and PFS for each treatment comparison. Unadjusted HR was preferred over adjusted HR. If the HR and its variance were not directly extractable, we calculated them from reported statistics using a prespecified algorithm of preferred calculations ( Appendix ). 20 For TRM we extracted crude numbers of death events that the paper authors reported as TRM (no time-to-event data available). Discrepancies were resolved by consensus. A third investigator (BD or TAT) adjudicated any unresolved discrepancies.

We contacted by email authors of trials reported in conference abstracts for additional data. When there was no response after 3 weeks, a second and final attempt was made.

Assessment of methodological quality

Two investigators (TTe, NAT) independently assessed seven methodological validity items from full publications. 21 These referred to random sequence generation, allocation concealment, blinding, intention-to-treat analysis, completeness of follow-up, salvage strategies at progression, and similarity in clinical characteristics between treatment arms. Assessment criteria and additional quality items are provided in the Appendix . We did not derive an overall quality grade for each trial.

Meta-analyses of direct evidence

We used typical random effects meta-analysis 22 to obtain summary HRs of the direct evidence for each comparison with two or more head-to-head trials available. We used the Peto method to combine odds ratios (ORs) for the rare outcome of TRM, as suggested elsewhere. 23 We tested for heterogeneity with Cochran’s Q and quantified its extent with I2. 24 We considered I2 as suggestive of intermediate or high heterogeneity when >50% or >70%, respectively.

Multiple Treatment Meta-analyses (MTM) of combined direct and indirect evidence

We then examined the networks formed by the treatments of interest for OS and PFS, and analyzed them with network meta-analysis, or MTM. (See results section on why we did not perform MTM for TRM.) With MTM we combined the totality of direct and indirect evidence in a single analysis.13, 14, and 25 The main MTM analyses assumed consistency between treatments. In sensitivity analyses we relaxed this assumption.

The Appendix lists the details on our MTM methods and their assumptions, models, assessment of model fit, and software. Briefly, we performed all MTM analyses in the Bayesian framework following previously published approaches.25 and 26 We calculated effect sizes and 95% credible intervals (95% CrI), the Bayesian “analogues” of 95% confidence intervals. Additionally, for each treatment we calculated the probability that it would be ranked first, second, etc., through last among the alternative treatment regimens. 27

Subgroup and sensitivity analyses

We had planned subgroup analyses according to abnormal cytogenetic frequencies, rates of crossover to comparator treatment arms, quality items, whether the trials were available only as a conference abstract, and whether additional data were provided by the authors or not. However, the small number of trials in most subgroups rendered such analyses noninformative.

In sensitivity analyses we repeated MTM analyses allowing the direct and the corresponding indirect effects to differ by an “inconsistency factor”. 26 We found no evidence of inconsistency in the OS and PFS networks (see Supplementary Appendix ) and thus report only results from the main analyses in the text. Finally, we repeated the MTM in an ad hoc set of sensitivity analyses by aggregating some of the predefined treatments to form six categories, as described in Table 1 .

Results

Eligible trials

After abstract level screening and full text review of selected papers, we reviewed the full text of 24 potentially eligible published articles and 18 conference abstracts (describing 40 unique RCTs) ( Appendix Fig. 1). Of these, we excluded 8 non-first-line trials, and 7 trials without data on survival or TRM. Unpublished data were available from one abstract author. A complete list of excluded studies along with reasons for exclusion is provided in the Appendix . Finally, 25 RCTs reported in 27 publications (19 full publications and 8 conference abstracts; 7926 patients total) were included in our meta-analyses ( Appendix Fig. 1).28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, and 54 No eligible trial included high-dose chemotherapy with autologous stem cell transplantation, leaving 10 out of the 11 prespecified treatment categories in our analyses ( Table 1 ). Four trials compared three treatment categories.38, 39, 40, 46, and 54 Only one trial 53 also included previously treated patients (20% of total enrollees), which was excluded post-hoc in sensitivity analysis.

Study and clinical characteristics

The median number of randomized patients was 278 (range, 33–938). The mean or median follow-up duration ranged from 22 to 180 (median, 38) months. Fifteen of 25 trials (64%) clearly reported the primary endpoint, which generally was OS for studies published before year 2000, and response rates or PFS thereafter ( Table 2 and Appendix Table 1).

Table 2 Included randomized clinical trials of treatment for chronic lymphocytic leukemia.

Trial ID (year)Ref Regimen N per arm Median age (range), years Median follow-up, months Median survival time per arm, months
Chlorambucil versus conventional combination regimen          
PETHEMA 1982 (1988) 28 Chlorambucil plus prednisone 48 63 nd 49.1
Cyclophosphamide, melphalan, plus prednisone 48 (46–84)   32.7
FCG-CLL 80B (1990) 29 Chlorambucil 151 63 (nd) 53 58
Cyclophosphamide, vincristine, plus prednisone 140     57
PETHEMA 1988 (1991) 30 Chlorambucil plus prednisone 14 nd nd nd
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 19      
LGCS 1982 (1991) 31 Chlorambucil plus prednisone 57 69 (nd) 38 44
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 56     52
UK MRC CLL1 (1991) 32 Chlorambucil 123 nd nd 44
Cyclophosphamide, vincristine, plus prednisone 115     49
ECOG EST 2480 (1991) 33 Chlorambucil plus prednisone 60 nd 84 57.6
Cyclophosphamide, vincristine, plus prednisone 62     46.8
Danish CLL-2 (1991) 34 Chlorambucil plus prednisone 80 60 (29–75) nd 63.6 b
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 77      
FCG-CLL 85B (1994) 35 Chlorambucil plus prednisone 140 61 (nd) 27 NR
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 147     NR
IGCI CLL-02 (1997) 36 Chlorambucil 116 62 (nd) 37 68
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 112     47
UK MRC CLL3 (2000) 37 Chlorambucil 208 61 (nd) 60 54
Chlorambucil plus epirubicin 210     58
 
Chlorambucil versus single-agent fludarabine          
CALGB 9011 (2000,2009) a 38 and 39 Chlorambucil 193 63 ∼180 56
Fludarabine 179 (32–89)   66
Fludarabine plus chlorambucil 137     55
UK NCRI LRF CLL4 (2007) a 40 Chlorambucil 387 65 41 NR
Fludarabine 194 (35–86)   NR
Fludarabine plus cyclophosphamide 196     NR
GCLLSG CLL5 (2009) 41 Chlorambucil 100 70 43 64
Fludarabine 93 (65–78)   46
 
Chlorambucil versus single-agent cladribine          
PALG CLL1 (2000) 42 Chlorambucil 103 61 nd NR
Cladribine plus prednisone 126 (31–92)   NR
 
Chlorambucil versus alemtuzumab          
CAM 307 (2007) 43 Chlorambucil 148 59 23 NR
Alemtuzumab 149 (35–86)   NR
 
Chlorambucil versus bendamustine          
Knauf (2009,2010)44 and 45 Chlorambucil 157 64 35 65.4
Bendamustine 162 (35–78)   NR
 
Conventional combination regimen versus single-agent fludarabine          
FCG-CLL 90 (2001) 46 Cyclophosphamide, doxorubicin, plus prednisone 240 62 70 70
Cyclophosphamide, doxorubicin, vincristine, plus prednisone 357 (53–71)   67
Fludarabine 341     69
 
Conventional combination regimen versus fludarabine-based regimen          
Abdelhamid (2006) 47 Cyclophosphamide, vincristine, plus prednisone 31 54 nd NR
Fludarabine plus cyclophosphamide 31 (33–65)   NR
Ionita (2009) 48 Cyclophosphamide, vincristine, plus prednisone 43 66   NR
Fludarabine plus cyclophosphamide 44 (37–78) nd NR
Single-agent fludarabine versus fludarabine-based regimen          
GCLLSG CLL4 (2006) 49 Fludarabine 175 59 22 NR
Fludarabine plus cyclophosphamide 176 (42–65)   NR
US intergroup E2997 (2007) 50 Fludarabine 137 61 24 NR
Fludarabine plus cyclophosphamide 141 (33–86)   NR
 
Fludarabine-based regimen versus fludarabine-rituximab-based regimen          
GCLLSG CLL8 (2010) 51 Fludarabine plus cyclophosphamide 409 61 nd nd
Fludarabine, cyclophosphamide, plus rituximab 408 (30–81)   nd
 
Fludarabine-based regimen versus cladribine-based regimen          
PALG CLL3 (2010) 52 Fludarabine plus cyclophosphamide 212 59 38 NR
Cladribine plus cyclophosphamide 211 (27–81)   NR
 
Fludarabine-based regimen versus pentostatin-rituximab-based regimen          
Reynolds (2008) 53 Fludarabine, cyclophosphamide, plus rituximab 92 nd nd nd
Pentostatin, cyclophosphamide, plus rituximab 92      
 
Single-agent cladribine versus cladribine-based regimen          
PALG CLL2 (2006) 54 Cladribine 162 61 nd 51.2
Cladribine plus cyclophosphamide 166 (28–81)   NR
Cladribine mitoxantrone, plus cyclophosphamide 151     NR

a This study also compares chlorambucil with fludarabine-based regimen, and single-agent fludarabine with fludarabine-based regimen.

b Median survival for the entire study participants.

nd = No data; NR = not reached.

Overall, trials enrolled relatively young patients (median age, 54–70 years) without major comorbidities ( Appendix Table 1). In 5 studies reporting cytogenetic information,40, 41, 50, 51, and 52 deletions of the short arm of chromosome 17 were found in up to 15% of patients.

Duration of treatment varied substantially across different regimens ( Appendix Table 2). Chlorambucil monotherapy and conventional combinations had a wide range of treatment period ranging from 3 to 36 months, whereas for newer agents like fludarabine or their combinations duration typically ranged between 6 and 12 months. Median survival times varied across treatments ranging from 33 to 70 months ( Appendix Table 1). Typically, the median survival time was not reached in those studies with a median follow-up period less than 30 months.

In general, methodologic quality items were not explicitly reported ( Appendix Tables 2 and 3). Patient adherence to the assigned treatment strategies were reported in 9 of 25 trials (36%), and information on salvage treatment algorithms or crossover was available in only 8 of 25 trials (32%). The reported per-arm crossover rates ranged from 0% to 71% (median, 13%).

Topology of treatment networks

From 10 treatment categories, it is possible to form 45 pairwise treatment contrasts. Fig. 1 shows the treatment networks formed by the included trials for OS, PFS and TRM. For OS a total of 24 RCTs, 4 of which had 3 arms, provided data on 30 comparisons (i.e., 30 meta-analysis entries, 7741 total patients) for 13 contrasts between pairs of treatments.28, 29, 30, 31, 32, 33, 35, 36, 37, 39, 40, 41, 42, 43, 45, 46, 47, 49, 50, 51, 52, 53, and 54 As shown in Fig. 1 a most comparisons for OS are between single-agent chlorambucil, conventional combinations, single-agent fludarabine, and fludarabine-based combinations. Chlorambucil monotherapy was one of the most compared treatments in 18 meta-analysis entries.

Evidence on PFS and TRM was more limited. For PFS, there were no direct comparisons of single-agent cladribine to other treatments ( Fig. 1 b), resulting in 13 comparisons between 9 treatment groups (3997 patients total).39, 40, 41, 43, 45, 48, 49, 50, 51, 52, and 53 TRM data were available for only 6 treatments ( Fig. 1 c; 4834 patients).33, 34, 38, 40, 43, 46, 48, 49, 50, and 51

gr1

Fig. 1 Topology of treatment networks of chronic lymphocytic leukemia. Overall survival (a), progression-free survival (b), and treatment-related mortality (c) are shown. Treatments are depicted by nodes (ovals) and their links indicating the number of head-to-head comparison.

Direct evidence

Statistically significant difference was found in only 1 (from a single trial) out of 13 treatment contrasts with data on OS: fludarabine-rituximab-based immunochemotherapy was superior to fludarabine-based combination (summary HR = 0.67, 95% CI: 0.48, 0.92). Evidence for statistical heterogeneity was found in 1 of 6 contrasts with 2 or more comparisons (chlorambucil versus fludarabine, 3 comparisons, I2= 75%) ( Appendix Table 4 and Appendix Fig. 2a).

For PFS, 7 out of 9 treatment contrasts showed statistically significant differences, but most were based on data from a single trial. ( Appendix Table 5 and Appendix Fig. 2b). In meta-analyses of head-to-head studies fludarabine monotherapy, fludarabine-based combination, alemtuzumab, and bendamustine were superior to chlorambucil; fludarabine-based combination was better than conventional combination or fludarabine alone; and fludarabine-rituximab-based chemoimmunotherapy was superior to fludarabine-based combination. There was no statistically notable heterogeneity in 2 contrasts with 2 or more comparisons.

Meta-analyses for TRM found significant difference in 2 out of 8 contrasts: chlorambucil had lower TRM than fludarabine-based combination or fludarabine alone; however, these two comparisons were based on only two studies each ( Appendix Table 6 and Appendix Fig. 2c). There was evidence of statistical heterogeneity only in the contrast between chlorambucil versus conventional combination regimens.

Combined direct and indirect evidence with MTM

In all our MTM analyses on OS and PFS results from analyses with and without inconsistency factors were almost identical. This is because the posterior means of the inconsistency factors in the respective models were zero or very near zero. For parsimony, we report data from models that assume consistency.

For OS, the 95% CrI of the HRs of all contrasts included 1 ( Fig. 2 and Appendix Table 4). However, the 95% CrI of most pairwise contrasts could not exclude clinically important effects: a HR as small as 0.70 or as large as 1.43 (=1/0.70) could not be excluded in 38 of 45 possible pairwise contrasts. The two treatments with the highest probability of being most effective were single-agent bendamustine and fludarabine-rituximab-based combinations (posterior cumulative probability to rank among the top two treatments was 70% and 56%, respectively; Appendix Fig. 3a and b).

gr2

Fig. 2 Multiple treatment meta-analysis of overall survival. Closed circles (blue) or diamonds (red) display a summary hazard ratio estimated based on direct meta-analysis or multiple treatment meta-analysis, respectively. Horizontal lines respectively indicate 95% confidence or credible intervals for the summary estimates.

For PFS the 95% CrI of the HRs for 20 out of 36 possible contrasts did not include 1, suggesting that in the pairwise 20 contrasts there is more than 95% probability that one of the two comparators has better PFS ( Fig. 3 , Appendix Table 5). For example, MTM analyses suggested that 6 of 8 newer regimens were superior to chlorambucil. Compared to chlorambucil, the two agents with the largest PFS benefit were bendamustine (summary HR = 0.23, 95% CI: 0.14, 0.36) and fludarabine–rituximab-based combination (summary HR = 0.24, 95% CI: 0.15, 0.40). There was no evidence of difference between these two treatments (summary HR = 0.95, 95% CI: 0.48, 1.79; with HR < 1 favoring bendamustine). Again, bendamustine and fludarabine-rituximab-based combination had the highest probability of being among the two most effective treatments for PFS (86% and 90%, respectively; Appendix Fig. 3c and d).

gr3

Fig. 3 Multiple treatment meta-analysis of progression-free survival. Closed circles (blue) or diamonds (red) display a summary hazard ratio estimated based on direct meta-analysis or multiple treatment meta-analysis, respectively. Horizontal lines respectively indicate 95% confidence or credible intervals for the summary estimates.

The network for TRM was inconsistent and self-contradictory. For this reason MTM analyses are not interpretable and are thus not reported.

Sensitivity analyses

MTM results using the alternative grouping suggested in Table 1 were congruent with the main analyses. ( Appendix Figs. 4–6, Appendix Tables 7 and 8). Alternative models allowing for inconsistency between direct and indirect data were practically identical to the main analyses.

Excluding one trial enrolling also previously treated patients (fludarabine, cyclophosphamide, rituximab versus pentostatin, cyclophosphamide, rituximab) did not change the MTM results (except for deleting the pertinent comparisons involving pentostatin–rituximab-based chemoimmunotherapies).

Discussion

In this study, we reported a 30-year overview of 25 RCTs evaluating 10 treatment categories for untreated CLL in nearly 7800 relatively young and uncomplicated patients. There was no evidence that any treatment is better than any other for OS. Although the credible intervals for most contrasts were wide, no newer therapy was shown to have a survival advantage over chlorambucil monotherapy beyond what is expected by chance. There were differences in PFS between several specific treatment contrasts, but these were based on limited evidence from three or fewer small trials making direct (head-to-head) comparisons. Sparse, limited, and inconsistent data precluded conclusions for the rare outcome of TRM.

Our MTM results are in agreement with previous meta-analyses of direct comparisons6, 7, 8, and 16 in that no particular treatment group is shown to be better than any other in OS. For PFS, we found no meta-analysis of direct data that grouped treatments in the same way as we did here, and thus we cannot assess the congruence of our findings with theirs (see for example references6, 7, 8, and 16). Nevertheless, we deem that our results are congruent with existing meta-analyses of PFS in that newer regimens are found to be better than older regimens: for example, PFS is better for purine analogues alone (i.e., fludarabine or cladribine) compared with alkylator-based regimens (i.e., chlorambucil alone or conventional combination regimens), and for purine analogue-based combinations compared with alkylator-based regimens or purine analogues alone).

Our results contribute to current knowledge because we were able to compare and quantify the efficacy of treatment regimens that have never been compared head-to-head in RCTs or meta-analyses.6, 7, 8, 16, and 17 For example, combined direct and indirect data show no evidence for better OS for any available treatment regimens including bendamustine and fludarabine–rituximab-based chemoimmunotherapy compared to chlorambucil. The 95% CrIs of the MTM results are still wide despite combining direct and indirect evidence, and cannot exclude a HR as small as 0.70 or as large as 1.43 in 38 out of 45 possible pairwise contrasts between the 10 treatments. This indicates that we cannot exclude clinically important effects in either direction and that more trials are needed to draw reliable conclusions. Treatment rankings convey the same information: fludarabine–rituximab-based chemoimmunotherapy or bendamustine have the highest probability of being the two most effective regimens, but neither has very high probability (e.g., more than 95%) of being most effective. This is an important point to highlight when it comes to interpretation of the available evidence for making recommendations.

Our results do not contradict the current expert guidelines9 and 10 that fludarabine-based chemoimmunotherapy be considered over chlorambucil for young uncomplicated patients, given the calculated wide credible intervals around the effect estimates for OS. However, these data highlight the fragility of the evidence base in the recommendations. Indeed, only a single trial 51 has shown that FCR improves survival over FC regimen, and another single trial 39 has shown that fludarabine alone (F) is superior to chlorambucil. No trials have found significant differences for the comparison between FC and F. A naïve argument would claim that if FCR is more effective than FC, FC and F are not significantly different between them, and F is more effective than chlorambucil, then FCR should be more effective than FC, F, and chlorambucil. However, this argument does not account for the uncertainty accompanying the hazard ratio of each comparison and does not include the totality of the available evidence. Our MTM is a theoretically motivated way to perform such calculations, and indicates that randomized evidence is insufficient to show superiority of any particular treatments including FCR.

We also calculated HRs for PFS between treatments that have never been compared head-to-head. For example, no trial compared PFS for bendamustine versus fludarabine–rituximab-based therapies. The MTM analyses suggest that the summary indirect HR between them is close to 1 (with wide 95% CrI). These calculations are based on a path (chain) of indirect comparisons that are not corroborated by direct comparisons or by other independent indirect paths. In Fig. 1 b, the indirect chain is bendamustine versus chlorambucil versus fludarabine-based versus fludarabine–rituximab-based (or bendamustine versus chlorambucil versus fludarabine versus fludarabine-based versus fludarabine–rituximab-based). The chain is as vulnerable as each of its links, and chance findings or systematic errors in a single link can dramatically affect the overall findings.

Our meta-analysis has several limitations. Although we searched multiple sources to locate RCTs, publication bias and selective reporting bias are always a concern.55 and 56 We excluded some otherwise relevant studies that did not have numerical data. Our summary estimates can be biased if the primary studies selectively report only statistically significant findings. MTM results can mislead if the assumption of consistency does not hold, that is, if direct and indirect data are in disagreement. Although there was no evidence that our networks were inconsistent, there were few independent cyclical paths in the networks for OS and PFS (4 and 1, respectively), and this diminishes our ability to detect inconsistencies between direct and indirect data.

Meta-analysis of multiple treatments inherits all the challenges of ordinary meta-analysis, and introduces more of its own. 57 Between trial heterogeneity is probably more of a concern in MTM, as the included trials, while they consistently enrolled otherwise healthy and relatively young patients, span several decades and thus vary in their eligibility criteria, typical regimen dosages, supportive treatments (antimicrobials, hematopoietic growth factors), as well as methodological characteristics such as sample sizes, and follow-up durations. For example, diagnostic criteria for CLL have evolved over the last three decades, which can affect the applicability of our results.

One explanation for not finding differences in OS may be that follow-up of the included trials is not sufficiently long. CLL is a chronic progressive disease, and most deaths occur after the completion of short and medium term trials. To improve statistical power, recent trials use PFS (a composite outcome comprising disease progression and death). However, there is little robust empirical evidence that PFS always predicts OS in CLL. Few included RCTs report both PFS and OS, and we cannot reliably evaluate the concordance of OS and PFS. While better treatment response may result in improvement in PFS and also OS especially in non-frail uncomplicated populations, differences in PFS may only mirror differences in disease progression rates (e.g., mere increase in lymphocyte counts) in other contexts, which may or may not be related to symptoms or quality of life. Another commonly referred explanation of lack of difference in OS is that crossover to comparator arms or other salvage treatments rescue patients progressing on one arm of therapy, 5 and PFS may thus be preferably selected as the primary outcome to remedy this problem. Because few trials reported crossover rates, we could not perform any relevant analysis.

Not withstanding these limitations, physicians and guideline panels still have to make their recommendations, and the same limitations apply to their interpretation of the body of evidence. When resources allow, it is desirable to consider a systematic approach and quantitative MTM methods to inform practice guidelines. 58 Current practice guidelines9 and 10 and expert opinions11 and 12 consistently recommend aggressive and costly chemoimmunotherapy for non-frail younger and chlorambucil or other less aggressive chemoimmunotherapy for older patients depending on presence or absence of 11q or17p deletion. Yet, although PFS seems to have improved in a stepwise manner, there is no robust evidence suggesting improved OS in any newer treatment regimens compared with chlorambucil. In addition, fludarabine-rituximab combination without cyclophophamide, one of the recommended chemoimmunotherapies, has not even been compared with other treatments in RCTs.

One may argue that future research should compare newer treatments that have not been contrasted head-to-head. For young uncomplicated patients, based on our ranking analysis, bendamustine and fludarabine–rituximab-based chemoimmunotherapies such as FCR may be good candidates of priority to be compared if improving PFS is of particular importance. However, future trials should also evaluate OS as the ultimate outcome in any treatment comparisons, given the limited comparative data on OS. Along this line of thinking, any older regimens including chlorambucil monotherapy should not automatically be discarded. For typical frail elderly CLL patients, randomized evidence is generally insufficient and thus any treatment comparisons are informative. Because chlorambucil-based regimens cost less and are less toxic, and as recommended by the current practice guidelines9 and 10 are still a realistic first-line therapy choice, future pragmatic trials assessing newer treatments targeting typical CLL patients should contrast the older regimens as the referent treatment. Given the current emphasis upon the need to establish comparative evidence,59 and 60 more useful data regarding relative effects and harms coupled with the information of cost on, for example, chlorambucil compared with newer regimens in typical patients in clinical practice deserve exploration.

Conflict of interests statement

The authors declare no conflict of interest.

Author Contributions

Authors TTe and TAT had the initial idea and designed the study, which was revised by all authors. All authors acquired the data. Authors TTe and TAT did the statistical analyses, and all authors interpreted the findings. Authors TTe and TAT drafted the first version of the report, which was critically revised by all authors. Author TTe had full access to all of the data in the study and takes responsibility for the integrity of the data. Authors TTe and TAT are guarantors of the accuracy of the data analysis.

Funding

Authors TTe and TAT were supported in part by R01 HS018574 from the United States Agency for Healthcare Research and Quality.

Acknowledgments

Authors TTe and TAT were supported in part by R01 HS018574 from the United States Agency for Healthcare Research and Quality. The funders had no role in the study design, in the collection, analysis and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication. We thank Feng Zhan, PhD (US Oncology Research, The Woodlands, TX) and Craig Reynolds, MD (Ocala Oncology Center, Ocala, FL) for providing the data in their original study. Drs. Zhan and Reynolds did not receive any compensation for their contributions.

Appendix A. Supplementary data

 

Download file

Supplementary data 1

References

  • 1 A. Jemal, R. Siegel, J. Xu, E. Ward. Cancer statistics, 2010. CA Cancer J Clin. 2010;60:277-300 Crossref.
  • 2 L.M. Morton, S.S. Wang, S.S. Devesa, P. Hartge, D.D. Weisenburger, M.S. Linet. Lymphoma incidence patterns by WHO subtype in the United States, 1992–2001. Blood. 2006;107:265-276 Crossref.
  • 3 H. Brenner, A. Gondos, D. Pulte. Trends in long-term survival of patients with chronic lymphocytic leukemia from the 1980s to the early 21st century. Blood. 2008;111:4916-4921 Crossref.
  • 4 G. Dighiero, T.J. Hamblin. Chronic lymphocytic leukaemia. Lancet. 2008;371:1017-1029 Crossref.
  • 5 T.D. Shanafelt, N.E. Kay. Comprehensive management of the CLL patient: a holistic approach. Hematology (Am Soc Hematol Educ Program). 2007;2007:324-331 Crossref.
  • 6 CLL Trialists’ Collaborative Group. Chemotherapeutic options in chronic lymphocytic leukemia: a meta-analysis of the randomized trials. J Natl Cancer Inst. 1999;91:861-868
  • 7 M. Steurer, G. Pall, S. Richards, G. Schwarzer, J. Bohlius, R. Greil. Single-agent purine analogues for the treatment of chronic lymphocytic leukaemia: a systematic review and meta-analysis. Cancer Treat Rev. 2006;32:377-389 Crossref.
  • 8 CLL Trialists’ Collaborative Group. Systematic review of purine analog treatment for chronic lymphocytic leukemia: lessons for future trials. Haematologica. 2012;97:428-436
  • 9 National Comprehensive Cancer Network. Non-Hodgkin’s Lymphomas, Version 2.2012. NCCN Clinical Practice Guidelines in Oncology. 2012.
  • 10 B. Eichhorst, M. Dreyling, T. Robak, E. Montserrat, M. Hallek. Chronic lymphocytic leukemia: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2011;22(Suppl. 6):vi50-vi54 Crossref.
  • 11 M. Hallek, N. Pflug. Chronic lymphocytic leukemia. Ann Oncol. 2010;21(Suppl. 7):vii154-vii164 Crossref.
  • 12 J.G. Gribben, S. O’Brien. Update on therapy of chronic lymphocytic leukemia. J Clin Oncol. 2011;29:544-550 Crossref.
  • 13 D.M. Caldwell, A.E. Ades, J.P. Higgins. Simultaneous comparison of multiple treatments: combining direct and indirect evidence. BMJ. 2005;331:897-900 Crossref.
  • 14 T. Lumley. Network meta-analysis for indirect treatment comparisons. Stat Med. 2002;21:2313-2324 Crossref.
  • 15 K.A. Robinson, K. Dickersin. Development of a highly sensitive search strategy for the retrieval of reports of controlled trials using PubMed. Int J Epidemiol.. 2002;31:150-153 Crossref.
  • 16 M. Steurer, G. Pall, S. Richards, G. Schwarzer, J. Bohlius, R. Greil. Purine antagonists for chronic lymphocytic leukaemia. Cochrane Database Syst Rev. 2006;3:CD004270
  • 17 E. Kimby, L. Brandt, P. Nygren, B. Glimelius. A systematic overview of chemotherapy effects in B-cell chronic lymphocytic leukaemia. Acta Oncol. 2001;40:224-230 Crossref.
  • 18 Q. Zhu, D.C. Tan, M. Samuel, E.S. Chan, Y.C. Linn. Fludarabine in comparison to alkylator-based regimen as induction therapy for chronic lymphocytic leukemia: a systematic review and meta-analysis. Leuk Lymphoma. 2004;45:2239-2245 Crossref.
  • 19 M. Hallek, B.D. Cheson, D. Catovsky, F. Caligaris-Cappio, G. Dighiero, H. Dohner, et al. Guidelines for the diagnosis and treatment of chronic lymphocytic leukemia: a report from the International Workshop on Chronic Lymphocytic Leukemia updating the National Cancer Institute-Working Group 1996 guidelines. Blood. 2008;111:5446-5456 Crossref.
  • 20 J.F. Tierney, L.A. Stewart, D. Ghersi, S. Burdett, M.R. Sydes. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials. 2007;8:16 Crossref.
  • 21 Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.0. 2008.
  • 22 R. Dersimonian, N. Laird. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177-188 Crossref.
  • 23 M.J. Bradburn, J.J. Deeks, J.A. Berlin, L.A. Russell. Much ado about nothing: a comparison of the performance of meta-analytical methods with rare events. Stat Med. 2007;26:53-77 Crossref.
  • 24 J.P. Higgins, S.G. Thompson, J.J. Deeks, D.G. Altman. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557-560 Crossref.
  • 25 G. Lu, A.E. Ades. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med. 2004;23:3105-3124 Crossref.
  • 26 G. Lu, A.E. Ades. Assessing evidence inconsistency in mixed treatment comparisons. J Am Stat Assoc. 2006;101:447-459 Crossref.
  • 27 G. Salanti, A.E. Ades, J.P. Ioannidis. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial. J Clin Epidemiol. 2011;64:163-171 Crossref.
  • 28 E. Montserrat, A. Alcala, C. Alonso, J. Besalduch, J.M. Moraleda, J. Garcia-Conde, et al. A randomized trial comparing chlorambucil plus prednisone vs cyclophosphamide, melphalan, and prednisone in the treatment of chronic lymphocytic leukemia stages B and C. Nouv Rev Fr Hematol. 1988;30:429-432
  • 29 French Cooperative Group on Chronic Lymphocytic L. A randomized clinical trial of chlorambucil versus COP in stage B chronic lymphocytic leukemia. The French Cooperative Group on Chronic Lymphocytic Leukemia. Blood. 1990;75:1422-1425
  • 30 E. Montserrat, M. Fontanilles, J. Estape. Treatment of chronic lymphocytic leukemia: a preliminary report of Spanish (PETHEMA) trials. Leuk Lymphoma. 1991;5:89-91
  • 31 E. Kimby, H. Mellstedt. Chlorambucil/prednisone versus CHOP in symptomatic chronic lymphocytic leukemias of B-cell type: a randomized trial. Leuk Lymphoma. 1991;5:93-96 Crossref.
  • 32 D. Catovsky, S. Richards, J. Fooks, T.J. Hamblin. CLL trials in the United Kingdom: the Medical Research Council CLL Trials 1, 2 and 3. Leuk Lymphoma. 1991;5:105-111 Crossref.
  • 33 B. Raphael, J.W. Andersen, R. Silber, M. Oken, D. Moore, J. Bennett, et al. Comparison of chlorambucil and prednisone versus cyclophosphamide, vincristine, and prednisone as initial treatment for chronic lymphocytic leukemia: long-term follow-up of an Eastern Cooperative Oncology Group randomized clinical trial. J Clin Oncol. 1991;9:770-776
  • 34 M.M. Hansen, E. Andersen, H. Birgens, B.E. Christensen, T.G. Christensen, C. Geisler, et al. CHOP versus chlorambucil + prednisolone in chronic lymphocytic leukemia. Leuk Lymphoma. 1991;5:97-100 Crossref.
  • 35 French Cooperative Group on Chronic Lymphocytic L. Is the CHOP regimen a good treatment for advanced CLL? Results from two randomized clinical trials. French Cooperative Group on Chronic Lymphocytic Leukemia. Leuk Lymphoma. 1994;13:449-456
  • 36 B. Jaksic, M. Brugiatelli, I. Krc, H. Losonczi, J. Holowiecki, A. Planinc-Peraica, et al. High dose chlorambucil versus Binet’s modified cyclophosphamide, doxorubicin, vincristine, and prednisone regimen in the treatment of patients with advanced B-cell chronic lymphocytic leukemia. Results of an international multicenter randomized trial. International Society for Chemo-Immunotherapy, Vienna. Cancer. 1997;79:2107-2114 Crossref.
  • 37 D. Catovsky, T.J. Hamblin, S. Richards. UK Medical Research Council CLL3 – A randomized trial of the addition of epirubicin to standard chlorambucil treatment. ASH Annual Meeting Abstracts. 2000;
  • 38 K.R. Rai, B.L. Peterson, F.R. Appelbaum, J. Kolitz, L. Elias, L. Shepherd, et al. Fludarabine compared with chlorambucil as primary therapy for chronic lymphocytic leukemia. N Engl J Med. 2000;343:1750-1757 Crossref.
  • 39 K.R. Rai, B.L. Peterson, F.R. Appelbaum, M.S. Tallman, A. Belch, V.A. Morrison, et al. Long-term survival analysis of the North American Intergroup Study C9011 comparing fludarabine (F) and chlorambucil (C) in previously untreated patients with Chronic Lymphocytic Leukemia (CLL) [abstract]. ASH Annual Meeting Abstracts. 2009;
  • 40 D. Catovsky, S. Richards, E. Matutes, D. Oscier, M.J. Dyer, R.F. Bezares, et al. Assessment of fludarabine plus cyclophosphamide for patients with chronic lymphocytic leukaemia (the LRF CLL4 Trial): a randomised controlled trial. Lancet. 2007;370:230-239 Crossref.
  • 41 B.F. Eichhorst, R. Busch, S. Stilgenbauer, M. Stauch, M.A. Bergmann, M. Ritgen, et al. First-line therapy with fludarabine compared with chlorambucil does not result in a major benefit for elderly patients with advanced chronic lymphocytic leukemia. Blood. 2009;114:3382-3391 Crossref.
  • 42 T. Robak, J.Z. Blonski, M. Kasznicki, M. Blasinska-Morawiec, E. Krykowski, A. Dmoszynska, et al. Cladribine with prednisone versus chlorambucil with prednisone as first-line therapy in chronic lymphocytic leukemia: report of a prospective, randomized, multicenter trial. Blood. 2000;96:2723-2729
  • 43 P. Hillmen, A.B. Skotnicki, T. Robak, B. Jaksic, A. Dmoszynska, J. Wu, et al. Alemtuzumab compared with chlorambucil as first-line therapy for chronic lymphocytic leukemia. J Clin Oncol. 2007;25:5616-5623 Crossref.
  • 44 W.U. Knauf, T. Lissichkov, A. Aldaoud, A. Liberati, J. Loscertales, R. Herbrecht, et al. Phase III randomized study of bendamustine compared with chlorambucil in previously untreated patients with chronic lymphocytic leukemia. J Clin Oncol. 2009;27:4378-4384 Crossref.
  • 45 W.U. Knauf, T. Lissitchkov, A. Aldaoud, A.M. Liberati, J. Loscertales, R. Herbrecht, et al. Bendamustine induces higher remission rates, prolongs progression free survival as well as time to next treatment, and improves overall survival for patients in complete remission without compromising quality of life when compared to chlorambucil in first line treatment of chronic lymphocytic leukemia [abstract]. ASH Annual Meeting Abstracts. 2010;116:2449
  • 46 M. Leporrier, S. Chevret, B. Cazin, N. Boudjerra, P. Feugier, B. Desablens, et al. Randomized comparison of fludarabine, CAP, and ChOP in 938 previously untreated stage B and C chronic lymphocytic leukemia patients. Blood. 2001;98:2319-2325 Crossref.
  • 47 T.H. Abdelhamid, N. Al-lahloby, A. Kamel, N. Elsharkawy, A. Zaher. Fludarabine and cyclophosphamid (Fc) Versus cyclophosphamide, vincristine and prednisone (Cvp) As first line treatment in chronic lymphocytic leukaemia (Cll) [abstract]. Haematologica. 2006;:290-291
  • 48 H. Ionita, L. Chreveresan, M. Ionita, M. Delamarian, M. Racz. Fludarabine and cyclophosphamide (fc) versus cyclophosphamide, vincristine and prednisone (cvp) as first line therapy in Chronic Lymphocytic Leukemia (CLL) [abstract]. Haematologica. 2009;94:599-600
  • 49 B.F. Eichhorst, R. Busch, G. Hopfinger, R. Pasold, M. Hensel, C. Steinbrecher, et al. Fludarabine plus cyclophosphamide versus fludarabine alone in first-line therapy of younger patients with chronic lymphocytic leukemia. Blood. 2006;107:885-891
  • 50 I.W. Flinn, D.S. Neuberg, M.R. Grever, G.W. Dewald, J.M. Bennett, E.M. Paietta, et al. Phase III trial of fludarabine plus cyclophosphamide compared with fludarabine for patients with previously untreated chronic lymphocytic leukemia: US Intergroup Trial E2997. J Clin Oncol. 2007;25:793-798 Crossref.
  • 51 M. Hallek, K. Fischer, G. Fingerle-Rowson, A.M. Fink, R. Busch, J. Mayer, et al. Addition of rituximab to fludarabine and cyclophosphamide in patients with chronic lymphocytic leukaemia: a randomised, open-label, phase 3 trial. Lancet. 2010;376:1164-1174 Crossref.
  • 52 T. Robak, K. Jamroziak, J. Gora-Tybor, B. Stella-Holowiecka, L. Konopka, B. Ceglarek, et al. Comparison of cladribine plus cyclophosphamide with fludarabine plus cyclophosphamide as first-line therapy for chronic lymphocytic leukemia: a phase III randomized study by the Polish Adult Leukemia Group (PALG-CLL3 Study). J Clin Oncol. 2010;28:1863-1869 Crossref.
  • 53 C. Reynolds, N. Di Bella, R.M. Lyons, W.J. Hyman, G.L. Lee, D.A. Richards, et al. Phase III trial of fludarabine, cyclophosphamide, and rituximab vs. pentostatin, cyclophosphamide, and rituximab in b-cell chronic lymphocytic leukemia [abstract]. ASH Annual Meeting Abstracts. 2008;
  • 54 T. Robak, J.Z. Blonski, J. Gora-Tybor, K. Jamroziak, J. Dwilewicz-Trojaczek, A. Tomaszewska, et al. Cladribine alone and in combination with cyclophosphamide or cyclophosphamide plus mitoxantrone in the treatment of progressive chronic lymphocytic leukemia: report of a prospective, multicenter, randomized trial of the Polish Adult Leukemia Group (PALG CLL2). Blood. 2006;108:473-479 Crossref.
  • 55 A.W. Chan, A. Hrobjartsson, M.T. Haahr, P.C. Gotzsche, D.G. Altman. Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles. JAMA. 2004;291:2457-2465 Crossref.
  • 56 K. Dickersin, Y.I. Min. Publication bias: the problem that won’t go away. Ann N Y Acad Sci. 1993;703:135-146
  • 57 F. Song, Y.K. Loke, T. Walsh, A.M. Glenny, A.J. Eastwood, D.G. Altman. Methodological problems in the use of indirect comparisons for evaluating healthcare interventions: survey of published systematic reviews. BMJ. 2009;338:b1147 Crossref.
  • 58 Finding What Works in Health Care: Standards for Systematic Reviews. Washington, DC: The National Academies Press; 2011.
  • 59 P.H. Conway, C. Clancy. Comparative-effectiveness research–implications of the Federal Coordinating Council’s report. N Engl J Med. 2009;361:328-330 Crossref.
  • 60 K.G. Volpp, A. Das. Comparative effectiveness–thinking beyond medication A versus medication B. N Engl J Med. 2009;361:331-333 Crossref.

Footnotes

a Department of Internal Medicine, Nanakuri Sanatorium, Fujita Health University School of Medicine, Tsu, Japan

b Center for Clinical Evidence Synthesis, Institute for Clinical Research and Health Policy Studies, Tufts Medical Center, Boston, MA, USA

c Clinical Research Center for Blood Diseases, National Hospital Organization Nagoya Medical Center, Nagoya, Japan

d University of Maryland, Marlene and Stewart Greenebaum Cancer Center, Baltimore, MD, USA

e Clinical Translation Science Institute, Center for Evidence-based Medicine and Health Outcome Research, University of South Florida, Tampa, FL, USA

f Department of Hematology and Department of Health Outcomes Behavior, H Lee Moffitt Cancer Center and Research Institute, Tampa, FL, USA

lowast Corresponding author at: Fujita Health University Nanakuri Sanatorium, 424-1 Odoricho, Tsu, Mie, 514-1295 Japan. Tel.: +81 59 252 1555; fax: +81 59 252 1383.