A Systematic Review and Meta-Analysis of Front-line Anthracycline-Based Chemotherapy Regimens for Peripheral T-Cell Lymphoma

Anthracycline-based chemotherapy remains standard treatment for peripheral T-cell lymphoma (PTCL) although its benefits have been questioned. We performed systematic literature review and meta-analyses examining the complete response (CR) and overall survival (OS) rates for patients with PTCL. The CR rate for PTCL patients ranged from 35.9% (95% CI 23.4–50.7%) for enteropathy-type T-cell lymphoma (ETTL) to 65.8% (95% CI 54.0–75.9%) for anaplastic large cell lymphoma (ALCL). The 5-year OS was 38.5% (95% CI 35.5–41.6%) for all PTCL patients and ranged from 20.3% (95% CI 12.5–31.2%) for ETTL to 56.5% (95% CI 42.8–69.2%) for ALCL. These data suggest that there is marked heterogeneity across PTCL subtypes in the benefits of anthracycline-based chemotherapy. While anthracyclines produce CR in half of PTCL patients, this yields reasonable 5-year OS for patients with ALCL but not for those with PTCL-NOS or ETTL. Novel agents and regimens are needed to improve outcomes for these patients.


Introduction
Peripheral T-cell lymphoma (PTCL) is a heterogeneous group of non-Hodgkin's lymphomas (NHL) characterized by poor treatment outcome with conventional chemotherapy. Anthracycline-based chemotherapy remains the standard treatment for patients with PTCL although such regimens have failed to induce sustained remissions for most patients. The role of anthracyclines in the treatment of PTCL remains debatable. The International PTCL Clinical and Pathologic Review Project retrospectively demonstrated no difference in overall survival (OS) comparing patients who did or did not receive an anthracycline for PTCL [1]. Prior studies have established worse outcome for PTCL compared to aggressive B-cell NHL treated with anthracycline-based chemotherapy, in terms of response, relapse, and OS rates [2][3][4]. Since there are no large randomized prospective studies that compare the benefits of anthracycline-based therapies to other strategies, we conducted a systematic literature review and metaanalysis of first-line therapy for PTCL patients to elucidate the role of anthracyclines and examine the complete response (CR) and OS rates associated with anthracycline-based regimens.
Given the well established favorable outcomes for patients with anaplastic lymphoma kinase (ALK) positive anaplastic large cell lymphomas (ALCL) [5], along with the heterogeneity in response and survival rates across PTCL subgroups, we focused our analyses on non-ALCL PTCL and performed subgroup meta-analyses on the outcomes of anthracycline-based regimens for patients with PTCL-not 2 ISRN Hematology otherwise specified (NOS), angioimmunoblastic T-cell lymphoma (AITL), natural-killer/T-cell (NK/T-cell) NHL, and enteropathy-type T-cell lymphoma (ETTL) subtypes.

Systematic Literature
Review. Studies were identified by searching Medline and Google Scholar databases through 2010 and the conference proceedings of the American Society of Hematology and the American Society of Clinical Oncology for the years 2003 to 2010. Each search used combinations of the terms "Peripheral T-Cell Lymphoma," "T-Cell Lymphoma," "Anthracyclines," "CHOP," "Doxorubicin," "Mitoxantrone," "Daunorubicin," "CVAD," and "Adriamycin." Two reviewers (A. N. AbouYabis and P. J. Shenoy) performed study selection, quality assessment, and data extraction independently using standardized forms. Any disagreement was resolved by a third reviewer (C. R. Flowers or M. J. Lechowicz).

Meta-Analysis Inclusion Criteria, Study Selection, and Data Extraction.
Criteria for including studies in the metaanalysis were (1) studies involving patients with untreated PTCL (studies involving relapsed/refractory PTCL patients were included only if they provided separate outcome data for untreated PTCL patients), (2) treatment with anthracycline-based regimen, (3) reporting in English, and (4) reporting of CR rates and/or 5-year OS. Only full text reports were included as most abstracts presented preliminary results with a short followup. The primary outcome measures were OS and CR. Extracted data also included type of study (prospective/retrospective), PTCL subtype, pretreatment disease status, and median follow-up time. Studies were carefully screened for possible duplication of study population based on the participating institutions and period of presentation of patients. Additional studies not included in the meta-analysis were discussed in the narrative review.

Data Analysis and Statistical Methods.
Studies included in the subtype-specific and combined PTCL meta-analyses were evaluated for heterogeneity as described below and evaluated for suitability for pooling. Pooled estimates of the CR rate and the 5-year OS for patients treated with anthracycline-containing regimens were computed. DerSimonian and Laird random effects [6] and Mantel-Haenszel fixed effect models [7] were used to combine subgroups to determine the overall effects. For each analysis, a forest plot was generated to display results. The study-to-study variance (tau-squared) was not assumed to be the same for all subgroups; this value was computed within subgroups and not pooled across subgroups. The consistency of results (effect sizes) among studies was investigated by means of two heterogeneity tests, the χ 2 -based Cochran's Q test, and the I 2 statistic. To evaluate heterogeneity across reported results, we performed visual inspection of forest plots and compared the fixed effects and random effects models.
We considered that heterogeneity was present when the P value of the Cochran's Q test was <.1 and I 2 statistic was >50%. Potential publication bias was estimated with the Begg and Mazumdar rank correlation test [8], Egger's test of intercept [9], and Duval and Tweedie's trim and fill test [10]. We also evaluated whether our estimates of CR and 5-year OS were influenced by publication bias by assessing funnel plots of the logit of the estimate versus its standard error and by comparing pooled estimates of CR and 5-year OS rates for full text reports and studies reported as abstracts only. Tests for publication bias were repeated after inclusion and exclusion of abstracts published during that timeframe but not later published as full manuscripts to determine if adding abstracts in the meta-analyses changed the results and added significant value.
Univariate metaregression analyses were conducted to identify patient characteristics (age > 60 years, male gender, stage III/IV, B symptoms, extranodal disease, LDH >upper limit of normal, ECOG performance status ≥ 2, bone marrow involvement, and high/high intermediate International prognostic index score) that were significant predictors of CR rate and 5-year OS. A two-sided alpha error of 0.05 was used to declare statistical significance. Statistical analyses were performed using SAS software, Version 9.1.3 (SAS Institute Inc., Cary, NC, USA) and Comprehensive Meta Analysis, Version 2 (Biostat Inc.).

Results
Overall, 389 potentially relevant references describing initial treatments for PTCL were identified and screened for retrieval. Among these, 44 studies involved first-line therapy with anthracyclines; six studies were excluded due to possible duplication of study population, and six studies published as abstracts only were excluded. Figure 1 depicts the exclusion of articles. Thirty-one studies (13 prospective and 18 retrospective) that involved 2,815 patients and met all the inclusion and exclusion criteria were included in the meta-analyses (Tables 1 and 2).
NK/T-cell NHL demonstrated CR rates ranging between 40% and 70% in localized disease and a CR rate of 25% in advanced disease [22,26,33]. In patients with stage I/II disease, radiotherapy (RT) improved CR rates to 52-100% [25,30,31,35,37,[41][42][43][44]. Given the high rate of disease progression during standard CHOP chemotherapy (35%) as well as the previously documented superiority of CHOP-14 over standard CHOP by the German high-grade NHL study group [45,46], two cycles of dose-intense CHOP every two weeks (DI-CHOP-14) followed by RT and consolidation   [18,47]. Overall, this regimen improved the CR rate to 76%, as compared to the 58% achieved by standard CHOP-21. The authors believed that this approach also reduced locoregional failure, likely due to early institution of RT [18,47].

CHOP versus More Intensive Anthracycline-Based
Regimens. There have been conflicting data regarding the benefits of intensive anthracycline-based chemotherapy regimens compared with CHOP in the first-line treatment of PTCL and other aggressive NHLs. A randomized phase 3 trial demonstrated no significant difference in the partial response, CR rates, or 3-year OS comparing CHOP and three intensive chemotherapy regimens for all patients with aggressive NHL [48]. The prospective "LTP95" protocol showed no difference in event-free survival (EFS) or OS between CHOP and vinblastine-ifosfamide-cisplatin and doxorubicin-bleomycin-vinblastine-dacarbazine regimen [49]. In a retrospective study comparing cohorts of PTCL patients, excluding ALCL, the CR rates were similar between CHOP and more intensive regimens, including hyper-CVAD (58% versus 59%, P = .99), with an estimated 3-year OS rates of 43% and 49%, respectively [34]. There has, however, been one randomized trial showing an 18% difference in disease-free survival (DFS) at 5 years (P = .0002) favoring intensive chemotherapy over CHOP [50].

Complete Remission Rates.
In order to determine if pooling the studies for subtype-specific meta-analysis and as an entire PTCL group was appropriate, we first determined the P value for the Cochran's Q test and the I 2 statistic for the studies to identify and quantify the level of heterogeneity [53,54]. When combining studies to estimate CR, the P value for the Cochran's Q test was <.01 and the I 2 statistic was 59% indicating substantial heterogeneity [53]; hence, a pooled estimate of the CR rate for all PTCL patients was not calculated. In the subtype-specific analysis, AITL, ALCL, and NK/T nasal demonstrated no evidence of heterogeneity (all I 2 < 50%) and hence were pooled for subtype-specific metaanalysis.
Analyses by PTCL subtype (Figure 2, black diamonds) revealed the following CR rates: ALCL 65   The funnel plot (Figure 3) of studies included in the meta-analysis of CR rates appeared symmetrical. Begg and Mazumdar's rank correlation test (P = .16), Egger's test of intercept (P = . 16), and Duval and Tweedie's trim and fill test (no studies added) all indicated that there was no clear evidence of publication bias. Pooled CR estimates of abstracts and full text reports were not significantly different (P = .49). Tests for publication bias repeated after inclusion of abstracts showed that adding abstracts did not add any significant informational value. The univariate metaregression analysis of prospective studies revealed that age at diagnosis > 60 (P = .03), the presence of stage III/IV (P = .046), B-symptoms (P = .012), and high/high-intermediate IPI (P = .056) were significant predictors of lower CR rates (Table 3).
The funnel plot ( Figure 5) of studies included in the meta-analysis of 5-year OS rates showed wide variance but was predominantly symmetrical. The Begg and Mazumdar's rank correlation test (P = .50), Egger's test of intercept (P = .26), and Duval and Tweedie's trim and fill test (no studies added) all indicated that there was no clear evidence of publication bias. Pooled 5-year OS estimates of abstracts and full text reports were not significantly different (P = .87). Tests for publication bias repeated after inclusion of abstracts showed that adding abstracts did not add any significant value.

Discussion
Despite the established worse outcome of anthracyclinebased chemotherapy in PTCL patients compared to aggressive B-cell NHL [2][3][4], such regimens, especially CHOP, have remained the standard treatment for PTCL. In this meta-analysis, the CR rate achieved with anthracycline-based regimens ranged from 36% in ETTL to 66% in ALCL. Fiveyear OS across PTCL subtypes also ranged widely from 20% in ETTL to 57% in ALCL. As has already been established [4,40], ALCL patients had a markedly better 5-year OS than other PTCL patients (57% versus 37%, P < .001). However, it is worth noting here that the majority of studies did not report on the ALK status of ALCL patients. Results on ALK-negative and ALK-positive ALCL patients were grouped together in most studies. Only three studies that reported on ALK-negative ALCL had sufficient data to be included in the meta-analysis [1,15,17] and revealed a CR rate of 62% and a 5-year OS of 49% [1] in this subgroup. Although ALK-negative ALCL has poorer outcomes when compared to ALK-positive ALCL, both subsets of ALCL appear to have superior outcomes when compared to other PTCL subtypes. Note: only prospective studies were included in the metaregression.

Study Limitations.
Until recently, the majority of our knowledge about treatment outcomes in PTCL emerged from small phase 2 studies. Hence, the data included in this meta-analysis comes from phase 2 clinical trials with all the inherent drawbacks of such studies. Additionally, the low prevalence of PTCL and its heterogeneity makes it difficult to draw conclusions from most of these small individual studies. It should also be noted that there were no consistent standards for reporting results. This limited the number of studies eligible for inclusion. Moreover, most studies did not specify the timing and criteria for CR determination and utilized investigator determination of CR rather than central review. This might lead to overestimate of the CR rate. Some studies did not report CR and OS for PTCL subtypes separately. These contributed to the review of overall CR and 5-year OS rates but not to subtype estimates.
High-quality meta-analyses often utilize data from randomized, double-blinded, controlled trials with descriptions of dropouts and withdrawals [55]. We did not perform Jadad scoring of individual trials because PTCL and its subtypes are rare diseases that precluded performance of large randomized trials. While, in an ideal setting, a metaanalysis for PTCLs would utilize data from randomized phase 3 trials, our meta-analyses represent a review of the best available data from existing publications on the treatment of PTCL.
Another concern raised by previous studies is the heterogeneity of PTCL populations in the published literature. This was addressed by separating out PTCL subtypes before computing estimates of CR and 5-year OS rates. Moreover, there was heterogeneity in reporting of data across studies. The meta-analysis of CR rate for PTCL, NOS, and PTCL as a whole group was associated with moderate heterogeneity (I 2 71% and 59%, resp.) as was the meta-analysis of 5-year OS rate for PTCL and NOS (I 2 = 54%), while for NK/T-cell, nasal lymphoma, and PTCL as a whole group was associated with high degree of heterogeneity (I 2 > 75%). To address this, random effects models were used but this degree of inconsistency may yield instability in the results. Nevertheless, these analyses represent the best available data for these rare disease entities and may only be improved by larger randomized controlled trials which have not been performed to date and may not be practical for some PTCL subsets.
We attempted to estimate the PFS rates achieved with anthracycline-based regimens, but while some authors reported PFS rates, others presented DFS or EFS rates without consistent definitions of events precluding combination. Unfortunately, the international prognostic index was not reported for the majority of patients participating in the included trials. However, whenever available, we extracted patient data and included it in the metaregression of predictors of CR shown in Table 3. Moreover, some studies included in the meta-analysis of CR rates were presentations of early results and had insufficient followup to be included in the meta-analysis of 5-year OS.
As in any meta-analysis, there was concern for publication bias. We addressed this issue using funnel plots and statistical testing and did not find evidence suggestive of overt publication bias. We also compared analysis with the addition of meeting abstracts that were not published as full manuscripts and determined that adding these studies did not influence our results.

Is There a Benefit for Anthracyclines in the Treatment of PTCL Patients outside of the ALCL Subsets?
To answer this question it may be useful to look at the results of anthracycline-based regimens in the treatment of other NHL subtypes where their benefit has already been established. While the CR rate of 52% from this meta-analysis achieved with anthracyclines in the treatment of PTCL patients (excluding ALK-positive ALCL) seems comparable to the 44%-63% CR rates obtained with CHOP chemotherapy in diffuse large B-cell lymphoma (DLBCL) patients in the pre-rituximab era [48,56], the estimated 5-year OS for PTCL patients was only 35% compared to the 41% in intermediate-grade NHL in general [57] and 45%-70% in DLBCL [58][59][60]. Moreover, the International PTCL Clinical and Pathologic Review Project retrospectively demonstrated no difference in OS comparing patients who did and did not receive an anthracycline for any subtype of PTCL [1].   CHOP-type regimens (like DI-CHOP) may overcome the high rate of disease progression and induce better long-term responses in PTCL [18,47]. The German high-grade NHL study group has previously shown that reducing the interval between CHOP cycles from 21 days to 14 days improves EFS [45] as well as OS [45,46]. Another recent approach to improve efficacy of CHOP regimen was substituting doxorubicin with pegylated liposomal doxorubicin, which was suggested to persist in the blood circulation significantly longer than doxorubicin [61].

What Modifications to the Current Anthracycline-Based Regimens Could Improve Treatment Outcomes in PTCL
Addition of Other Active Agents? Other modifications to the CHOP-based chemotherapy include the addition of etoposide [45,46], bortezomib, and purine analogs [17,[62][63][64][65][66][67][68]. Although the addition of etoposide to CHOP-based regimens improved CR in T-NHL patients in some studies, EFS and OS rates have been disappointing when compared to B-NHL [13,69,70]. On the other hand, encouraging results have been achieved adding gemcitabine to CHOPetoposide in the front-line setting [17,68]. Another approach to overcome drug resistance in PTCL is the combination of CHOP with the proteasome inhibitor bortezomib. This currently is being investigated in a phase 1/2 study in patients with advanced stage PTCL [71]. Another option for improving the outcome of anthracycline-based chemotherapy in PTCL is combining the anti-angiogenic agent bevacizumab with CHOP (A-CHOP). The basis of this combination emerges from the finding that serum concentrations of the vascular endothelial growth factor (VEGF) have an independent prognostic influence on survival in NHL [72]. Moreover, significant expression of VEGF transcripts was observed in PTCL, particularly in AITL [73]. This combination has been studied in the Eastern Cooperative Oncology Group (ECOG).
Incorporation of Immunotherapy? The significant improvements in response rates and survival achieved by adding rituximab to CHOP-type chemotherapy in DLBCL [56,60,74] triggered the investigation of similar chemoimmunotherapy combinations in PTCL. The interim analysis of the CONCEPT phase 2 study of denileukin diftitox in combination with CHOP revealed a 68% overall response rate and 57% CR in patients with PTCL [75]. Longterm results are still awaited. Alemtuzumab-chemotherapy combinations also have been evaluated in several studies [76][77][78]. Alemtuzumab-CHOP combinations in the frontline treatment of non-ALCL PTCL patients resulted in 60-80% CR rate with estimated 2-year OS of 53-75% [79]. The estimated 2-year EFS in those studies ranged between 40% and 50%. Moreover, the combination of alemtuzumab with fludarabine-cyclophosphamide-doxorubicin induced remission in 63% (58% CR) of ALK-negative PTCL patients treated in the front-line setting [80]. However, infections have been a concern in all alemtuzumab-combination studies.

Incorporation of Radiation Therapy?
The addition of RT also may be important particularly in NK/T cell lymphoma. Our meta-analysis revealed that anthracycline-based chemotherapy alone resulted in a CR rate of 49% (95% CI 38.7-59.5%) while chemoradiotherapy resulted in CR of 59.6% (95% CI 42-75%; P = .02). However, these results cannot be used as strong evidence to the benefits of combination of chemoradiotherapy in NK/T-cell lymphoma over chemotherapy alone, since most studies included in this meta-analysis did not report treatment results in relation to the extent and stage of the disease [18,26,30,31,33,35,37,43,47,81]. Clarifying the benefits of chemoradiotherapy over chemotherapy in NK/T cell NHL will require larger randomized controlled clinical trials.

Consolidating Responses with Hematopoietic Stem Cell Transplant?
Since the major difficulty in the treatment of PTCL with conventional chemotherapy has been sustaining CR [3,23,27,29,39], autologous and allogeneic hematopoietic stem cell transplant have been explored as ways to consolidate remissions and prolong survival among PTCL patients treated with anthracycline-based regimens. However, there has been conflicting data regarding the advantage of performing transplantation in first remission in PTCL [82][83][84][85][86][87][88][89][90].

What Are the Alternatives to Anthracyclines in PTCL Management?
In the search for nonanthracycline alternatives, some agents have demonstrated promise as single agents in NHL in general and PTCL in particular. Of those are the anti-folate agents, such as pralatrexate, which showed 28% overall response rate in patients with relapsed/refractory PTCL and 49% rate of disease control leading to its FDA approval in this setting [91]. Moreover, the activity of purine nucleoside analogues such as pentostatin, nelarabine, and gemcitabine appears promising as well [62][63][64][65][66][67]92]. Histone deacetylase (HDAC) inhibitors, such as vorinostat and romidepsin, are another promising group of agents that are being tested as single agents or in combination in treatment of PTCL both front-line and relapsed/refractory settings [93,94]. Although nonanthracycline-based chemotherapy seems  an attractive alternative, comparative studies are needed to establish its benefit over anthracyclines in PTCL.

Conclusion
Despite achieving CR with front-line anthracycline-based therapy in more than half of PTCL patients, 5-year OS remains poor for most PTCL subtypes. Although CHOP remains the standard front-line therapy for PTCL, given these poor outcomes, most PTCL patients (with the exception of ALK-positive ALCL) should be considered for clinical trials of initial therapy. Future clinical trials need to focus on subtype-specific treatment, incorporation of newer agents, and nonanthracycline-based combinations to improve the long-term outcome for PTCL patients. Moreover, strategies capable of sustaining responses such as maintenance therapy and transplant consolidation should be actively investigated in prospective clinical trials. Given the small number of PTCL patients seen by any single institution, such trials need to be multicentered and community oncologists should encourage their patients to enroll in such studies whenever available.

Disclosure
Dr. M. J. Lechowicz has been a consultant for the Allos Therapeutics and Eisai in the past. Dr. M. J. Lechowicz has also participated in clinical trials using pralatrexate, romidepsin, and denileukin diftitox. The authors alone are responsible for the content and writing of the paper.