Failing Forward in Peripheral T-Cell Lymphoma

Peripheral T-cell lymphomas (PTCLs) are an uncommon, heterogeneous group of non-Hodgkin lymphomas. The most frequent subtypes are T-follicular helper (TFH) cell, peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS), and systemic anaplastic large cell lymphoma (sALCL). Standardly used frontline therapeutic strategies are borrowed from other aggressive lymphomas with historic 5-year overall survivals (OSs) of 30%-40%.^1,2 In the landmark ECHELON-2 study, brentuximab vedotin (BV) in combination with cyclophosphamide, doxorubicin, and prednisone (CHP) resulted in significant improvements in progression-free survival (PFS) and OS.³ The greatest benefits were seen in ALCL, setting the stage for subtype-specific therapy in PTCL.³ In the past decade, we have learned that PTCL derived from TFH cells, such as nodal TFH lymphoma (TFHL) angioimmunoblastic-type (AITL), nodal TFHL follicular-type, and nodal TFHL not otherwise specified, share genomic and gene expression signatures and appear to have greater sensitivity to newer classes of therapies.^4-6

THE TAKEAWAY

• In the article that accompanies this editorial, Camus et al⁷ have updated and reanalyzed their negative study of frontline romidepsin therapy in peripheral T-cell lymphomas and demonstrate that patients with T-follicular helper phenotypes benefit from romidepsin combined with cyclophosphamide, doxorubicin, vincristine, and prednisone chemotherapy. This observation regarding differential subtype-specific efficacy in peripheral T-cell lymphoma is now seen multiple studies and should be accounted for in future studies.

In the article that accompanies this editorial, Camus et al⁷ updated and reanalyzed their negative study of frontline therapy in PTCL. In the aspirational spirit of learning from failure, their post hoc analysis teaches important lessons for improving outcomes in PTCL. Many previous attempts to improve results for PTCL have failed to make headway.^8-11 We have recurrently underestimated the impact of heterogeneity in PTCL, in part due to lack of understanding. In this international phase III study conducted at 98 centers between 2013 and 2017, 421 patients with untreated PTCL were randomly assigned to cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) with or without romidepsin.¹² The study sought to improve PFS as a necessary step toward improving OS. It failed to meet its primary end point with PFS 12.0 months for romidespin, cyclophosphamide, doxorubicin, vincristine, and prednisone (Ro-CHOP) versus 10.2 months for CHOP (P = .096). In this update, the authors reclassified patients into TFH phenotype versus non-TFH showing a near doubling of PFS for those with a TFH phenotype (19.5 v 10.6 months, hazard ratio, 0.703 [95% CI, 0.502 to 0.985], P = .0395).

The analysis builds on the increasing recognition of the unique biology of TFHL.^13,14 TFHLs are characterized by mutations in genes involved in chromatin modification and the microenvironment such as TET2, RHOA, DNMT3A, and IDH2 and frequently develop on a background of clonal hematopoiesis.¹⁵ Concurrently, multiple studies suggest that TFHLs are more responsive to therapies targeting epigenetics, including the histone deacetylase (HDAC) inhibitor, romidepsin. This differential response was only suggested in the phase II study that led to accelerated approval of romidepsin, with an overall response rate of 25% for all patients but 33% in AITL.^16-18 The registration study of another HDAC inhibitor, belinostat, similarly showed preferential activity in AITL compared with PTCL-NOS or ALCL.¹⁹ Subsequent combination studies with romidepsin in the relapsed/refractory setting repeatedly demonstrated that those with TFHL have higher response rates, complete response (CR) rates, and longer durations of response.^20-23 In a retrospective multicenter study of HDAC inhibitor–based therapy, those with TFHL had better responses (objective response rate, 58% v 30%) and higher CR rates (29% v 12%).²⁴ Given these observations from smaller studies, it is unsurprising that if there were benefits of Ro-CHOP over CHOP, which would be within the TFH subset. Importantly, this understanding was not known before initiating the Ro-CHOP study.

Compare this with another large prospective randomized study, ECHELON-2, which demonstrated improved PFS and OS for BV-CHP versus CHOP in patients with CD30-expressing PTCL.³ ECHELON-2 was not only enriched for ALCL, which is most likely to benefit from BV, but also designed to evaluate the overall study population and ALCL specifically. At the time of study design, the significant potency of BV in relapsed/refractory ALCL was well understood and the impetus to independently power the ALCL subgroup was a regulatory requirement. It was similarly fortuitous that the addition of BV did not lead to a reduction in the dose intensity of chemotherapy, as occurred in the Ro-CHOP study. The result was a new standard of care with a higher cure rate for a subset of patients with PTCL.

This leads us to other lessons learned: chemotherapy may cure more patients with T-cell lymphoma than we often estimate. Retrospective series in PTCL show low rates of long-term survival with combination frontline chemotherapy. In the frequently referenced International T-cell project, failure-free survival rates at 5 years are as low as 20%, 18%, and 36% for PTCL-NOS, AITL, and ALK-negative ALCL, respectively.² This has led some to conclude that chemotherapy is ineffective in PTCL, and chemotherapy-free approaches should be piloted in untreated patients.^22,25 However, in recent large prospective trials, CHOP chemotherapy cures a higher proportion of patients^3,12 In the Ro-CHOP study, the median survival for those treated with CHOP alone was 50.4 months. Similarly, in the CHOP arm of the ECHELON-2 study, the 5-year PFS and OS were 43% and 61%, respectively, showing significant numbers of patients cured with time-limited chemotherapy and it remains a platform to thoughtfully build on.

A second lesson from this trial is the importance of not just understanding the heterogeneity within these rare diseases but also embracing this knowledge in trial design. This obstacle is also a path to move forward in more meaningful ways. As above, HDAC inhibitors and BV show subtype-specific efficacy.^{17,20,21,23,24,26} As we gain experience with other newer targeted agents in relapsed/refractory PTCL, we continue to observe biologically driven and histology-specific activities (Tables 1 and 2). However, even in the standard CHOP arm of recent randomized studies, outcomes vary by histology: ALCL is cured more frequently than AITL, which is more favorable than PTCL-NOS (Table 3). We know from experience that certain rare subtypes of PTCL such as adult T-cell leukemia/lymphoma and hepatosplenic T-cell lymphoma are rarely, if ever, cured with chemotherapy alone and alternate approaches are used.³¹ Within more common PTCL subtypes, we are now identifying factors predicting outcomes with chemotherapy. Alterations in p53, CDKN2A, DNMT3A, rearrangements of TP63, and expression patterns consistent with the GATA-3 subtype appear to identify groups less likely to benefit from chemotherapy, whereas DUSP22 rearrangements and the TBX21 subtype may be more chemosensitive.^6,32 None of these observations yet define therapy, but when studying novel chemotherapy-based approaches, we need to recognize and account for these predictive factors in study design and interpretation.

TABLE 1.

Single Agents

Agent	No.	ORR	ORR by Subtype
Romidepsin16,17	130	25%	PTCL-NOS—29% AITL—33% ALK-ALCL—24%
Belinostat19	120	26%	PTCL-NOS—23% AITL—45% ALK-ALCL—15%
Brentuximab vedotin26,33	92	69%	PTCL-NOS—33% AITL—54% sALCL—86%
Cerdulatinib36	58	36%	PTCL-NOS—0% AITL/TFH—52% Other—32%
Duvelisib27	101	48.5%	PTCL-NOS—48.1% AITL—66.7% ALCL—13.3%
Golidocitinib28	88	44%	PTCL-NOS –46% AITL/TFH—56.3% ALCL—10%
Ruxolitinib29	25	25%	By histology  PTCL-NOS—18%  AITL/TFH—33%  ALCL—25% JAK/STAT status  Activating mutations—20%  Functional evidence with pSTAT3—44%  None—0%
Valemetostat30	119	43.7%	PTCL-NOS—31.7% PTCL TFH—50% AITL—54.8% ALCL—33.3%

Abbreviations: AITL, angioimmunoblastic-type lymphoma; ALCL, anaplastic large cell lymphoma; ORR, objective response rate; PTCL-NOS, peripheral T-cell lymphoma, not otherwise specified; sALCL, systemic anaplastic large cell lymphoma; TFH, T-follicular helper.

TABLE 2.

Combinations

Agent	No.	ORR	ORR by Subtype
Romidepsin + azacitidine20	25	61%	TFH—80% Non-TFH—25%
Romidepsin + duvelisib34	55	58%	PTCL-NOS—50% AITL/TFH—70% ALCL—100%
Romidepsin + gemcitabine/oxaliplatin21	17	58%	PTCL-NOS—40% AITL—100%
Romidepsin + lenalidomide22,23	2	65%	PTCL-NOS—50% AITL—79%

Abbreviations: AITL, angioimmunoblastic-type lymphoma; ALCL, anaplastic large cell lymphoma; ORR, objective response rate; PTCL-NOS, peripheral T-cell lymphoma, not otherwise specified; TFH, T-follicular helper.

TABLE 3.

Front-Line Studies

Regimen	No.	PFS	PFS by Subtype
Ro-CHOP7,12	421
Ro-CHOP	211	Median PFS—12.0 months	Median PFS TFH—19.5 months Non-TFH—8.7 months Median OS TFH—65 months PTCL-NOS—25.8 months
CHOP	210	Median PFS—10.2 months	Median PFS TFH—10.6 months Non-TFH—9 months
ECHELON-23	452
BV CHP	226	Median PFS—62.3 months 5-year PFS—51%	Median PFS PTCL-NOS—32.3 months AITL—21.7 months sALCL—NR 5-year PFS PTCL-NOS—26.5% AITL—26.6% sALCL—60%
CHOP	226	Median PFS—23.8 months 5-year PFS—43%	Median PFS PTCL-NOS—10.7 months AITL—47.6 months sALCL—54.2 months 5-year PFS PTCL-NOS—25.7% AITL—48.1% sALCL—48.4%
Nordic CHOEP-ASCT35	160	5-year PFS 44%	5-year PFS PTCL-NOS—38% AITL—49% ALCL—61%

Abbreviations: AITL, angioimmunoblastic-type lymphoma; ALCL, anaplastic large cell lymphoma; BV, brentuximab vedotin; CHP, cyclophosphamide, doxorubicin, and prednisone; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; PFS, progression-free survival; PTCL, peripheral T-cell lymphoma; PTCL-NOS, peripheral T-cell lymphoma, not otherwise specified; Ro-CHOP, romidespin, cyclophosphamide, doxorubicin, vincristine, and prednisone; sALCL, systemic anaplastic large cell lymphoma; TFH, T-follicular helper.

A third lesson from this study is that there are rarely second chances in rare diseases. With all the benefits of hindsight, it would have been feasible to power and prospectively assess the TFHL subset in the Ro-CHOP study. Even without preplanning, patients with AITL accounted for almost half the accrual. Had that been done, the study might have been positive, and we could have a new standard up-front regimen for TFHL with most patients alive at 5 years. As conducted, the data do not support adding romidepsin to CHOP and the negative results led to the voluntary withdrawal of romidepsin in PTCL, which has affected its availability for relapsed PTCL and hampered future clinical trials. We do not currently incorporate romidepsin in frontline therapy, but studies to explore other HDAC inhibitors and other newer agents with combination chemotherapy are underway (ClinicalTrials.gov identifiers: NCT06072131, NCT04803201).

The cost and logistics of pursuing studies in rare diseases make such subtype-specific or subtype-enriched randomized studies challenging. Yet, studies that are underpowered using overly optimistic goals for the experimental arm or overly pessimistic predictions for the control arm are unlikely to move the field forward. Similarly, studies that do not refine the patient population or preplan for subgroup analyses are less likely to move us forward. In early-phase single-arm studies, we accept what we do not know and build in correlatives that allow us to generate hypotheses to inform future trials.^29,34 Applying this is more challenging when advancing to randomized studies because of their size, cost, and need to generate statistically conclusive results. One approach being used in the current US intergroup study (A051902) randomly assigns patients to standard chemotherapy (CHOP or CHOEP) versus duvelisib plus standard chemotherapy versus azacitidine plus standard chemotherapy with TFHL as a stratification factor in the random assignment, resulting in those populations being equally distributed among the arms (ClinicalTrials.gov identifier: NCT04803201). Future randomized studies should enrich for the subtype of most biologic and clinical interest, prospectively powering specific cohorts whenever possible.

Learning from failure is a way forward. It is very unlikely that a one-size-fits-all approach will transform therapy in PTCL. With multiple promising agents being developed in T-cell lymphoma, it is incumbent on us to explore subtype-specific activity to enrich for those who will benefit and minimize those who will not in the subsequent studies of these rare and difficult-to-treat diseases.^27,28,30 Simply understanding the heterogeneity will not be enough. We must concurrently apply that knowledge, as imperfect as it may be, to future study design. By capitalizing on the increasingly understood vulnerabilities of specific subsets of PTCL, we can work smarter. By focusing the patient populations for confirmatory studies, we will have to work harder requiring collaborations across investigators, industry, regulatory bodies, and patients. We do not see another way.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Collection and assembly of data: All authors

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Failing Forward in Peripheral T-Cell Lymphoma

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).