Progress and pitfalls with the use of image-guided personalised approaches in lymphoma

Abstract

The use of ¹⁸F-FDG PET CT has become an essential part of the management of patients with lymphoma. The last decade has seen unrivalled progress in research efforts to personalise treatment approaches using PET as a predictive imaging biomarker. Critical to this success has been the standardisation of PET methods and reporting, including the 5-point Deauville scale, which has enabled the delivery of robust clinical trial data to develop response-adapted treatment approaches.(1, 2) The utility of PET as a predictive imaging biomarker in assessing treatment success or failure has been investigated extensively in malignant lymphomas. Considerable progress has been made over the last decade, in using PET to direct more personalised “risk-adapted” approaches, as well as an increased understanding of some of the limitations. Arguably the greatest success has been in Hodgkin Lymphoma (HL) where PET was initially demonstrated to be a powerful predictive biomarker (3) and is now routinely used in both early-stage and advanced HL to reduce or escalate the use of chemotherapy as well as guiding the delivery of more selective radiotherapy to patients.

The successful treatment of early-stage HL has been one of the greatest successes in haemato-oncology with freedom from treatment failure of over 90% at 5 years using combined modality treatment (CMT) with brief chemotherapy (two cycles of ABVD) and Involved Field Radiation Therapy (IFRT) 20 Gy in 10 treatments.⁴ Given that it is difficult to further improve these excellent tumour control rates, increasingly the focus in clinical trials has been instead been on improving the quality of long-term survival by aiming to decrease treatment‐induced mortality and morbidity whilst maintaining long-term disease control.

The overarching aim of all of the large international trials (UK NCRI RAPID trial, EORTC H10, US intergroup 5,0604 Phase II trial and GHSG HD16 trials) has been to investigate the utility of PET CT to omit radiotherapy in those that achieve a complete metabolic remission (CMR) after initial chemotherapy^5–8 and these are well reviewed by Phiilips et al in this issue. For those that achieve a CMR, chemotherapy alone confers excellent overall survival rates and is a valid option for selected patients. This is particularly the case where the RT induced late toxicity of cardic damage and risk of a second cancers, for example, breast cancer in younger females and lung cancers potentially outweights the benefit of improved local control.

In advanced HL, the two most common established treatment approaches use either doxorubicin, Bleomycin, vinblastine, and dacarbazine (ABVD) or the more intensive escalated bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine and prednisolone (BEACOPPesc). Both approaches have been extensively investigated using PET CT to risk adapt management and are reviewed by Phillips et al. The NCRI RATHL trial risk adapting treatment using interim PET/CT after of 2 cycles ABVD (PET2) and deescalating therapy for patients achieving complete metabolic response (CMR) based on a DS of 1–3 omiting bleomycin in subsequent four cycles has been practice changing ref 10, maintaining good tumour control and decreasing Bleomycin lung injury. This trial also confirmed that PET could replace the bone marrow biopsy during staging also changing clinical practice ref 10.

The results from those patients that were PET positive (DS 4 and 5) after two cycles of ABVD and who were subsequently escalated to BEACOPP still leave plenty of room for clinical improvement and the question of whether these patients with higher risk disease should be treated initially with a more intensive approach such as BEACOPP remains an open one. The German Hodgkin study group have long advocated an initial treatment approach using BEACOPPesc. In the HD15 trial, a PET-directed approach was used to determine whether patients required consolidation radiotherapy.¹³ An option of starting therapy with BEACOPP_esc and deescalating to ABVD for patients with a PET-determined CMR is being explored by the French–Belgian Lymphoma Study Association.¹¹ PET after two cycles of induction BEACOPP_esc chemotherapy safely guided treatment in patients with advanced Hodgkin lymphoma and allowed the use of ABVD in early responders without impairing disease control and reduced toxicities with excellent disease control of over 90%. (ref 12 please change order of references in this paragraph)

The use of PET CT in non-Hodgkin lymphoma (NHL) has recently been expertly reviewed by Barrington and Trotman.¹⁴ The initial studies confirmed that PET predicts response in diffuse large B-cell lymphoma (DLBCL), but unfortunately more intensive chemotherapy approaches have failed to improve outcomes for patients with interim PET-positive scans.¹⁵ Therefore, interim PET has not established clinical utility in the management of DLBCL. Whether consolidation RT can be omitted in patients with bulky DLBCL who achieve a CMR after immunochemotherapy remains an important unanswered research question that must be addressed in appropriately designed randomised prospective trials such as the Phase 3 IELSG 37 study primary mediastinal large B-cell lymphoma where patients achieving CMR after immunochemotherapy are randomised to receive mediastinal radiotherapy. This important trial has completed recruitment and the first reports awaited.

The initial retrospective of analysis of PET-CT responses in a subset of the PRIMA trial suggested achieving CMR after induction immunochemotherapy is highly predictive of patient outcome in follicular lymphoma,¹⁶ an observation confirmed in the GALLIUM trial and a subsequent pooled analysis.^17,18 However whether PET CT can lead to personalised risk adapted approaches in Follicular lymphoma requires further investigation with randomised data. The initial presentation of the FIL FOLL12 study concluded that for patients with intermediate‐high risk Follicular lymphoma PET/CT cannot be used to stratify management, as despite the attainment of a post‐induction CMR, omission of rituximab maintenance resulted in a significantly lower 3‐year PFS, 68% vs 84%.¹⁹ Further studies are ongoing including the UK NCRI PETReA study which is a Phase 3 evaluation of PET guided, response adapted therapy in patient with previously untreated, advanced stage, high tumour burden FL. In this study those who achieve a CMR (DS 1‐3) are randomised 1:1 to Rituximab maintenance versus no further treatment, whereas those who remain PET +ve (DS 4‐5) are randomised 1:1 to R maintenance with or without lenalidomide.

Truly personalised therapy needs to integrate individual patient characteristics such as age, sex, site and size of initial disease into clinical decision making.¹⁰ To do this will require more accurate modelling of risks of disease recurrence in omitting RT versus late risks associated with the radiation field for individual patients. Simulation models are being developed that may be able to provide individualised risk quantification in future, but require further validation. Patient preferences and goals are a key consideration. Patients’ priorities tend to vary with age, gender, parity and co-morbidities, but are highly individual and cannot be dictated by algorithms. One of the main challenges with current HL treatment options is finding a way to effectively communicate the benefits and risks of different treatment strategies to empower patients to make informed decisions. So, whilst early-stage HL and DLBCL trials have substantially informed both clinician and patient in decision-making regarding the omission of radiotherapy, there is still much work to be done to realise the ultimate goal of a truly personalised approach. In this regards, the concept of shared decision-making is emerging and may play an increasingly important part of personalised care.²⁰

Despite notable successes in HL, trials of PET-directed therapy in HL have not been able to resolve longstanding debates about efficacy versus toxicity, particularly with respect to consolidative radiotherapy in ES-HL and intensity of frontline treatment in unfavourable-risk/advanced HL. The latter highlights one of the main limitations of response-adapted treatment in that it cannot inform upfront treatment. Overall, these studies reveal a clear need to improve risk stratification in order to better identify individuals that can safely receive low-intensity therapy and facilitate early detection of treatment resistance.

One dilemma in everyday clinical management is the ongoing utility of pre-treatment staging and clinical risk stratification in the modern era of response-adapted approaches. For decades, clinical parameters such as nodal sites, erythrocyte sedimentation rate (ESR), and the presence of mediastinal bulk or B-symptoms to differentiate early-stage HL into favourable- and unfavourable-risk groups with different treatment approaches. Are these still valid in the modern era of PET and can they still be incorporated in clinical management to facilitate personalised approaches? Data from the UK NCRI “RAPID” trial suggest that the PET response after three cycles of ABVD was more important than preclinical risk factors; however, patients with mediastinal bulk and B-symptoms were excluded from this trial.⁹ Similarly, the International Prognostic Score in advanced HL may lose prognostic value when combined with early PET assessment.

Looking forward, there are multiple ways that risk stratification may be improved. Firstly, there is potential to harness developments in PET technology and digital image analysis to improve the accuracy of response assessment, beyond the semi-quantitative Deauville scale. Novel quantitative PET assessment and artificial intelligence techniques are currently under evaluation. Secondly, additional imaging biomarkers can be employed to enhance risk stratification, including baseline PET metrics that reflect tumour burden, tumour heterogeneity and spatial dissemination. Metabolic tumour volume is the most widely studied of these and can improve prognostication when combined with interim PET response in both HL and NHL. Thirdly, circulating and tissue biomarkers can be combined with PET response assessment to refine risk prediction and/or detect minimal residual disease. Several biomarkers show promise, including circulating tumour DNA, molecular classifiers in DLBCL, immunoglobulin gene rearrangements in low-grade NHL and serum cytokines in HL. Importantly, many of these imaging and molecular biomarkers can inform baseline risk assessment and thus provide a platform for upfront risk-adapted treatment.

Despite recent advances, outcomes for high-risk NHL and very high-risk HL remain suboptimal (see Phillips et al review). Novel approaches are required to overcome treatment resistance. There has been notable recent success with immunotherapy for relapsed/refractory lymphomas, such as chimeric antigen receptor T-cell therapy, immune checkpoint inhibitors and bispecific antibody therapy. These agents are under evaluation in the frontline setting and have the potential to revolutionise the treatment over the coming decades. However, their ability to activate the immune system can confound PET response assessment and increase the risk of false-positive or indeterminate findings. One of the big challenges will be incorporating these targeted agents within current response-adapted treatment approaches.

In conclusion, there remain many challenges still to address in using PET CT as a reliable predictive imaging biomarker in personalising therapy in lymphoma. Some of these challenges are technical and include the robust, reproducible, application of the response criteria and their change over time, the timing of PET CT relative to therapy, and pitfalls associated with interpretation including “brown fat” and the increasing use of the immune check-point inhibitor anti-PD1 monoclonal antibody (mAb) especially in HL.^21,22 This observation emphasises the need for further evaluation in larger series and close collaboration between imaging and oncology specialists on a per-patient basis. Attempts to intensify chemotherapy in aggressive non-Hodgkin lymphomas have, however, proved ineffective to date. Trials are underway to determine whether PET can obviate consolidation radiotherapy in patients with diffuse large B-cell lymphoma and primary mediastinal B-cell lymphoma.

In moving forward to personalised approaches to treatment, PET-adapted trials provide important information regarding the potential increased risk of disease relapse with chemotherapy alone omitting radiotherapy. However, consideration of the risks of early and late toxicity are equally as important in determining the optimum treatment approach for individual patients. This is perhaps best highlighted in achieving the goal of personalised therapy in early-stage HL where there remain many unmet challenges in integrating individual patient characteristics (age, sex, site and size of initial disease) and modelling risks of omitting or delivering RT.^23,24 Looking to the future, evidence is emerging that baseline metabolic tumor burden may improve response prediction.²⁵ Combining PET with non-imaging biomarkers, such as tumour gene expression, circulating biomarkers such as serum thymus activation-related cytokine (TARC) and circulating tumour DNA, are all areas for further research.^26,27

In NHL, PET CT has become established as an important predictive imaging biomarker of response to therapy and predicting survival²⁸ but has as yet, failed to establish a role in risk adapted or personalising therapy decisions. In DLBCL, some of the limitations of PET have been observed in escalating systemic therapy in those PET positive. Further well-designed randomised clinical studies are needed to fully inform the personalised decision-making of omitting RT especially in bulky nodes and extra nodal disease. The lower sensitivity and specificity of PET CT for indolent lymphomas has meant that the number of large randomised studies using PET CT are currently limited. The relative paucity of randomised data underly the importance of completing ongoing studies such as the UK NCRI led international PETReA study, which is the first trial in FL to stratify patients for separate questions based on post‐induction PET status. Whilst much progress has been made in developing PET CT, this field is still in its infancy and integration with other prognostic and predictive biomarkers is likely to be required before personalised approaches can be routinely robustly applied to the management of the majority of common lymphomas.