Recommendations for managing PD-1 blockade in the context of allogeneic HCT in Hodgkin lymphoma: taming a necessary evil

PD-1 blockade is associated with high response rates and durable benefit in patients with relapsed/refractory (R/R) classical Hodgkin lymphoma (cHL) who have relapsed after or are ineligible for autologous hematopoietic cell transplantation (auto-HCT).^1-5  Despite this, a majority of patients treated with PD-1-blocking antibodies (nivolumab and pembrolizumab) will eventually progress on these therapies. For patients with highly chemorefractory disease or those who relapse after auto-HCT, allogeneic HCT (allo-HCT) offers a potentially curative treatment option.^6-8  Therefore, allo-HCT is a consideration for selected patients with R/R cHL who achieve a remission (partial remission [PR] or complete remission [CR]) with PD-1 blockade, or for those who progress on those agents and subsequently are able to achieve remission. However, prior immune modulation with PD-1 blockade could significantly alter allo-HCT outcomes, with the theoretical potential for both increased immune toxicity and increased graft-versus-tumor (GVT) efficacy, raising many questions about the appropriate use of allo-HCT in those patients.⁹  Moreover, relapse remains a problem after allo-HCT, with reported progression-free survival (PFS) ranging from 20% to 60% at 3 years.^10,11  For post–allo-HCT relapses, treatment options are very limited,^12-15  and the use of PD-1 blockade in this setting is attractive given the activity of the drugs in the pre–allo-HCT R/R setting and the theoretical enhancement of the graft-versus-lymphoma effect. Here again, there is significant concern about the risk of attendant immune toxicity.⁹  Given the complex and dynamic nature of these questions, specifically the use of allo-HCT after PD-1 blockade and that of PD-1 blockade for post–allo-HCT relapse, the authors of this review convened to form a working group of transplanters and nontransplanters with published experience in the treatment of patients receiving both PD-1 blockade and allogenic hematopoietic stem cell transplantation. As there is currently a dearth of prospective data in this field, the aim of this working group was to provide a set of consensus recommendations to inform clinical practice until more definitive data are obtained.

The PD-1 pathway serves as a checkpoint to limit T-cell-mediated immune responses.¹⁶  Both PD-1 ligands, PD-L1 and PD-L2, engage the PD-1 receptor, induce PD-1 signaling and associated T-cell tolerance.¹⁶  Weak cytotoxic T lymphocyte (CTL) activity in lymph nodes promotes local tumor escape.¹⁷  PD-1 blockade reverses this inhibition of T-cell activation and proliferation and restores CTL sensitivity to antigen and killing in lymph nodes. In the pre–allo-HCT setting, residual anti-PD-1 antibodies as well as the immunologic effects may linger, whereas in the post–allo-HCT setting, donor CTLs may be activated by PD-1 blockade. In both cases, PD-1 blockade can enhance T-cell responses against cHL and/or bystander cells. Moreover, PD-1 ligands may play a role in maintenance of tolerance following allo-HCT, and blocking this pathway may induce or exacerbate graft-versus-host disease (GVHD).^18,19  When blocking the PD-1 pathway in a murine model of acute GVHD (aGVHD), there was increased GVHD-related mortality.^20,21  In a sublethally irradiated minor antigen mismatched model of sclerodermatous chronic GVHD (cGVHD), blockade of the PD-1 pathway worsened GVHD.²²  With respect to GVT effect, preclinical studies have shown that host PD-1 ligand (PD-L1) expansion can inhibit GVT activity because of ligation of PD-1 on donor T cells. Moreover, blocking PD-1/PD-L1 expression permitted GVT activity without severe GVHD.^17,23,24  The use of PD-1 blockade before or after allo-HCT may therefore be associated with increased toxicity and/or increased efficacy, and the balance of those 2 effects is likely to significantly affect patient outcomes.

A 28-year-old woman with primary refractory cHL after doxorubicin, bleomycin, vinblastine, and dacarbazine frontline chemotherapy has a complete response with ifosfamide, carboplatin, and etoposide and undergoes auto-HCT. She receives brentuximab vedotin (BV) consolidation, which is stopped after 6 months because of grade 3 peripheral neuropathy. Three months later, her lymphoma relapses with nodal and bony disease. She is treated with single-agent nivolumab and achieves PR after 4 doses of therapy. Repeat imaging after 4 additional doses of nivolumab shows no significant change with continued PR. How should this patient be managed?

Patients with cHL who are ineligible for auto-HCT or relapse after auto-HCT have historically had a poor prognosis with a median survival of 2 years or less.^25-28  Brentuximab achieves high response rates (75%) in this setting, and a subset of patients who achieve a CR with BV (∼33%) will experience lengthy remissions (estimated 5-year PFS 52%).²⁹  Even so, a large majority of patients treated with BV will relapse; moreover, given the increasing use of BV in earlier lines of therapy,³⁰  many patients with post–auto-HCT relapse will have already received BV. The anti-PD-1 monoclonal antibodies nivolumab and pembrolizumab are both approved in this setting and are associated with overall response rates (ORRs) of ∼70%.^1-5  Here too a subset of patients can achieve durable remissions.^1-4,31,32  However, based on the phase 1/2 data available to date, it appears that a majority of patients will relapse within 1 to 2 years of starting PD-1 blockade, and it is not clear yet what proportion of patients, if any, can attain long-term remission. Thus, for patients responding to anti-PD-1 mAb’s, allo-HCT represents the only modality with known curative potential; indeed, allo-HCT with reduced intensity conditioning (RIC) is associated with long-term disease-free survival in 25% to 40% of patients.^8,33,34  This raises the challenging question of how to optimally manage patients who achieve a response to PD-1 blockade. One option is to proceed to allo-HCT to capitalize on the response. Alternatively, patients may remain on PD-1 therapy, in the hope of maintaining a durable remission without incurring the risks of allo-HCT. This approach would also allow patients to undergo other novel or experimental therapies following PD-1 blockade, which could themselves possibly serve as a bridge to allo-HCT. Remission quality with PD-1 blockade is not as clear a predictor of outcome as it is with BV, and patients with PRs can still have durable remissions. Further complicating matters, initial reports suggest that patients undergoing allo-HCT following PD-1 blockade may be at increased risk for severe immune-related complications early after transplantation. The largest series of cHL patients who received anti-PD-1 mAb’s prior to allo-HCT included 39 patients and reported higher than expected rates of grade 4 aGVHD and veno-occlusive disease (VOD), as well as a noninfectious febrile syndrome requiring prolonged steroid treatment. Compared with matched controls, patients treated with anti-PD-1 mAb’s had a prolonged and severe reduction in circulating PD-1 positive T cells and a relative reduction in T regulatory cells after transplantation, which could underlie the observed early immune complications.³⁵  Despite this, the relapse rate appears low (16% at 1 year), lower than historical series of cHL patients who underwent allo-HCT without prior anti-PD-1 therapy, and PFS was encouraging in this cohort (74% at 1 year). Subsequent accumulated experience seemed to support these results.^2,36-38  However, larger studies are warranted to determine whether the efficacy of allo-HCT is enhanced following anti-PD-1 therapy.

The following discussion only applies to patients who would otherwise be good candidates for allo-HCT based on age, comorbidities, and donor availability (recognizing that haploidentical donors can be successfully used in older patients with cHL³⁹  and that therefore most patients will have an available donor). In the absence of a clearly documented large difference in outcome between patients in PR and CR after PD-1 blockade, we do not recommend using depth of remission as the sole criterion for determining which patient should go to allo-HCT after PD-1 blockade. In general, given the favorable safety profile of these agents and the toxicity of allo-HCT in this setting, we favor keeping responders on therapy instead of stopping this therapy to proceed to allo-HCT. In our experience, it is frequently possible to reinduce remission after PD-1 blockade failure. Supporting this approach, a recent retrospective analysis showed that initial salvage chemotherapy following PD-1 blockade resulted in an overall response in 10/19 (53%) of patients.⁴⁰  Nevertheless, we strongly recommend a referral to transplant center to discuss considerations around allo-HCT, and to at least prepare for a potential transplant at time of PD-1 blockade failure. We would consider early allo-HCT for patients in remission after PD-1 blockade with heavily pretreated refractory disease and no viable post–PD-1 salvage options. If an allo-HCT is planned, we typically hold PD-1 therapy for 6 weeks before allo-HCT, although there are no data to suggest a minimal or optimal delay. However, PD-1 induced changes (ie, alterations in T-cell subsets) are long lasting, persisting 6 months or more after completion of PD-1 therapy,³⁵  so a delay of a few weeks may be immunologically inconsequential.

At present, there is no consensus regarding the optimal transplant strategy for patients previously treated with immune checkpoint blockade. Given the biological mechanism of immune checkpoint blockade and observed early posttransplant immune toxicity, we favor using a transplant strategy that minimizes GVHD and VOD risk. Given the absence of data suggesting a benefit to myeloablative conditioning, the frequent heavy pretreatment (including auto-HCT) in these patients, and the increased risk of VOD and GVHD with ablative conditioning, we favor RIC irrespective of depth of response. Graft source, HLA match, and GVHD prophylaxis are the additional key considerations for reducing GVHD risk. In the largest published series of allo-HCT after PD-1 blockade, aGVHD risk was significantly lower among 11 patients receiving a bone marrow graft compared with 28 patients receiving peripheral blood graft (0% vs 32%, P = .04).³⁵  Based on this preliminary data, we prefer to use bone marrow grafts in this setting. GVHD prophylaxis with posttransplant cyclophosphamide (PtCy) has been a very effective strategy to reduce GVHD in other transplant settings associated with an elevated risk of GVHD.^41-43  Merryman et al found no association between GVHD prophylaxis regimen and GVHD rates in their study of 39 patients including the 14 patients who received PtCy. However, a recent retrospective analysis of 11 patients who underwent allogeneic allogenic hematopoietic stem cell transplantation with PtCy GVHD prophylaxis after treatment with a checkpoint inhibitor demonstrated low rates of both aGVHD and cGVHD.⁴⁴  Based on these results and the favorable anecdotal experience using this strategy at our centers, we consider transplantation using a bone marrow graft followed by a PtCy-based GVHD prophylactic regimen (eg, PtCy/tacrolimus/mycophenolate mofetil) for patients with previous exposure to immune checkpoint agents. We also use ursodiol for VOD prophylaxis in all patients.⁴⁵  We encourage clinicians to have heightened awareness of early transplant complications (early severe GVHD, VOD, noninfectious febrile syndrome), which are likely more common in this population and may require early immunosuppressive treatments. If GVHD occurs, we recommend treating as outlined in the management of treatment emergent GVHD (teGVHD) section. In our experience, early intervention may be critical and second-line treatments of GVHD could be considered earlier than in a context without PD-1 blockade.

A 27-year-old female with relapsed cHL post–auto-HCT achieved a partial response to brentuximab and underwent a matched sibling allo-HCT, receiving fludarabine and melphalan RIC and posttransplant methotrexate with tacrolimus for GVHD prophylaxis. She tapered off immunosuppression at day 180 without evidence of GVHD but relapsed at day 220. Is treatment with single-agent PD-1 blockade a good treatment option? After weighing the risks and benefits, the physician decides to treat this relapse with nivolumab 3 mg/kg. Within 14 days of her first dose, she develops stage IV hepatic aGVHD (peak bilirubin 20.4 mg/dL). How should this GVHD be managed?

Many case reports and case series have been published regarding PD-1 blockade therapy in cHL patients with relapsed disease after allo-HCT.^37,38,46-54  All 12 patients reported in these publications experienced at least a partial response, and 4 developed GVHD. In our 2 retrospective studies,^55,56  we observed an ORR of 85% (N = 41 of 48 assessable patients) to anti-PD-1 mAb’s in a heavily pretreated cHL cohort relapsing after allo-HCT, including 48% complete responses (N = 23). In the phase 2 studies using nivolumab and pembrolizumab for R/R cHL, ORR was 69% and 73% (CR rate of 29% and 28%), respectively.^1,4  Although lacking the rigor of a clinical trial, we observed an unexpectedly high response rate in our post–allo-HCT cohort at least comparable to that observed in previous anti-PD-1 cHL trials, despite the advanced stage of disease and short duration of therapy in many patients. Median PFSs in our retrospective studies were 20 months in the US cohort and not reached (at a median follow up 12 months) in the French cohort, which also compares favorably with the experience in the phase 2 studies on nonallografted patients.

teGVHD post–PD-1 blockade has been reported in a variety of clinical scenarios including various graft sources, conditioning intensity, and prophylactic GVHD strategies.^46-50,55,56  Several trends are emerging from the current body of published cases that may be associated with risk of teGVHD. Table 1 highlights the baseline patient characteristics and frequency of teGVHD from the 2 largest retrospective series of patients from the United States and France receiving anti-PD-1 post–allo-HCT. Of the 23 patients who developed GVHD after anti-PD-1 in the 2 largest series,^55,56  16 were classified as developing aGVHD (or overlap), occurring after 1 to 2 doses of anti-PD-1, and in 11 it was severe (grade 3 or 4). Eleven of 16 experienced hepatic GVHD. Severe hepatic aGVHD is a rare occurrence in allo-HCT, yet we observed 8 patients with stage ≥III aGVHD (ie, bilirubin >6). Moreover, grade 3 to 5 hepatic (>3× upper limit of normal) toxicity from anti-PD-1 mAb’s was observed in <5% of patients in phase 1 and 2 trials.^1-4,57  In a separate retrospective series, Schoch et al reported no evidence of teGVHD in 7 patients treated with anti-PD-1 mAb’s after allo-HCT.⁴⁴  Donor source varied but all 7 patients received PtCy. In the US retrospective series, there were 5 patients who received PtCy (4 haploidentical and 1 MUD); 1 patient (MUD) developed severe teGVHD 14 days after anti-PD-1.⁵⁶  In the French series, 13 of 20 patients received anti-thymocyte globulin (ATG), and 3 of 13 developed grade 2 to 3 teGVHD.⁵⁵  Thus, we hypothesize that there may be less risk of teGVHD in patients receiving T-cell modulating or depleting strategies such as PtCy or ATG. PtCy is proposed to selectively destroy primarily alloreactive activated T cells, spare non–cross-reactive memory and naïve T cells, and promote T regulatory cells and long-term tolerance by intrathymic clonal deletion of anti–host T cells.^58-61  Thus, PtCy might be a safer platform after which to administer anti-PD-1 for relapsed disease. Currently, we do not routinely alter our preparative conditioning on account of possible relapse and anti-PD-1 administration. However, these data are intriguing and should be considered especially when designing clinical trials evaluating checkpoint blockade after allo-HCT; results from Blood and Marrow Transplant Clinical Trials Network 1203 may help in decision making (#NCT02208037). In both series, teGVHD occurred more frequently in patients with a prior history of GVHD. As illustrated in the case, aGVHD can occur in patients off immunosuppression and without a history of GVHD. The US study reported that 5 of the 12 patients without a history of GVHD developed teGVHD post–anti-PD-1, which is contrary to the French series where only patients with a history of GVHD developed teGVHD post–anti-PD-1. Additionally, in the French series, teGVHD occurred more frequently in patients who were administered anti-PD-1 earlier post–allo-HCT; there was no statistically significant difference in the US series, although the median time from allo-HCT to anti-PD-1 was shorter in the 17 patients developing teGVHD. In some murine models, checkpoint blockade administered before and within days after transplant was shown to accelerate GVHD lethality.^20,21,62  Thus, it remains possible that administration during the earlier phase of transplantation would trigger severe GVHD and lethality after conditioning-related inflammation of host tissues similar to that observed in murine models. So, although the definition of “earlier” remains ill-defined, we would encourage restraint in administering anti-PD-1 mAb’s within the first 180 days post–allo-HCT outside the context of a clinical trial. As shown in a recent study, PD-1-expressing T-cells were increased early after transplantation and this was associated with worse survival.⁶³  We speculate that anti-PD-1 mAb’s upregulate major histocompatibility antigens via activation of the interferon-γ pathway,^2,21,64  decrease T regulatory cells,⁶⁵  and increase activated T and NK cells, thereby decreasing peripheral tolerance and increasing the incidence of GVHD. Well-designed correlative studies assessing T-cell subsets, PD-1/PD-L1 expression, and GVHD predictive markers before and after anti-PD-1 administration will be key to better understanding risks and identifying patients suitable to receive anti-PD-1 mAb’s post–allo-HCT.

Others have suggested that initial low doses of anti-PD-1 postallograft or concomitant immunosuppression may be a safer approach.^47,49  In a recently reported prospective study,⁶⁶  8 patients were treated with nivolumab after allograft (6 patients at 1 mg/kg and 2 patients at 0.5 mg/kg). In the 1 mg/kg cohort, there were 2 deaths (1 patient with sepsis and fatal acute respiratory distress syndrome, and 1 patient who developed fatal thrombotic cerebral vascular accidents) as well as a number of additional adverse events including cGVHD, grade 3 pneumonitis, and transaminitis (N = 1 each). Because of the toxicity observed at 1 mg/kg, a 0.5-mg/kg cohort recently opened, and no significant toxicities have been observed in the 2 patients accrued to date.⁶⁶  Consequently, if anti-PD-1 mAb’s are given post–allo-HCT outside the context of a clinical trial, we would strongly consider starting at a low dose and consider escalating the dose if there is no apparent efficacy and no significant toxicity.

Table 2 highlights systemic treatments used for teGVHD in our 2 series. In the US study, 3 of 17 patients had complete responses to GVHD directed treatment. One patient received topical treatments only for mild skin cGVHD. One patient with stage I hepatic aGVHD had a complete response to high-dose steroids alone. The last patient, similar to the case presentation, failed to respond to high-dose steroids and mycophenolate mofetil (MMF), but had a complete response to ATG. One additional patient in the US study received ATG and had a partial response to grade 3 aGVHD (stage III hepatic, stage II skin) and was alive at last follow-up. These anecdotal cases support the idea that significantly T-cell depleting or modulating strategies may be necessary to temper teGVHD. In the French cohort, 2 patients with teGVHD responded well to standard first-line treatment of aGVHD, 2 mg/kg of methylprednisolone alone (N = 2). Two patients did not respond to high-dose steroids but were controlled with cyclosporine and basiliximab/ECP, respectively. Thus, although the ideal treatment of GVHD post–anti-PD-1 remains far from clear, early intervention with significantly T-cell modulating and/or depleting drugs may be useful to treat severe teGVHD. We recommend consideration of ATG, or alternatively calcineurin inhibitor + ECP, as second-line strategy if not responsive to high-dose steroids.

Today, the majority of patients with post–allo-HCT cHL relapse will have received PD-1 blockade before allo-HCT. Given the expected response rates and the scarcity of other options in this setting, retreatment with PD-1 blockade could be chosen, especially if a patient has no history of GVHD, relapse >180 days after transplant, and no reasonable alternative options. Nevertheless, in the 2 published retrospective series, teGVHD was common and often refractory to GVHD directed treatment. Currently, dose exploration of anti-PD-1 mAb’s is being explored in clinical trials for relapse after allo-HCT. Although these and other strategies to reduce the risk of teGVHD are being tested, we would strongly consider starting anti-PD-1 mAb’s at a low dose (eg, 0.5 mg/kg nivolumab) when given outside the context of a clinical trial. We recommend close monitoring for evidence of GVHD. In the event of teGVHD, we recommend immediate cessation of therapy and rapid initiation of high-dose corticosteroids (2 mg/kg per day). Further immunosuppression should be given promptly if the patient does not respond to steroids, with a preference for calcineurin inhibitor + ECP or ATG.

With the growing role of PD-1 blockade in cHL, most patients who consider allo-HCT will have already received an anti-PD-1 mAb. Careful considerations of the risk and benefit of transplantation is important, and we propose recommendations for managing this decision (Table 3) and for choosing a transplant method that best accounts for downstream effects of prior PD-1 blockade. In addition, given the efficacy of PD-1 blockade in post–allo-HCT relapse and the lack of good alternative options, anti-PD-1 therapy is likely to be a consideration for most if not all patients in this setting. Here again, we provide recommendations (Table 4) for the timing and dosing of those agents, as well as for the management of GVHD-related toxicity. We acknowledge the substantial gaps in our knowledge and look to future carefully designed retrospective and prospective studies (Table 5) to shed further light on the optimal handling of the PD-1/allo-HCT interface. We strongly advocate that patients be treated on a clinical trial whenever possible and encourage continued efforts to collect off-trial data through international collaborations and registries.

The authors recognize the courage of patients who risked their lives on decisions made under heavy uncertainty, some with excellent and some with devastating outcomes, all of whom provided the experience on which we now all rely.

Contribution: C.H., R.M., S.D., P.A., R.H., F.M., and B.H. wrote the manuscript.

Conflict-of-interest disclosure: C.H., R.H., and F.M. received consulting fees from Bristol-Myers Squibb. P.A. received consulting fees from Merck and Bristol-Myers Squibb. The remaining authors declare no competing financial interests.

Correspondence: Charles Herbaux, Service des Maladies du Sang, Centre Hospitalier Régional Universitaire de Lille, Lille, France; e-mail: charles.herbaux@chru-lille.fr; and Brad Haverkos, Division of Hematology, University of Colorado, 1665 Aurora Ct., Mail Stop F754, Aurora, CO 80045; e-mail: bradley.haverkos@ucdenver.edu.