Introduction
Treatment of chronic lymphocytic leukaemia (CLL) in previously untreated patients has improved remarkably over the past 10 to 15 years [
1]. Specifically, in fit patients, chemoimmunotherapy combining fludarabine, cyclophosphamide and rituximab remarkably increased overall survival [
2] and became a standard of care for induction. However, toxicities may limit this approach to younger and fitter patients. In addition, recently gained insights into the pathobiology of CLL suggest that targeted treatment of CLL may be feasible by either attacking specific intracellular pathways or using substances that modify the microenvironment more globally [
3]. Examples of the former (such as ibrutinib, idelalisib, or venetoclax) have now entered the clinical field in relapsed/refractory patients [
4,
5] and have also entered initial therapy. While ibrutinib is licenced for the treatment of previously untreated patients regardless of the physical fitness, the use of idelalisib/rituximab or venetoclax in front-line treatment is restricted to CLL patients with del17p that are not suitable for other therapies. Among the options for microenvironment modification, lenalidomide has attracted interest for a number of years (reviewed in Kater et al.) [
6].
Initial findings of efficacy of lenalidomide [
7,
8] were accompanied by reports of so-called tumour flare reactions and a number of observed and sometimes life-threatening tumour lysis reactions [
7,
9]. Since these observations were made at the 25-mg dose, previously established for multiple myeloma [
10,
11], this led to a reassessment of lenalidomide dosing in CLL and the development program for the substance mandated a dosing schedule following slowly increasing doses in trials started from 2007 on. In monotherapy, this schedule proved tolerable [
12], but showed much less efficacy than expected from trials presented with higher doses [
7,
8].
A combination of a chemoimmunotherapy backbone (fludarabine and rituximab) with lenalidomide may offer a treatment option that allows the use of three effective principles with different mechanisms of action. In addition, we hypothesised that an initial phase of chemoimmunotherapy for the debulkment of CLL might reduce or even eliminate lenalidomide-induced tumour lysis by reducing the tumour burden and possibly decrease the probability of tumour flares. At the same time, this triple combination may allow a slow increase in lenalidomide doses to a maximum tolerable dose without running the risk of missing lenalidomide efficacy due to slow response and early progress. A further question was the feasibility of a prolonged treatment period with lenalidomide and rituximab after remission induction with chemoimmunotherapy plus lenalidomide.
We designed the RevliRit trial to (a) determine individual maximum tolerated doses for lenalidomide in combination with fludarabine and rituximab; (b) explore toxicity and efficacy of the regimen, also using a panel of parallel laboratory investigation for exploratory analyses and (c) to determine the feasibility of a lenalidomide/rituximab consolidation/maintenance treatment.
Patients and methods
This study protocol was approved by the ethics committee of the state of Salzburg on July 30, 2008 (protocol number: 415-E/966) and by the ethics committee of each individual participating centre.
All patients had given their written informed consent. Patients with previously untreated CLL were enrolled onto this phase I/II trial. Patients had a diagnosis of CLL and active disease as defined by the 1996 NCI WG criteria [
13]. Further inclusion criteria consisted of age older than 18 years, ECOG performance status 0–2 and understanding of the need of effective contraception. Exclusion criteria were positivity for HIV or hepatitis B or C, other active infections, a creatinine clearance of below 30 ml/min and a history of other malignancy within the previous 2 years, with exception of localised skin, cervix, prostate and breast cancer, likely to be cured by surgery. Also excluded were patients with severe heart disease or other conditions severely limiting life expectancy.
Study design and treatment
This was a non-randomised, multicentre, open-label, single-arm phase I/II trial of the Arbeitsgemeinschaft Medikamentöse Tumortherapie (AGMT) trial consortium. All patients were included in this trial at seven cancer centres of the AGMT between September 2008 and November 2010. The phase I part evaluated the maximum tolerated lenalidomide dose level in combination with fludarabine and rituximab (FR) chemoimmunotherapy. Phase II of the study was planned to determine efficacy using the combination with the previously defined maximum tolerated dose (MTD) as plateau dose for dose escalation. An induction phase was followed by a maintenance phase evaluating the tolerability and possibility to further improve response quality (Supplementary Fig.
S1).
In the induction phase, an MTD of lenalidomide in combination with FR was to be determined during six cycles of fludarabine (40 mg/m
2 per os (po) days 1–3, repeated every 4 weeks) and rituximab (375 mg/m
2 intravenously (iv) day 4 at cycle 1; 500 mg/m
2 iv day 1 at cycles 2–6, repeated every 4 weeks). In cycle 1, lenalidomide was added from days 8 to 21 at 2.5 mg. Toxicity permitting, lenalidomide dose was escalated to 5, 10, 15, 20 and 25 mg on days 1–21 over cycles 2–6. Dose-limiting toxicities were defined as serious infections or limiting grade 3/4 toxicities, other than neutropenia. The latter was excluded from the definition of dose-limiting toxicities (DLT) because the FR backbone was expected to lead to a significant rate of grades 3–4 neutropenia by itself [
14]. However, prolonged neutropenia (neutropenia not resolving to grade ≤ 2 within 4 weeks) was defined as DLT. Patients were allowed to continue on the last tolerated dose or enter a stepwise de-escalation should a new DLT appear at a previously tolerated dose level. Because of a notion of an individual component in lenalidomide tolerability in reported trials [
7,
8,
12], an individual per patient dose escalation was favoured over a 3 + 3 design for dose escalation. Initially, the stepwise lenalidomide escalation was planned for the first 10 patients in the trial, using 24-h reporting of all grade 3/4 adverse events (AEs) as a safety measure for these patients to define a MTD for the remainder of the trial. Due to the findings in the first 10 patients (see “
Results” section), an amendment specified that all patients could be treated with the escalation schedule to their individual MTD.
In the maintenance phase, lenalidomide was planned to continue for six cycles continuously at the individually determined MTD, changing the schedule to a continuous 28-day cycle. De-escalation to a 21-day cycle and a stepwise dose de-escalation of lenalidomide were allowed for toxicity. Rituximab (375 mg/m2 iv) was administered after maintenance cycles 2, 4 and 6. For patients going off lenalidomide for any reason, completion of chemoimmunotherapy and rituximab maintenance was allowed for intention to treat (ITT).
Antimicrobial prophylaxis and growth factor support were not mandatory and were allowed per institutional practice. Antithrombotic prophylaxis with aspirin at a dose between 70 and 100 mg per day was mandated during lenalidomide treatment as an amendment after 10 patients.
Study endpoints and assessment
The primary endpoint was to determine the MTD per patient of lenalidomide in combination with FR. Secondary endpoints were efficacy and safety, risk factor dependency and minimal residual disease (MRD) analyses as well as T cell subset analyses.
Clinical and laboratory assessment was performed before beginning of every cycle during induction and maintenance phase. Computed tomography scans to assess nodal disease and MRD analyses from peripheral blood were performed after three and six cycles of induction treatment and after three cycles and completion of maintenance. MRD analyses were performed in a central laboratory (LIMCR at the IIIrd Medical Department in Salzburg) using a predefined 4-colour panel (according to Rawstron 2001) [
15] and using a cutoff of 10
−3 to define MRD negativity. Response was assessed using the 1996 National Cancer Institute guidelines including computed tomography (CT) scan results and bone marrow cytology and histology [
13]. Molecular risk assessment was performed in the central laboratory (LIMCR) as previously described [
16]. All patients that had received at least one dose of the study medication have been followed for adverse events (AEs) for at least 28 days after discontinuing study treatment or completion of study treatment. AEs have been assessed according to Common Terminology Criteria for Adverse Events v3.0 (CTCAE) by the investigator.
T cell analyses
T cell analyses were performed in purified peripheral blood mononuclear cells (PBMCs). A primary stain with unlabelled murine anti-human antibodies directed against PD-1 (clone EH12.2H7) or isotype control followed with secondary phycoerythrin (PE)-labelled goat-anti-mouse immunoglobulin antibody (Dako, Glostrup, Denmark) was followed by primary labelled surface stains as follows: CD4 (Pacific Blue); CD5 (allophycocyanin (APC)-AlexaFluor 700); CD25 (fluorescein isothiocyanate, FITC); CD45RA (energy-coupled dye, ECD or FITC); CD62L (PE or phycoerythrin-cyanin 5, PC5); CD127 (PE) (all from Beckman Coulter, CA, USA); CD8 (Pacific Orange) (Invitrogen, MD, USA); CD183 (APC) and CD194 (phycoerythrin-cyanin 7, PC7) (all from BD Biosciences, CA, USA). Analyses were done on a Gallios flow cytometer and T cell subsets were defined as follows: naïve (CD62L+CD45RA+), central memory (CM, CD62L−CD45RA+), effector memory (EM, CD62L−CD45RA−), TEMRA (CD62L+CD45RA−), TH1 (CD183+CD194−), TH2 (CD183−CD194+) and Treg (CD25+CD127−). Statistical analyses were performed using unpaired t tests for all subgroup results. A simple Bonferroni correction was applied to adjust for multiple testing, leading to a necessary p value of < 0.0025 for the definition of significance.
Statistical analyses
Statistical analyses were performed using IBM® SPSS® statistics software, version 24. Survival was estimated using Kaplan-Meier curve analysis, with statistical comparison using the log-rank statistic. For progression-free survival (PFS), analysis events were progression or death to any reason. Study dropouts were censored at the time of dropout. A two-tailed significance level of 0.05 was considered statistically significant. Cox regression was used for univariate analyses, where appropriate. Median follow-up was calculated by the reverse OS method using Kaplan-Meier curve analysis.
Discussion
Combination chemoimmunotherapy remains a standard of care in first-line treatment of CLL [
30,
31] and a randomised comparison between FCR and bendamustin/rituximab (BR) suggests that FCR is still the benchmark for the treatment of fit, untreated CLL patients [
32]. However, recent years have seen attempts at integrating novel targeted drugs into such chemoimmunotherapy regimens. Most recently, combinations of bendamustine and rituximab with the kinase inhibitors ibrutinib and idelalisib were presented [
33,
34]. In the case of idelalisib, the added toxicity presented a relevant problem for the patients and their management and in respect to both trials, there is an ongoing discussion about what the added value of the chemotherapy on top of the kinase inhibitor may be. The integration of novel drugs into combination regimens thus remains a major challenge. We aimed to define the feasibility of integrating the immunomodulatory drug lenalidomide into a chemoimmunotherapy regimen.
We found that in combination, chemoimmunotherapy and lenalidomide can be safely dose escalated to 10 mg or above in 65% of the patients. The combination of FR with lenalidomide was effective in terms of inducing responses and the overall strategy yielded a long PFS. In fact, the presented median PFS of 60 months compares very well with the 57 months reported for the FCR arm in the long-term observation of the German CLL8 trial [
35], despite a somewhat more comorbid population. Thus, the combination of chemoimmunotherapy replacing cyclophosphamide by lenalidomide yielded an encouraging disease control. We also demonstrate an effective control of expected lenalidomide-associated problems with tumour lysis and flare phenomena in the cohort. In our experience, the maintenance phase with the rituximab/lenalidomide combination led to a limited improvement in responses (five patients (12% of the maintenance population) improved their clinical response during maintenance (from PR to CR)), but this showed significant myelotoxicity. We conclude that the maintenance regimen, based on the established dose from the dose escalation part of the trial, was too ambitious, and suggest that maintenance may be more efficient at lower, more tolerable doses, which is in line with the observed doses achievable in two randomised lenalidomide maintenance trials recently presented [
36,
37].
In exploratory analyses, we found that achieving an MRD-negative state after induction was predictive of longer PFS, as has been reported for a number of other treatment modalities [
38‐
40]. Similar to most reported chemoimmunotherapy experiences, we detected a significant difference in outcome by IgVH mutation state. IgVH-mutated cases had a particularly long PFS with a majority of patients remaining in remission past the 5-year mark. In addition, high-risk cytogenetics remained a negative prognostic factor, although the impression is that the negative effect may be smaller than in previously reported chemoimmunotherapy trials with FCR or BR [
2,
41]. Despite the limitations inherent to a phase II trial, we aimed to analyse outcomes by mutations in a distinct set of eight genes that had been reported to affect outcomes in large sequencing studies [
23‐
26] to present hypothesis generating findings.
TP53 and
NOTCH1 were associated with the strongest risk for reduced overall survival, and the combined presence of more than one of these eight mutations also proved as negative factor in this respect (Supplementary Table
S1).
In our mind, the most surprising findings concern the pattern of toxicities, which we observed in our cohort. While in the majority of patients the major dose-limiting effect was a seemingly dose-dependent myelotoxicity, a subgroup of 35% did not tolerate doses above 5 mg lenalidomide throughout induction. In the respective patients, this became apparent very early on and at the lowest doses. Compared with 33% of patients that tolerated 25 mg lenalidomide, a dose up to tenfold of these patients, this phenomenon suggested an individual predisposition for a non-dose-dependent pattern of toxicities (Supplementary Fig.
S2). Hereby, the most prominent toxicities were skin toxicities, also contributing the largest cause of dose limitation or study termination by the patients.
Our results may contain some more general lessons for the development of combination approaches with novel substances. Some potential lessons stem from the comparison of our trial with a trial performed in parallel at the Dana Farber Cancer Institute. Brown et al. [
42] reported a trial with a highly similar treatment design but two potentially crucial differences: first lenalidomide was started at day 1 of the first cycle (together with the chemoimmunotherapy) and second, the trial was performed in a 3 + 3 design. The trial had to be stopped after nine patients at dose level − 1 (i.e. 2.5 mg every other day) for toxicity and futility reasons. How can we explain the fact that two thirds of our patients could tolerate more than five times the dose determined as a MTD in the trial by Brown et al.? We would like to propose two explanations: an argument of scheduling and one of trial design (individual dose escalation vs 3 + 3 design). A similar rate of early skin toxicities has not been reported in FR-treated or in lenalidomide monotherapy patients, suggesting a specific interaction of lenalidomide with the FR backbone in this respect. Since our previous work suggested significant changes in T cell reactivity early after fludarabine treatment, we speculate that a concomitant start of lenalidomide and fludarabine may lead to T cell-mediated toxic effects. Such effects may be mitigated by a delay in the start of the lenalidomide treatment, as was hypothesised in our trial design, since the parallel T cell depletion by fludarabine may limit such effects. Interestingly, a similar trial by Flinn et al., presented as meeting abstracts [
43,
44], also showed overall tolerability of a combination of lenalidomide with FR and also suggested using a 1-week delay before adding lenalidomide. Confirming our experience, the authors also found rashes to be a main characteristic side effect of the combination. Another small trial, introducing lenalidomide in addition to a dose-reduced FCR regimen, also opted to start lenalidomide treatment at day 8 of the first cycle and reported on 20 evaluable patients [
45]. Close to 5% of the patients had a grade 3/4 skin reaction and the authors described the overall toxicity profile acceptable. Comparing these results to those from Brown et al., we thus believe that the sequencing of the components of novel drug combinations may propose significant challenges for the development.
A second aspect relates to the trial design. Our experience suggested that a subpopulation of patients may have a predilection to suffer early non-dose-dependent toxicities leading to dose limitation and this group was significant in size. It is easy to see how such behaviour of a population could lead to a dismal outcome in a 3 + 3 design, since a relatively likely chance-recruitment of two patients from such a population may stop the dose escalation early and lead to a frustrating trial experience after a short interval. By comparison, our design of individual dose escalation in the first 10 patients allowed us to recognise the pattern, amend the trial accordingly and finish the trial with the presented results. We present these speculative interpretations as a note of caution regarding trial designs for substances that may have very complex mechanisms of action like lenalidomide and a potential to have significant toxicities in relevant subgroups. This may be especially important for combination approaches.
Finally, we asked ourselves whether we could identify such a subpopulation with increased toxicity upfront. We speculated that the T cell compartment of patients’ pretreatment may contain relevant information to this end. This was based on our earlier observations of T cell changes after fludarabine treatment [
29], the fact that the mode of action of lenalidomide is thought to include relevant modulation of T cell function [
6,
29,
46,
47] and the observation of excess skin toxicity, a phenomenon often associated with T cell activities [
48]. Indeed, studies on the original Imid, thalidomide, had shown that it changed the recruitment of T cells to skin lesions in the treatment of cutaneous sarcoidosis and leprosy [
49,
50]. Regarding a potential mechanism of T cell activation, lenalidomide has been shown to modulate exhaustion phenotypes of T cells in CLL [
46,
47,
51] and we hypothesised that a reversal of T cell exhaustion may be associated with the clinical side effects profile observed in the combination. In order to analyse such a phenomenon, we devised a panel of immunostains to define T cell subpopulations and their state of exhaustion via PD-1 staining and compared the individual patient data with a combined clinical endpoint, devised as a surrogate for intolerability. When analysing the dataset using a multiple testing correction, it became apparent that the PD-1 expression levels on the memory T cell fraction of the CD4 positive T cells contained predictive information. Indeed, while overall PD-1 expression on T cells was similar between patients with or without significant trouble with lenalidomide dose escalation, the best correlation was found with the fraction of PD-1 positive CD4 memory cells, a T cell subpopulation we had previously found to be important in CLL progression [
52]. In our current trial, this T cell subset was, however, not associated with either response to treatment or PFS. Overall, we propose that, while the mechanism of the interaction currently remains unclear, such a marker population may be useful to select patients for lenalidomide combination regimens in the future. This may be of special interest since recently, a number of trials containing both lenalidomide and agents targeting PD-1 were put on hold due to observed toxicities. Given our observation from this trial, it might be useful to include similar T cell analyses in the workup of trials investigating lenalidomide combination treatments.
In summary, the addition of lenalidomide to an FR chemoimmunotherapy backbone seemed feasible and effective, given a relevant attention to individual tolerability profiles of patients. However, based on the recent advances in the field, including novel combinations with high efficacy, but still short follow-up, it is unlikely that lenalidomide chemoimmunotherapy combinations will have a significant role in the treatment paradigm of previously untreated CLL.