Discontinuation of tyrosine kinase inhibitors in CML patients in real-world clinical practice at a single institution

Chronic myeloid leukemia (CML) is a myeloproliferative neoplasia caused by the fusion of the BCR and ABL1 genes, most frequently as the result of the reciprocal translocation t(9;22)(q34;q11). ABL1 tyrosine kinase inhibitors (TKI) have dramatically improved the outcome for CML patients. Indeed, when TKI therapy is addressed appropriately, it can led to an optimal molecular response in the majority of CML patients and a life expectancy that approaches that of the general population [1, 2].

Although the ELN (European LeukemiaNet) recommendations on the management of CML propose indefinitely continuation of TKI therapy in all responding patients [3], the latest update of the NCCN (National Comprehensive Cancer Network) guidelines [4], based on the most recent studies on treatment-free remission, support the view that treatment can be interrupted in a selected group of patients. However, due to the inability of all TKIs to eliminate quiescent leukemic stem cells, the majority of the patients will require TKI therapy to be continued indefinitely, since most will relapse if treatment is withdrawn [5–7]. Nevertheless, lifelong TKI therapy may have consequences, including chronic, mostly low-grade, adverse events that can substantially impact patients’ quality of life, adherence to therapy and, consequently, success of treatment [8]. In the last few years, several clinical discontinuation trials have demonstrated that 40–60% of chronic phase CML patients (CP-CML) who have achieved a stable deep molecular response (DMR), defined as a sustained molecular response of at least 4.5 (MR^4.5), can stop therapy without relapsing [reviewed in 9,10]. In addition to DMR, other variables that have been associated with a successful treatment-free remission (TFR) include low Sokal risk group at diagnosis, chronic-phase patients, optimal response to TKI therapy, longer duration of TKI therapy (> 8 years), and longer duration of DMR (> 2 years) [9]. Recently, natural killer (NK) cell number and function at the time of discontinuation were associated with outcome after imatinib discontinuation in chronic-phase CML patients in deep molecular response [10]. In all published trials, the majority of patients who experienced relapse did so within 6 months of TKI cessation and, with the exception of one case that progressed to blast crisis, relapsed patients remained responsive to retreatment and regained at least a major molecular response (MMR) [9,10). However, most information on treatment cessation was obtained from clinical trials with strict recruiting criteria. In this study, we retrospectively aimed to assess persistence of TFR in 25 CML patients treated at our institution that discontinued therapy due to several causes, including DMR and TKI intolerance, and to identify factors that could be associated with TFR.

In the present study, we evaluated the outcomes of CP-CML patients who discontinued TKI therapy in our institution due to any reason including SDMR and/or adverse events. Our data show that 44% of the patients remained in TFR at 12 months, with an estimated SWMR of 45.4%, a value lower than that observed in clinical trials that used MMR loss as a trigger to restart TKI therapy, which is around 50–60% [12–14]. This may be explained by the fact that our series includes patients that discontinued TKI therapy for any reason, namely patients without SDMR, whose molecular relapse rate was 83%, with an estimated SWMR of only 5.0%. If we use more strict criteria, such as patients in SDMR at the time of therapy discontinuation, the TFR rate rises to 53%, more in line with the results observed in clinical trials. These results support the hypothesis that a sustained deep molecular response, defined as a MR^4.5 for at least 2 years, is associated with a greater probability of TFR. Using this criterion, the proportion of imatinib-treated CML patients eligible to a TFR attempt is approximately 30–40% after 5 years of therapy [15]. However, using more potent TKIs, such as nilotinib and dasatinib, the rates can raise up to 42–54% [16, 17]. The rates seem to slowly but steadily increase over time, raising the possibility that with very prolonged treatment the level of minimal residual disease may continue to fall. This can in turn increase the pool of CML patients that can access treatment discontinuation. However, it is a matter of intense debate whether a sustained MR^4.5 should be a strict requirement for TKI discontinuation, with preliminary results from the EURO-SKI trial suggesting that it could be safe to stop treatment in patients in MR^4.0, and that TKI treatment duration could be more important than the deepness of molecular response [18]. In previous clinical trials, imatinib treatment duration has emerged as a continuous variable, with increasing length of therapy correlated to increased success of TFR [13, 19, 20]. Supporting this hypothesis, we also found that a duration of TKI therapy both, over 60 months (5 years) and 96 months (8 years), was significantly associated with a lower probability of molecular relapse after TKI discontinuation: 45 and 38% in the group of patients treated with TKI > 60 months and TKI > 96 months, respectively. In addition, when both SDMR and longer TKI treatment duration were used as TFR selection criteria, the rates of molecular relapse were even lower: 37.5 and 88.9% in SDMR patients treated with TKI > 60 months or ≤ 60 months, and 25.0, and 84.6% in SDMR patients treated with TKI > 96 months or ≤ 96 months, respectively. Our data suggest that both the duration of DMR and time on TKI treatment should be carefully evaluated when TFR is considered. Importantly, it is not known if attempting TFR after a shorter duration of TKI treatment may reduce the likelihood of a successful outcome and eventually compromise the long-term success of TFR.

In addition to the degree of molecular response and treatment duration, several studies have suggested that other patient characteristics, including age, gender and Sokal score, could be associated with successful TFR, but results have not been consistent between studies [18, 21]. In our series, we did not find any statistically significant association between TFR with age, gender, and Sokal score. Furthermore, it has been suggested that the BCR-ABL1 transcript type could predict for optimal ELN responses and for longer event-free and transformation-free survival [22]. Therefore, we also evaluated the potential impact of transcript type (e14a2, e13a2 and e13a2 + e14a2) on TFR, but again no association could be found. We nevertheless recognize that much larger series are needed to conclusively evaluate other variables besides DMR and treatment duration.

In our series, five out of six patients that interrupted therapy in UDMR relapsed. The reasons for discontinuation included personal option (two cases), intolerance or toxicity issues (three cases) and pulmonary tuberculosis (one case), which clearly illustrates that following strict criteria in the selection of patients eligible for drug discontinuation is not always feasible outside clinical trials. Three patients under treatment with dasatinib discontinued therapy due to PE. Interestingly, two of them remain free of treatment, one that interrupted treatment in SDMR and the other in UDMR. PE is a common side-effect of dasatinib treatment with an incidence ranging from 14 to 39% [16, 23]. Although, most cases are mild or moderate, with grade 3/4 reported in only 3% of patients, PE represents one of the leading causes of dasatinib discontinuation [16, 23, 24]. It has been suggested that an inflammatory immune response involving NK cells underlies the development of PE in patients receiving dasatinib, a mechanism that could be associated with tumor regression [25, 26]. Interestingly, evidence implicating the role of immune response in TFR is increasing. The abundance and activity of NK cells, T-regulatory cells and dendritic cells has been correlated with likelihood of TFR [27–29], with recent data suggesting that NK cells may play a role controlling leukemia-initiating cells responsible for relapse after TKI cessation [10]. This can explain fluctuations in BCR-ABL1 transcript levels below the level of MMR, observed in most of our patients that remain in TFR. This phenomenon implies persistence of molecular disease due to a residual pool of leukemia stem cells (LSC) and that NK cells might contribute to prevent overt CML relapse [7]. These data highlight a potential role of the immune system in controlling residual LSCs, which may not be BCR-ABL1 addicted and therefore may be resistant to TKI treatment and represent a potential risk of relapse.

In our series, most relapses occurred early in the first 6 months after TKI discontinuation and late relapses were a rare event, which is in agreement with most of the studies published to date [9, 30]. All our relapsed patients responded to TKI treatment and regained at least a MMR with no progression to advanced phase, also in accordance to previously published data [9, 30]. This seems to support the hypothesis that the TFR period does not favor the emergence of resistant clones, as long as TKI treatment is reintroduced soon after molecular relapse is identified. Only two exceptions to this scenario were reported in the literature: in one patient MMR loss was accompanied by the detection of a nilotinib-resistant F359 V mutation, whereas in another patient, who was in MMR after imatinib resumption, CML progressed to lymphoid blast crisis [12, 31]. Indeed, it would be surprising if TFR was completely innocuous, with a very small risk of subsequent progression impossible to exclude.

Despite the excitement over TKI discontinuation, only a minority of newly diagnosed CML patients are likely to enter successful TFR, assuming relapse is defined as loss of MMR. For the remaining majority, lifelong TKI therapy may be required. An important practical question is when to consider TFR outside clinical trials. Our data, reflecting “real-world” clinical practice without selection of patients, support the hypothesis that patients in long-term TKI treatment and sustained deep molecular response are clearly the best candidates. However, access to high quality internationally standardized qRT-PCR, rapid turn-around of qRT-PCR test results, and structured follow-up established to enable rapid intervention in cases of molecular relapse are also mandatory.

Despite the small and heterogeneous group of patients, our results are in agreement with recent published data regarding TKI discontinuation success in patients with CML. We here show that TKI discontinuation is feasible and safe outside clinical trials, being particularly effective in CML patients in SDMR and with longer duration of TKI treatment.