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

The medical records of all CP-CML patients who were treated with TKIs in our institution between 1997 and 2015 were reviewed to identify patients that discontinued TKI therapy due to any reason. This retrospective study was in agreement of our institutional protocol for the management of patients with chronic myeloid leukemia. The eligibility criteria were CP-CML, treated with any of the first-line approved TKIs (imatinib, nilotinib, and dasatinib), and with typical transcripts [that is, b3a2 (e14a2) and/or b2a2 (e13a2)] determined by RT-PCR at the time of diagnosis. Recorded data included patient characteristics at diagnosis, the treatment received prior to discontinuation, the reasons for treatment discontinuation and the course of disease before and after discontinuation. Informed consent was registered in the clinical record of the patients in agreement with the Declaration of Helsinki.

Cytogenetic analysis was performed on bone marrow cells using standard G-banding methods on at least 20 metaphases from 24 h cell cultures, at diagnosis and at 3 months intervals during the first year, until a complete cytogenetic response (CCyR) was achieved. Molecular monitoring was performed by reverse transcriptase quantitative PCR (RT-qPCR) with an assay sensitivity of ≥4.5 log and scoring of deep molecular responses (DMR; MR^4.0, MR^4.5 and MR^5.0) was made according to published recommendations [3, 11]. After stopping TKI treatment, patients were, whenever possible, monitored by RT-qPCR monthly during the first year, then every 2 months for the second year, and every 3 months thereafter. Sustained DMR (SDMR) was defined as at least a MR^4.5 sustained during at least 2 years, and unsustained DMR (UDMR) was defined as at least a MR^4.5 with less than 2 years. Molecular relapse was defined as loss of major molecular response [MMR; BCR-ABL1/ABL1 internationally standardized (IS) ratio ≤ 0.1%], which triggered TKI resumption.

All statistical analyses were performed using SPSS software (SPSS Inc., Chicago, Ill., USA). Study variables were summarized using standard descriptive statistics and measures of central tendency, including median, ranges, frequencies and percentages. Time to molecular relapse was measured from the date of TKI discontinuation to the date of molecular relapse or the date of last molecular evaluation for patients who did not relapse. Treatment-free survival was estimated using the Kaplan-Meier method. To establish factors associated with TFR, patients with a follow-up of at least 16 months after TKI discontinuation were analyzed. We assessed age, sex, Sokal risk group, BCR-ABL1 transcript type, duration of TKI therapy, and DMR duration as potential prognostic factors. Differences in molecular relapse between groups were compared using the log-rank test. A two-tailed p-value of < 0.05 was considered to indicate statistical significance.

Twenty-five patients with chronic phase CML diagnosed and/or treated at our institution agreed to participate in the study and their baseline characteristics are detailed in Table 1. Fifteen patients (60%) were female and median age of all patients was 54 years (range, 15–78 years). Sokal score was low, intermediate and high in 10 (40%), 10 (40%) and 5 (20%) patients, respectively.

All patients were in chronic phase and none had a previous history of allogeneic stem cell transplantation (Table 2). Median age at TKI discontinuation was 64 years (range, 24–84). Imatinib as first line therapy was given to 23 (92%) patients, either soon after CML diagnosis (n = 20) or after intolerance to IFN-α (n = 3). Two patients (8%) received nilotinib as first line therapy. Median duration of TKI treatment before discontinuation was 100 months (range, 25–202) and, with the exception of one patient that was treated for only 25 months, all patients received TKI treatment during at least 36 months. Median time to DMR was 56 months (range, 13–207) in 24 patients, excluding the patient treated for 25 months that failed to achieve DMR. The median duration of DMR before stop was 41 months (range, 0–100). At the time of therapy discontinuation, 20 (80%) patients were receiving their initial TKI (19 imatinib and one nilotinib), four (16%) had changed to a second TKI due to intolerance (three from imatinib to dasatinib and one from nilotinib to imatinib), and one (4%) had changed from imatinib to dasatinib due to resistance.

Nineteen patients (76%) interrupted TKI therapy in SDMR and, with the exception of one case that developed pleural effusion (PE) under dasatinib treatment, none presented major adverse events at the time of discontinuation. The remaining six patients interrupted treatment in UDMR: two due to shared clinician and patient decision (both under imatinib treatment), three due to adverse events (one under treatment with dasatinib that developed PE, one under treatment with dasatinib that developed both PE and pulmonary hypertension, and one under treatment with nilotinib that developed acute renal failure) and one that discontinued imatinib due to onset of pulmonary tuberculosis. Molecular response at TKI cessation was MR^5.0 in 12 patients (48%), MR^4.5 in 12 patients (48%), and MR^4.0 in one patient (4%).

After TKI discontinuation, patients were followed for a median of 24 months (range, 15–97) (Table 3). At the time of this analysis, 14 patients (56%) had a molecular relapse after a median of four months (range, 2–9). Their characteristics, treatment and response are summarized in Additional file 1: Table S1. The estimated survival without molecular relapse (SWMR) for all patients at 12 months was 45.4% (95% CI 27.5–63.3; Fig. 1). The majority of relapses (79%) occurred within the first six months after TKI discontinuation, with the earliest relapse occurring after two months and the latest at nine months. Relapsed patients promptly resumed TKI therapy, and all regained a MMR after a median duration of TKI treatment of two months (range, 1–10). Of these, at the time of the last evaluation, 10 achieved MR^4.5 or better, after a median time of five months (range, 2–10) (Table 3, Additional 1: Table S1). None of the two patients with a relapse within two months of TKI discontinuation was in SDMR. The first (patient 3, Additional file 1: Table S1) was treated with imatinib for 10 months and was in DMR for 10 months at the time of TKI suspension. After relapse, imatinib was restarted and the patient was in MMR at the last follow-up. The second (patient 5, Additional file 1: Table S1) was treated for a total of 126 months: first with imatinib for 24 months but, due to treatment intolerance, he was changed to second–line dasatinib. Under dasatinib, this patient developed PE and pulmonary hypertension and, for this reason, a TKI discontinuation was attempted when the patient was in DMR for only eight months. After molecular relapse, imatinib treatment was initiated and the patient achieved a MMR. Patient 6 developed resistance under first-line imatinib after 68 months of treatment. BCR-ABL1 mutational analysis was negative and the patient was changed to second-line dasatinib. He was treated for an additional 44 months, achieved a DMR and was in SDMR for 33 months at the time of TKI discontinuation. He relapsed after six months, which prompted dasatinib resumption and was in MR^5.0 at the last evaluation. Of the three patients that received initial treatment with IFN-α before imatinib, one relapsed four months after TKI discontinuation (patient 2, Additional file 1: Table S1). This patient was treated with IFN-α for eight months, followed by imatinib for 171 months and, at the time of TKI discontinuation, was in DMR for only 10 months. At relapse, treatment with dasatinib was instituted and the patient regained a DMR (MR^5.0). Patient 10 developed acute renal failure under treatment with nilotinib and treatment was interrupted. The patient was not in SDMR and relapsed after seven months. Nilotinib therapy was reinstituted and, at the last follow-up, the patient was in MR^4.0 with a normalized renal function. Patient 14 was in treatment with imatinib for 25 months and in MR^4.0 when he acquired pulmonary tuberculosis. TKI treatment was discontinued and the patient relapsed after three months. At this point, imatinib therapy was resumed, with the patient achieving a MR^4.5 at the last follow-up. Patient 9 was diagnosed with CML outside our institution and first-line therapy with nilotinib was started. After six months, due to patient preference, he was transferred to our institute. Molecular evaluation showed a MR^4.0 and, according to our institute protocol for first-line treatment of CML patients, he was changed to imatinib. The patient was in SDMR for 56 months when TKI discontinuation was attempted but relapse occurred after four months. Imatinib therapy was resumed and at the last follow-up the patient was in MR^4.0.

At the last evaluation, 11 patients (44%) remain treatment-free with a median follow-up from treatment discontinuation of 27 months (range, 15–97), including two patients treated with dasatinib that discontinued therapy due to PE (patients 16 and 17, Additional file 2: Table S2). Of note, patient 16 was previously treated with IFN-α for 36 months but developed intolerance. For that reason, he was switched to imatinib but, due to persistent eosinophilia, was changed to dasatinib. At the time of TKI interruption both patients were in DMR: patient 16 in UDMR for 10 months and patient 17 in SDMR for 41 months and, at the last evaluation, they were relapse-free, after 17 and 27 months of follow-up, respectively. At the data cut-off, eight non-relapsing patients (73%) are in MR^5.0, two (18%) are in MR^4.5, and one (9%) is in MMR. In all cases, fluctuations in BCR-ABL1 levels were observed after treatment discontinuation. Three patients (27%), showed an improvement of their molecular response from therapy discontinuation to the last follow-up (MR^4.5 to MR^5.0), and one patient (9%) showed a detriment of their molecular response (MR^4.5 to MMR). In seven patients (63%), five in MR^5.0 and two in MR^4.5, no change in their molecular response level was observed at the last follow-up when compared to the discontinuation date. Of note, patient 24 showed a particularly interesting pattern of variation of BCR-ABL1 levels after TKI discontinuation (Additional file 2: Table S2 and Fig. 2). This patient was a female treated for 77 months with imatinib who achieved a DMR after 30 months and was in SDMR (MR^4.5, undetectable disease) at the time of TKI discontinuation. Five months later, increasing BCR-ABL1 levels lead to loss of MR^4.5 (BCR-ABL1 = 0.0050%; MR^4.0) but, subsequently declining BCR-ABL1 levels resulted in a MR^5.0 (BCR-ABL1 = 0.0008%) seven months after TKI interruption. This was again followed by increasing levels of BCR-ABL1, leading to loss of DMR to a MMR (BCR-ABL1 = 0.0447%) at the 9th month evaluation. She regained DMR (MR^4.5) at 10 months but again, increasing transcript levels, resulted in a MMR (BCR-ABL1 = 0.0106%) 13 months after TKI discontinuation. In the following months she showed a decreasing trend of BCR-ABL1 levels, with acquisition of DMR and, at the last follow-up, 50 months after imatinib discontinuation, she is in MR^5.0 with undetectable disease. In this patient, no therapeutic intervention directed to CML disease was performed since TKI discontinuation.

Statistical analysis was performed to identify factors associated with survival without molecular relapse (Table 4). In our series, the molecular relapse rate was 47.4% (nine patients) in SDMR patients, after a median of five months (range, 4–9). Most of the relapses occurred within the first six months (79%), with two patients relapsing at eight and nine months, respectively. In this group, the estimated SWMR was 53.7% (95% CI 33.1–74.2). Five out of six patients (83.3%) in the UDMR group relapsed after a median of three months (range, 2–7), with an estimated SWMR of 6.3% (95% CI 1.3–11.4). The only patient in the UDMR group that did not relapse interrupted therapy as a result of PE. Also in this group most of the relapses (80%) occurred in the first six months. The probability of SWMR was statistically different between patients in SDMR when compared with patients in UDMR (p = 0.016, Table 4). SWMR was not only significantly associated with SDMR but also with the duration of TKI therapy. Indeed, therapy duration time > 60 months (p = 0.026) and > 96 months (p = 0.040) before treatment interruption was significantly associated with the probability of molecular relapse: 45 and 100% in the group of patients treated with TKI > 60 months or ≤ 60 months, and 38 and 89% in the group of patients with TKI > 96 months or ≤ 96 months, respectively. The estimated SWMR at 12 months was 55.4% (95% CI 35.2–75.6) and 5.4% (95% CI 3.6–7.2) in the group of patients treated with TKI > 60 months or ≤ 60 months, and 62.0% (95% CI 39.9–84.1) and 10.7% (95% CI 1.5–19.8) in the group of patients with TKI > 96 months or ≤ 96 months, respectively.

When both the depth of molecular response and TKI therapy duration were considered, the probability of molecular relapse was 37.5% in SDMR patients with TKI treatment > 60 months compared to 88.9% in UDMR patients with TKI treatment ≤60 months (p = 0.003), with an estimated SWMR at 12 months of 62.7% (95% CI 41.1–84.4) in the former compared with only 6.1% (95% CI 2.7–9.6) in the later (Fig. 3a). With longer treatment duration, the probability of molecular relapse was 25.0% in SDMR patients with TKI treatment > 96 months patients and 84.6% in UDMR patients with TKI treatment ≤96 months (p = 0.003). The estimated SWMR at 12 months was 73.9% (95% CI 51.3–96.5) in SDMR patients with TKI treatment > 96 months compared with 11.9% (95% CI 2.9–20.8) in UDMR patients with TKI treatment ≤96 months (Fig. 3b).

We did not find any statistically significant association between SWMR with age, gender, Sokal score, and transcript type.