Assessment of Outcomes After Stopping Tyrosine Kinase Inhibitors Among Patients With Chronic Myeloid Leukemia

Key Points

Question

Are there characteristics associated with a successful discontinuation of a tyrosine kinase inhibitor for patients with chronic myeloid leukemia, and do patient symptoms improve after discontinuation?

Finding

In this nonrandomized clinical trial of 172 patients with chronic myeloid leukemia (171 with available molecular data) from 14 US sites, only detectable BCR-ABL1 by real-time quantitative polymerase chain reaction or droplet digital polymerase chain reaction at the time of tyrosine kinase inhibitor discontinuation was associated with molecular recurrence. Among all 172 included patients, tyrosine kinase inhibitor discontinuation was associated with improvements in patient-reported fatigue, diarrhea, depression, and sleep disturbance.

Meaning

This nonrandomized clinical trial suggests that tyrosine kinase inhibitor discontinuation is safe and feasible and is associated with improvements in patient-reported outcomes.

Abstract

Importance

Tyrosine kinase inhibitors (TKIs) have been associated with improved survival of patients with chronic myeloid leukemia (CML) but are also associated with adverse effects, especially fatigue and diarrhea. Discontinuation of TKIs is safe and is associated with the successful achievement of treatment-free remission (TFR) for some patients.

Objective

To evaluate molecular recurrence (MRec) and patient-reported outcomes (PROs) after TKI discontinuation for US patients with CML.

Design, Setting, and Participants

The Life After Stopping TKIs (LAST) study was a prospective single-group nonrandomized clinical trial that enrolled 172 patients from 14 US academic medical centers from December 18, 2014, to December 12, 2016, with a minimum follow-up of 3 years. Participants were adults with chronic-phase CML whose disease was well controlled with imatinib, dasatinib, nilotinib, or bosutinib. Statistical analysis was performed from August 13, 2019, to March 23, 2020.

Intervention

Discontinuation of TKIs.

Main Outcomes and Measures

Molecular recurrence, defined as loss of major molecular response (BCR-ABL1 International Scale ratio >0.1%) by central laboratory testing, and PROs (Patient-Reported Outcomes Measurement Information System computerized adaptive tests) were monitored. Droplet digital polymerase chain reaction (ddPCR) was performed on samples with undetectable BCR-ABL1 by standard real-time quantitative polymerase chain reaction (RQ-PCR).

Results

Of 172 patients, 89 were women (51.7%), and the median age was 60 years (range, 21-86 years). Of 171 patients evaluable for molecular analysis, 112 (65.5%) stayed in major molecular response, and 104 (60.8%) achieved TFR. Undetectable BCR-ABL1 by either ddPCR or RQ-PCR at the time of TKI discontinuation (hazard ratio, 3.60; 95% CI, 1.99-6.50; P < .001) and at 3 months (hazard ratio, 5.86; 95% CI, 3.07-11.1; P < .001) was independently associated with MRec. Molecular recurrence for patients with detectable BCR-ABL1 by RQ-PCR was 50.0% (14 of 28), undetectable BCR-ABL1 by RQ-PCR but detectable by ddPCR was 64.3% (36 of 56), and undetectable BCR-ABL1 by both ddPCR and RQ-PCR was 10.3% (9 of 87) (P ≤ .001). Of the 112 patients in TFR at 12 months, 90 (80.4%) had a clinically meaningful improvement in fatigue, 39 (34.8%) had a clinically meaningful improvement in depression, 98 (87.5%) had a clinically meaningful improvement in diarrhea, 24 (21.4%) had a clinically meaningful improvement in sleep disturbance, and 5 (4.5%) had a clinically meaningful improvement in pain interference. Restarting a TKI resulted in worsening of PROs.

Conclusions and Relevance

In this study, TKI discontinuation was safe, and 60.8% of patients remained in TFR. Discontinuation of TKIs was associated with improvements in PROs. These findings should assist patients and physicians in their decision-making regarding discontinuation of TKIs. Detectable BCR-ABL1 by RQ-PCR or ddPCR at the time of TKI discontinuation was associated with higher risk of MRec; clinical application of this finding should be confirmed in other studies.

Trial Registration

ClinicalTrials.gov Identifier: NCT02269267

This nonrandomized clinical trial evaluates molecular recurrence and patient-reported outcomes after tyrosine kinase inhibitor discontinuation in US patients with chronic myeloid leukemia.

Introduction

Chronic myeloid leukemia (CML) is caused by the BCR-ABL1 (OMIM: ABL1, 189980; and BCR, 151410) tyrosine kinase, the product of the t(9;22) translocation visible as the Philadelphia chromosome. Since the introduction of tyrosine kinase inhibitors (TKIs), the survival of patients with CML has improved.¹ Although TKIs have fewer toxic effects than previous treatments,^2,3 they are associated with fatigue, depression, sleep disturbances, and diarrhea.⁴

After initial studies^5,6 suggested that TKI discontinuation is safe and feasible for patients with a sustained deep molecular response and is associated with treatment-free remission (TFR), several studies have confirmed the feasibility of this approach.^{6,7,8,9,10,11} To our knowledge, the Life After Stopping TKIs (LAST) study is the first prospective US-only trial designed to evaluate the proportion of US patients with CML who achieve TFR and describe patient-reported outcomes (PROs) after TKI discontinuation. We report the results of the LAST study with a minimum follow-up of 3 years.

Methods

A description of the study design has been published¹² (trial protocol in Supplement 1). In brief, the LAST study was a prospective nonrandomized clinical study of patients with chronic-phase CML. From December 18, 2014, to December 12, 2016, 173 patients from 14 US centers were enrolled (Figure 1). Key eligibility criteria included chronic-phase CML and either the b3a2 (e14a2) or b2a2 (e13a2) variants that give rise to the p210 BCR-ABL protein; aged 18 years or older; taking imatinib, dasatinib, nilotinib, or bosutinib for 3 or more years with continuous documented BCR-ABL less than 0.01% by polymerase chain reaction (PCR) for at least 2 years; and no previous TKI resistance. Patients were assessed by real-time quantitative polymerase chain reaction (RQ-PCR) by the Central Molecular Diagnostics Core Laboratory at the Fred Hutchinson Cancer Research Center (the central laboratory in Seattle, Washington) twice, at least 21 days apart, to confirm BCR-ABL1 less than 0.01% (ie, >4-log reduction [MR4] using the International Scale [IS]). The study was approved by the institutional review boards of Medical College of Wisconsin, Milwaukee; Karmanos Cancer Institute, Wayne State University School of Medicine, Detroit, Michigan; Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington; H. Lee Moffitt Cancer Center and Research Institute, Tampa, Florida; Huntsman Cancer Institute, The University of Utah, Salt Lake City; University of Chicago, Chicago, Illinois; Memorial Sloan Kettering Cancer Center, New York, New York; Duke Cancer Institute, Durham, North Carolina; Weill Medical College of Cornell University, New York, New York; University of California at San Francisco; Dana-Farber Cancer Institute, Boston, Massachusetts; Roswell Park Cancer Institute, Buffalo, New York; Emory University, Atlanta, Georgia; and MD Anderson Cancer Center, Houston, Texas. All patients provided written informed consent.

Figure 1. Patient Flow Diagram.

MMR indicates major molecular response; PRO, patient-reported outcome; and TKI, tyrosine kinase inhibitor.

Monitoring BCR-ABL1

Monitoring for molecular recurrence (MRec) using RQ-PCR was performed on peripheral blood samples monthly for the first 6 months, every 2 months until 24 months, then every 3 months until 36 months. Molecular recurrence was defined as a single value greater than 0.1% BCR-ABL1 based on the IS ratio (loss of major molecular response [MMR]). For patients who restarted TKI therapy, RQ-PCR was performed approximately every 3 months at the Central Molecular Diagnostics Core Laboratory at Fred Hutchinson Cancer Center until the patient’s BCR-ABL1 IS ratio was less than 0.01% (MR4) 2 consecutive times. Clinical data were recorded using the OnCore clinical trial management system.

Real-time quantitative polymerase chain reaction monitoring was performed on the IS scale (Cepheid). The RQ-PCR detects only the major breakpoint (p210) transcripts. All samples with an undetectable BCR-ABL1 IS ratio (<MR4.5 with adequate ABL1 control amplification) were examined using droplet digital PCR¹³ (ddPCR; Bio-Rad), which offers approximately 0.5 to 1 log greater sensitivity owing to a single-copy distribution of the target.

PRO Assessments

Patient-reported outcomes were assessed monthly for the first 6 months, at 8 and 12 months, then every 6 months until 36 months. After restarting a TKI, PROs were assessed in conjunction with RQ-PCR, approximately every 3 months for 1 year, then every 6 months thereafter. Missing PRO assessments were reported as deviations. Because of the approximate RQ-PCR schedule after restarting a TKI, during this phase deviations associated with PROs were reported if a patient completed fewer than 4 PRO assessments during the 18 months after restarting a TKI. Assessments of PROs were primarily administered via REDCap¹⁴ on a study iPad at the clinic or laboratory, but alternatives were also allowed (eg, via REDCap outside the clinic or on paper). The selection of PRO domains was based on CML experts’ opinions and literature review.¹⁵ We used PROMIS (Patient-Reported Outcomes Measurement Information System) measures, which are scored on a standardized scale, where 50 corresponds to the mean in the US general population (SD ±10), and approximately 20% of people report symptoms of at least moderate severity (≥1 SD higher or worse than the mean).¹⁶ Mean scores for adults without any chronic conditions are lower (eg, 45 on fatigue, 46 on pain interference, and 47 on depression).¹⁷

Statistical Analysis

Statistical analysis was performed from August 13, 2019, to March 23, 2020. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05. Statistical analyses were conducted using SAS, version 9.4 (SAS Institute Inc).

We powered the study to detect differences by patient characteristics in estimating MRec. Specifically, 172 patients provided 90% power to detect a difference of 25% between groups of equal size in relapse-free survival at 18 months. For the PRO measures, we considered 3 points to be clinically meaningful (corresponding to a standardized effect size of 0.3)¹⁸ and hypothesized that, for patients in TFR at 6 months, fatigue, depression, sleep disturbance, and diarrhea would improve by a mean of 3 points or more. Relative to the null hypothesis of no difference in PROs with TKI discontinuation, 172 patients provided greater than 90% power to detect a change of 3 points. Given reports of a TKI withdrawal syndrome of musculoskeletal pain in some patients,¹⁹ pain was an additional outcome of interest.

Analysis of Clinical End Points

Patient- and disease-related factors were compared between groups using the χ² test for categorical variables and the Mann-Whitney test for continuous variables. The probability of molecular relapse-free survival and TFR was calculated using the Kaplan-Meier estimator. Molecular relapse-free survival was defined as the time from TKI discontinuation to loss of MMR. Treatment-free remission was defined as the time from TKI discontinuation to restarting TKI for any reason. The probability of MRec was calculated using the cumulative incidence estimator, with time to MRec and time to death without losing MMR as competing risks. First, the maximized partial likelihood method was used to find the optimal threshold cutoff point for PCR at 3, 5, and 9 months. Cutoffs from 0.001 to 0.9 were considered. Using the −2 log likelihood fit statistic, we identified no cutoff. Modeling proceeded using detectable vs undetectable at 3 months for the landmark analysis to estimate MRec. To adjust for differences in baseline characteristics, a stepwise Cox proportional hazards regression model was used to identify all significant risk factors associated with time to MRec. The variables considered in the multivariable models included sex, age, Sokal Risk score at diagnosis, duration of TKI, and BCR-ABL1 transcript levels measured by RQ-PCR or ddPCR. A Fine-Gray subdistribution hazards model was used to evaluate the association of PCR with time to MRec. A 3-month landmark analysis was performed for patients who were alive and had not restarted a TKI.

Analysis of PROs

The primary objective was to describe patient symptoms after TKI discontinuation. A secondary objective was to describe symptoms after restarting TKI. For each PRO domain, we estimated a polynomial piecewise linear mixed-effects model that specified 1 nonlinear trajectory after TKI discontinuation and, for those who restarted treatment, another trajectory after restarting TKI. The models included patient-level random effects for the intercepts and linear slopes. This modeling approach offered several advantages; the likelihood-based estimation allowed for inclusion of all available data from each patient,²⁰ correlations within individual patients over time were addressed, and any missing data were considered ignorable conditional on the observed data.²¹ We plotted the trajectories of model-estimated changes in PRO measures after TKI discontinuation and after restarting TKI, including box plots showing the full distribution of patient scores at each time point.

Results

From December 18, 2014, to December 12, 2016, 173 patients from 14 US centers were enrolled (Figure 1). Of the 172 evaluable patients, 89 (51.7%) were women, and the median age was 60 years (range, 21-86 years) (eTable 1 in Supplement 2).The median time taking TKI prior to enrollment was 82.7 months (range, 36.1-199 months), which is similar to other studies.^6,22,23 The TKIs prior to discontinuation were imatinib (102 [59.3%]), nilotinib (39 [22.7%]), dasatinib (27 [15.7%]), or bosutinib (4 [2.3%]).

MRec and TFR

Of the 171 patients evaluable for molecular analysis, 112 (65.5%) stayed in MMR, and 104 (60.8%) achieved TFR; 59 patients (34.5%) had MRec, and 67 (39.2%) restarted therapy (Figure 2A). Median follow-up was 41.6 months (range, 4.1-61.8 months). Reasons for restarting therapy were MRec as determined by the results of central (n = 54) or local (n = 4) laboratory tests, physician or patient decision (n = 6), and withdrawal syndrome (n = 3). Of the 59 patients who had MRec, 16 (27.1%) had MRec at less than 3 months after discontinuation, 23 (39.0%) had MRec at 3 to 6 months after discontinuation, 9 (15.3%) had MRec at 6 to 12 months after discontinuation, 6 (10.2%) had MRec at 12 to 24 months after discontinuation, 3 (5.1%) had MRec at 24 to 36 months after discontinuation, and 2 (3.4%) had MRec at more than 36 months after discontinuation. The median time to MRec was 4 months (range, 1.5-41.3 months). Of the patients who had MRec, 55 achieved a BCR-ABL1 IS ratio of 0.01% or less, 2 patients were lost to follow-up, 1 did not restart treatment, and 1 died prior to achieving a BCR-ABL1 IS ratio of less than 0.01%. Four patients died during the study of causes not associated with CML (suicide [n = 1], unknown [n = 2], and stroke [n = 1]).

Figure 2. Probability of Molecular Relapse-Free Survival, Treatment-Free Remission (TFR), and Molecular Recurrence (MRec).

A, Kaplan-Meier plot for probability of molecular relapse-free survival and TFR. B, Probability of MRec, by BCR-ABL1 detectable by either real-time quantitative polymerase chain reaction (RQ-PCR) or droplet digital polymerase chain reaction (ddPCR) vs undetectable at baseline. C, Landmark analysis at 3 months: probability of MRec, by detectable or undetectable baseline PCR and 3-month PCR detectable or undetectable. MMR indicates major molecular response.

At the time of study enrollment, 28 patients (16.4%) had detectable BCR-ABL1 transcripts by RQ-PCR, and 143 (83.6%) had undetectable BCR-ABL1 transcripts by RQ-PCR. Of the 143 patients with undetectable BCR-ABL1 transcripts by RQ-PCR, 56 (39.2%) had BCR-ABL1 transcripts detected by ddPCR. Molecular recurrence for patients with detectable BCR-ABL1 transcripts by RQ-PCR was 50.0% (14 of 28), MRec for patients with undetectable BCR-ABL1 transcripts by RQ-PCR but detectable by ddPCR was 64.3% (36 of 56), and MRec for patients with undetectable BCR-ABL1 transcripts by both ddPCR and RQ-PCR was 10.3% (9 of 87) (P ≤ .001) (Figure 2B). Undetectable BCR-ABL1 transcripts by either ddPCR or RQ-PCR at the time of TKI discontinuation (hazard ratio, 3.60; 95% CI, 1.99-6.50; P < .001) and at 3 months (hazard ratio, 5.86; 95% CI, 3.07-11.1; P < .001) was independently associated with MRec. In addition, a landmark analysis demonstrated that detectable BCR-ABL1 transcripts by RQ-PCR at 3 months was independently associated with MRec later compared with patients with undetectable BCR-ABL1 transcripts by RQ-PCR (68.6% [35 of 51] vs 11.5% [12 of 104]; hazard ratio, 7.6; 95% CI, 4.19-13.8; P < .001).

On univariable analysis, detectable BCR-ABL1 by either method prior to discontinuation and a shorter duration of TKI therapy were associated with a higher risk of MRec (eTable 2 in Supplement 2). By multivariable analysis, age, sex, Sokal risk score, duration of TKI, type of TKI, time to MMR, and lymphocyte count at the time of discontinuation had no significant associations with MRec (eTable 3 in Supplement 2). Detectable BCR-ABL1 by either method at discontinuation was associated with a significantly higher risk of MRec (hazard ratio, 10.11; 95% CI, 3.35-30.49; P < .001).

The MRec rate was calculated for the 4 possible categories of detectable and undetectable (by either assay) at baseline and 3 months. The MRec rate was 4.0% (3 of 75) for undetectable BCR-ABL1 at both baseline and 3 months, 31.0% (9 of 29) for detectable baseline and undetectable at 3 months, 44.4% (4 of 9) for undetectable baseline and detectable at 3 months, and 73.8% (31 of 42) detectable at both time points (Figure 2C; eTable 4 in Supplement 2).

The sensitivity of a detectable BCR-ABL1 by RQ-PCR at 3 months to estimate MRec at 6 months was 93%, and the specificity was 81%. The sensitivity of a detectable BCR-ABL1 by RQ-PCR at 6 months to estimate MRec at 12 months was 100%, and the specificity was 82%. All 13 patients who had detectable BCR-ABL1 at 6 months had lost MMR by 12 months.

Patient-Reported Outcomes

We analyzed 1883 PRO assessments and reported 90 PRO-related deviations, representing 4.8% missing assessments. Changes in PROs are summarized in Figure 3. After TKI discontinuation, there was a statistically significant mean improvement in fatigue, depression, diarrhea, and sleep disturbance and no mean change in pain interference. Improvements were sustained over time. Restarting a TKI resulted in mean worsening of PROs.

Figure 3. Mean Changes in Patient-Reported Outcomes After Tyrosine Kinase Inhibitor (TKI) Discontinuation and TKI Restart at 6 and 12 Months.

Vertical lines indicate 95% CIs. PROMIS indicates Patient-Reported Outcomes Measurement Information System.

Fatigue

Before TKI discontinuation, the mean fatigue score was 3 points worse than the mean fatigue score in the US general population (eTable 5 in Supplement 2); 15.7% of patients (27 of 172) had at least moderately severe fatigue. Fatigue improved after TKI discontinuation, and the improvement was sustained over time (Figure 4A). For patients in TFR at 6 months, 25.6% (33 of 129) had a clinically meaningful (ie, ≥3 points) improvement in fatigue; at 12 months, 80.4% of patients in TFR (90 of 112) had a clinically meaningful improvement in fatigue. At 12 months, 18.8% of patients in TFR (21 of 112) had a larger (ie, ≥5 points) improvement. The mean fatigue score worsened after restarting a TKI, peaking about 10 months after restarting and then improving again (Figure 4B).

Figure 4. Fatigue After Tyrosine Kinase Inhibitor (TKI) Discontinuation and After Restarting a TKI.

A, Trajectories (gray sold line) with 95% CIs (gray dotted lines) of model-estimated changes in Patient-Reported Outcomes Measurement Information System (PROMIS) fatigue scores after TKI discontinuation (n = 172). B, Trajectories (gray sold line) with 95% CIs (gray dotted lines) of model-estimated changes in PROMIS fatigue scores after restarting a TKI (n = 57). The distributions of patient scores over time are shown as box plots (minimum excluding outliers, first quartile, median, third quartile, and maximum excluding outliers, with outliers shown as dots). Higher scores indicate greater severity of fatigue.

Depression

Before TKI discontinuation, the mean depression score was similar to the mean depression score in US general population (eTable 5 in Supplement 2); 19.8% of patients (34 of 172) had at least moderately severe depression. The mean depression score improved after TKI discontinuation (eFigure 1A in Supplement 2). At 12 months, 34.8% of patients in TFR (39 of 112) had a clinically meaningful improvement in depression. There were minimal changes in the mean depression score after restarting a TKI (eFigure 1B in Supplement 2).

Diarrhea

Before TKI discontinuation, mean diarrhea scores were 6 points better than the mean diarrhea score in the US general population (eTable 5 in Supplement 2); 17.4% of patients (30 of 172) had at least moderately severe diarrhea. Mean diarrhea scores improved after TKI discontinuation (eFigure 2A in Supplement 2). By 12 months, 87.5% of patients in TFR (98 of 112) had a clinically meaningful improvement in diarrhea scores. Mean diarrhea scores worsened after restarting a TKI, peaking about 12 months after restarting and then improving again (eFigure 2B in Supplement 2).

Sleep Disturbance

Before TKI discontinuation, the mean sleep disturbance score was similar to the mean sleep disturbance score among the US general population (eTable 5 in Supplement 2); 14.0% of patients (24 of 171) had at least moderately severe sleep disturbance. The mean sleep disturbance score improved after TKI discontinuation, and the improvement was sustained (eFigure 3A in Supplement 2). By 12 months, 21.4% of patients in TFR (24 of 112) had a clinically meaningful improvement in sleep disturbance score. Mean sleep disturbance scores worsened after restarting a TKI, although they appeared to improve again over time (eFigure 3B in Supplement 2).

Pain Interference

Before TKI discontinuation, the mean score for pain interference (ie, the extent to which pain affects daily life) was similar to the mean pain interference score among the US general population (eTable 5 in Supplement 2); 20.3% of patients (35 of 172) had at least moderately severe pain interference. The mean pain interference score worsened slightly after TKI discontinuation, but the change was not statistically or clinically significant (eFigure 4A in Supplement 2). Very few patients reported a clinically meaningful worsening in pain interference score (none at 6 months and 2 of 112 [1.8%] at 12 months), whereas 5 of 112 patients (4.5%) reported a clinically meaningful improvement at 12 months. The mean pain interference score worsened after restarting a TKI, although it appeared to improve again over time (eFigure 4B in Supplement 2).

Discussion

Patients in the LAST study were representative of the current TKI prescribing patterns in the US, with 59.3% of patients receiving imatinib.²⁴ Although 65.5% of patients maintained MMR at 3 years, the rate of successful TFR was lower at 60.8%, highlighting the other reasons patients restart treatment in the absence of MRec, such as withdrawal syndrome and physician or patient anxiety. The rate of MRec in our study was similar to that in other studies (range, 32%-67%),²⁵ and 67.1% of MRec occurred by 6 months.

We found that detectable BCR-ABL1 at the time of TKI discontinuation was associated with MRec. This finding contrasts with the EURO-SKI (European Stop Kinase Inhibitor) study that demonstrated no difference in MRec for patients with a BCR-ABL1 IS ratio of less than 0.0032% (MR4.5) vs a BCR-ABL1 IS ratio less than 0.01% (MR4.0).²³ In addition, in the LAST study, BCR-ABL1 detection by ddPCR at the time of discontinuation was associated with MRec. Based on BCR-ABL1 detectability by either RQ-PCR or ddPCR at the time of TKI discontinuation and by RQ-PCR at 3 months, we were able to classify patients into 4 distinct groups with regard to risk of MRec. Molecular recurrence occurred in 4% of patients with neither risk factor vs 74% of patients with both risk factors. At this time, ddPCR is not widely available in the US. However, if available and based on our results and those of others,^26,27,28 a less-intensive monitoring approach for patients with undetectable BCR-ABL1 at TKI cessation by ddPCR and with undetectable BCR-ABL1 at 3 months may be justified (eg, every 3 months) given the low risk for MRec. For others, the current intensive (eg, every month) monitoring approach remains justified. In addition, the use of ddPCR is associated with better risk stratification and may help patients make decisions regarding discontinuation. In our ancillary study examining patients’ decision-making regarding TKI discontinuation, patients were hesitant to discontinue for multiple reasons, 2 of which were concerns about relapse and the need for frequent monitoring.²⁹ Given these patient concerns, ddPCR results could be used to counsel patients more accurately on the risks of MRec and the need for frequent monitoring, which may mitigate anxiety.

It remains unclear why many patients with detectable BCR-ABL1 transcripts remain without clinical disease progression. One explanation is that BCR-ABL1 positivity arises from memory B lymphocytes, as shown by Pagani et al.³⁰ Another hypothesis is that stem cell exhaustion over time is associated with a decrease in residual leukemia stem cells,³¹ perhaps more so in patients with shorter telomeres.³²

To our knowledge, the LAST study is the first TKI discontinuation study to include a comprehensive PRO assessment and the first to use the rigorously developed National Institutes of Health PROMIS measures. Previous studies using Functional Assessment of Cancer Therapy measures have reported that US adults with CML and other cancers report PROs similar to the general adult population,^33,34 which was confirmed in our study. Likewise, in a European study using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Chronic Myeloid Leukemia 24 measure, 12% of patients taking dasatinib and 19% taking imatinib reported moderate to severe fatigue, similar to patients in the LAST study (16%).³⁵ We found that TKI discontinuation was associated with improvements in patient-reported fatigue, depression, sleep disturbance, and diarrhea. Most patients in TFR at 1 year experienced clinically meaningful improvements in fatigue and diarrhea. For patients who restarted a TKI, PROs worsened again.

Despite previous reports of increased musculoskeletal pain after TKI discontinuation (TKI withdrawal syndrome) for a substantial minority of patients,³⁶ very few patients in our study reported an increase in how much pain interfered with their daily life. In previous studies, the adverse event of musculoskeletal pain was reported by clinicians, and associations between clinician-reported symptomatic adverse events and PROs are moderate at best.³⁷ We will investigate pain and TKI withdrawal syndrome using additional types of data in future analyses.

Limitations

This study has some limitations. It was not a randomized clinical trial and therefore may provide less evidence for causal conclusions. Conducting a randomized clinical trial in this context presented several difficulties, chiefly that we believed patient compliance would be problematic. Patients who enrolled because they were inclined to stop therapy would have little incentive to take their assigned (blinded) pill, especially because they knew their blood was being frequently monitored, ensuring they would be told to restart a TKI if necessary.

Conclusions

The LAST study was the largest and only prospective US-only TKI discontinuation study, to date, and the first to include comprehensive PRO measurement, to our knowledge. Our data demonstrate the applicability of other trials’ conclusions to the US population and may improve US clinicians’ compliance and awareness regarding TKI discontinuation practices.³⁸ The LAST study demonstrates that patients with undetectable BCR-ABL1 by ddPCR had a very high chance of maintaining MMR, with modest improvements in PROs. Our results provide important new information to inform the discussion between patients with CML and clinicians about the risks and benefits of TKI discontinuation and the potential to assess TKI cessation more accurately.

Supplement 1.

Trial Protocol

Supplement 2.

eTable 1. Patient Demographics

eTable 2. Univariable Analysis for Loss of Major Molecular Response

eTable 3. Multivariable Analysis for Loss of Major Molecular Response

eTable 4. Three-Month Landmark Analysis for Baseline and 3-Month BCR-ABL1 Results for Risk of MMR Loss

eTable 5. Baseline PROs and Clinically Meaningful Changes for Patients in TFR at 6 and 12 Months

eFigure 1. Trajectories of Model-Predicted Changes in PROMIS Depression Scores After TKI Discontinuation (A, n = 172) and After Restarting a TKI (B, n = 57), With Box Plots Showing the Distribution of Patient Scores Over Time

eFigure 2. Trajectories of Model-Predicted Changes in PROMIS Diarrhea Scores After TKI Discontinuation (A, n = 172) and After Restarting a TKI (B, n = 57), With Box Plots Showing the Distribution of Patient Scores Over Time

eFigure 3. Trajectories of Model-Predicted Changes in PROMIS Sleep Disturbance Scores After TKI Discontinuation (A, n = 172) and After Restarting a TKI (B, n = 57), With Box Plots Showing the Distribution of Patient Scores Over Time

eFigure 4. Trajectories of Model-Predicted Changes in PROMIS Pain Interference Scores After TKI Discontinuation (A, n = 172) and After Restarting a TKI (B, n = 57), With Box Plots Showing the Distribution of Patient Scores Over Time