Timing of Allogeneic Hematopoietic Cell Transplantation (AlloHCT) for Chronic Myeloid Leukemia (CML) Patients

Abstract

While TKI are the preferred first-line treatment for chronic phase (CP) CML, alloHCT remains an important consideration. The aim is to estimate residual life expectancy (RLE) for patients initially diagnosed with CP CML based on timing of alloHCT or continuation of TKI in various settings: CP1 CML, CP2+ [after transformation to accelerated phase (AP) or blast phase (BP)], AP, or BP. Non-transplant cohort included single-institution patients initiating TKI and switched TKI due to failure. CIBMTR transplant cohort included CML patients who underwent HLA sibling matched (MRD) or unrelated donor (MUD) alloHCT. AlloHCT appeared to shorten survival in CP1 CML with overall mortality hazard ratio (HR) for alloHCT of 2.4 (95% CI 1.2–4.9; p=0.02). In BP CML, there was a trend towards higher survival with alloHCT; HR=0.7 (0.5–1.1; p=0.099). AlloHCT in CP2+ [HR=2.0 (0.8–4.9), p=0.13] and AP [HR=1.1 (0.6–2.1); p=0.80] is less clear and should be determined on a case-by-case basis.

INTRODUCTION:

Tyrosine kinase inhibitors (TKIs) have changed the therapeutic landscape of chronic myeloid leukemia (CML) since the introduction of imatinib mesylate nearly 20 years ago¹. Tolerable side effect profiles and high efficacy of the currently five FDA approved TKIs have improved both quality of life and life expectancy of patients with CML^1–4; consequently, TKIs have replaced interferon and allogeneic stem cell transplant (alloHCT) as first line treatment of patients with CML.

Due to efficacy of available TKIs and the ability to sequence them when one is unsuccessful, life expectancy of newly diagnosed chronic phase (CP) CML patients approaches that of the general population^5,6. However, a small subset of CP CML patients never achieve optimal response to TKIs, and secondary resistance rates in patients initially responding to TKI therapy occur at a rate of 2–4% per year⁷. Second-line therapy in imatinib-resistant patients with second generation TKIs dasatinib, nilotinib and bosutinib yield lower complete cytogenetic response (CCyR) rates of approximately 40–50% ⁸, with even lower rates of major molecular response (MMR). Treatment beyond second TKI failure, although effective with ponatinib, is limited by safety concerns whereas prospective data with other TKI options is limited to bosutinib with major cytogenetic response (MCyR) rates of approximately 40% ^9,10. Additionally, majority of patients with CML cannot be considered cured with TKIs as approximately 40–60% patients relapse after elective TKI discontinuation ^11–13.

Although use of alloHCT has declined significantly since the introduction of TKI, it remains an effective and curative option, and mortality rates have decreased considerably in the last two decades making it an important consideration for some patients with TKI failure ¹⁴. Yet, with increased availability of multiple TKI, timing of alloHCT is undefined. There is decreasing appetite for performing alloHCT before a patient experiences failure of multiple TKI but waiting too long to perform alloHCT may decrease the success rates of transplant as the time from diagnosis to transplant is a recognized adverse prognostic feature. In addition, outcomes in alloHCT are clearly worse in patients with advanced stage CML at the time of transplantation than when performed in CP.¹⁵ In this analysis we aimed to explore the timing of alloHCT using two large databases: one based on TKI therapy and one for alloHCT.

MATERIALS AND METHODS:

Data Source

The CIMBTR (Center for International Blood and Marrow Transplantation Research) includes data from more than 450 international transplant centers. They contribute detailed data on consecutive autologous and allogeneic stem cell transplants. Compliance is monitored by on-site audits and patients are followed longitudinally. Computerized checks for discrepancies, physicians’ review of submitted data, and on-site audits of participating centers ensure data quality. Studies conducted by the CIBMTR are performed in compliance with all applicable federal regulations pertaining to the protection of human research participants. Protected Health Information used in the performance of such research is collected and maintained in CIBMTR’s capacity as a Public Health Authority under the HIPAA Privacy Rule.

CIBMTR collects data at two levels: Transplant Essential Data (TED) level and Comprehensive Report Form (CRF) level. TED-level data is an internationally accepted standard data set that contains a limited number of key variables for all consecutive transplant recipients. TED and CRF level data are collected pre-transplant, 100 days and six months post-transplant, annually until year 6 post-transplant and biannually thereafter until death.

MD Anderson Cancer Center (MDACC) Leukemia Department database includes all leukemia patients treated at MDACC and includes demographics, disease characteristics, treatment modalities with start and stop dates, response, and dates of evaluation. Longitudinal follow-up is conducted for patients per protocol and institutional guidelines. Data are recorded, submitted and checked by physician. Clinical trials are performed in compliance with all applicable federal regulations pertaining to the protection of human research participants. Protected Health Information used in the performance of such research is collected and maintained in MDACC’s capacity under the HIPAA Privacy Rule.

Patient population

We compiled data from these two large databases: MDACC for patients treated with TKI and CIBMTR for patients treated with alloHCT. Patients ages 18–65 were included if they had been diagnosed with Philadelphia chromosome or BCR-ABL1-positive CP CML between 1/1/2001 and 12/31/13 and initiated on TKI therapy within 6 months of diagnosis. Patients with prior alloHCT or auto-HCT were excluded. Patients who switched TKIs for any reason other than failure (as defined by European Leukemia Net 2013 criteria) were excluded in order to have a more objectively defined population. Patients who remained on their initial TKI were also excluded from analysis as they were not deemed to have experienced TKI failure; thus alloHCT would not have been a consideration in the treatment of their CML. Dates and reasons of TKI switch were gathered through chart review. Staging of CML into CP, accelerated (AP), or blast phase (BP) CML was based on standard criteria ^16,17 (Supplemental Table 1).

The alloHCT cohort was extracted from the CIBMTR. Patients ages 18–65 at the time of transplant were included if they had a diagnosis of Philadelphia chromosome or BCR-ABL1-positive CML and underwent HLA sibling matched (MRD) or matched unrelated donor (MUD) alloHCT (myeloablative or reduced intensity/non-myeloablative) between 1/1/2001 and 12/31/2013. Myeloablative regimens are defined as a regimen that produces pancytopenia within 1–3 weeks from administration that would be expected to be irreversible and fatal without hemopoietic stem cell infusion, as previously defined¹⁸. Last follow up date was 12/31/2016. Staging of CP, AP, BP CML was based on WHO Criteria ¹⁹ (Supplemental Table 1).

Study Definitions and Statistical analysis

The decision to treat with TKI or HSCT was made by the treating physician according to practice guidelines and/or institutional algorithms. We used Cox proportional hazards model on the data set combining patients who received and did not receive alloHCT, to estimate the hazard ratios. We used Kaplan-Meier estimator to evaluate survival time distributions, and used integration of Kaplan Meier curves to estimate the residual life-expectancy (RLE) of patients with CP CML at initial diagnosis based on timing of alloHCT in various settings: alloHCT at CP1, at CP2+ (after transformation to AP or BP), at AP, or at BP, or continuation of 2^nd or 3^rd TKI at CP1, CP2+, AP or BP, see Figure 1. That is to say, for a particular group of patients, their RLE (all restricted to 10 years throughout this article) is estimated by the area under their survival curve from 0 to 10. The restriction of 10 years is chosen to be slightly less than the maximum follow-up time. Death from any cause was counted as an event.

Figure 1:

Disease States and Possible Pathways for CML patient

CP=chronic phase

CML=chronic myeloid leukemia

TKI=tyrosine kinase inhibitor

AP=accelerated phase

SCT=stem cell transplant

RESULTS:

A total of 1361 patients were included in the analysis: 138 in the MDACC cohort and 1223 in CIBMTR cohort (Table 1). In the MDACC cohort, 77 (56%) were male and median age was 46 years (range 19–64). In the CIBMTR cohort, 703 (57%) were male and median age was 39 years (range 18–69), P<0.01 for both gender and age. Median follow up time was 96.4 months for the MDACC cohort and 59.6 months for the CIBMTR cohort. Most transplanted patients (n=1029 or 84%) underwent a myeloablative alloHCT.

Table 1:

Patient characteristics of the TKI and SCT cohorts

	MDACC (n=138)	CIBMTR (n=1223)	p-value
Age
Median (range)	46.0 (19, 64)	38.8 (18, 65)	<0.01
18–49	88 (64%)	974 (80%)
50–65	50 (36%)	249 (20%)
Gender			0.72
Male	77 (56%)	703 (57%)
Female	61 (44%)	520 (43%)
Year of Transplant
2001–2004	N/A	534 (44%)
2005–2008		515 (42%)
2009–2013		174 (14%)
Type of Transplant
Matched Related	N/A	686 (56%)
Matched Unrelated		537 (44%)
Nonmyeloablative	N/A	154 (12%)
Myeloablative		1029 (84%)
Missing		40
Number of TKIs used
Median (range)	2 (1–4)	1 (0–3)
Last known stage of			<0.0001
CML
CP1	53 (38%)	738 (60%)
CP2+	12 (9%)	255 (21%)
AP	20 (14%)	148 (12%)
BP	53 (38%)	82 (7%)
Progressed to AP/BP	73 (52%)	230 (19%)

Abbreviations: CML (chronic myeloid leukemia), CP1 (first chronic phase), CP2+ (return to chronic phase after progression to accelerated or blast phase CML), AP (accelerated phase), BP (blast phase)

Reasons for transplant in CP1 are listed in Supplemental Table 2. Half of the patients (n=370) underwent alloHCT due to limited access to TKI (mostly patients from outside United States). Only 1% (n=8) underwent alloHCT in CP1 due to failure of 3 or more lines of therapy. Causes of death were mainly from primary disease (42%) followed by GVHD (17%) and infection 14%), Supplemental Table 3.

Table 2 depicts the effect of transplant and age on survival by univariate analysis. In CP1 alloHCT was associated with worse overall survival compared to change to a 2^nd TKI when there is failure to the first TKI. Multivariate analysis (MVA) confirmed that compared to non-HCT, alloHCT was associated with inferior OS (hazard ratio (HR) of 2.4 (95% CI 1.2–4.9; p=0.02)]). Increased mortality was more pronounced in MUD HCT patients [HR of MUD HCT vs. non-HCT =3.3 (95% CI 1.6–6.7, p=0.001)] than MRD HCT [HR of MRD HCT vs. non-HCT =1.9 (95% CI 0.9–3.8, p=0.1)], Table 3. As compared to non-HCT, alloHCT had a HR of 1.5 (95% CI 1.1–2.1, p=0.02) in the subgroup of patients aged ≥50. Mean RLE was 7.1 years (95% CI 6.8–7.5) if patients underwent alloHCT in CP1 CML vs. 8.5 years (95% CI 7.6–9.5) if they did not (Table 4 and Figure 2a). CP1 CML patients on their 2^nd TKI who developed failure also had an improved survival if they switched to a 3^rd TKI than if they went on to alloHCT. In this setting the HR for alloHCT compared to non-HCT was 2.1 (95% CI 0.5–9.2; p=0.3; Table 3). The corresponding mean RLE was 6.8 years (5.6–8.3) for CP1 CML patients on 2^nd TKI if they underwent alloHCT vs. 8.5 years (6.6–10) continuing on TKI therapy (Table 4).

Table 2:

Univariate Analysis for overall survival

CP 1 (n=791)	Hazard Ratio	95% CI	P-value
Transplant Status
No transplant (n=53)	1.000
Transplant (n=738)	2.247	1.109, 4.552	0.025
Matched Related Transplant (n=459)	1.730	0.841, 3.556	0.136
Matched Unrelated Transplant (n=279)	3.052	1.489, 6.254	0.002
Age≥50 (n=114)	1.389	0.988, 1.950	0.058
CP 1 with 1 TKI (n=477)	Hazard Ratio	95% CI	P-value
Transplant Status
No transplant (n=41)	1.000
Transplant ((N=436)	2.490	1.099, 5.643	0.028
Matched Related Transplant (n=231)	2.169	0.936, 5.025	0.136
Matched Unrelated Transplant (n=205)	2.818	1.226, 6.473	0.014
Age≥50 (n=69)	1.330	0.863, 2.050	0.196
CP 1 with 2 TKI (n=59)	Hazard Ratio	95% CI	P-value
Transplant Status
No transplant (n=12)	1.000
Transplant (n=47)	2.180	0.501, 9.487	0.299
Matched Related Transplant (n=12)	0.859	0.120, 6.107	0.88
Matched Unrelated Transplant (n=14)	2.792	0.634, 12.291	0.175
Age≥50 (n=17)	0.704	0.231, 2.138	0.536
CP 2+ (n=267)
Transplant Status
No transplant (n=12)	1.000
Transplant (n=255)	2.037	0.833, 4.981	0.119
Matched Related Transplant (n=119)	1.862	0.745, 4.651	0.183
Matched Unrelated Transplant (n=136)	2.183	0.883, 5.398	0.091
Age≥50 (n=89)	0.915	0.644, 1.300	0.620
AP (n=168)
Transplant Status
No transplant (n-=20)	1.000
Transplant (n=148)	1.074	0.569, 2.028	0.825
Matched Related Transplant (67)	0.977	0.491, 1.947	0.948
Matched Unrelated Transplant (n=81)	1.159	0.596, 2.256	0.662
Age≥50 (n=47)	1.085	0.685, 1.715	0.729
BP (n=135)
Transplant Status
No transplant (n=53)	1.000
Transplant (n=82)	0.717	0.490, 1.050	0.087
Matched Related Transplant (n=41)	0.756	0.476, 1.199	0.235
Matched Unrelated Transplant (n=41)	0.686	0.440, 1.070	0.096
Age≥50 (n=51)	1.405	0.956, 2.066	0.083

Table 3:

Multivariate Analysis for overall survival

CP1 (n=791)	Hazard Ratio	95% CI	P-value
Transplant (n=738)	2.401	1.182, 4.879	0.015
Matched Related Transplant (n=459)	1.850	0.897, 3.812	0.095
Matched Unrelated Transplant (n=279)	3.274	1.592,6.730	0.001
Age≥50 (n=114)	1.484	1.055, 2.088	0.023
CP1 with 1 TKI (n=477)
Transplant (n=436)	2.623	1.154, 5.963	0.021
Matched Related Transplant (n=231)	2.279	0.981, 5.291	0.055
Matched Unrelated Transplant (n=205)	2.991	1.297, 6.897	0.010
Age≥50 (n=69)	1.444	0.934, 2.230	0.097
CP1 with 2 TKI (n=59)
Transplant (n=47)	2.116	0.484,9.241	0.319
Matched Related Transplant (n=14)	0.862	0.120,6.145	0.882
Matched Unrelated Transplant (n=33)	2.773	0.619, 12.418	0.182
Age≥50 (n=17)	0.966	0.306, 3.047	0.954
CP2+ (n=267)
Transplant (n=255)	2.010	0.817, 4.942	0.128
Matched Related Transplant (n=119)	1.841	0.733, 4.623	0.193
Matched Unrelated Transplant (n=136)	2.156	0.867, 5.365	0.098
Age≥50 (n=89)	0.959	0.674, 1.367	0.819
AP (n=168)
Transplant (n=148)	1.098	0.576, 2.093	0.776
Matched Related Transplant (n=67)	0.998	0.498, 2.001	0.996
Matched Unrelated Transplant (n=81)	1.193	0.606, 2.349	0.610
Age≥50 (n=47)	1.111	0.697, 1.772	0.657
BP (n=135)
Transplant (n=82)	0.726	0.496, 1.062	0.099
Matched Related Transplant (n=41)	0.793	0.498, 1.261	0.327
Matched Unrelated Transplant (n=41)	0.677	0.434, 1.056	0.085
Age≥50 (n=51)	1.414	0.957, 2.089	0.082

Table 4:

Mean Residual Life by treatment in the various CML stages

		SCT				TKI
	n	Mean Residual Life	95% CI		n	Mean Residual Life	95% CI
CP1	738	7.136	6.789, 7.502		53	8.484	7.560, 9.520
CP1 with 1 TKI	436	7.044	6.604, 7.518		41	8.579	7.588, 9.681
CP1 with 2 TKI	47	6.825	5.621, 8.305		89	8.453	6.646, 10
CP2+	255	4.668	4.084, 5.346		12	7.382	5.557, 9.709
AP	148	5.313	4.501, 6.305		20	5.608	3.956, 8.245
BC	82	2.571	1.807, 3.756		53	1.536	0.914, 2.826

Figure 2:

Overall survival for various stages of CML

CML = chronic myeloid leukemia

CP1= chronic phase 1

CP2+ = chronic phase 2+

AP=accelerated phase

BC=blast crisis

For patients with CML in CP2+ there is also an advantage for those receiving TKI. On MVA, The HR was 2.0 (95% CI 0.8–4.9, p=0.13) for alloHCT compared to non-HCT, Table 3 and Figure 2d. HR was 1.8 (95% CI 0.7–4.6, p=0.19) and 2.2 (95% CI 0.9–5.4; p=0.10) for MRD and MUD HCT, respectively as compared to non-HCT. Mean RLE was 4.7 (95% CI 4.1–5.3) years for patients undergoing alloHCT in CP2+ CML compared to 7.4 (95% CI 5.6–9.7) years with TKI continuation, Table 4.

The outcome for patients with AP CML was not significantly different for patients treated with TKI or with alloHCT, Figure 2e. The HR was 1.1 for alloHCT (95% CI 0.6–2.1; p=0.8) compared to non-HCT. The corresponding mean RLE was 5.3 (95% CI 4.5–6.3) years for AP patients undergoing alloHCT and 5.6 years (95% CI 4.0–8.2) for those continuing with TKI, Table 4.

Among patients in BP CML, there was a trend towards improved survival with alloHCT compared to continuation of TKI therapy. By MVA the HR was 0.7 (95% CI 0.5–1.1; p=0.099) for alloHCT compared to non-HCT for the total population. The corresponding HR was 0.8 (95% CI 0.5–1.3; p=0.33) for MRD HCT and 0.7 (95% CI 0.4–1.0; p=0.08) for MUD HCT compared to non-HCT. The mean RLE for patients with BP CML was 2.6 (95% CI 1.8–3.8) years for the patients who underwent alloHCT in BP compared to 1.5 (95% CI 0.9–2.8) years for those who continued their TKI. There seems to be more long term survivors with transplant as the tail end of the survival curve is higher for the alloHCT arm compared with TKI continuation (Figure 2f).

DISCUSSION:

While alloHCT offers a curative option for a significant percentage of patients with CML, risks of early treatment related morbidity and long-term complications such as graft-versus-host disease have tampered its use in recent years, considering the excellent results achieved with TKI as initial therapy and the availability of several inhibitors that may still offer clinical benefit after failure of an initial TKI attempt. The ELN recommendations offer suggestions of the recommended timings for searching for a donor and performing HCT. However, these recommendations are based on personal experience and small series. To date, there have been no studies to evaluate the optimal timing of performing alloHCT in patients with CML who fail one or more TKIs.

Our analysis suggests that in early stages of the disease TKI therapy offers a survival benefit compared to alloHCT. This benefit remains nearly constant even after failure of a second TKI as long as the patient remains in CP. The survival curves in Figure 2 seem to favor TKI continuation over both MRD and MUD transplant for patients in CP1 and there is a trend of favoring TKI continuation in CP2+ CML. This survival advantage observed with non-HCT therapies appears to diminish as the disease progresses to AP (Figure 2), suggesting that a multitude of factors including sensitivity to remaining TKIs, affordability and availability of TKIs (significantly influenced by country of residence), comorbidities, toxicity of transplant regimens and subsequent complications from transplant influence the outcomes and survival. There seems to a survival advantage with HCT for those who progressed to BP CML as evidenced by higher proportion of survivors who underwent alloHCT at the tail end of the survival curve (Figure 2f), but perhaps given the small sample size this finding was not statistically significant. Both MRD and MUD alloHCT seem to benefit survival for those with BP CML over TKI continuation.

Since benefit for patients with AP is less clear, the choice of HCT versus TKI continuation should be determined on a case-by-case basis, dependent on mutation status, age and comorbidities of the patient, type and availability of donor used, and remaining TKI options available. Interestingly, the survival curves for alloHCT and TKI continuation (Figure 2e) in AP cross. Initially, survival is lower for the alloHCT cohort, likely due to transplant-related mortality. However, there is a subset who achieve long term survival likely due to disease control with alloHCT, as evidenced by the tail end of the curve. These results are similar to those reported in a series by Nicolini et al specifically for patients with T315I treated with ponatinib ²⁰. Patients with CP treated with ponatinib had a superior overall survival compared to those who received HCT. For patients in AP there was no difference between the two cohorts, whereas for the BP patients receiving HCT, while still doing very poorly, had a better overall survival compared to patients treated with ponatinib.

Our analysis suffers from limitations that should be taken into account when considering the results. First, this was a retrospective study and there may have been inherent biases between the non-transplant and transplant group that we are unable to determine. The patients referred to alloHCT may have had more aggressive disease by variables that are difficult to account for in a registry series. Additionally, availability and affordability of TKIs may not be equal in all countries and account for differences in transplant referrals in CP1. Prioritization therefore should be adapted to available options and financial considerations that may affect desirability of TKI and/or SCT in specific circumstances. In addition, patients in both cohorts may be inherently different than the patients treated in the community, who may have more comorbidities or a less favorable disease biology (e.g., Sokal risk score)²¹.

Furthermore, the number of patients in the non-transplant cohort were small and thus comparisons between TKI continuation vs. alloHCT for the CP2+, AP and BP groups were not statistically significant and the comparison may have been underpowered to detect a noticeable difference in overall survival. This cohort is also derived from a single institution with a dedicated CML clinic and considerable expertise where the results may not be representative of what can be expected in other settings. The reported median survival for CML patients in the SEER database for example is considerably shorter than what is achieved at MDACC ^6,22. Although some of this may be related to access to TKI, factors such as patient characteristics and the local expertise may also contribute.

It is notable that transplant results were quite good for match related donors and inferior survival for transplant may be the result of decreased survival in matched unrelated donor transplants. Moreover, 84% of patients in the alloHCT cohort underwent a myeloablative regimen. More recent retrospective studies have demonstrated that nonmyeloablative regimens for CML are well tolerated with less transplant related mortality and non-inferior for disease related mortality, particularly in CP^23,24. The survival outcomes for AP and BP CML with a nonmyeloablative regimen compared with TKI continuation will need to be studied in the future.

Additionally, the staging in the non-transplant and transplant cohort were different. Stage of CML in non-transplant cohort was determined using the standard criteria used in all of the pivotal TKI studies, while the stage of CML in the transplant cohort more closely aligned with the WHO criteria. We did not change the staging of the nontransplant cohort based on WHO criteria as treatment decisions were made based on standard rather than WHO criteria. In addition, the subjectivity of some of the WHO criteria (e.g., increasing spleen size and increasing WBC unresponsive to therapy) make it difficult to apply retrospectively. This introduces an uneven variable that cannot be resolved in an analysis such as ours. This “stage migration” phenomenon has been described in the literature whereby classifying patients based on WHO criteria improves their outcomes without any therapeutic intervention due to differences in definition of stage of CML^25–27. However, when we reclassified the CIBMTR cohort based on the percentage of bone marrow blasts to better align with the standard criteria, only 6 of 137 patients were reclassified from BP to AP and only 10 of 279 patients were reclassified from AP to CP, representing less than 5% of the entire cohort.

Of note, we first aimed to use Markov Decision Analysis model using the TreeAge decision modeling software program. However, the Markov model is inadequate for this analysis due to the fundamental assumption in Markov decision analyses that any outcome of a particular disease state in any cycle is only dependent on the properties of the current cycle and has no “memory” of prior cycles²⁸. Hence, the Markov model underestimated life expectancy in the alloHCT patient as the transition probabilities calculated for each cycle length (e.g. 3-, 6-, 12- month cycles) were based on survival data in the first year following alloHCT. Because transplant-related mortality is greater during the first year (10–20% chance of death), and chance of death decreased significantly after the first year with at least 50% of patients cured of CML, the probability of death of alloHCT patients was overestimated as the Markov Decision Analysis does not account for the decreasing probability of death overtime after alloHCT. We thus used the integration of Kaplan Meier curves to calculate life expectancy, which may provide a more realistic representation of expected outcomes in this setting.

In conclusion, our analysis may help guide treatment decisions in the earlier and late stages of CML. TKI may result in better outcome in CP1 and HCT in BP. Decisions for intermediate stages (CP2+ and AP) are less clearly guided. Despite a trend for CP2+ in favor of TKI and null effect in AP, the caveats mentioned earlier suggest that decisions should be made considering in addition factors such as mutation status, availability of TKI and donor, comorbidities, patient characteristics, conditioning regimen and cost. Although prospective randomized trials would ideally help resolve these questions, such studies are unfortunately unlikely to be conducted.