Does a rise in the BCR-ABL1 transcript level identify chronicphase CML patients responding to imatinib who have a high riskof cytogenetic relapse?

Imatinib induces a complete cytogenetic response (CCyR) in

the majority of chronic myeloid leukaemia (CML) patients

who receive this drug as first line therapy but, despite initial

good responses, some patients relapse (Druker et al, 2006; de

Lavallade et al, 2008). Because real-time quantitative polymer-

ase chain reaction (RQ-PCR) can measure the leukaemia cell

burden up to three orders of magnitude below the level that

corresponds to CCyR, it is normally used to monitor patients

who have achieved CCyR. Thus, with RQ-PCR it should be

possible to identify a rise in BCR-ABL1 transcript levels before

a patient loses his/her cytogenetic response, which should

allow earlier and more effective intervention (Cross et al,

1993). The use of RQ-PCR was well established for monitoring

patients after stem cell transplantation (Kaeda et al, 2006), and

is now widely used to monitor patients treated with imatinib

(Druker et al, 2006; de Lavallade et al, 2008). Here we present

our experience with the use of molecular monitoring in 161

patients who started imatinib early after diagnosis and

achieved CCyR.

Subjects and methods

Between June 2000 and May 2007, 224 consecutive adult

patients with BCR-ABL1-positive CML in chronic phase (CP)

received imatinib 400 mg daily at the Hammersmith Hospital

as first line therapy; 161 achieved CCyR during follow up and

constitute the study population. The imatinib dose was

adjusted according to tolerance and response (de Lavallade

et al, 2008). BCR-ABL1 transcripts were measured in the blood

at 6–12 week intervals using RQ-PCR (de Lavallade et al,

2008). Results were expressed as per cent ratios relative to an

ABL1 internal control or as log10 reductions compared with a

standardised median value for the 30 untreated patients that

were originally treated in the IRIS (International Randomized

Study of IFN and STI571) study (Druker et al, 2006; de

Lavallade et al, 2008). Major molecular response (MMR) and

complete molecular response (CMR) were computed as

defined elsewhere (Druker et al, 2006; de Lavallade et al,

2008).

Changes in the transcript level were only considered in this

analysis (see below) if they were also seen in a subsequent

sample, thereby constituting a ‘confirmed increase’ above the

nadir. In analysing fold differences between time points we

excluded any contribution of intra-laboratory variations by

showing that changes in ABL1 and BCR-ABL1 measurements

(as total numbers of molecules per test) never accounted for

more than a 10% difference (equivalent to 0Æ1–0Æ2 of Ct values)

between runs and tests. Samples obtained for RQ-PCR were

also analysed for kinase domain (KD) mutations on a routine

basis (Khorashad et al, 2008; de Lavallade et al, 2008).

David Marin, Jamshid S. Khorashad,

Letizia Foroni, Dragana Milojkovic,

Richard Szydlo, Alistair G. Reid,

Katayoun Rezvani, Marco Bua, John M.

Goldman and Jane F. Apperley

Department of Haematology, Hammersmith

Hospitals Trust, Imperial College London,

London, UK

Received 22 December 2008; accepted for

publication 29 January 2009

Correspondence: Dr David Marin, Department

of Haematology, Imperial College London, Du

Cane Road, London, London W12 0NN, UK.

E-mail: d.marin@imperial.ac.uk

Summary

BCR-ABL1 transcript numbers were monitored in 161 patients who started

treatment with imatinib early after diagnosis of chronic myeloid leukaemia

in chronic phase and achieved complete cytogenetic responses (CCyR).

A confirmed doubling in BCR-ABL1/ABL1 transcript levels was found to be a

significant factor for predicting loss of CCyR [relative risk (RR) 8Æ3,

P < 0Æ0001] and progression to advanced phase (RR 0Æ07, P = 0Æ03) provided

that the eventual BCR-ABL1/ABL1 transcript level exceeded 0Æ05%; increases

that never exceeded 0Æ05% had no predictive value. The finding of a kinase

domain mutation in a patient in CCyR, though rare, also predicted for loss of

CCyR.

Keywords: chronic leukaemia, imatinib, molecular studies.

short report

First published online 12 March 2009ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 145, 373–375 doi:10.1111/j.1365-2141.2009.07646.x

Results

The median follow up from starting imatinib was 48 months

(range 12Æ5–95). The 5-year overall survival (OS) and progres-

sion-free survival (PFS) were 96Æ2% and 94Æ4%, respectively. The

median time to achieve CCyR was 6Æ5 months (range 2Æ5–55Æ2);

145 patients (90%) had achieved CCyR by 19Æ5 months. Ninety-

four patients achieved MMR and 15 achieved CMR. During

follow up the CCyR, MMR and CMR were lost by 18 (11%), 10

(10Æ6%) and 6 (40%) patients, respectively. Six patients

developed KD mutations (T315I, F359V, M351T, S417F,

L387M, E459K) while in CCyR. These six patients all lost their

CCyR and one progressed to blastic phase (E459K). The median

time between the detection of mutation and loss of CCyR was

16Æ5 months (range 1–38Æ5). The median transcript level at the

time of first detection of the mutation was 0Æ4% (0Æ3–18%). The

5-year cumulative incidence of KD mutations only recognised

after achieving CCyR was 4Æ0%.

In 36 patients the transcript level continued to decline during

the follow up. In the remaining 125 patients there was at least one

confirmed increase in the transcript level during the follow up.

We wanted to determine the smallest increase in the transcript

level that predicted for loss of cytogenetic response and

progression to advanced phase. Because all patients in CCyR

who lose their response will have a rise in the transcript levels, we

considered only the first confirmed increase in the BCR-ABL1/

ABL1 ratio. Only RQ-PCR results obtained during CCyR were

considered in this analysis and results obtained after dose

reduction or imatinib interruption were excluded.

First, we hypothesised that any increase in the transcript

level that did not reach a BCRBCR-ABL1/ABL1 ratio of 0Æ05%

(e.g. from 0Æ008% to 0Æ03%) might have no prognostic value.

This was the value below which increases in transcripts

numbers in the transplant setting seemed to have no clinical

significance (Kaeda et al, 2006). Thus the increases in the

transcript levels were classified according to whether or not the

actual increase exceeded this 0Æ05% threshold. Patients in

whom the BCRBCR-ABL1/ABL1 ratio had doubled and had

risen above the 0Æ05% threshold (e.g. from 0Æ03% to 0Æ06%)

had a relative risk (RR) for loss of CCyR of 8Æ3 (P < 0Æ0001),

whereas patients with increases by a factor of two in the ratio

that did not exceed this threshold had a RR for loss of CCyR

that was no different from that of patients with no increase at

all, namely RR = 0Æ1 (P = 0Æ98). Similar results were found

when we considered rises by a factor of three (crossing the

0Æ05% threshold RR = 12Æ8, P < 0Æ0001, not crossing

RR = 0Æ1, P = 0Æ99) and rises by a factor of five (crossing the

0Æ05% threshold RR = 21Æ0, P = <0Æ0001, not crossing

RR = 0Æ1, P = 0Æ98). Thus for the following analysis we

considered only increases in the ratio that crossed the 0Æ05%

threshold (defined as significant increases).

The increases in transcript levels that we observed preceded

the loss of CCyR by a considerable period of time (Table I).

Significant increases in the BCR-ABL1 transcript levels pre-

dicted for PFS (Table I). Patients with a significant increase in

transcript levels by factors of two, three and five had RRs for

PFS of 0Æ07 (P = 0Æ03), 0Æ05 (P = 0Æ01) and 0Æ02 (P = 0Æ002),

respectively. However, significant increases in BCR-ABL1

transcripts did not predict for OS (data not shown).

In our multivariate analysis the development of KD

mutations (RR = 12Æ0 P < 0Æ0001) and a threefold increase

in the transcript level (RR = 10Æ2, P < 0Æ0001) were the only

independent predictors for loss of CCyR. In order to obtain a

predictive model that could be used earlier in a patient’s

clinical course we repeated the multivariate analysis including

a twofold increase in the transcript level. The development of

KD mutations (RR = 14Æ4, P < 0Æ0001) and a twofold increase

(crossing the 0Æ05% boundary) in the transcript levels

(RR = 6Æ9, P = 0Æ001) were the only independent predictors

for loss of CCyR.

Discussion

We have shown that the emergence of KD mutations in

patients with no other signs of resistance and a tripling (or

doubling) in the transcript level are the only independent

predictors for loss of CCyR and progression to advanced

phase. Systematic screening for the presence of KD mutations

might therefore be useful but its cost effectiveness is dubious.

Increases in the transcript levels that did not cross the 0Æ05%

threshold were found to have no prognostic value, while

increases that did cross that level strongly predicted for loss of

response and progression (Table I). We tried to identify the

magnitude of increase in transcript levels that better predicted

for loss of response. An increase by a factor of at least three was

the most predictive rise for loss of CCyR and progression to

advanced phase, but an increase by a factor of two was also

Table I. Relative risks for loss of CCyR and PFS according to a sig-

nificant increase in the BCR-ABL1 transcript level.

Fold increase

in BCR-ABL1

transcript level

RR* for loss

of CCyR (P)

Time from first

transcript increase

to loss of CCyR,

median (months)

(range)

RR* for

PFS (P)

·2 8Æ3 (<0Æ0001) 19 (4Æ5–55Æ7) 0Æ07 (0Æ03)

·3 12Æ8 (<0Æ0001) 17Æ4 (4Æ5–55Æ7) 0Æ05 (0Æ01)

·5 21Æ0 (<0Æ0001) 9Æ5 (1Æ5–24Æ7) 0Æ02 (0Æ002)

This increase was by a factor of at least two in 87 patients, by a factor of

at least three in 67 patients and by a factor of at least five in 42 patients.

In our definition of increase we considered both abrupt increases and

small steady increases from the nadir (both cases requiring confirma-

tion, see text); for example we classified both the following hypothe-

tical sequences of transcript numbers as a threefold increase:

(A): 0Æ2% fi 0Æ2% fi 0Æ7% fi 0Æ8% and (B): 0Æ2% fi 0Æ2% fi0Æ4% fi 0Æ35% fi 0Æ45% fi 0Æ55% fi 0Æ6% fi 0Æ7%.

*The risk is relative to patients who had no increase in transcript levels

or who did have an increase that never exceeded the 0Æ05% threshold

(see text).

Short Report

374 ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 145, 373–375

found to be an independent predictor when the variables

defined by higher increases were excluded from the multivar-

iate analysis, allowing for an earlier identification of patients

who may lose the CCyR.

We confirmed our previous finding that the failure to achieve

a MMR predicted for loss of CCyR (Marin et al, 2008) in the 12-

and 18-month landmark analyses (data not shown), but the

prognostic benefit of molecular responses was lost when the

occurrence of rises in the transcript level was taken into account.

This is consistent with the fact that some patients may improve

their molecular responses very slowly or achieve a stable plateau

with a relatively high transcript level (Marin et al, 2005).

In our series, patients were not changed to second gener-

ation TK inhibitors or subjected to stem cell transplantation

until they had lost a CCyR. In 24 of the 63 patients with a

doubling in the transcript level the dose of imatinib was

increased (15 patients had already had their dose increased to

600 or 800 mg daily before achieving CCyR, and in the

remaining 24 the dose could not be increased or was not

tolerated on account of toxicity). Four of the 24 patients lost

their CCyR and one progressed to advanced phase. The

European LeukemiaNet recommendations (Baccarani et al,

2006) considered an increase in transcript levels as a ‘warning’,

which meant that such patients should be followed more

closely, but a change in therapy was not necessarily warranted.

We have shown that increases in the transcript level identified

those patients with a higher risk of progression some time

before loss of CCyR. It could be argued that once a high risk

patient has been identified treatment should be adjusted

accordingly, particularly if alternative agents with toxicity or

cost little different from those of imatinib are now available.

Acknowledgement

We are grateful for support from the NIHR Biomedical

Research Centre Funding Scheme (UK).

References

Baccarani, M., Saglio, G., Goldman, J., Hochhaus, A., Simonsson, B.,

Appelbaum, F., Apperley, J., Cervantes, F., Cortes, J., Deininger,

M., Gratwohl, A., Guilhot, F., Horowitz, M., Hughes, T.,

Kantarjian, H., Larson, R., Niederwieser, D., Silver, R. &

Hehlmann, R. (2006) Evolving concepts in the management of

chronic myeloid leukemia. Recommendations from an expert

panel on behalf of the European LeukemiaNet. Blood, 108, 1809–

1820.

Cross, N.C., Feng, L., Chase, A., Bungey, J., Hughes, T.P. & Gold-

man, J.M. (1993) Competitive polymerase chain reaction to

estimate the number of BCR-ABL transcripts in chronic myeloid

leukemia patients after bone marrow transplantation. Blood, 82,

1929–1936.

Druker, B., Guilhot, F., O’Brien, S., Gathmann, I., Kantarjian, H.,

Gattermann, N., Silver, R., Goldman, J., Cervantes, F., Hochhaus,

A., Powell, B., Gabrilove, J., Rousselot, P., Reiffers, J., Cornelissen,

J., Hughes, T., Valent, T., Fischer, T., Verhoef, G., Shepherd, J.,

Saglio, G., Nielsen, J.L. & Larson, R. (2006) Five-year follow-up

of imatinib therapy for newly diagnosed chronic myelogenous

leukemia in chronic-phase shows sustained responses and high

overall survival. New England Journal of Medicine, 355, 2408–

2417.

Kaeda, J., O’Shea, D., Szydlo, R.M., Olavarria, E., Dazzi, F., Marin,

D., Saunders, S., Khorashad, J.S., Cross, N.C., Goldman, J.M. &

Apperley, J.F. (2006) Serial measurement of BCR-ABL tran-

scripts in the peripheral blood after allogeneic stem cell trans-

plantation for chronic myeloid leukemia: an attempt to define

patients who may not require further therapy. Blood, 107, 4171–

4176.

Khorashad, J., de Lavallade, H., Apperley, J., Milojkovic, D., Reid,

A., Bua, M., Szydlo, R., Olavarria, E., Kaeda, J., Goldman, J.M. &

Marin, D. (2008) The finding of kinase domain mutations in

chronic phase CML patients responding to imatinib may identify

those at high risk of disease progression. Journal of Clinical

Oncology, 26, 4806–4813.

de Lavallade, H., Apperley, J.F., Khorashad, J.S., Milojkovic, D., Reid,

A.G., Bua, M., Szydlo, R., Olavarria, E., Kaeda, J., Goldman, J.M. &

Marin, D. (2008) Imatinib for newly diagnosed patients with

chronic myeloid leukaemia: incidence of sustained responses in an

intention-to-treat analysis. Journal of Clinical Oncology, 26, 3358–

3363.

Marin, D., Kaeda, J., Szydlo, R., Saunders, S., Fleming, A., Howard,

J., Andreasson, C., Bua, M., Olavarria, E., Rahemtulla, A., Dazzi,

F., Kanfer, E., Goldman, J.M. & Apperley, J.F. (2005) Monitoring

patients in complete cytogenetic remission after treatment of

CML in chronic phase with imatinib: patterns of residual leu-

kaemia and prognostic factors for cytogenetic relapse. Leukemia,

19, 507–512.

Marin, D., Milojkovic, D., Olavarria, E., Khorashad, J.S., de Lavallade,

H., Reid, A.G., Foroni, L., Rezvani, K., Bua, M., Dazzi, F., Pavlu, J.,

Klammer, M., Kaeda, J., Goldman, J.M. & Apperley, J.F. (2008)

European LeukemiaNet criteria for failure or sub-optimal response

reliably identify patients with CML in early chronic phase treated with

imatinib whose eventual outcome is poor. Blood, 112, 4437–4444.

Short Report

ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 145, 373–375 375