Minimal residual disease is a significant predictor of treatment failure in non T-lineage adult acute lymphoblastic leukaemia: final results of the international trial UKALL XII/ECOG2993


Professor Letizia Foroni, Department of Haematology, Imperial College London, Hammersmith Hospital Campus, London W12 0NN, UK. E-mail:


The predictive value of molecular minimal residual disease (MRD) monitoring using polymerase chain reaction amplification of clone-specific immunoglobulin or T-cell Receptor rearrangements was analysed in 161 patients with non T-lineage Philadelphia-negative acute lymphoblastic leukaemia (ALL) participating in the UK arm of the international ALL trial UKALL XII/Eastern Cooperative Oncology Group (ECOG) 2993. MRD positivity (≥10−4) in patients treated with chemotherapy alone was associated with significantly shorter relapse-free survival (RFS) at several time-points during the first year of therapy. MRD status best discriminated outcome after phase 2 induction, when the relative risk of relapse was 8·95 (2·85–28·09)-fold higher in MRD-positive (≥10−4) patients and the 5-year RFS 15% [95% confidence interval (CI) 0–40%] compared to 71% (56–85%) in MRD-negative (<10−4) patients (P = 0·0002) When MRD was detected prior to autologous stem cell transplantation (SCT), a significantly higher rate of treatment failure was observed [5-year RFS 25% (CI 0–55%) vs. 77% (95% CI 54–100%) in MRD-negative/<10−4, P = 0·01] whereas in recipients of allogeneic-SCT in first complete remission, MRD positivity pre-transplant did not adversely affect outcome. These data provide a rationale for introducing MRD-based risk stratification in future studies for the delineation of those at significant risk of treatment failure in whom intensification of therapy should be evaluated.

Studies investigating the prognostic importance of minimal residual disease (MRD) in adults with acute lymphoblastic leukaemia (ALL) have focused mainly on its value in the protracted consolidation/maintenance chemotherapy treatment setting (Vidriales et al, 2003; Bruggemann et al, 2006). However, consolidation of remission in adult patients has frequently involved autologous (auto)-haematopoietic stem cell transplantation (SCT) or allogeneic (allo)-SCT), as indicated by published trials (Sebban et al, 1994; Attal et al, 1995; Ribera et al, 1998; Hunault et al, 2004; Thomas et al, 2004). Despite recurrent investigation of these therapeutic approaches by adult cooperative groups worldwide, few studies have examined the significance of MRD in these therapeutic settings. Those reported are hampered by the inclusion of heterogeneous patient/disease populations, which limits interpretation (Miglino et al, 2002; Spinelli et al, 2007).

The largest study to examine the role of high dose therapy for consolidation remission in adult ALL was conducted by the UK Medical Research Council (MRC) and the Eastern Cooperative Oncology Group (ECOG) study groups. A comparative analysis of outcome for the three assigned post-remission therapies (consolidation/maintenance chemotherapy, auto-SCT and allo-SCT) in the Philadelphia chromosome (Ph)-negative group showed a significantly lower event-free survival (EFS) in those randomised to auto-SCT compared to conventional protracted consolidation/maintenance chemotherapy whilst a donor versus no donor analysis disclosed >8% superior survival at 5 years in those assigned allo-SCT in first complete remission (CR) (Goldstone et al, 2008). Despite the intensity of stem cell transplantation, relapse occurred in over 20% of patients following allotransplantation and over half of auto-SCT recipients (Goldstone et al, 2008).

An interim report of molecular MRD studies on UKALL XII indicated the importance of sensitive measures of disease response in predicting relapse in patients with B lineage ALL (Mortuza et al, 2002). In the study presented herein we extend these initial observations in a much larger cohort of patients and further define the predictive times for MRD testing that will form the basis for prospective assignment of risk groups in future trials.


Patients and treatment protocol

UKALL XII entrants registered between 1993 and 2006 were selected for study on the basis of (i) non T lineage ALL, (ii) Ph-negative disease and (iii) availability of an adequate sample. Those patients entering the MRD study but without a clonal Ig/TCR rearrangement identified at diagnosis or if the only rearrangement identified achieved a sensitivity of <10−3 were subsequently excluded. The MRD study participants were entered from 84 of the 106 participating UK centres. An outline of the UKALL XII/ECOG 2993 study, eligibility criteria and specific details of treatment components have been published elsewhere (Rowe et al, 2005; Goldstone et al, 2008).

Sample timing

Scheduling of bone marrow samples for MRD assessment followed haematopoietic recovery at pre-defined stages during the patient’s therapy: after phase 1 and 2 induction and after completion of the intensification treatment block (Fig 1). Patients assigned to protracted consolidation/maintenance chemotherapy were scheduled for further MRD surveillance at 28 and 39 weeks from start of therapy and thereafter 6 monthly during maintenance treatment. In auto-SCT recipients a bone marrow harvest specimen was requested for MRD analysis.

Figure 1.

 MRD sampling schedule and schema of UKALL XII treatment protocol. TP, time-point.

Molecular evaluation of MRD

Diagnostic specimens (bone marrow and peripheral blood) were processed for DNA extraction following mononuclear cell preparation. Where the material available was a slide preparation, DNA was extracted according to established methods (Mortuza et al, 2002). DNA amplifiability was assessed by ACTB amplification as previously described (Mortuza et al, 2002) or real time quantitative polymerase chain reaction (RQ-PCR) amplification of the ALB control gene.

Clonal rearrangements of the T cell Receptor (TCR) and immunoglobulin (Ig) genes were identified in diagnostic samples using standard primers and cold or radionucleotide-incorporated PCR techniques (Chim et al, 1996). Complete rearrangements of the IGH@ in addition to rearrangements involving the TRG@ and TRD@ were screened throughout the study. Clonal rearrangements of IGKDEL (van der Velden et al, 2002) and incomplete rearrangements of IGH (IGHD@–IGHJ@) genes (Pongers-Willemse et al, 1999) were added to the screening panel from the beginning of 2005. Following gene sequencing, allele specific oligonucelotides (ASOs) were designed complementary to the junctional regions of each identified Ig/TCR clone and sensitivity and specificity for clone detection determined by logarithmic dilution of diagnostic DNA in normal mononuclear cell DNA. Evaluation of MRD in follow-up samples was performed by clone specific PCR amplification of Ig and TCR gene rearrangements using either a semi-quantitative radionucleotide fingerprinting approach (Mortuza et al, 2002) (n = 94) or ASO based RQ-PCR (n = 67). RQ-PCR detection of Ig/TCR clones was performed on the Lightcycler using SYBR green I (Roche Diagnostics Ltd. Burgess Hill, UK) or on the Taqman Applied Biosystem 7500 platform (Applied Biosystems, Foster City, CA, USA) using Taqman dual labelled probe chemistry. Concordance between both RQ-PCR systems and ASO-PCR for molecular MRD quantification has already been demonstrated (Nakao et al, 2000; Eckert et al, 2003).

Samples that were negative by DNA fingerprinting were confirmed with the more sensitive (1:10000) ASO technique (Chim et al, 1996). Individual ASO quantitative thresholds could be assigned for 57 of the 67 patients tested by RQ-PCR technique. In 56 of 57 patients (98%) the quantitative range (QR) of at least one marker reached 10−3 or less with a QR of 10−4 reached in 37/57 (65%) cases. Evaluation of MRD in follow-up samples was performed only if ALB/ACTB amplification was equivalent to 10 000 cells and if CR1 had been confirmed. RQ-PCR generated MRD data was interpreted according to the European Study Group on MRD detection in ALL (ESG-MRD-ALL) guidelines (van der Velden et al, 2007) with the interpretation of MRD positivity based on ESG criteria for clinical protocols aimed at therapy reduction [The cycle threshold (Ct) value of at least one of the replicates ≥1 lower than the lowest Ct of the background].

Real time quantitative polymerase chain reaction-determined MRD levels in follow-up samples were corrected according to RQ-PCR amplification of the ALB. A level of ≥10−4 was selected as the threshold value for defining MRD positivity in line with proposals for European consensus guidelines (MRD symposium, Kiel, September 2008, unpublished observations). In the case of MRD RQ-PCR assays with a QR <10−4, MRD positivity was also assigned where a reproducible PCR signal was detected outside the assigned QR. Where MRD was quantified using two ASOs, the higher level was considered for clinical correlation. MRD negative results were assigned where no PCR amplification was observed regardless of the sensitivity of the applied target or where RQ-PCR determined disease levels were <10−4.

Clinical data

Clinical data including patient and disease characteristics at diagnosis, treatment progress, allocated therapy and clinical outcomes for patients entered into the MRD study were obtained from the Clinical Trial Service Unit computerised patient database held in Oxford.

Risk classification

High risk was defined as age >35 years and/or diagnostic white blood cell (WBC) count ≥30 × 109/l in line with validated adverse prognostic factors for relapse on UKALL XII/ECOG 2993 (Rowe et al, 2005; Moorman et al, 2007).

Statistical analysis

The relationships between grouped variables were analysed using the chi-square test or Fisher’s exact test (for 2 × 2 tables with small expected numbers), whereas the Mann–Whitney test was used to compare continuous variables such as age and WBC count. The main analysis was of relapse-free survival (RFS) defined as the time from diagnosis to relapse censoring at death in remission. For patients without an event, observation was censored at the cut off date of 31 October 2008 or, for those lost to follow-up before this date, the last contact date. Actuarial event percentages were calculated by the Kaplan–Meier method and comparisons between groups used the log rank test. The relative risk for relapse in patients with residual disease compared to those with undetectable residual disease was estimated by the one step odds ratio. A P value of <0·05 was used as the cut-off for statistical significance. All analysis include the 84 UKALL XII entrants previously reported (Mortuza et al, 2002).


Patient characteristics and response to therapy

One hundred and sixty-one patients with non T-lineage ALL were monitored for MRD evaluation during the study period. The median follow-up of the 79 study patients in continuous clinical remission was 58 months. Sixty-one patients relapsed during the observation period (isolated bone marrow in 53 cases, isolated extramedullary sites in four cases, and combined in four). The median time to relapse was 15 months (range 3–61 months). There were 21 deaths in remission. Following remission induction, 94 patients received protracted consolidation/maintenance chemotherapy, 25 underwent auto-SCT and 40 received allo-SCT. Two patients received off protocol treatment and were therefore excluded from subsequent analyses. A median of two samples per patient were analysed in the six time-periods (post-phase 1 induction, post-phase 2 induction, post-intensification, 6–9, 9–12 and 12–24 months).

Comparison with non-MRD studied cohort

Since the study cohort was a sub-section (161/813) of the non-T lineage patients with Ph-negative disease who entered CR, comparison of patient characteristics and distribution of prognostic factors between the MRD study subjects and the remaining UKALL XII patients without MRD studies was carried out (Table I). Although there was no difference in the median age between groups a notable finding was the higher participation of patients aged under 35 years in the MRD study compared to the 652 non-participants (P for heterogeneity = 0·003). This could be attributed to greater participation in the 20–29 years age group (42% MRD study vs. 25% non-participants) and probably reflects a selection bias. Both overall survival and EFS were significantly higher for MRD study participants, a finding that was not fully explained by the younger age constitution of this group because outcome analysis accounting for age was still significant for EFS (P = 0·03) with overall survival being marginally significant. The improved outcome of the MRD group was thus likely to reflect a bias for surviving patients in whom MRD could be performed.

Table I.   Clinical characteristics in Ph-negative non-T lineage UKALL XII patients.
CharacteristicUKALL XII patients with MRD studies
= 161
UKALL XII patients without MRD studies
N = 652
P value
  1. *Chi-square test.

  2. †Mann–Whitney test.

  3. ‡Where immunophenotyping was unknown B lineage disease was assigned on the basis of a complete IGH and/or IGK rearrangement (n = 9) for the current analysis.

  4. §Includes mixed immunophenotype.

  5. ¶Number of patients with specific abnormality out of number with successful cytogenetic FISH or PCR studies.

  6. **Other abnormalities may occur in conjunction with the above abnormalities.

  7. ††Actuarial % at 5 years (95% CI).

  8. ‡‡Log rank test.

Gender (%)
 Male93 (58%)397 (61%)0·5*
 Female68 (42%)255 (39%) 
Median age, years (range)25 (15–55)28 (14–55)0·1†
 <20 years36 (22%)180 (28%)0·0001*
(chi-square test for trend = 0·08)
 ≥20 years68 (42%)164 (25%)
 ≥30 years28 (17%)116 (18%) 
 ≥40 years29 (18%)192 (29%) 
Median WBC (×109/l)
(25th, 75th percentiles)
11 (5,40)9 (3,36)0·02†
WBC (×109/l)
 <1074 (46%)340 (52%)0·3*
 10–4953 (33%)178 (27%)
 ≥5034 (21%)134 (21%)
 Precursor B-ALL‡156 (97%)580 (89%)0·4*
 Other§5 (3%)17 (3%)
 Unknown055 (8%)
 t(4;11) 15/13737/5040·2*
 Complex (>5 abnormalities)  5/12024/4360·6*
 Hypodiploidy/near triploidy  2/12025/4450·07*
Adverse factors
 Age ≤ 35 years126 (78%)411 (63%)0·0003*
 Age > 35 years35 (22%)241 (37%) 
 WBC < 30 × 109/l116 (72%)475 (73%)0·8*
 WBC ≥ 30 × 109/l45 (28%)177 (27%) 
 Overall survival51% (43–60%)43% (39–47%)0·02‡‡
 Event-free survival47% (39–55%)39% (35–43%)0·007‡‡
 Risk of relapse43% (34–51%)50% (46–54%)0·07‡‡
 Risk of death in remission18% (10–26%)22% (18–26%)0·03‡‡

Correlation between MRD and clinical outcome

Chemotherapy and auto-SCT treatment groups.  There was a statistically significant correlation between MRD status and RFS at most time-points measured (Fig 2). At completion of Phase 1 induction, 5-year RFS was 69% [95% confidence interval (CI) 55–83%] in MRD negative/<10−4 patients and 42% (23–61%) in MRD positive patients; P = 0·03 (Fig 2A) with a 2·36-fold (95% CI 1·11–5·04) increase in relative relapse rates compared with MRD negative/<10−4 patients. The relationship between MRD status and RFS became progressively stronger at later time-points (Fig 2B,C). The relative risk (RR) of relapse (95% CI) of MRD positive patients compared to MRD negative/<10−4 patients at later time-points were: post-induction 2, 4·98 (95% CI 1·96–12·65); post-intensification, 5·18 (95% CI 2·15–12·48). Although there was a trend towards a shorter RFS in MRD positive patients at the 6–9 month time-point (Fig 2D) this was only of marginal statistical significance [RR 3·36 (95% CI 1·02–11·05)]. Due to small numbers of samples for the 9–12 and 12–24 time-points, log rank analysis was not possible: Of the two patients with a MRD positive sample during the 9–24 month period, one relapsed, compared to six of the 32 MRD negative/<10−4 patients. Relapse in four of six evaluable cases occurred at a median of 249 d from the last MRD negative/<10−4 test (range 28–894) days.

Figure 2.

 Kaplan–Meier estimates of RFS according to MRD results at four time-points in auto-SCT/chemotherapy treated patients. (A) End of phase 1 induction, (B) end of phase 2 induction, (C) post-intensification therapy, (D) 6–9 months. Obs, observed; Exp, expected.

To assess whether MRD differentiated outcomes in standard-risk patients (age ≤35 years, WBC count < 30 × 109/l) data were re-analysed by subgroups (Table II). MRD status in standard risk patients remained a strong adverse predictor for relapse at defined time-points resulting in significantly lower 5-year RFS estimates in MRD positive patients compared to those who were MRD negative/<10−4 [Post-induction 2: 14% (95% CI 0–38) vs. 80% (62–98%); post-intensification: 25% (95%CI 1–50) vs. 73% (95% CI 56–90) respectively]. Although the differences in RFS by MRD status were not significant in the high risk patients, probably due to the smaller numbers of patients in this group, there was no evidence of a heterogeneity of effect by risk group; thus there was no indication that MRD differentiated outcomes to a different degree in standard-risk compared to high risk patients (P > 0·05 in each case).

Table II.   Outcome according to risk group and MRD results (chemotherapy and auto-SCT treatment groups).
Risk groupMRD-positiveMRD-negative/<10−4Log rank
P value
RFS at 5 years (95% confidence interval)Relapses/ patientsRFS at 5 years (95% confidence interval)Relapses/ patients
  1. Het P = the P value for heterogeneity in the effect of MRD between standard and high risk patients.

Post-induction 1
 Standard54% (26–82%)6/1669% (50–87%)9/290·5
 High29% (5–52%)10/1466% (41–90%)5/160·07 (het P = 0·5)
Post-induction 2
 Standard14% (0–38%)9/1280% (62–98%)4/240·0002
 High50% (15–85%)4/863% (41–85%)7/190·3 (het = 0·09)
 Standard25% (1–50%)10/1273% (56–90%)7/290·0008
 High33% (0–69%)5/873% (50–96%)4/160·2 (het = 0·3)
6–9 months
 Standard25% (0–68%)3/463% (37–89%)5/140·07
 High25% (0–65%)4/671% (38–100%)2/70·2 (het P = 0·7)
9–24 months
 Standard0%1/181% (64–98%)4/220·06
 High100%0/167% (26–100%)2/90·7 (het P = 0·2)

Predictive value of MRD in patients treated with chemotherapy

A possible explanation for the improved EFS in those randomised to protracted consolidation/maintenance chemotherapy versus auto-SCT on UKALL XII/ECOG 2993 could be the success of prolonged chemotherapy in eliminating residual disease. Accordingly, the prognostic significance of MRD positivity for relapse at early time-points in those receiving protracted consolidation/maintenance chemotherapy might be lost. To investigate this possibility the analysis was repeated excluding auto-SCT recipients (Table III). The findings were not substantially changed. Presence of MRD at the completion of phase 2 induction and intensification time-points, as previously seen with the addition of the 6–9 month time-point, predicted a significantly shorter RFS when compared to patients without MRD in the chemotherapy treatment group. The results were in the same direction for the post-induction 1 time-point, however the difference in outcome according to MRD status did not reach statistical significance.

Table III.   RFS and relative relapse risks in patients receiving chemotherapy according to MRD results.
MRDNumberNo with relapsesRelative risk of relapse (95% CI)*5-year RFS (95% confidence interval)Log rank
P value
  1. *Relative risk of relapse of MRD positive compared to MRD negative patients.

Post-induction 1
 Positive26142·13 (0·95–4·79)42% (22–62%)0·07
 Negative/<10−4361170% (55–86%)
Post-induction 2
 Positive13108·95 (2·85–28·09)15% (0–40%)0·0002
 Negative/<10−4391171% (56–85%)
 Positive14103·65 (1·33–10·02)33% (8–59%)0·02
 Negative/<10−4341070% (55–86%)
6–9 months
 Positive855·27 (1·15–24·09)33% (0–69%)0·03
 Negative/<10−417476% (55–97%)
9–24 months
 Negative/<10−427577% (59–96%)

Predictive value of MRD in those receiving stem cell transplantation

Prior to auto-SCT.  Twenty-five patients had MRD evaluation performed prior to auto-SCT. This included molecular testing of nine bone marrow harvests. In the remaining patients the MRD status of the last sample taken prior to auto-SCT was used for analysis (median time from sample testing to auto-SCT 35 d, range 7–89 d). The median time from the start of induction treatment to auto-SCT was 6·2 months (range 4·6–18·1 months). Despite small numbers of patients, MRD at the time of stem cell collection or just prior to autograft was associated with a higher rate of treatment failure due to relapse [log rank P = 0·01, 5-year RFS 25% (95% CI 0–55) MRD positive vs. 77% (95% CI 54–100) in MRD negative/<10−4 negative patients]. Notably, four of the nine harvest samples were MRD positive and three relapsed at 3, 6 and 9 months after the auto-SCT. None of the five with a MRD negative/<10−4 harvest sample relapsed at a median follow-up of 8·4 years after time of transplant.

Prior to allo-SCT.  Molecular evaluation of MRD was carried out in 36 patients prior to unrelated or sibling allo-SCT in first CR. The median time to allo-SCT from start of ALL therapy was 5·9 months (range 3·8–15·4 months). Twenty-three of the 36 allograft recipients were MRD negative/<10−4; of these, four relapsed [at 3·5, 7 (n = 2) and 10·3 months from transplant]. Of the 13 patients that were MRD positive prior to allo-SCT, only two relapsed at 10 months and 2·7 years. In contrast to the findings in the auto-SCT setting, MRD status taken at a median time of 8 weeks (range 1–23·7 weeks) prior to bone marrow transplant did not significantly impact RFS. Five-year RFS was 79% (95% CI 52–100) vs. 79% (95% CI 61–97), P > 0·1 in MRD positive vs. MRD negative/<10−4 patients. EFS at 5 years was also similar between the MRD groups [52% (95% CI 24–80) in MRD positive vs. 50% (26–75%) in MRD negative/<10−4 patients; P > 0·1] as was the actuarial 5-year risk of death in remission [34% (95% CI 6–62) in MRD positive vs. 36% (9–64%) in MRD negative/<10−4 patients; P > 0·1].


This study sought to determine the prognostic importance of molecular MRD evaluation in a large group of adult patients with Ph-negative ALL in three therapy settings, thereby defining the role of this factor in current treatment decision algorithms for adult ALL.

In patients receiving chemotherapy, molecularly determined MRD post-induction was a significant predictor for treatment failure resulting in ≥3·65-fold higher rates of relapse. Importantly, the adverse outcome associated with MRD positivity was also demonstrated in patients with standard risk disease and led to the identification of a previously uncharacterised high risk population in whom the 5 year RFS at significant time-points was ≤25%. Thus, in agreement with Bruggemann et al (2006), we show that MRD evaluation in patients not characterised by known prognostic factors enables further discrimination of relapse risk.

The strongest prediction for relapse from a MRD positive test was at later time-points – those times at or after completion of induction. This finding is in keeping with MRD studies on the German 06/99 adult ALL trial, which reported a particularly poor outcome for adult patients who were MRD positive at week 16 (Bruggemann et al, 2006). However, they contrast with findings in childhood ALL where high MRD levels at any time-point and, particularly after first induction, predict an increased risk of relapse (Cave et al, 1998; van Dongen et al, 1998; Coustan-Smith et al, 2000; Dworzak et al, 2002; Nyvold et al, 2002). The greater prognostic value of detectable MRD at later time-points in adults can be explained by the distinct pattern of MRD kinetic. Compared to paediatric ALL (Ajay Vora, UKALL 2003, personal communication), adults exhibit a higher frequency of MRD after first induction (40% this study vs. c.35% at day 28 on UKALL 2003); and slower rate of disease clearance (30% MRD positive post-intensification this study vs. c.6% at weeks 11–15 on UKALL 2003). Hence, MRD testing at later time-points in adults carries a higher predictive value. It is not clear whether the differing disease kinetic represents a biological or protocol-related phenomenon, evidence for the former might be indicated by the finding of higher rates of MRD positivity in adolescents treated on paediatric protocols compared to younger children (Ajay Vora, UKALL 2003, personal communication).

Selecting a predictive time-point on which to base a MRD-defined prospective assignment of risk group requires not only consideration of its prognostic relevance but also clinical utility. In this study, the presence of MRD after phase 2 induction or post-intensification delineated a subset of standard-risk patients with an extremely poor outcome, this finding represents a possible strategy for MRD-based risk stratification in future studies for the identification of a high risk patient group. Risk stratification using the earlier, post-induction 2 time-point has the advantage of providing a clinically useful timeframe for therapy intervention, particularly if allo-SCT is planned. Furthermore, because relapse is rarely salvageable (Fielding et al, 2007) and overall success of therapy limited by cumulative treatment toxicity, prediction of treatment failure at the earliest opportunity would seem an appropriate strategy.

Notably, the outcome of patients with standard risk disease not treated with allo-SCT and attaining molecular remission or MRD levels of <10−4 by phase 1 or 2 induction was very favourable (5-year RFS: 69% and 80%, respectively). These findings raise the question: could sibling allografting be avoided in this subgroup? A definitive answer to this question will be hard to obtain due to the pragmatic impossibility of randomising MRD negative/<10−4 patients to allo-SCT versus no allo-SCT. An important finding nonetheless is the relapse rate for the group outlined above which is not substantially higher than the 20% mortality after allo-SCT in standard risk patients reported on UKALL XII/ECOG 2993 (Goldstone et al, 2008). Further enhancement of an MRD-defined low risk categorization should come from the investigation of earlier time-points, as indicated in the studies by Vidriales et al, 2003 and Bruggemann et al, 2006. Reported rates of relapse in adult patients clearing MRD by 11 or 14 d after start of therapy in these studies were 0% and 10%, respectively.

In the auto-SCT treatment setting, detectable MRD pre-transplant was a significant adverse predictor for relapse in the small number of patients studied. These results are in agreement with other studies (Tiley et al, 1993; Uckun et al, 1993; Mizuta et al, 1999) and indicate the potential value of MRD evaluation for aiding patient selection in circumstances when autografting for ALL might be considered an appropriate therapy, e.g. where reducing the duration of therapy is considered necessary.

Interestingly, molecularly determined MRD was not a risk factor for relapse in patients receiving an allo-SCT in first CR. Whilst interpretation of these data might be limited by a high number of treatment-related deaths (10/36) and/or small patient numbers studied, an alternative explanation of this finding is that allo-SCT in first CR overcomes the negative prognostic impact of MRD in adult ALL. There is a potent graft versus leukaemia effect (Ribera et al, 1998; Hunault et al, 2004; Thomas et al, 2004; Goldstone et al, 2008) in patients with high risk ALL and this might be sufficient to eliminate MRD. Previous studies (Knechtli et al, 1998; Sanchez et al, 2002; Spinelli et al, 2007) that reported higher rates of post-transplant relapse in adults entering allo-SCT with detectable MRD had included those with advanced (>first CR) disease. Hence, a differential predictive role of MRD according to the stage of disease at the time of allo-SCT may exist.

Limitations of MRD assessment are recognised: first, lack of prediction of extramedullary relapse. MRD assessment did not predict any of the four cases of extramedullary relapses where testing occurred 2·6–10·2 months prior to the event. Second: clonal evolution of Ig and TCR rearrangements during the disease course leading to false negative results. These events were not documented in any of the cases where relapse material was available for study (n = 5). It is noteable that events occur less frequently in adult ALL (Szczepanski et al, 1998).

In summary, the present study has shown that molecular investigation of MRD differentiates relapse risk in non T-lineage ALL enabling the identification of clinically relevant subgroups. The detection of MRD post-induction is a significant risk factor for therapy failure, indicating a requirement for treatment intensification in this group if outcomes are to improve. The findings of this study provide the basis for incorporating molecular MRD assessment into current treatment protocols for non T-lineage adult ALL for prospective identification of relapse risk and further define the informative time-points for MRD-based risk assignment. These results form the basis for a planned MRD-response based assignment of therapy in future UK trials of adult ALL.