PCR-based Minimal Residual Disease Detection Research Articles

Molecular characteristic of acute leukemias with t(16;21)/FUS-ERG.

T(16;21)(p11;q22)/FUS-ERG is a rare but recurrent translocation in acute leukemias and in some types of solid tumors. Due to multiple types of FUS-ERG transcripts, PCR-based minimal residual disease detection is impeded. In this study, we evaluated a cohort of pediatric patients with t(16;21)(p11;q22)/FUS-ERG and revealed fusion gene breakpoints. We implemented next-generation sequencing (NGS) on long PCR amplicons for the detection of fusion genes with unknown partners or DNA breakpoints. That allowed us to describe different fusion variants of FUS/ERG in different patients and to detect MRD on both RNA and DNA levels. We also found several accompanying mutations in epigenetic regulators (DNMT3A, ASXL1, BCOR) by targeted NGS approach in AML cases. These mutations preceded full transformation by t(16;21)(p11;q22)/FUS-ERG and allowed us to trace clonal evolution on all steps of therapy. As a casual observation, the ASXL1 mutation was found in the unrelated donor hematopoietic cells.

Next-Generation Sequencing and Real-Time Quantitative PCR for Minimal Residual Disease (MRD) Detection Using the Immunoglobulin Heavy Chain Variable Region: A Methodical Comparison in Acute Lymphoblastic Leukemia (ALL), Mantle Cell Lymphoma (MCL) and Multiple Myeloma (MM)

AbstractAbstract 788 Background.Real-Time Quantitative (RQ) PCR-based MRD detection using tumor-specific primers derived from the immunoglobulin heavy chain variable region (IGH) is an established disease monitoring tool with high predictive value in ALL, MCL and MM. It is highly sensitive and has been standardized in the context of international cooperative groups such as the European Scientific Foundation for Laboratory Hematooncology (ESLHO). However it has some limitations, including marker identification failure, particularly in hypermutated tumors and false negatives due to clonal evolution, particularly in ALL. IGH-based next-generation sequencing (NGS) might overcome some of these limitations and further increase sensitivity, specificity, accuracy and reproducibility. We have thus performed a head to head comparison of the two methods on diagnostic (DG) and post-treatment follow-up (FU) samples on a panel of 55 patients. Patients and Methods.Overall, 381 samples (215 bone marrow, 166 peripheral blood; 62 DG, 319 FU) were collected from 55 patients (15 ALL, 30 MCL, 10 MM) in which RQ-PCR based MRD analysis had been performed in the context of prospective clinical trials. IGH-based RQ-PCR was carried out as previously described [Ladetto et al, BBMT 2000; Brüggemann et al, Blood 2006], according to the criteria of the European Study Group on MRD detection [van der Velden et al, Leukemia 2007], at two experienced MRD laboratories (Kiel, DE, 238 samples; Torino, IT, 143 samples). NGS was performed at the Sequenta facilities in San Francisco, CA, USA. Using universal primer sets, we amplified IGH variable, diversity, and joining gene segments from genomic DNA. Amplified products were sequenced to obtain a high degree of coverage (14 reads per each IGH molecule) and analyzed using standardized algorithms for clonotype determination. Tumor-specific clonotypes were identified for each patient based on their high-frequency in DG sample and then quantitated in FU samples. A quantitative and standardized measure of clone level among all leukocytes was determined using internal reference DNA. NGS analysis was performed independently and data were blinded until comparison. Comparability of MRD results by RQ-PCR and NGS was assessed by means of bivariate correlations between methods analysis (software R 2.15.1 package irr). Discordances were classified as follows: a positive/negative discordance was defined as major when the positive result was >1 E-05 and minor when ≤1 E-05; a quantitative discordance was defined as the presence of two positive results with a quantitative discrepancy >1 log. Results.51 pts (93%) were evaluable with at least one tool (RQ-PCR 45, NGS 49), 43 (78%) with both and 4 (7%) with none. Disease-specific success rates are shown in Tab. 1. Overall, 333 samples (87%) were evaluated with at least one tool (RQ-PCR 282, NGS 319) and 268 (70%) with both. The latter group was thus comparable in terms of MRD output. A correlation analysis between RQ-PCR and NGS results is shown in Fig. 1. Overall a significant concordance was observed (p<0.001, R=0.791). 210/268 (78%) samples had an optimal level of concordance. Of these 99 were MRD negative and 111 MRD positive. Major discordances were 17 (6,3%) with NGS overestimating compared to RQ-PCR in 11 samples and underestimating in 6; minor discordances were 23 (8,6%) with NGS overestimating in 5 samples and underestimating in 18; quantitative discordances were 18 (6,7%) with NGS overestimating in 14 cases and underdestimating in 4. In 2 ALL clonal evolution hampered straightforward MRD assessment in one of both methods. In one case IGH RQ-PCR underestimated MRD while a second RQ-PCR marker (TCRD) led to results comparable to NGS. In a second case NGS did not detect the tumor clone identified at diagnosis due to loss of the complete IGHV at relapse whereas the preceeding IGHDJ was preserved and detected by RQ-PCR targeting the preserved DJ region. Conclusions.NGS is a feasible tool for IGH-based MRD monitoring in ALL, MCL and MM, in selected cases reaching similar sensitivities compared to standardized RQ-PCR. Good concordance was seen in the vast majority of cases. However, disease-specific pitfalls (clonal evolution, somatic hypermutations, frequency of complete IGH rearrangements) have to be considered for both methods and prospective comparative analysis of unselected cases is required to verify the clinical impact of NGS-based MRD assessment. [Display omitted] [Display omitted] Disclosures:No relevant conflicts of interest to declare.

PCR-Based Minimal Residual Disease (MRD) Detection Is a Strong Independent Outcome Predictor Also in Rituximab-Intensive Non-ASCT-Based Programs: Results From the ML17638 Multicenter Randomised Phase III Trial for Elderly Follicular Lymphoma (FL) Patients of the Fondazione Italiana Linfomi (FIL)

AbstractAbstract ▪787▪This icon denotes a clinically relevant abstract Background:Many studies support the value of PCR-based MRD detection using the bcl-2-IgH translocation as an outcome predictor in FL but some failed to confirm this observation. Concerns have been raised particularly for programs which are highly Rituximab (Rtx) intensive (with or without maintenance) and non-ASCT-based. The ML17638 study, contained an extensive centralized MRD monitoring program, whose results are here presented. Patients and methods:Clinical results of study have been already reported (Vitolo et al, ASH 2011). The program consisted of 4 R-FND courses (Rtx, fludarabine, mitoxantrone, dexamethasone) followed by 4 doses of weekly Rtx. Patients (pts) achieving 3partial response (PR) were randomized to Rtx maintenance (arm A) or observation (arm B). A total of 234 untreated elderly (age 60–75 years) pts at diagnosis were enrolled. With a median follow-up from randomization of 34 months, 3-year PFS and OS were 66% (95%CI:59-72%) and 89% (95%CI:85-93%), with a clear trend in favor of arm A for 2-year PFS (81% vs 69%).At enrolment, pts were screened for a molecular marker based on the bcl-2/IgH MBR or mcr. If found, pts were tested at 8 fixed timepoints: at month 5 (M5) after 4-R-FND, at the end of induction therapy (M8) and during maintenance/observation and follow-up (M12,M18,M24,M30,M36 and M42) or until relapse. MRD was assessed by both nested PCR (n-PCR) and real time quantitative PCR (RQ-PCR) on BM cells. Methods have been already reported (Ladetto Exp Hematol 2001). RQ-PCR was performed and analyzed according to the Euro-MRD guidelines (Van der Velden Leukemia 2007). The lab performs routine quality controls in the context of Euro-MRD and was blinded to clinical results and radomization arm. The impact of MRD on PFS was evaluated by log-rank tests and Cox models including age, sex, FLIPI, ECOG PS and complete remission (CR). In addition, the effect of PCR negativity on PFS during the whole follow-up period was evaluated by a time-varying covariate included in the models, also considered in a cumulative way (0, 1, 2, 3 or more consecutive PCR-negative timepoints). Results:229 of 234 enrolled pts (98%) were screened at study entry. A molecular marker was found in 118 (51.5%). Of these, 9 were excluded due to withdrawal before M5 (7) or inadequate sampling (2). Overall, 800 follow-up samples were expected. Of these, 707 (88%) were received and analysed: 98% of pts were evaluable for 350% of timepoints and 87% for 375%. Pts with and without a marker had identical PFS (61% at 42m for both). Sixty six per cent of pts achieved PCR-negativity after R-FND and 81% at the end of treatment, with a mean tumor burden reduction of 11 natural logaritm after R-FND and a further decrease of 1.6 after the 4 weekly Rtx. At randomization, PCR-positivity rate was similar in the two arms while during and after maintenance pts in Arm A had a lower rate of PCR-positivity (9% vs 17% p=0.02).The achievement of PCR-negativity by both n-PCR and RQ-PCR at timepoints M8,12,18 and 24 predicted a better PFS (M5 not predictive, M30, 36, 42 have too early follow up for meaningful evaluation). After M8, 2-year PFS was significantly better in PCR-negative than PCR-positive pts: 72% vs 39% (p=0.007, Fig. 1). Achieving a double PCR-negativity at M8-M12 or triple molecular negativity at M8-M12-M18 was associated with a further increase of PFS (82% vs 46% for months 8–12, p=0.001 and 87% vs 53% for months 8–12-18, p=0.001). PCR-negativity at M8 ensured a subsequent better PFS both in CR (p=0.023, HR=0.33, 95%CI: 0.13–0.86) and PR (p=0.074, HR=0.28, 95%CI: 0.07–1.13) pts (Fig. 2). [Display omitted] [Display omitted] Disclosures:Ladetto:Hoffman-La Roche: Consultancy, Honoraria. Rossi:Roche: Honoraria. Musto:Roche: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding. Gamba:Roche: Employment. Vitolo:Celgene: Honoraria; Janssin-Cilag: Honoraria; Roche: Membership on an entity's Board of Directors or advisory committees.

Time Point-Dependent Concordance of Flow Cytometry and RQ-PCR in the MRD Detection in Childhood ALL: The Experience of the AIEOP-BFM- ALL MRD Study Group

In the AIEOP-BFM ALL2000 trial, childhood acute lymphoblastic leukemia (ALL) patients were stratified mainly according to minimal residual disease (MRD) levels at day 33 (Time Point 1, TP1) and 78 (TP2) of treatment, as detected by PCR amplification of clonotypic immunoglobulin and T-cell receptor gene rearrangements. Overall, the results confirmed that PCR-based MRD detection is an independent prognostic indicator which overrides classical risk factors. In this context, MRD measurement by flow cytometry (FCM) on day 15 has been evaluated as an earlier predictor of relapse-free outcome in children with ALL. To understand how these two methods could be applied in a MRD-based clinical protocol, we analyzed the correlation of methodologies in data sets in which PCR and FCM were simultaneously applied. We studied 3,618 BM samples from 1,570 patients (32.5% of all patients on trial) derived from day15 (n=478, 30.4%), 33 (n=1,570) and 78 (n=1,570) of remission induction therapy of the AIEOP-BFM ALL 2000 protocol. Patients were enrolled from September 2000 to June 2006 in Italy, Austria and Germany, and selection was based only on available samples for both PCR and FCM. As in most MRD-based protocols, RQ-PCR was performed on BM mononuclear cells after Ficoll gradient separation, while 4-colors FCM was done on 300,000 nucleated cells (NC) from the whole BM sample after red blood cells lyse-wash procedure. MRD levels ≥10−4 were considered as ‘positive’, while levels below this threshold, not-quantifiable or undetectable were classified as ‘negative’. Overall, qualitative concordance was observed in 2,704/3,618 samples (74.7%) measured at d15, d33 and d78. Concordances at each TP are indicated in the Table. [Display omitted] Around 70% of discordant samples in all TPs were cases with low-positivity by PCR (10−4 log range) and negative by FCM, or viceversa. Concordance was also evaluated according to PCR-based MRD risk subgroups. Of note, 519/571 (90.9%) of Standard Risk by PCR-MRD (negative at d33 by 2 markers with sensitivity ≥10−4) were also FCM negative at d33; whereas 59/121 (48.8%) of cases PCR ≥10−3 at d78 (High Risk by PCR-MRD) were FCM negative. In order to evaluate the correlation of methodologies more exactly, we investigated similar cell preparations (same MNC divided for PCR and FCM), and we enhanced the resolution of FCM by using 7-colors and analyzing 500,000 MNC. Samples were collected at days 15, 33, 52, and 78 within a single center. Among a total of 266 samples, the concordance increased up to 87%. More specifically, 100 samples (37.6%) resulted undetectable by both methods; 56 samples were FCM-undetectable and PCR-positive <10−4, while only 13 samples were PCR-positive ≥10−4 but undetectable by FCM. In summary, several methodological issues limit the concordance of FCM and PCR as used by the AIEOP-BFM MRD study group. These include differences in the material analyzed by PCR and FCM, as well as in the amount of sample input (number of cells analyzed). The recent implementation of further technical developments in the FCM procedure, are designed to increase the concordance rate. In conclusion, within BFM-based protocols, FCM cannot simply substitute the current PCR-based MRD risk stratification at the same TPs. Instead, a tailored FCM-based risk definition may be independently reliable. Moreover, the two methodologies applied at different TPs might be complementary in more advanced MRD-based patient stratification protocols. Overall, the choice of the MRD methodology largely depends on the aims of the study protocol, resources available, and treatment strategy.

Nucleophosmin Gene Mutations Are Predictors of Favourable Prognosis in Acute Myeloid Leukemia with a Normal Kayotype and Can Be Used as a New Marker for Quantitative PCR To Detect Minimal Residual Disease.

Nucleoplasmin (NPM) mutations have recently been described as a new kind of mutation in AML. In the presented work we have analyzed 977 AML cases from different prognostic subgroups. Fifteen different mutations (type A:n=245 (78.3%), B:n=21 (6.7%), D: n=28 (11.4%), I:n=3 (1%), J:n=4 (1.3%), K:n=2, G,H, L, M, N, O, P, Q, R: n=1, each (0.3%) were detected in 313 cases. No mutation was found in inv(16) (n=89), t(8;21) (n=98), t(15;17) (n=17), t(11q23)/MLL (n=8), inv(3) (n=8), other reciprocal translocations (n=3), and complex aberrant karyotypes (n=59). In the normal karyotype group the incidence was 299/603 (49.6%) and in all others 13/90 (14.4%). In the latter cohort cases with NPM+ were: − Y (1 case), +4 alone (1 case), +4 and +8 (2 cases), +4 with multiple trisomies (1 case),+8 (1 case), multiple other trisomies (1 case), del(5q) (1 case), − 7 alone (2 cases), del(9q) (2 cases), unbalanced translocation (1 case).Data for evaluation of prognosis were available in 401 AML patients with normal karyotype treated within the AMLCG99 study. Results were calculated in relation to MLL-PTD, FLT3-LM, FLT3-TKD, NRAS, KIT, and CEBPA mutations and correlated to further clinical characteristics. NPM-mutated cases were frequently associated with FLT3 mutations but rarely with other mutations. The NPM-mutated group had a higher CR rate (70.5% vs. 54.7%, p=0.003), a trend to longer OS (median 1012 vs. 549 days, p=0.0762), and significantly longer EFS (median 428 vs. 336 days; p=0.0121). There was a clear favourable impact of NPM+/FLT3-LM- on OS and EFS (264/395 cases; 66.8%). This positive effect was lost in the presence of a concomitant FLT3-LM, since survival of the NPM+/FLT3-LM+ double positive pts. was similar to NPM-/FLT3-LM+ cases. We further analysed stability for the NPM mutation at diagnosis and relapse and found stability in 14/15 paired samples. One case with AML M5 lost the NPM mutation at relapse. This case had also a shift from normal karyotype to 46,XX,del(7)(q22),der(7)t(7;21)(p11;q?) and a complete shift of immunophenotype indicating a therapy-related AML. In addition, we have analyzed the applicability of NPM mutations as targets for minimal residual disease (MRD) detection. In total, 82 samples of 22 selected cases were analyzed by use of quantitative real time PCR for the three most common mutation types. Fifteen of these cases had a type A, 3 a type B, and 4 a type D mutation. Samples at diagnosis had an NPM/ABL ratio of 267.6 (range: 54.9 to 3174.7) while each 10 NPM negative cases had a median ratio of 0.03 (assay for A), 0.0001 (assay for B), and 0.002 (assay for D), showing that these assays were nearly specific for the respective mutation. Using limited dilution series of NPM+ in NPM- cDNA a median sensitivity of 1:100.000 could be shown. In both analyzed cases with relapse these were early detectable by increasing NPM levels. Thus, NPM mutations can be used for highly sensitive PCR-based MRD detection in AML, especially if other markers are lacking, i.e. in normal karyotype.In conclusion, our data showa nearly exclusive association of NPM+ to the normal karyotype,a good prognosis of NPM+/FLT3-LM- in normal karyotype.an intermediate prognosis of NPM+/FLT3-LM+ in normal karyotypeNPM as a new marker for PCR-based MRD-detection.

Quantification of Minimal Residual Disease (MRD) in MLL-PTD Positive AML Can Be Used for Prognostication in AML with Normal Karyotype.

Partial tandem duplications of the MLL gene (MLL-PTD) occur with the same frequency of approximately 6% as PML-RARA, AML1-ETO, and CBFB-MYH11 that have previously been shown to be relevant markers for detection of minimal residual disease (MRD). MLL-PTD has a frequency of 11% in the normal karyotype and here it is associated with an unfavourable prognosis. In the normal karyotype AML PCR-based MRD detection was not applicable so far, and thus MLL-PTD may be a promising MRD marker for this large intermediate AML subtype. So far, the clinical significance of the expression level of MLL-PTD at diagnosis or MRD during the course of therapy have not been investigated. Using real time RT-PCR MLL-PTD expression levels were quantified relative to the control gene ABL. Quantitative data were available from 143 patients at diagnosis. The range of the MLL-PTD/ABL ratios at diagnosis was 9.6–1255 (median: 149). The levels of MLL-PTD expression in 50 healthy control samples were between 0 and 0.08 (median 0.02). In total 47 of these patients with a median follow up sample number of 4 per patient (range 2–17) were evaluated for MRD during and after therapy. Twenty-one patients were evaluable for MRD analysis after 2 months after start of therapy. Thirty-one patients were evaluable for MRD analysis after 4 and 6 months after start of therapy. A 2-log reduction of MLL-PTD expression after 2 months, 4 months and 6 months was a significant prognostic factor for overall survival (OS: p=0.0179, p=0.035, p=0.048, respectively) as well as for event free survival (EFS: p=0.0047, p=0.037, p=0.020, respectively). There is no evidence that a 3-log reduction after 2 and 4 months improves OS significantly (p=0.62 and 0.24 respectively). This was mainly due to the low number of patients that achieved a 3 log decrease (6/21 and 10/31 respectively). In contrast, at 6 months a 3-log decrease had a significant impact on OS and EFS (p=0.049 and p=0.030, respectively). Neither the expression levels at diagnosis nor fusion type (involved MLL exons) showed prognostic significance. However, age above 60 years and leukocytosis above 50 G/l were associated with a worse prognosis (p=0.028 and p=0.0035, respectively). Overall, the MLL-PTD expression levels correlated well to the course of the disease and a molecular relapse was detected before clinical manifestation in 2 patients based on increasing expression ratios. Another 15 cases were assessed at clinical relapse when expression levels were again in the range of the diagnostic sample. Here the median interval between the last follow up and relapse was too long (9.2 months) for early detection of relapse. Thus, more frequent follow up analyses at least every three months should improve the early detection rate. In conclusion, this analysis confirms the significance of MRD levels as a prognostic factor in this AML subtype. 1) MLL-PTD can be applied in 11% of intermediate risk group AML. 2) Patients with an apriori high risk for relapse can be identified on the basis of a 2 log reduction. 3) Relapses can be early detected based on increasing expression ratios.

Polymerase chain reaction (PCR)- and reverse transcription PCR-based minimal residual disease detection in long-term follow-up of childhood acute lymphoblastic leukaemia.

Minimal residual disease (MRD) was investigated in 52 children with acute lymphoblastic leukaemia (ALL), using antigen receptor gene rearrangements and reverse transcription polymerase chain reaction for fusion transcripts as molecular targets. Patients [treated according to the Medical Research Council United Kingdom ALL (MRC UKALL) XI protocol or Total XI and XIII protocols] were monitored for a median period of 45 months (range, 9-110 months). Among 17 patients who relapsed, MRD persisted for longer (66.7%, 47.1%, 53.8% and 41.7% at 0-2, 3-5, 6-9, 10-24 months respectively) than patients who remained in continuous clinical and immunological remission (n = 35) (27.3%, 11.1%, 4.3%, 8.0%). Association between MRD tests and outcome was assessed and found to be significant at all time-points. The difference in survival for MRD-positive and MRD-negative patients (using the log-rank test) was statistically significant at all time intervals, as was risk of relapse for MRD-positive patients (1.89, 2.20, 2.65 and 2.16) and MRD-negative patients (0.72, 0.82, 0.65 and 0.70). Sixteen of the 52 patients had an oligoclonal pattern at presentation but oligoclonality did not have an impact on outcome. Cox regression analysis revealed that MRD assessment is an independent and prognostically significant factor during treatment and should be used for patients' stratification in future studies.

Real-time quantitative PCR for the detection of minimal residual disease in acute lymphoblastic leukemia using junctional region specific TaqMan probes.

Analysis of minimal residual disease (MRD) can predict outcome in acute lymphoblastic leukemia (ALL). A large prospective study in childhood ALL has shown that MRD analysis using immunoglobulin (Ig) and T cell receptor (TCR) gene rearrangements as PCR targets can identify good and poor prognosis groups of substantial size that might profit from treatment adaptation. This MRD-based risk group assignment was based on the kinetics of tumor reduction. Consequently, the level of MRD has to be defined precisely in follow-up samples. However, current PCR methods do not allow easy and accurate quantification. We have tested 'real-time' quantitative PCR (RQ-PCR) using the TaqMan technology and compared its sensitivity with two conventional MRD-PCR methods, ie dot-blot and liquid hybridization of PCR amplified Ig/TCR gene rearrangements using clone-specific radioactive probes. In RQ-PCR the generated specific PCR product is measured at each cycle ('real-time') by cleavage of a fluorogenic intrinsic TaqMan probe. The junctional regions of rearranged Ig/TCR genes define the specificity and sensitivity of PCR-based MRD detection in ALL and are generally used to design a patient-specific probe. In the TaqMan technology we have chosen for the same approach with the design of patient-specific TaqMan probes at the position of the junctional regions. We developed primers/probe combinations for RQ-PCR analysis of a total of three IGH, two TCRD, two TCRG and three IGK gene rearrangements in four randomly chosen precursor-B-ALL. In one patient, 12 bone marrow follow-up samples were analyzed for the presence of MRD using an IGK PCR target. The sensitivity of the RQ-PCR technique appeared to be comparable to the dot-blot method, but less sensitive than liquid hybridization. Although it still is a relatively expensive method, RQ-PCR allows sensitive, reproducible and quantitative MRD detection with a high throughput of samples providing possibilities for semi-automation. We consider this novel technique as an important step forward towards routinely performed diagnostic MRD studies.

Rapid detection of gammaT cell receptor gene rearrangements in acute lymphoblastic leukemia by electrophoresis and silver staining: implications for detection of minimal residual disease.

Minimal residual disease (MRD) in acute lymphoblastic leukemia (ALL) was studied using polymerase chain reaction (PCR). GammaT cell receptor (TCRG) genes are ideal targets for PCR-based detection of MRD due to their molecular characteristics. Polyacrylamide gel electrophoresis (PAGE) analysis of PCR products followed by silver staining was performed for 72 children with ALL at the onset of disease. Silver staining is an effective technique to detect gene rearrangements without the use of ethidium bromide. Moreover, this method may show heteroduplex bands of a clonal nature when both TCRG alleles are rearranged. PCR products subjected to a rapid staining protocol were recovered from the gel, reamplified by a second PCR and directly sequenced. After sequencing, we identified the junctional region and obtained patient-specific probes. In more than half of the patients we detected TCRG rearrangements that were used as molecular markers for residual disease.

Long-term follow-up of minimal residual disease in childhood acute lymphoblastic leukemia patients by polymerase chain reaction analysis of multiple clonespecific or malignancy-specific gene markers

Two types of markers, namely the clone-specific markers including T-cell receptor (TCR) γ, TCR δ and Ig heavy-chain (IgH) gene rearrangements, and malignancy-specific fusion gene mRNA such as SIL-TAL-1, BCR-ABL, and HRX-partner genes, were investigated by molecular biology techniques in 65 Chinese patients with acute lymphoblastic leukemia (ALL). In combination, these markers were informative among 96% of patients. Minimal residual disease (MRD) was followed up in 23 of these patients with available materials over a period varying from 8 to 54 months with at least one leukemia-specific probe. In most children, MRD was decreased continuously to an ultimately undetectable level within 6 to 12 months after remission induction therapy. One patient exhibited low-level residual leukemic cells for 4 years before the MRD turned negative. Another patient remained in complete remission for 45 months, although a positive signal was detected at 34 months using TCR S probe, but was negative with a TCR y marker which was positive at presentation. In three patients who relapsed, MRD either persisted through the clinical course or became positive and eventually increased 3–11 months before clinical relapse. These data suggested that the combined use of multiple gene markers is a valuable tool for the PCR-based MRD detection, since it can cover most ALL patients. Furthermore, longterm follow-up of MRD is helpful for determining the dosage as well as the period of maintenance chemotherapy and for predicting impending relapse.