Statistics and measurable residual disease (MRD) testing: uses and abuses in hematopoietic cell transplantation

The area under the time-dependent ROC curve (AUC) may be used to quantify the ability of a marker to predict the onset of a clinical outcome in the future. For survival analysis with competing risks, two alternative definitions of the specificity may be proposed depending of the way to deal with subjects who undergo the competing events. In this work, we propose nonparametric inverse probability of censoring weighting estimators of the AUC corresponding to these two definitions, and we study their asymptotic properties. We derive confidence intervals and test statistics for the equality of the AUCs obtained with two markers measured on the same subjects. A simulation study is performed to investigate the finite sample behaviour of the test and the confidence intervals. The method is applied to the French cohort PAQUID to compare the abilities of two psychometric tests to predict dementia onset in the elderly accounting for death without dementia competing risk. The 'timeROC' R package is provided to make the methodology easily usable. Copyright © 2013 John Wiley & Sons, Ltd.

A marker strongly associated with outcome (or disease) is often assumed to be effective for classifying persons according to their current or future outcome. However, for this assumption to be true, the associated odds ratio must be of a magnitude rarely seen in epidemiologic studies. In this paper, an illustration of the relation between odds ratios and receiver operating characteristic curves shows, for example, that a marker with an odds ratio of as high as 3 is in fact a very poor classification tool. If a marker identifies 10% of controls as positive (false positives) and has an odds ratio of 3, then it will correctly identify only 25% of cases as positive (true positives). The authors illustrate that a single measure of association such as an odds ratio does not meaningfully describe a marker's ability to classify subjects. Appropriate statistical methods for assessing and reporting the classification power of a marker are described. In addition, the serious pitfalls of using more traditional methods based on parameters in logistic regression models are illustrated.

Minimal residual disease (MRD) is the most sensitive and specific predictor of relapse risk in children with acute lymphoblastic leukaemia (ALL) during remission. We assessed whether treatment intensity could be adjusted for children and young adults according to MRD risk stratification. Between Oct 1, 2003 and June 30, 2011, consecutive children and young adults (aged 1-25 years) with ALL from the UK and Ireland were recruited. Eligible patients were categorised into clinical standard, intermediate, and high risk groups on the basis of a combination of National Cancer Institute (NCI) criteria, cytogenetics, and early response to induction therapy, which was assessed by bone marrow blast counts taken at days 8 (NCI high-risk patients) and 15 (NCI standard-risk patients) after induction began. Clinical standard-risk and intermediate-risk patients were assessed for MRD. Those classified as MRD low risk (undetectable MRD at the end of induction [day 29] or detectable MRD at day 29 that became undetectable by week 11) were randomly assigned to receive one or two delayed intensification courses. Patients had received induction, consolidation, and interim maintenance therapy before they began delayed intensification. Delayed intensification consisted of pegylated asparaginase on day 4; vincristine, dexamethasone (alternate weeks), and doxorubicin for 3 weeks; and 4 weeks of cyclophosphamide and cytarabine. Computer randomisation was done with stratification by MRD result and balancing for sex, age, and white blood cell count at diagnosis by method of minimisation. Patients, clinicians, and data analysts were not masked to treatment allocation. The primary outcome was event-free survival (EFS), which was defined as time to relapse, secondary tumour, or death. Our aim was to rule out a 7% reduction in EFS in the group given one delayed intensification course relative to that given two delayed intensification courses. Analyses were by intention to treat. This trial is registered, number ISRCTN07355119. Of 3207 patients registered in the trial overall, 521 MRD low-risk patients were randomly assigned to receive one (n=260) or two (n=261) delayed intensification courses. Median follow-up of these patients was 57 months (IQR 42-72). We recorded no significant difference in EFS between the group given one delayed intensification (94·4% at 5 years, 95% CI 91·1-97·7) and that given two delayed intensifications (95·5%, 92·8-98·2; unadjusted odds ratio 1·00, 95% CI 0·43-2·31; two-sided p=0·99). The difference in 5-year EFS between the two groups was 1·1% (95% CI -5·6 to 2·5). 11 patients (actuarial relapse at 5 years 5·6%, 95% CI 2·3-8·9) given one delayed intensification and six (2·4%, 0·2-4·6) given two delayed intensifications relapsed (p=0·23). Three patients (1·2%, 0-2·6) given two delayed intensifications died of treatment-related causes compared with none in the group given one delayed intensification (p=0·08). We recorded no significant difference between groups for serious adverse events and grade 3 or 4 toxic effects; however, the second delayed intensification course was associated with one (<1%) treatment-related death, and 74 episodes of grade 3 or 4 toxic effects in 45 patients (17%). Treatment reduction is feasible for children and young adults with ALL who are predicted to have a low risk of relapse on the basis of rapid clearance of MRD by the end of induction therapy. Medical Research Council and Leukaemia and Lymphoma Research. Copyright © 2013 Elsevier Ltd. All rights reserved.