• To determine if AMD3100 given in a sequential dose escalation at 80, 160 and 240 µg/kg in combination with busulfan, fludarabine (and ATG for unrelated or HLA nonidentical donors) for treatment of AML/MDS:
  
• Stage 1 - Is Safe
  
• Stage 2 - Will improve progression free survival post allogeneic stem cell transplantation from an HLA compatible donor compared to historical controls.
  

• Stage 2 - Determine the time from stem cell transplant to PMN engraftment compared to historical controls.
  
• Determine the rate and severity grading of GVHD compared to historical controls.
  
• Determine immune reconstitution compared to historical controls.
  

This is a single centre phase I/II study which will be conducted in two stages. The ability of AMD3100 to augment the antileukemic effect of chemotherapy and stem cell transplantation in patients with MDS or with AML in first or subsequent relapse, second or greater remission or primary induction failure will be assessed.

The primary objective in Stage 1 is to determine an acceptably safe dose, utilizing the continual reassessment method (CRM).The CRM is a model-based statistical procedure for conducting phase I clinical trials that assigns doses to successive cohorts of patients based on the doses given and outcomes observed (toxicity or no toxicity) of all previous patients. In this sense, the method is "outcome-adaptive." The CRM relies on a simple Bayesian model for the probability of toxicity as a function of dose. As the dose-toxicity data in the trial accumulate the Bayesian model "learns" about the relationship between dose and probability of toxicity.

Numerous computer simulation studies (and the references therein) have shown that the CRM has greatly superior properties compared to conventional "3+3" algorithms. This is due to the facts that conventional algorithms are not model-based, they only use data from the most recent 6 patients, and they tend to stop the trial very early, with the consequences that they are very likely to give very unreliable estimates of the toxicity probability at each dose and provide an unreliable recommended maximum tolerated dose (MTD).

Under the particular version of the CRM being used in this trial, following treatment of the first cohort of three patients at the initial dose level 80 µg/kg, the following decisions are possible:

• If 0 of the first 3 patients (#1, 2 and 3) experience toxicity, then the method will escalate and the second cohort of patients (#4, 5 and 6) will be treated at the second dose level, 160 µg/kg.
• If either 1 or 2 of the first 3 patients experience toxicity, then the method will stay at the starting dose, so the second cohort will be treated at the lowest dose level, 80 µg/kg.
• If all 3 of the first 3 patients experience toxicity, then the posterior probability that the lowest dose is unacceptably toxic will be 0.951, and since this exceeds the decision cut-off 0.90 in the protocol the phase I trial will be terminated with the conclusion that the lowest dose level 80 µg/kg is excessively toxic.

If the trial continues, for all successive cohorts after the first, since the decision of which dose to assign utilizes all of the dose-toxicity data from all patients treated previously, there are too many possibilities to enumerate. For example, there are 12 possible outcomes for the second cohort, and this number increases exponentially as the trial progresses. However, at any point in the trial, the estimated probability of toxicity at each dose based on the most recent data may easily be estimated and this information made available to the investigators. Additional safety provisions are that the middle dose of 160 µg/kg may not be skipped when initially escalating from 80 µg/kg, and the trial will be terminated early with no dose chosen if the lowest dose is excessively toxic.

The primary objective in Stage 2 is to determine progression free survival post-allogeneic transplant, in terms of time to treatment failure and survival. Stage 1 will include both prognostic subgroups CR and not in complete remission (NCR). In Stage 2, different monitoring rules will be used in CR and NCR subgroups to reflect their very different historical failure rates. Additionally, all patients treated at the dose selected in Stage 1 will be counted as members of the Stage 2 sample.