|
|
|
|
|
|
|
Fibonacci scheme. Three patients are treated at the starting dose and, in the absence of toxicity, a series of dose escalations is applied in successive cohorts until toxicity appears. Thus if level 1, the starting dose, is X1, level 2 is X1 + 100%, level 3 is X2 + 67%, level 4 is X3 + 50%, and levels 5 and above are Xn + 30%. Three patients are usually treated at each dose level but additional patients are entered at a given dose if significant toxicity is documented. More conservative dose escalation steps (i.e., 30% or less) proceed after the observation of WHO grade 2 toxicity [7]. This scheme can be time consuming and wasteful of precious resources, particularly when the MTD is significantly higher than predicted by animal studies and if sporadic events in early doses are wrongly attributed to drug effects, thus slowing down escalation. |
|
|
|
|
|
|
|
|
Two recent innovations are finding increasing application in an attempt to improve our current methodology, these are known as PGDE (pharmacokinetically guided dose-escalations) and CRM (continuous reassessment method). |
|
|
|
|
|
|
|
|
PGDE was originally proposed by Collins et al. [15]. The European Organisation for Research and Treatment of Cancer (EORTC) produced guidelines for its use in Phase I in 1987 [16] and since then it has been prospectively evaluated in a number of Phase I studies (reviewed by Graham and Workman [17]). The aim of PGDE is to reach a target AUC (area under the plasma concentration-time curve) in humans in 47 steps. This target AUC is derived from the AUC at the LD10 in mice, and the assumption is made that discrepancies between the LD10 in mice and the MTD in man are similar. Implementation of PGDE is highly dependent on suitable (rapid) analytical methodology and expertise, but these are almost always available on site or at the pharmaceutical research establishment. Although PGDE does have limitations it has the potential to reduce escalation steps. It is only through continued prospective evaluation that the true worth of PGDE will be revealed. |
|
|
|
|
|
|
|
|
CRM is based on use of all available data to predict the MTD. In other words the preclinical data are modified and reassessed with each patient's experience to update the estimate of MTD. Prior to patient dosing a probability distribution for the MTD and dose toxic-response model are selected. These may be modified by pharmacokinetic or other data. The new MTD is calculated after each individual patient has been treated with the new agent using the methods of O'Quigley et al. [18] and Faries [19]. This allows for just one patient to be treated at each dose level, which may not save time but will reduce the number of patients exposed to low (inactive) doses. The method is also sufficiently flexible to re-accelerate escalation if sporadic events at lower dose levels are not repeated in individuals given higher doses as the trial progresses. This methodology is in its infancy but again could radically alter our current practice. |
|
|
|
|
|
|
|
|
One further approach is to use pharmacokinetic and pharmacodynamic information to define the maximal tolerated systemic exposure of a drug [20]. |
|
|
|
|
|