Model-guided determination of maximum tolerated dose in phase I clinical trials: evidence for increased precision.

BACKGROUND

A widely used phase I design in clinical trials of chemotherapy for cancer and for AIDS (acquired immunodeficiency syndrome) allows for dose escalation in cohorts of three to six patients. Escalation continues until a predefined percentage of patients experience unacceptable toxic effects at a given dose level. A safe and maximum tolerated dose (MTD) for phase II study is then determined. This standard phase I study design has serious inadequacies. MTD is not a model-based estimate of the true dose that would yield the targeted dose-limiting toxicity rate. Moreover, this simplistic study design allows some patients in the phase I study to be treated at doses unlikely to have therapeutic efficacy.

PURPOSE

We constructed a novel quantitative assessment design that repetitively evaluates accumulating dose-toxicity data by repeatedly fitting and updating a pharmacodynamic model after small cohorts of patients are treated. The goal was to more accurately estimate the MTD.

METHODS

One hundred phase I studies were simulated by both the standard and quantitative assessment phase I designs. We compared determination of MTD, frequency of grade 0 leukopenia (no toxicity), and study size in the studies simulated using the standard design with those in the studies simulated using the quantitative assessment design.

RESULTS

The median MTD determined from the 100 studies was nearly identical for the two designs: 100 and 95 mg/m2 per day for standard and quantitative assessment designs, respectively. However, the interstudy variation in the MTD was decreased in the quantitative assessment design. Moreover, the study size was significantly reduced (P < .0001), and the median percentage of patients treated at subtoxic doses (no leukopenia) was significantly lower for the quantitative assessment design (44% versus 48%; P < .0001).

CONCLUSION

Our results show clear evidence that a phase I study design using dose and toxicity data in a repetitive and quantitative manner can identify the MTD with more accuracy than the standard design.

IMPLICATIONS

New approaches must be explored to improve our ability to identify the optimal dose for phase II studies of chemotherapy for cancer and for AIDS. There is evidence that the quantitative assessment design will identify the MTD with fewer patients, more precision, and fewer patients exposed to suboptimal doses.