• © 2003 by American Society of Clinical Oncology

In reply:

Drs Sparreboom and Verweij raise a number of issues related to their article¹ and my recent editorial.² Their initial concern deals with the appropriate model for relating the pharmacokinetics of paclitaxel to the pharmacodynamic consequences associated with its administration. They argue correctly that knowledge of paclitaxel pharmacokinetics has increased and that alternative models for relating the pharmacokinetics and pharmacodynamics of that agent have been proposed. However, their statement that the threshold model for relating paclitaxel pharmacokinetics to paclitaxel-induced neutropenia lacks usefulness when applied to other conditions is not entirely correct. Data from numerous investigators have demonstrated the utility of threshold models for describing the pharmacokinetic-pharmacodynamic relationships of paclitaxel administered at 1-, 3-, and 24-hour infusions. Moreover, those threshold models and relationships have proven useful in evaluating perturbations of paclitaxel pharmacokinetics and paclitaxel pharmacokinetics-neutropenia relationships when paclitaxel has been combined with other agents, including cytotoxins and resistance-modulating agents. Furthermore, these models have proven applicable when applied to patients with hepatic and renal dysfunction who were treated with paclitaxel. While Drs Sparreboom and Verweij were correct in stating that additional models have been proposed and that several of those relate the area-under-the-curve of unbound paclitaxel as a pharmacokinetic parameter to describe exposure-neutropenia relationships, those models have not been demonstrated to be statistically superior in a randomized comparison with threshold-based models. Moreover, determination of unbound paclitaxel currently involves use of radiolabeled paclitaxel and equilibrium dialysis, which are reagents, representing a procedure not likely to be available or implemented in the majority of laboratories performing clinical pharmacology studies of paclitaxel. While debates concerning the appropriate models could be continued ad infinitum or ad nauseam, it is important to understand that a model is nothing more than a simplification of a complex system that allows one to work with that complex system. More importantly, it is well recognized that all models are wrong; it is just that some models are more wrong than others.

The second concern raised by Drs Sparreboom and Verweij relates to the question of the generalizability of their results. Unfortunately, the points that they raised do nothing to substantiate the generalizability of their results. Data derived from retrospective analyses must be viewed with caution, particularly when such retrospective analyses may involve determiniation of unbound drugs in samples that have been stored for unspecified periods under unspecified conditions.

The final issue raised by Drs Sparreboom and Verweij neglects consideration of their own clinical data. Irrespective of the number of hypotheses and models to be considered, the clinical reality of their trial was that body-surface area–based dosing did not produce a reduction in clinically significant neutrophil toxicity, which should be the determinant of clinical practice, rather than pharmacokinetic modeling. The concern of Drs Sparreboom and Verweij, that weekly determination of blood counts was insufficient to detect a clinical difference, rings hollow. If prolonged neutropenia were present, it would certainly be detected with this standard-of-practice clinical assessment of hematologic toxicity. Drs Sparreboom and Verweij should be congratulated on planning the appropriate prospective trial wherein they might detect a statistically significant difference in the time-course of blood cell counts. Nevertheless, whether a statistically significant difference in neutropenia would actually represent a clinically significant increase in duration of neutropenia is a separate issue.

As stated in my editorial, Smorenburg et al¹ conducted a well planned clinical trial. Their original article, my editorial, and this subsequent exchange of letters have provided an extensive opportunity for discussion of the results of that trial. It is hoped that the readers of this journal will interpret the pharmacologic and clinical results presented and form their own opinions as to the necessity for body-surface area–based dosing of paclitaxel and as to whether any of the evidence presented outweighs the significant logistical advantages associated with fixed dosing of paclitaxel that were indicated in my editorial.

The letter from Dr Extermann raises an interesting point. The concern about defining a flat dose based on an obese American population and applying that to other populations may be a valid concern. However, the 4% difference between 1.8 m² and 1.73 m² is certainly far less of a concern than other well recognized variability in physiologic and pharmacogenetic parameters that are totally disregarded when international studies are designed. Examples of such variables are differences in N-acetyl transferase, differences in diet (including foods such as cole crops, which are known to induce both phase I and phase II drug-metabolizing enzymes), and differences in standards-of-practice regarding administration of H₂ antihistamines and proton-pump inhibitors. Additional potentially important considerations that are rarely, if ever, addressed in planning international studies involve cultural differences in use of complementary medicine, such as St. John’s wort, which can have profound effects on the metabolism and disposition of the agent under study. When one considers variables such as these, a 4% difference in body-surface area seems trivial.

I agree with Dr Extermann’s concerns that flat dosing may not be the best answer; however, I believe that her concerns should be further refined to reflect an understanding that each agent administered to patients is different and that the most appropriate dosing scheme cannot be generalized from one drug to another. Unfortunately, I believe that Dr Extermann’s recommendation of assessing individual pharmacokinetic and pharmacodynamic patient parameters so as to prescribe a patient-specific drug dose is an impossible task. In order to do so, the relative pharmacokinetic parameter would need to be defined for each agent under consideration, as would the relationship to both therapeutic and toxic pharmacokinetic consequences. Then, before a patient could be treated, test doses of the drug to be used or surrogate probe drugs would have to be administered to define a specific patient’s pharmacokinetic behavior. Not only is this a daunting task, but if the relevant pharmacokinetic parameter involves determination of free drug, such as that proposed by Drs Sparreboom and Verweij, the methodology necessary for such studies would be so restricted as to limit the number of patients and institutions that could participate in drug-development studies, let alone administration of a drug in routine clinical practice.