Can pharmacokinetic and pharmacodynamic studies improve cancer chemotherapy?

Annals of Oncology 5 (Suppl. 4): S9-S15, 1994. O 1994 Kluwer Academic Publishers. Printed in the Netherlands.

Symposium article Can pharmacokinetic and pharmacodynamic studies improve cancer chemotherapy? D. R. Newell Cancer Research Unit, University of Newcastle upon Tyne, Newcastle, U.K.

Summary

Key words: pharmacokinetics, pharmacodynamics, dose optimization, adaptive dosing

Introduction

Pharmacokinetic and pharmacodynamic terminology

In general, two approaches are being taken to the development of improved cancer chemotherapies: optimization of the use of existing drugs and the development of new agents. In theory, both of these approaches should benefit from the use of pharmacokinetic and pharmacodynamic information, and the purpose of this article is to explore how this benefit might be achieved and where future developments might lead. Attention will focus primarily on the exploitation of pharmacokinetic and pharmacodynamic relationships to improve the use of existing drugs; phase I trials are considered elsewhere in this issue [1]. In certain areas of clinical pharmacology, the importance of therapeutic drug monitoring is recognized. Indeed, Aronson and Hardman [2] concluded that although the value of plasma drug measurement was proven for aminoglycoside antibiotics, phenytoin, carbamazepine, digoxin/digitoxin, lithium, theophylline, cyclosporin, and thyroid hormones, such was not the case for methotrexate, the only widely used anticancer agent mentioned. The usefulness of pharmacodynamic and pharmacokinetic measurements in oncology is controversial [3,4] but the consensus is that at present, the potential benefits of pharmacokinetic and pharmacodynamic studies in cancer therapy have not been realised [5]. Before considering why this should be and what might be done to change matters, pharmacokinetic and pharmacodynamic terminology will be described briefly.

Pharmacokinetics is traditionally described as the mathematical description of the in vivo fate of a drug (i.e. the rates and extent of drug absorption, distribution, metabolism and excretion). Central to this description is the plasma concentration v.s. time profile from which simple and important primary pharmacokinetic parameters are derived: plasma drug concentrations at defined times, peak concentration, area under the plasma concentration v.s. time curve (AUC) and half-lives. Secondary pharmacokinetic parameters (e.g. clearance, volumes of distribution and microscopic rate constants for the partition of the drug between compartments) can then be defined by either compartmental or non-compartmental analyses. For the practising clinician these secondary parameters can largely be ignored beyond the simple relationships between dose, AUC and clearance: Clearance = Dose rate/Plasma concentration. For drugs given at a constant rate via intravenous infusion, once the steady state plasma concentration has been achieved, the relationship between clearance, plasma concentration and dose rate is as follows: Plasma concentration D Dose rate/Clearance. With knowledge of the plasma clearance of a drug, the dose required for a given AUC or steady-state plasma level may readily be calculated. It should be noted, however, that the clearance of a drug may depend on

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

Background: The exploitation of pharmacokinetic-pharmacodynamic relationships is worthwhile for drugs where there is difficulty or delay in obtaining clinical evidence of therapeutic or toxic effects, where there is a clear relationship between the pharmacokinetics and pharmacodynamics, and where the drug has a small therapeutic index. These criteria are amply met by cytotoxic anticancer agents. In particular, an extensive literature for all classes of cancer chemotherapeutics shows that the correlation between pharmacokinetic parameters (peak concentration, area under the plasma concentration-time curve, clearance or steady-state levels) and

toxicity is in many cases better than the relationship between dose and toxicity. Adaptive dosing: Prospective studies of dose adaptation on the basis of pharmacokinetic or pharmacodynamic information, with or without feedback control, have shown that pharmacologically guided dosing is feasible. Adaptive dosing results in reduced pharmacokinetic variability and more consistent toxicity. Currently, there are insufficient prospective studies to allow conclusions concerning the efficacy of such a dosing system.

10 100 i -

lability of effect

80 60 40

o 20 0 l o g ^ (drug dose or concentration)

Fig. 1. Sigmoidal dose-effect or concentration-effect relationships in cancer chemotherapy.

the marriage of pharmacokinetic and pharmacodynamic studies in the clinical setting. Thus, in Fig. 1, dose level on the abscissa is replaced by drug concentration at the site of drug action, or in a body compartment where concentration bears a direct and constant relationship to the level at the site of action. Effect on the ordinate is either clinical response, toxicity or the molecular events underlying these effects. Approaches towards achieving this level of sophistication will be reviewed later in this article.

Pharmacokinetic—pharmacodynamic relationships in cancer chemotherapy

When considering candidate drugs for pharmacokinetically guided dosing, Aronson and Hardman [2] identified three major criteria: - Interpretation of clinical evidence of therapeutic or toxic effects should be difficult. If the required therapeutic effect or unwanted toxicity is rapidly elicited and easily measured, pharmacokinetic studies may have little to contribute. For example, the effects of a hypotensive agent, which rapidly reduces blood pressure, can be readily monitored and the dose regulated on the basis of achieved blood pressure. In this setting, pharmacodynamics alone may be used to guide dosing. - There should be evidence of a relationship between drug concentrations and therapeutic or toxic effects. If there is no relationship between pharmacodynamics and a readily determined pharmacokinetic parameter, it will not be possible to implement pharmacokinetically-guided dosing. This may be the case if pro-drug metabolism to an active species takes place, when dosing on the basis of active metabolite levels may be indicated. Alternatively, due to transport barriers, plasma drug concentrations may not reflect levels at the site of drug action. - The ratio between toxic and effective doses should be small. All patients can be safely treated

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

plasma concentration (i.e. it may exhibit dose-dependent pharmacokinetics). Hence, the AUC or steady state concentrations used to calculate the plasma clearance of a drug should always be considered before using the above equations to calculate doses. Pharmacokinetics may be seen as describing the effect of the body on the drug, while pharmacodynamics describe the effect of the drug on the body. The effects of any drug are two-fold (i.e. activity and toxicity), and may be described either by discontinuous parameters (e.g. response v.s. no response) or by continuous data (e.g. blood counts). Underlying both the activity and toxicity of a drug are molecular events, which take place within the target cell. At this point, pharmacokinetics and pharmacodynamics converge (e.g. at the inhibition of a target enzyme by the intracellular metabolite of the drug), hi routine clinical oncology practice, this convergence is not an issue, as data on events in the target cell (e.g. tumour or bone marrow stem cell) are seldom available. Mathematical functions have been used to describe pharmacodynamics and the most commonly used function is the modified Hill equation, which describes a sigmoidal curve (Figure 1). The sigmoidal shape of the curve has a number of important consequences. First, beyond a certain drug level (dose or concentration), no increase in the probability of an effect may be expected, unlike a linear relationship, in which extrapolation leads to a meaningless effect of more than 100%. Secondly, as the maximal effect is approached, the benefit achieved by increasing the drug level decreases (depending on the steepness of the sigmoid curve). With the introduction of improved haematopoietic support measures, dose level escalation with established drugs is being widely studied, and it should be remembered that a two-fold increase in dose or dose intensity may not always yield a corresponding increase in efficacy. Finally, the sigmoidal curve implies a no-effect region at low drug levels, and this needs to be born in mind in dose escalation studies when patients treated in this non-therapeutic area will not be receiving even potentially effective doses. The sigmoidal response curve shown in Fig. 1 is a stylized representation, with the exact shape of the curve depending upon the agent concerned and the disease being treated. Thus, if 'standard' therapy produces an effect that is already above the mean effect level, high dose or dose intensity therapy is unlikely to be worthwhile as the non-linear portion of the dose response relationship will rapidly be encountered. Conversely, where standard therapy produces effects at the lower end of the curve, dose or dose intensity escalation may be particularly fruitful. A discussion of the role of high dose therapy with regard to individual drug and disease types is beyond the scope of this article. However, both study design and interpretation of the results obtained benefit greatly from pharmacodynamic analyses. The ultimate in pharmacological sophistication is

11 Table 2. Examples of pharmacokinetic-toxicity relationships for established agents used in cancer chemotherapy. Toxicity

Drug

Pharmacokinetic parameter

Reference

Cardiotoxicity

Cyclophosphamide

AUC

19

Urotoxicity

Cyclophosphamide

Metabolism to acrolein

20

Cardiotoxicity

Anthracyclines

Peak level

21

Neurotoxicity

Vincristine

AUC

22

Nephrotoxicity

Cisplatin

Peak level

23,24

Haematological

Carboplatin

AUC

14, 25-28

Haematological

Methotrexate

48-hour plasma level

29

Haematological

6-Mercaptopurine

Red cell thioguanine nucleotides

15

Haematological

Doxorubicin

Steady-state level

30

Haematological

Etoposide

AUC, steadystate level

31-34

relates. Although there are more examples of the latter, the relevance of pharmacokinetic-toxicity v.s. pharmacodynamic-activity relationships should be viewed in the context of the use of the drugs. For palliative drug treatment, often used by the oncologist, manageable and predictable toxicity is a primary requirement. PharTable 1. Examples of pharmacokinetic-response relationships for macokinetic-pharmacodynamic relationships are valuestablished drugs used in cancer chemotherapy. able if they can be exploited in order to achieve this. In Tumour type Pharmacokinetic Reference Drug contrast, if curative therapy is being attempted, more parameter severe toxicity may be accepted, provided that the drug is being used to optimal therapeutic effect, and to Clearance Methotrexate Acute lympho9, 10 achieve this, pharmacokinetic studies may be useful. cytic leukemia Particularly encouraging examples of pharmacoki11 Blast methoMethotrexate Acute lymphonetic-pharmacodynamic relationships are those for trexate polycytic leukemia teniposide [12] and mercaptopurine [15] in the treatglutamates ment of childhood tumours, and carboplatin in the Steady-state 12 Teniposide Paediatric management of both adult and paediatric neoplasms solid level [13, 14, 25-28]. In the work with teniposide [12], a 13,14 AUC Carboplatin Teratoma/ relationship between steady-state plasma teniposide Ovary concentration and response was observed, but there Red cell Acute 15 6-Mercaptowas no clear relationship between dose and response. thioguanine lymphocytic purine In the case of mercaptopurine [15], both event-free and nucleotides leukemia relapse-free survival was improved in children with Lung Steady-state 16 Etoposide acute lymphocytic leukemia who had higher red blood level cell thioguanine nucleotide levels. Similarly, for carbo17 AUC Epimbicin Nasopharynplatin, both AUC-toxicity [25-28] and AUC-activity geal relationships [13,14] have been reported. Relapsed 18 Cytosine Blast cytosine The final criterion for pharmacologically-guided triphosphate arabinoside leukaemia dosing, a low therapeutic index, is perhaps the one arabinoside most convincingly met by anticancer agents. Despite AUC Breast 19 Cyclophosextremely common toxicity, curative activity, particuphamide larly in the common adult neoplasms, is seldom exhib-

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

with a relatively high dose of a drug that has a large therapeutic index. Although some may be overdosed, the large safety margin means that this is of no consequence. If the therapeutic ratio is small, however, overdosing is not an option, and hence dose optimization is necessary for each patient. On the basis of these criteria, anticancer drugs are clear candidates for pharmacokinetically-guided dosing. In cancer therapy, though therapeutic and toxic effects may be easy to measure, they are often elicited some time after drug administration. In the case of response and delayed toxicity, effects may only be seen after multiple courses of therapy. Indeed, if the endpoint of a trial is patient survival, it may take years to define the pharmacodynamics of the agent; this information will therefore be of limited value in dose optimization. Thus, in cancer chemotherapy, it may not be possible to use conventional pharmacodynamic endpoints to determine doses for individual patients. It is conceivable that modern diagnostic techniques (e.g. positron emission tomography or magnetic resonance spectroscopy) could be used to provide real-time information on tumour or normal tissue responses to drugs, but such techniques will not be widely available in the immediate future. With regard to pharmacokinetic-pharmacodynamic relationships in cancer therapy, the extensive literature has been widely reviewed [5-8]. Table 1 illustrates pharmacokinetic-activity relationships and Table 2 gives some examples of pharmacokinetic-toxicity cor-

12 ited, and hence the therapeutic ratio of antitumour agents as a group is poor. Thus, on all three counts the use of anticancer drugs should benefit from the implementation of pharmacologically-guided dosing. To date, attempts to do so have been few and have been largely limited to experimental drugs.

Table 4. Pharmacogenetic variability in cytotoxic drug metabolism. Drug

Enzyme

Effect

Reference

5-Fluorouracil

Dihydropyrimidine dehydrogenase (DPD)

Inactivation

40

6-Mercaptopurine

Thiopurine methyltransferase (TPMT)

Inactivation

41

Oxazaphosphorenes

Aldehyde dehydrogenase (ALDH)

Inactivation

42

Amonafide

Af-acetyltransferase (NACT)

Activation

43

Methods of dose optimization in cancer chemotherapy

Table 3. Patient characteristics used for the adaptive control of cytotoxic drug dosing. Drug

Patient characteristic

Reference

Carboplatin

Glomerular filtration rate

25,37

Etoposide

Serum albumin

38

An thracyc lines

Hepatic function

39

Cisplatin

Glomerular filtration rate

39

agent both qualitatively and quantitatively is extremely exciting. In adaptive dosing, the patient characteristics used to determine the dose of a drug are all factors that are determined prior to treatment. The most sophisticated approach to dose optimization involves adaptive dosing with feedback control once information on the handling of the drug by each individual patient is available. Thus, once the drug has been administered, plasma concentrations are determined in each patient and subsequent doses, dose intervals or the current dose rate (for infusional therapy) adjusted to bring the plasma drug concentration to within the desired range for the required period. The logistical implications of feedback control of dosing are considerable, but a major advantage of the approach is that it reduces the need to define those pretreatment factors that influence the pharmacokinetics of the drug. As such, the approach is directly applicable to any agent for which a rapid drug assay exists. The implementation of adaptive dosing with feedback control has benefited from advances in population pharmacokinetics and the development of limited sampling strategies [35]. These approaches minimize the number of blood samples that have to be taken and analysed, and extract the maximum amount of information from these limited samples. At the moment, the hardware and software required for the implementation of population-based models and limited sampling strategies are not widely available; there is an urgent need to rectify this limitation.

Prospective studies of adaptive dosing in cancer therapy

The prospective evaluation of adaptive dosing strategies can have both pharmacokinetjc and pharmacodynamic endpoints. The pharmacokinen'c endpoint is usually whether the plasma levels of the drug fall within the desired range for the desired time and whether the

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

The goal of dose optimization is to maximize, in each individual patient, the chances of a therapeutic effect while, at the same time, minimizing the risks of toxicity. Possible methods of dosing have been outlined by Egorin [35], namely empiric or non-adaptive dosing, adaptive dosing and adaptive dosing with feedback control. Empiric or non-adaptive dosing is the method most frequently employed by oncologists. On the basis of phase I and phase n studies, a dose (usually scaled to body surface area), is selected and recommended for all patients. Dose reductions or delays are implemented in the face of unacceptable toxicity, but seldom are doses escalated in the absence of toxicity. Due to the poor therapeutic indices of anticancer agents, this undoubtedly means that patients not showing signs of toxicity are being underdosed. One recent example of this problem comes from the evaluation of carboplatin and chlorambucil therapy for the treatment of ovarian cancer [36]. The significant positive relationship between myelosuppression and response led to the recommendation that adaptive dosing should be used for carboplatin to ensure consistent myelosuppression and the maximum chance of activity. Adaptive dosing attempts to take into account the pretreatment patient characteristics known to affect the pharmacokinetics or pharmacodynamics of a drug when determining the most appropriate dose. Table 3 lists certain characteristics that have been recommended for adaptive dosing with specific drugs and, of these, the use of renal function (glomerular filtration rate) to guide carboplatin dosing is the most firmly established example. Recent developments in the field of pharmacogenetics should also be noted in the context of adaptive dosing. Examples relevant to cancer chemotherapy are listed in Table 4. The possibility of phenotyping each patient and hence predicting the metabolism of an

13

Conclusions and future directions The now extensive literature makes it clear that anticancer drugs are excellent candidates for therapeutic drug monitoring with adaptive control of dosing to optimize efficacy and minimize toxicity. Limited proTable 5. Prospective studies of adaptive dosing in cancer chemotherapy. Drug

Use of feedback control

Reduced pharmacokinetic variability

Reduced pharmacodynamic variability

Reference

Carboplatin

No

Yes

Yes

26,27 37,44

Hexamethylene bisacetamide

Yes

Yes

Yes

45

Fluorouracil

Yes

Not reported

Not reported

46

Methotrexatecytosine arabinosideteniposide

Yes

Yes

Not reported

47

Etoposide

Yes

Not reported

Yes

48

Teniposide

Yes

Yes

Yes

49

spective studies, both with and without feedback control, have shown that adaptive dosing is possible, though at present only in specialist centres. The accrual of more such prospective data remains the top priority. In addition, an increase in the number of centres where studies can be performed and improvements in drug assay methodology and pharmacokinetic data analysis are sorely needed. In particular, there is a real need to make computer programs for the analysis of pharmacokinetic data both user-friendly and more widely available. This article posed the question: can pharmacokinetic and pharmacodynamic studies improve cancer chemotherapy? Although definitive prospective clinical studies in common adult tumours have yet to be performed, a wealth of retrospective data and results of a smaller number of prospective studies strongly suggest that cancer chemotherapy will benefit from the exploitation of pharmacokinetic and pharmacodynamic relationships. Data from paediatric studies are particularly encouraging, and it is perhaps in this setting that initial success might be anticipated. References 1. Kerr DJ. Phase I clinical trials: Adapting methodology to face new challenges. Ann Oncol 1994; 5(Suppl. 4): 67-70. 2. Aronson JK, Hardman M. Measuring plasma drug concentrations. BMJ 1992; 305:1078-80. 3. Me Vie JG. The relevance of pharmacology in clinical oncology practice. Pharmacology is (as yet) not relevant ... Ann Oncol 1993; 4:465. 4. Gianni L. The relevance of pharmacology in clinical oncology practice. Pharmacology is relevant Ann Oncol 1993; 4: 463-4. 5. Workman P. The relevance of pharmacology in clinical oncology practice. The moderator thinks that ... Ann Oncol 1993; 4:466-9. 6. Newell DR. Pharmacokinetic determinants of the activity and toxicity of antitumour agents. Cancer Surv 1989; 8: 557-603. 7. Evans WE, Relling MV. Clinical pharmacokinetics-pharmacodynamics of anticancer agents. Clin Pharmacokinet 1989; 16: 327-36. 8. Ratain MJ, Schilsky RL, Conley BA, Egorin MJ. Pharmacodynamics in cancer therapy. J Clin Oncol 1990; 8:1739-53. 9. Evans WE, Crom WE, Abromowitch M et al. Clinical pharmacodynamics of high-dose methotrexate in acute lymphocytic leukaemia. N Engl J Med 1986; 314: 471-7. 10. Borsi JD, Moe PJ. Systemic clearance of methotrexate in the prognosis of acute lymphoblastic leukaemia in children. Cancer 1987; 60: 3020-4. 11. Whitehead VM, Rosenblatt DS, Vlchich M-J, Shuster JJ, Witte A, Beaulieu D. Accumulation of methotrexate and methotrexate polyglutamates in lymphoblasts at diagnosis of childhood acute lymphoblastic leukaemia: A pilot prognostic factor analysis. Blood 1990; 76:44-9. 12. Rodman JH, Abromowitch M, Sinkule JA, Hayes FA, Rivera GK, Evans WE. Clinical pharmacodynamics of continuous infusion etoposide: Systemic exposure as a determinant of response in a Phase I trial. J Clin Oncol 1987; 5:1007-14. 13. Horwich A, Dearnaley DF, Nicholls J et al. Effectiveness of carboplatin, etoposide and bleomycin combination chemotherapy in good prognosis metastatic testicular non-seminomatous germ cell tumours. J Clin Oncol 1991; 9: 62-9. 14. Jodrell DI, Egorin MJ, Canetta RM et al. Relationship between

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

variation in the concentrations is less than predicted for conventional non-adaptive dosing. In some studies, the latter point has been addressed by comparing patients randomized to receive either surface area or pharmacokinetically-guided dosing as part of the same trial. Pharmacodynamic endpoints have varied depending on the nature of the trial. Thus phase I AUC and plasma concentration escalation studies have used toxicity (e.g. leucocyte nadir) as an endpoint, while phase II and HI trials have used response as well as tolerance. Ultimately, for pharmacokinetically guided dosing to be widely implemented, there will have to be significant advantages in terms of pharmacodynamic endpoints, advantages that should be worthwhile in cost-benefit terms. At present, insufficient data are available to make either cost-benefit (monetary) or cost-effectiveness (patient-wellbeing) analyses; the accrual of more information should therefore be a priority. Those studies in which adaptive dosing has been used, either with or without feedback control, are listed in Table 5. In general, the pharmacokinetic goal of reduced interpatient variability in either AUC or plasma concentration has been achieved. Similarly, when toxicity has been a measurable endpoint, adaptive control has usually led to more reproducible toxicity than would have been predicted from historical data. Response data are too limited to allow conclusions at this stage.

14

15. 16. 17.

18. 19. 20.

22. 23. 24.

25.

26.

27.

28. 29. 30.

31.

32. 33.

34. Miller AA, Tolley EA, Niell HB, Stewart CF, Griffin JP. Pharmacodynamics of three daily infusions of etoposide in patients with extensive-stage small-cell lung cancer. Cancer Chemother Pharmacol 1992; 31:161-6. 35. Egorin MJ. Therapeutic drug monitoring and dose optimisation in oncology. In Workman P (ed): New Approaches in Cancer Pharmacology: Drug Design and Development. Berlin: Springer-Verlag, 1992: pp 75-91. 36. Rankin EM, Mill L, Kaye SB et al. A randomised study comparing standard dose carboplatin with chlorambucil and carboplatin in advanced ovarian cancer. Br J Cancer 1992; 65: 275-81. 37. Calvert AH, Newell DR, Gumbrell LA et al. Carboplatin dosage: prospective validation of a simple formula based on renal function. J Clin Oncol 1989; 7:1748-56. 38. Mick R, Ratain MJ. Modelling interpatient pharmacodynamic variability of etoposide. J Natl Cancer Inst 1991; 83: 1560-4. 39. Sulkes A, Collins JM. Reappraisal of some dosage adjustment guidelines. Cancer Treat Rep 1987; 71: 229-33. 40. Harris BE, Carpenter JT, Diasio RB. Severe 5-fluorouracil toxicity secondary to dihydropyrimidine dehydrogenase deficiency. A potentially more common pharmacogenetic syndrome. Cancer 1991; 68: 499-501. 41. Lennard L, van Loon JA, Lilleyman JS, Weinshilboum RM. Thiopurine pharmacogenetics in leukaemia: Correlation of erythrocyte thiopurine methyltransferase activity and 6-thioguanine nucleotide concentrations. Clin Pharmacol Ther 1987; 41:18-25. 42. Hadidi A-HFA, Coulter CEA, Idle JR. Phenotypically deficient urinary elimination of carboxyphosphamide after cyclophosphamide administration to cancer patients. Cancer Res 1988; 48: 5167-71. 43. Ratain MJ, Mick R, Berezin F et al. Paradoxical relationship between acetylator phenotype and amonafide toxicity. Clin Pharmacol Ther 1991; 50: 573-9. 44. Egorin MJ, Van Echo DA, Olman EA et al. Prospective validation of a pharmacologically based dosing scheme for the cis-diamminedichloroplatinum (n) analogue diamminecyclobutanedicarboxylatoplatinum. Cancer Res 1985; 45: 6502-6. 45. Conley BA, Forrest A, Egorin MJ, Zuhowski EG, Sinibaldi V, van Echo DA. Phase I trial employing adaptive control dosing of hexamethylene bisacetamide (HMBA, NSC 95580). Cancer Res 1989; 44: 3436-40. 46. Santini J, Milano G, Thyss A et al. 5-FU therapeutic drug monitoring with dose adjustment leads to improved therapeutic index in head and neck cancer. Br J Cancer 1989; 59: 287-90. 47. Relling MV, Rodman JH, Crom WR, Pui C-H, Evans WE. Individualised administration of anticancer drugs. In Kimura K, Saito H, Carter SK (eds): Cancer Chemotherapy: Challenges for the Future. Vol. 7. Tokyo: Excerpta Medica, 1992: pp 287-96. 48. Ratain MJ, Mick R, Schilsky RI, Vogelzang NJ, Berezin F. Pharmacologically based dosing of etoposide: A means of safely increasing dose intensity. J Clin Oncol 1991; 9:1480-6. 49. Rodman JH, Furman WL, Sunderland M, Rivera G, Evans WE. Escalating teniposide systemic exposure to increase dose intensity for pediatric cancer patients. J Clin Oncol 1993; 11: 287-93.

Correspondence to: Dr D. R. Newell Cancer Research Unit Medical School University of Newcastle upon Tyne Framlington Place Newcastle NE2 4HH, UK.

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

21.

carboplatin exposure and tumour response and toxicity in patients with ovarian cancer. J Clin Oncol 1992; 10: 520-8. Lennard I, Lilleyman JS. Variable mercaptopurine metabolism and treatment outcome in childhood lymphoblastic leukaemia. J Clin Oncol 1989; 7: 1816-23. Desoize B, Marechal F, Cattan A. Clinical pharmacokinetics of etoposide during 120 hours continuous infusions in solid tumours. Br J Cancer 1990; 62: 840-1. Hu OY-P, Cheng S-P, Jame J-M, Chen K-Y. Pharmacokinetic and pharmacodynamic studies with 4-epi-doxorubicin in nasopharyngeal cancer patients. Cancer Chemother Pharmacol 1989; 24: 332-7. Kantarjian HG, Estey EH, Plunkett W et al. Phase I-D clinical and pharmacological studies of high-dose cytosine arabinoside in refractory leukaemia. Am J Med 1986; 81: 387-94. Ayash LJ, Wright JE, Tretyakov O et al. Cyclophosphamide pharmacokinetics: Correlation with cardiac toxicity and tumour response. J Clin Oncol 1992; 10: 995-1000. Brock N, Stekar J, Pohl J, Niemeyer U, Scheffler G. Acrolein, the causative factor of urotoxic side-effects of cyclophosphamide, ifosfamide, trofosfamide and sufosfamide. Arzneimittelforschung 1979; 29:659-61. Bielack SS, Erttmann R, Winkler K, Landbeck G. Doxorubicin: Effect of different schedules on toxicity and anti-tumour efficacy. Eur J Cancer Clin Oncol 1989; 25: 873-82. Desai ZR, van den Berg HW, Bridges JM, Shanks RG. Can severe vincristine neurotoxicity be prevented? Cancer Chemother Pharmacol 1982; 8: 211-4. Campbell AB, Kalman SM, Jacobs C. Plasma platinum levels: Relationship to cisplatin dose and nephrotoxicity. Cancer Treat Rep 1983; 67: 169-72. Reece PA, Stafford I, Russel J, Khan M, Gill PG. Creatinine clearance as a predictor of ultrafiltrable platinum disposition in cancer patients treated with cisplatin: Relationship between peal ultrafiltrable platinum levels and nephrotoxicity. J Clin Oncol 1987; 5: 304-9. Egorin MJ, van Echo DA, Tipping SJ et al. Pharmacokinetics and dosage reduction of cisdiammine(l,l-cyclobutanedicarboxylato)platinum in patients with impaired renal function. Cancer Res 1984; 44: 5432-8. Sorensen BT, Stomgrem A, Jakobsen P, Jakobsen A. Dosetoxicity relationship of carboplatin in combination with cyclophosphamide in ovarian cancer patients. Cancer Chemother Pharmacol 1991; 28: 397-401. Marina NM, Rodman J, Shema SJ et al. Phase I study of escalating targeted doses of carboplatin combined with ifosfamide and etoposide in children with relapsed solid tumours. J Clin Oncol 1993; 11: 554-60. Newell DR, Pearson ADJ, Balmanno K et al. Carboplatin pharmacokinetics in children: The development of a paediatric dosing formula. J Clin Oncol 1993; (In press). Stoller RG, Hande KR, Jacobs SA, Rosenberg SA, Chabner BA. Use of plasma pharmacokinetics to predict and prevent methotrexate toxicity. N Engl J Med 1977; 297:630-4. Ackland SP, Ratain MJ, Vogelzang NJ, Choi KE, Ruana M, Sinkule JA. Pharmacokinetics and pharmacodynamics of longterm continuous-infusion doxorubicin. Clin Pharmacol Ther 1989; 45: 340-7. Bennett CL, Sinkule JA, Schilsky RL, Senekjian E, Choi KE. Phase I clinical and pharmacological study of 72-hour continuous infusion of etoposide in patients with advanced cancer. Cancer Res 1987; 47:1952-7. Miller AA, Stewart CF, Tolley EA. Clinical pharmacodynamics of continuous-infusion etoposide. Cancer Chemother Pharmacol 1990; 25: 361-6. Stewart CF, Arbuck SG, Fleming RA, Evans WE. Relation of systemic exposure to unbound etoposide and hematologic toxicity. Clin Pharmacol Ther 1991; 50: 385-93.

15 Discussion Question: In your presentation you said a lot about interpatient variability, but what about intrapatient variability? Dr Newell: Of the few studies to have looked at this area, most suggest that within-patient variability is substantially less than interpatient variability by a factor of about three. Of course, if you take drug measurements during therapy and use that information to guide treatment for that course, intrapatient variability becomes irrelevant.

Question: You talked about the adaptive control strategies for renally-cleared drugs. What about adaptive control strategies for other agents? Dr Newell: I think the solution is to measure drug levels in real time in the patient. If you cannot do that, a low, non-toxic test dose might be appropriate. However, this might be problematical as you might not want to wait to give your patient a full effective dose, and you might not be able to measure the drug at low levels.

Question: What parameters, other than glomerular filtration rate, are useful in this context? Dr Newell: We are back to non-renal clearance. Carboplatin is very rare in that it is largely cleared by the kidney; most drugs have a component of hepatic clearance. As we understand more and more about the particular isozymes responsible for the metabolism of cytotoxic drugs, it may well become possible to predict how a patient is going to handle a particular agent. The most likely approach to that will be to use innocuous model substrates. Some work has been done on that already, but not systematically. Question: How long will it be before we can determine how a tumour reacts to a drug? Dr Newell: That depends on progress in non-invasive techniques and surrogate measures of drug action. For example, an initiative I very warmly welcome is the one at the Hammersmith Hospital using PET with experimental drugs coming out of the CRC phase I/H clinical trials' committee. Cost is also important; we have to prove that new techniques are cost-effective and result in improved response rates.

Downloaded from http://annonc.oxfordjournals.org/ at New York University on June 29, 2015

Question: There is an adaptive response to cancer chemotherapy at the molecular level. So are we really measuring the right thing? Dr Newell: It may now be possible to use methods like nuclear magnetic resonance spectroscopy and positron emission tomography (PET) to measure what is actually going on in the tumour cell, but whether that can be done in real time and exploited is still very much a research question. However, that is what I would look to in the future. In addition, more studies of molecular pharmacodynamics may enable us literally to titrate a patient to get exactly the levels required.

Question: Is population pharmacokinetics useful here, perhaps as a supplement? Dr Newell: Yes. Basically, you are trying to use information about the way the whole human race handles a drug, in conjunction with a little bit of information about the individual patient, to work out what the pharmacokinetic parameters of that drug are likely to be in that patient.