Therapeutic Drug Monitoring of 21-Day Oral Etoposide in Patients with Advanced Non-Small Cell Lung Cancer’

Vol. 4, 1705-1710, July /998 Clinical Cancer Research 1705

Antonius A. Miller,2 Elizabeth A. Tolley, and but poor in patients whose dosages were increased. The Harvey B. Niell actual observed ANC was compared with the predicted ANC based on the pharmacodynamic model. The predic- Department of Veteran Affairs Medical Center [A. A. M., H. B. N.] and University of Tennessee [A. A. M., E. A. I., H. B. N.], Memphis, tion was considered accurate if the predicted and actual Tennessee 38163 ANC values were within 500/pt of each other. Using this margin, the ANC was accurately predicted in 10 of 22 patients. Etoposide concentrations >0.3 pig/ml on this ABSTRACT schedule were significantly correlated with combined grades The purpose of this study was to prospectively test a 3 and 4 neutropenia (P < 0.0001). In conclusion, the phar- pharmacodynamic model for therapeutic drug monitoring macodynamic model is statistically sound when applied to a of 21-day oral etoposide. In our previous studies, etoposide population of patients. However, when applied to individual trough concentrations on this schedule were related to the patients for therapeutic drug monitoring, the model lacks hematological toxicity, expressed as WBC and neutrophil precision and accuracy. counts at the nadir. The following pharmacodynamic model estimated the absolute neutrophil count at the nadir (ANC) based on the etoposide concentration (Er) and the pretreat- INTRODUCTION ment count (ANC): Measurements of drugs like aminoglycosides, digoxin, phenytoin, or theophylline for making decisions during treat- ANC = 0.32(1 + ANC x e 2.47 X Ec) ment are accepted practice in internal medicine. Useful strate- Patients were treated with 40 mg/m2/day etoposide p.o. x 21 gies exist for these drugs such that concentrations can be deter- days and 100 mg/m2 cisplatin i.v. on day 1. All patients had mined at steady-state (i.e., theophylline) or at specified times non-small cell lung cancer stage 11111 or IV, had a perfor- (i.e., peak and trough for arninoglycosides) and doses adjusted, mance status of 0-2, and had a median age of 66 (range, depending on how far off the measured concentration is from 42-80). Etoposide was measured in the plasma on day 8 by the target (1). In cancer therapy, however, therapeutic drug high-performance liquid chromatography, and dosage ad- monitoring has not found a standard role except for methotrex- justments were made for the remainder of the course. We ate measurements to guide leucovorin rescue after high-dose targeted for grade 3 neutropenia (ANC, 500 to 999/pI) and therapy (2). Clinical oncologists are well aware of the narrow attempted to avoid grade 4 neutropenia (ANC0, <500/pA). therapeutic index of antineoplastic drugs and make dose adjust- of 25 patients entered, 22 were evaluable for therapeutic ments after toxicity is encountered. The problem of making drug monitoring in the first course. Three patients devel- correct dosing decisions is at least in part attributed to our oped grade 3 neutropenia, and seven patients developed incomplete clinical assessment of the pharmacokinetic variabil- grade 4 neutropenia. Etoposide concentrations were signif- ity between patients, which in turn determines the pharmacody- icantly correlated with ANC in the first course (r -0.50, name variability (3). Pharmacodynamics is defined as the rela- P < 0.02). For those patients whose dosages were not tionship between drug concentrations and therapeutic or toxic changed, the estimated correlation between predicted and responses. The problem of interpatient variability and the in- ability to predict the severity of toxicity cannot be overcome actual ANC was 0.77 (P < 0.01). No evidence of significant bias of the pharmacodynamic model was detected. The eto- unless the relationships between pharmacological measurements poside dosages were increased in 12 patients and were not and clinical effects are explored. The toxicity of anticancer changed in the remaining patients. The precision of the chemotherapy is mainly due to growth inhibition of rapidly model was good in patients whose dosages were not changed dividing normal tissues exemplified by bone marrow suppres- sion. Troublesome consequences of myelosuppression are neu- tropemic fever and sepsis. These adverse consequences of chemotherapy are not immediately apparent, and the time interval and degree of toxicity are poorly predictable. Received 1/16/98; revised 4/3/98; accepted 4/10/98. Lung cancer is the leading cause of death from cancer in The costs of publication of this article were defrayed in part by the the United States (4). The majority of patients have non-small payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to cell lung cancer. Patients with advanced non-small cell lung indicate this fact. cancer commonly receive chemotherapy for palliation of their Supported by a Merit Review Grant from the Office of Research and disease. Our previous Phase II study demonstrated that a 2 1 -day Development, Medical Research Service, Department of Veteran Af- schedule of oral etoposide in combination with iv. cisplatin was fairs. active in patients with advanced non-small cell lung cancer (5, 2 To whom requests for reprints should be addressed, at University of Tennessee, 3 North Dunlap, Memphis, IN 38163. Phone: (901) 448- 6). Etoposide was administered at a dosage of 50 mg/rn2 for 21 5157; Fax: (901) 448-5033; E-mail: [email protected]. days, and 100 mg/rn2 cisplatin were given on day 1 of a 28-day

cycle. Etoposide plasma concentrations were measured weekly PATIENTS AND METHODS just before the daily dose (trough levels). The etoposide con- Patients had biopsy-proven non-small cell lung cancer centrations on this schedule were related to the hematological stage IIIB or IV, measurable or evaluable disease, and a per- toxicity, expressed as WBC and neutrophil counts at the nadir formance status of 0-2 (Eastern Cooperative Oncology Group). (5). The following pharmacodynamic model was developed to Staging procedures were the same as in our previous studies estimate the absolute neutrophil count at the nadir (ANC) based (5-7). Prior chemotherapy and bone marrow or brain metastases on the etoposide concentration (Er) and the pretreatment count were reasons for exclusion. Required laboratory values for entry (ANC): on study were: ANC l500/i.l; hemoglobin 10 g/dl, platelets l00,000/iil; serum creatinine 1.5 mg/dl; and bilirubin 1 .5 X normal. Patients had to be able to swallow capsules so ANC = 0.32(1 + ANC X e247 X EC) they could comply with the oral etoposide regimen. Patients with another malignancy and other serious medical or psychi- Subsequently, this model was tested prospectively and per- atric disease were excluded. All patients gave written informed formed reliably (7). consent. A similar regimen consisting of 21-day oral etoposide (50 The treatment consisted of 40 rng/m2/day etoposide p.o. X mg/m2/day) with cisplatin given at a dosage of 33 mg/rn2 i.v. on 21 days and ]00 mg/rn2 cisplatin iv. on day 1. Treatment cycles days 1, 2, and 3 was used by the CALGB3 for extensive stage were repeated every 28 days. Up to three cycles of therapy were small-cell lung cancer (8). Multiple linear regression analysis of given with therapeutic drug monitoring. After that, patients were this CALGB trial led to the following pharrnacodynamic model off protocol but could receive a maximum of six cycles. Therapy (9): was given in the outpatient clinic. Bristol-Myers Oncology (Princeton, NJ) provided 10-mg capsules of etoposide, which log10(ANC) = - 0.153 - 0.017 X age - 0.363 were similar in formulation to the marketed 50-mg capsules

x E + 0.530 x log10 (alkaline phosphatase) (VePesid). The trial was conducted under Investigational New Drug 43,71 1 from the Food and Drug Administration (Rock- ville, MD). The commercially available formulation of cisplatin However, this latter pharmacodynamic equation was not avail- was used. able at the beginning of the trial that is described here. It is All patients had laboratory studies performed before study important to note that the above equations were derived from entry and before each treatment cycle: complete blood cell count groups of patients (population pharmacology), and that their with differential; electrolytes (including calcium and magnesi- performance when applied to individual patients (therapeutic urn); liver and renal function chemistries; lactate dehydrogem- drug monitoring) remained unknown. ase; total protein; and albumin. While on treatment, patients In previous studies of 50 mg/m2/day X 21 days, we ob- returned weekly to clinic for monitoring of compliance with oral served grades 3 and 4 neutropenia in 73- 83% of all patients etoposide, complete blood cell counts, and etoposide plasma (5-9). Therefore, the starting dose was reduced from 50 to 40 concentrations. For therapeutic drug monitoring, weekly etopo- mg/m2/day X 21 days for this trial. Etoposide is marketed in the side samples were obtained just before the daily dose (trough United States only as a 50-mg capsule. This capsule size was used in the previous studies (5-9). For therapeutic drug moni- concentrations). The etoposide concentration on day 8 of cycle toting, the 50-mg capsules were not useful because dose adjust- 1 was used to predict the neutrophil count at the nadir based on ments were not feasible. For the sake of prospective therapeutic the following equation: ANC, 0.32 (1 + ANC X e247 X Ec) dose adjustments, 10-mg capsules were obtained from Bristol- Based on this prediction, the dose of etoposide was adjusted for Myers Oncology. Otherwise, the treatment was the same as in the remainder of the cycle. The goal was to avoid grade 4 the previous trials that led to the development of the pharrna- neutropenia (ANC <500/pA; Common Toxicity Criteria by the codynamic model in patients with mom-small cell lung cancer National Cancer Institute) by means of dose reduction. Grade 3 was deemed acceptable, and (5-7). neutropenia (ANC 500-999/pA) Our hypothesis was that therapeutic drug monitoring was no dose change was made. For grades 0, 1 , and 2 projected useful in predicting hematological toxicity and thereby in- neutropema (ANC 1000/pA), dose increases were planned. creased the safety of treatment with etoposide. To test this Patients were seen in the clinic in the morning, the blood sample hypothesis, we treated patients with advanced non-small cell was taken to the laboratory immediately for analysis of the lung cancer with the regimen of 21-day oral etoposide and iv. etoposide concentration the same day, and patients were called cisplatin on day 1 and measured etoposide trough concentrations in the afternoon with dosage instructions. Dose adjustments were made in 20-mg increments. Predicted and actual nadir on day 8 of cycle 1 . The etoposide dosage was adjusted prospectively, depending upon the drug concentration. The purpose counts were recorded. Etoposide plasma concentrations were was to determine whether this drug monitoring of etoposide was measured by high-performance liquid chromatography accord- clinically useful in treating individual patients. ing to a method published previously (5). The lower limit of detection was 0.05 .tg/ml. The day-to-day coefficients of van- ation for etoposide measurements were less than 10%, as in the previous studies (5-9). Statistical Methods. Pearson correlation coefficients

3 The abbreviation used is: CALGB, Cancer and Leukemia Group B. were estimated for evaluating the relationships between etopo-

Table 1 Results in the first treatment cycle

Neutrophil nadir” Neutropen ia grade Patient Etoposide concentration No. Age (p.g/ml plasma) Predicted Actual Predicted Actual Dose

1 58 Interference on HPLC 3.07 0 No change 2 62 0.21 1.22 4.40 2 0 Increased 3 49 0.10 1.27 2.80 2 0 No change 4 80 0.42 0.53 0.24 3 4 No change S 72 0.22 0.70 0.48 3 4 No change 6 71 0.10 1.27 1.46 2 2 Increased 7 59 0.21 1 .00 1 .65 2 1 Increased 8 67 Noncompliant 1.75 1 No change 9 69 0.41 1.83 0.11 1 4 Increased 10 49 0.28 1.45 5.09 2 0 Increased I 1 69 0.73 0.54 0.87 3 3 No change 12 66 0.10 0.93 1.59 3 1 No change 13 66 0.48 0.84 0.32 3 4 No change 14 72 0.38 0.93 0.81 3 3 No change 15 66 0.85 0.55 0.43 3 4 No change 16 45 Noncompliant 17 80 0.31 1.31 0.10 2 4 Increased 18 63 0.56 1.86 0.39 1 4 Increased 19 74 0.53 1.05 0.70 2 3 No change 20 57 0.15 1.09 1.67 2 1 No change 21 42 0.16 3.41 3.60 0 0 Increased 22 51 0.28 1.71 5.40 1 0 Increased 23 77 0.11 2.65 3.04 0 0 Increased 24 51 0.10 3.19 3.98 0 0 Increased 25 47 0.09 1.64 2.09 1 0 Increased a Counts times 103/j.a.l.

side concentration and ANC, predicted and actual ANC,1, age 13 were African-American. The performance status was 0 in 4 and ANC, and age and etoposide concentration. The t test was patients, 1 in 14 patients, and 2 in 7 patients. The tumor stage applied to evaluate the relationships between race and etoposide was IIIB in 3 patients and IV in 22 patients. Table 1 shows the concentration and race and ANC. Differences between actual results for the first treatment cycle. The predicted values for the and predicted values of nadir counts were obtained by subtract- neutrophil nadir in this table were based on the equation: ANC,, ing the predicted values from the actual ones. The mean differ- = 0.32 (1 + ANC x C247 Ec) No patient had a dosage ence between actual nadir counts and predicted nadir counts is reduction. The etoposide dosages were increased in 12 patients. a measure of bias. Whether bias equaled zero was evaluated Three patients (Table 1, numbers 3, 19, and 20) did not have the with a paired t test. The SD of the difference between actual and planned dose increases due to problems with communications. predicted nadir counts is a measure of precision. Whether this The observed hematological toxicity was substantial, with three value was 500/jJ was tested with a x2 test. We classified the patients developing grade 3 neutropenia and 7 patients devel- prediction as accurate if the actual and predicted nadir counts oping grade 4 meutropemia in the first treatment cycle (Table 1). were within 500/pA of each other. The risk of combined grade 3 The total number of cycles administered on this protocol was 37 and 4 neutropenia for E >0.3 pg/m1 versus E <0.3 p.g/ml (median per patient, 1; range, 1-3). was estimated by the two-tailed Fisher’s exact test. Whether Correlations and Associations. In the whole population dose change was related to the development of neutropenia was of 22 patients, etoposide concentrations were significantly cor- t also evaluated with Fisher’ s exact test. The test was used to test related with ANC in the first treatment course (r = -0.50, P < whether the mean ANC,, values of the two groups of patients 0.02) and in all courses (r = -0.48, P < 0.01). Predicted and (those without dose adjustment versus those with dose adjust- actual ANC were correlated when tested for all patients in the ment) were similar. Unless otherwise stated, the mean ± SE is first course (r = 0.52, P < 0.02). The patients’ age was presented. correlated with ANC in the first course (r = -0.68, P < 0.0005) but not with the etoposide concentration (r = 0.37, P < RESULTS 0. 10). The patients’ race was neither related to ANC nor the Of 25 patients entered, all were eligible, and 22 were etoposide concentration (both P > 0.3). For those patients evaluable for therapeutic drug monitoring in the first course. whose dosages were not changed, the estimated correlation One patient was inevaluable because of a peak that interfered between predicted and actual ANC was 0.77 (P < 0.01). with the etoposide peak on the chromatogram, and two patients Bias and Precision. For those patients whose dosages were inevaluable because of noncompliance with the prescribed were not changed, no evidence of significant bias of the phar- etoposide treatment. All patients were males. Their median age macodynamic model was detected. The estimated bias was 148 was 66 years (range, 42-80). Twelve patients were white, and ± 199/pA (P < 0.5). The precision of the predictor equation was

Another way tojudge the clinical utility of therapeutic drug monitoring is to evaluate how often grade 3 or 4 toxicity was accurately predicted. Ten patients developed grade 3 or 4 meu- tropenia; grade 3 was accurately predicted in 2 of 3 patients, and grade 4 was predicted in 0 of 7 patients. However, Table 1 shows that E > 0.3 p.g/ml was associated with grades 3 and 4 U neutropenia (i.e., ANC < 1000 i.1). Nine of nine patients (numbers 4, 9, 11, 13-15, and 17-19) with E > 0.3 i.g/ml developed grade 3/4 neutropenia, whereas 1 of 13 patients (number I 5) with E < 0.3 p.g/ml had grade 3/4 neutropenia. The risk of grade 3/4 neutropenia was 13 times (confidence interval, 2-85; P < 0.0001) higher for E > 0.3 rig/mi than for E < 0.3 ig/mI. The pharmacodynamic model was developed for a 21-day 1 2 3 4 5 course of unaltered etoposide dosage. Because of study design, the predicted ANC values of those patients whose dosages were Actual ANC x 1OId increased were greater than the predicted ANC of those patients Fig. 1 Relationship between the actual and predicted ANC (ANC, whose doses were kept the same. When the dosage on day 8 of cells X l03/p.l) from cycle I of treatment based on the prediction cycle I was increased based on the prediction of ANC > equation ANC = 0.32 (1 + ANC X e247 X Ec) where ---- are ±5004a.l from the solid line of identity. #{149},patients whose dosages were 1000/il, this decision rendered the predicted ANC value po- not changed: 0, patients whose dosages were increased. tentially invalid. Therefore, as expected, the proportion of patients with incorrectly predicted ANC was larger in the group who had the dose increase compared with the group without dose change, but this was not statistically significant (P < 0.21). 6294a.I, a value that was not significantly different from 500/pi Inspection of the types of incorrectly predicted ANC revealed (P < 0.11). that increasing the dose did not significantly increase the pro- The actual ANC of three patients (Table 1, numbers 9, 17, portion of patients with actual ANC that were lower than and 1 8) whose doses were increased were lower than predicted, predicted (P < 0.6). Of those patients whose doses were in- which was an expected effect of dose adjustment. However, the creased, 5 of 12 had actual ANC values that were more than ANC of three other patients (numbers 2, 10, and 22) whose 500/il greater than predicted. The actual ANC values for the doses were increased were higher than predicted, which was two groups of patients were significantly different: 991 ± unexpected. Because these results tended to offset each other, no 254/il for those whose doses were not changed versus 2609 ± evidence of significant bias was detected for patients whose 552/il for those whose doses were increased (P < 0.02). These doses were increased. The estimated bias was 731 ± 5414U results suggest that on average the pharmacodynamic model (P < 0.4). However, precision was 1 874/pA, a value much correctly identified patients with less hematological toxicity higher than 500/pA (P < 0.0001). from etoposide. However, to be clinically useful the model must Accuracy. The actual observed ANC was compared not only predict who will not develop neutropenia, but more with the ANC that was predicted based on the pharmacody- importantly who will develop severe neutropenia. Of those namic model (Fig. I ). The prediction was considered accurate if patients whose dosages were kept the same, 4 of I 0 (40%) the predicted and actual ANC values were within 500/ji.l of developed grade 4 neutropenia compared with 3 of 12 (25%) each other. Using this criterion, the ANC was accurately pre- patients in whom the dosages were increased (P < 0.7). dicted in 10 of 22 patients (Table 1 , numbers 4-6, 1 1, 14, 15, All of the above work was performed prospectively based 19, 21, 23, and 25). For patients whose dosages were not on the previously validated equation ANC = 0.32 (1 + ANC changed, the accuracy of the predictor equation: ANC = 0.32 x e247 Ec) As stated in the introduction, another pharma- (I + ANC X e247 Ec) was 0.60, with actual ANC being codynamic equation was developed while this trial was in pro-

within 5004tl of predicted values for 6 of 10 patients; actual gress: log10 (ANC) = -0.153 - 0.017 X age - 0.363 X E + ANC were within 660/pi of the predicted values for three other 0.530 X log10 (alkaline phosphatase). This model was devel- patients. For those patients whose dosages were increased, the oped by multiple linear regression of pharmacodynamic data accuracy of the predictor equation was 0.33, with 4 of 12 from 83 patients treated on a CALGB study. Applying this patients having accurately predicted ANC. Inspection of Fig. 1 CALGB model retrospectively to the data available from the 22 shows that three patients each were either far above or far below patients in this trial did not result in any statistical improvement the line of identity. These patients can be identified in Table 1 in prediction (Fig. 2). The CALGB model also failed to identify as follows. Patients numbered 2, 10, and 22 were predicted to the three patients whose doses were increased and who devel- have ANC < 2000/pJ but actually had ANC > 4000/i.l (far oped grade 4 neutropenia (Table 1, patients 9, 17, and 18). below the line of identity). Patients numbered 9, 17, and 18 were predicted to have ANC above l000/p.l, but their actual ANC were below 500/p.l (far above the line in Fig. I). These three DISCUSSION patients, in whom the pharmacodynamic model failed to predict Both pharmacodynamic models appear to contain influen- ANC < 500/pJ, illustrate how a statistically significant model tial predictor variables such as E, ANC, age, and alkaline can have limited clinical utility. phosphatase. The relationship among these predictors is corn-

aspects: absorption, distribution, metabolism, and excretion. Pharmacokinetic variability may be captured by measuring plasma concentrations. For instance, a patient with impaired clearance should have increased plasma concentrations. Dose reduction in a patient like this should result in average plasma concentrations and reduced toxicity. On the other hand, a patient may have an enhanced sensitivity to the drug. This pharmacodynamic variability may be related to prior therapy, bone ma- row abnormalities, age, performance status, genetics, other mcd- ications, comorbid conditions, and other poorly understood reasons (3, 11, 12). In a patient with enhanced sensitivity,

I average plasma concentrations may result in severe toxicity. Lowering the dose in such a patient may result in lower plasma concentrations and average toxicity (12). The difference in the two pharmacodynamic models points Actual ANC z 1OId out that ANC is not a reliable predictor. In ANC = 0.32

Fig. 2 Relationship between the actual and predicted ANC (ANC, (1 + ANC X e247 Ec) ANC contributed significantly to he cells X l0/l) from cycle 1 of treatment based on the prediction model. In the CALGB study (9), no significant association equation log10(ANC) -0.153 - 0.017 X age - 0.363 X E + between ANC and the nadir count was demonstrated, and ANC 0.530 X log10 (alkaline phosphatase), where are ± 500/p.l from the was left out of the model: log10(ANC) -0.153 - 0.017 X solid line of identity. #{149},patients whose dosages were not changed; 0, patients whose dosages were increased. age - 0.363 x E + 0.530 X log10 (alkaline phosphatase). Two potential confounding problems should be acknowl- edged: (a) the pharmacodynamic models were developed for 50-mg etoposide capsules given at a dosage of 50 mg/m2/day. This plex. When applied to the whole group of patients, the models trial used 10-mg etoposide capsules of the same formulation given yielded statistically significant results; thus, the present study at a starting dosage of 40 mg/m2/day; and (b) cisplatin was given confirmed the concept of population pharmacology. However, iv. on day 1 in combination with the 21-day course of p.o. etopo- to be clinically useful the pharmacodymamic model must make side, but cisplatin was not measured. Therefore, the contribution of the transition to therapeutic drug monitoring. Here, a clinician is cisplatin to the neutropenia was not assessed. The previous studies dealing with an individual patient rather than with a population ofSO rng/m2/day etoposide X 21 showed an incidence of combined of patients. Applying the phamacodynamic models to imdivid- grades 3 and 4 neutropenia between 73 and 83% (5-9). The current ual patients for therapeutic drug monitoring lacked accuracy and study had an incidence of grades 3 and 4 neutropenia of 45% precision. Specifically, the etoposide trough concentration om day 8 (Table 1). Thus, the dose reduction to 40 mg/m2/day etoposide X of cycle 1 did not reliably predict the ANC for patients whose 21 had the desired effect. doses were increased. However, E > 0.3 g/m1 was significantly Fig. 1 illustrates the limitations of using our pharmacody- correlated with grades 3 and 4 neutropenia in spite of dose in- namic model in this prospective trial. Fig. 2 shows the retro- creases on day 8. This is consistent with work by Clark et a!. (10), spective application of the CALGB model, which performed who reported a significantly increased risk ofANC < 1000/il for worse. Overall, it appears doubtful that applying the models will exposure to E > 3 p.g/ml on a 5- or 8-day regimen of etoposide. result in a marked improvement in clinical care. On the other Grochow (1 1) has pointed out that the logic underlying the hand, dose adjustments based on the first model did not increase monitoring of plasma concentrations is based on six premises: the incidence of severe neutropemia; thus, patients were not (a) primary clinical end points (equivalent to measuring blood harmed by applying the model. sugar or blood pressure) are mot easily assessed as a basis for Our current understanding of the pharmacodynamics of eto- altering drug doses; (b) an accurate, precise, specific, and sen- poside was recently reviewed by McLeod and Evans (13). The sitive assay for drug measurements is available; (c) individual influence of etoposide systemic exposure on the hematological variations in drug elimination are the source of different drug toxicity has been clearly demonstrated. The equations that best concentrations; (d) reducing variations in drug disposition describe the relationship vary with each study and have included (pharmacokimetics) cam reduce variations in drug effect (phar- AUC, concentrations at steady-state (C), and trough concentra- macodynamics); (e) the relationship between drug concentration tions (Er) as in our study. As reviewed by McLeod and Evans (13), and effect is closer than the relationship between drug dose and two studies have attempted therapeutic drug monitoring of etopo- effect; and even when a drug satisfies all these logical criteria, side, one by Ratain et a!. (14) and one by Madden et a!. (15). (t) the utility of measuring drug concentrations in patients and In the prospective study of pharmacologically based dosing individualizing doses must be demonstrated to maximize useful by Rataim et al. (14), 45 patients with a variety of solid tumors effects and manage toxicity within acceptable limits. Our study were randomly assigned to receive a fixed dose of etoposide 125 fails to meet the sixth criterion. There may be a variety of rng/m2/day by 72 h continuous infusion or to receive individu- reasons for variable toxicity that can generally be categorized as alized dosing to a target WBC nadir of 1700/il (grade 3 leu- either pharmacokinetic or pharmacodymamic. Ratain (12) has kopenia corresponding to our grade 3 neutropenia). Dosage emphasized that it is important to understand the distinction. modification was determined using an exponential model that Pharmacokinetics is classically considered to consist of four included total etoposide dose, starting dose, 24-h etoposidc

concentration (C,, corresponding to our Er), patient perform- REFERENCES ance status, albumin, and bone marrow function (based on prior 1. Bauer, L. A. Individualization of drug therapy: clinical pharmacoki- transfusion requirements). Our patients were newly diagnosed netics and pharmacodynamics. In: J. T. DiPiro, R. L. Talbert, G. C. Yee, with non-small cell lung cancer without prior transfusion re- G. R. Matzke, B. G. Wells, and L. M. Posey (eds.), Pharmacotherapy-A Pathophysiologic Approach, Ed. 3, pp. 29-48. Stamford, CT: Appleton quirements, and albumin and performance status did not con- & Lange, 1997. tribute significantly to the model. In the study by Ratain et al. 2. Ratain, M. J., Schilsky, R. L., Conley, B. A., and Egorin, M. J. (14), the total dose was increased in patients with individualized Pharmacodynamics in cancer therapy. J. Clin. Oncol., 8: 1739-1753, dosing, and this was associated with a decrease in the mean 1990. WBC at the nadir (15 10 ± 950 versus 2500 ± 1420/il). Both 3. Miller, A. A., Ratain, M. J., and Schilsky, R. L. Principles of the study by Ratain et a!. (14) and our study resulted in an pharmacology. In: M. C. Perry (ed), The Chemotherapy Sourcebook, Ed. 2, pp. 27-41. Baltimore, MD: Williams and Wilkins, 1997. increase in the dose intensity without increasing the incidence of 4. Landis, S. H., Murray, T., Bolden, S., and Wingo, P. A. Cancer life-threatening toxicity. Although the pharmacodynamic model statistics, 1998. CA Cancer J. Clin., 48: 6-29, 1998. in the study by Rataim et al. (14) did appear to accurately predict 5. Miller, A. A., Tolley, E. A., Niell, H. B., Griffin, J. P., and Mauer, the WBC nadir, the degree of variability around the target value A. M. Pharmacodynamics of prolonged oral etoposide in patients with indicated that a more specific model was needed before large- advanced non-small-cell lung cancer. J. Clin. Oncol., 11: 1179-1188, scale implementation of this dosing approach (14). This limita- 1993. 6. Miller, A. A., Niell, H. B., and Griffin, J. P. Phase II study of tion also applies to our study [in fact, Fig. 4A in the publication prolonged oral etoposide in combination with intravenous cisplatin by Rataim et a!. (14) and our Fig. 1 are similar]. in advanced non-small cell lung cancer. Lung Cancer, 12: 59-65, In the trial of individualized etoposide therapy by Madden et 1995. aL (15), 17 children with acute nonlymphocytic leukemia received 7. Miller, A. A., and Tolley, E. A. Predictive performance of a phar- a 70 mg/rn2 loading dose, followed by 500 mg/m2/day for 96 h. A macodynamic model for oral etoposide. Cancer Res., 54: 2080-2083, 1994. blood sample was obtained 6 h after the start of infusion, and 8. Miller, A. A., Herndon, J. E., H, Hollis, D. R., Ellerton, J., Langle- dosage rates were adjusted to achieve a target plasma concentration ben, A., Richards, F., and Green, M. R. Schedule dependency of 21-day within 12 h. Ten of the 17 patients required a dosage increase. This oral versus 3-day intravenous etoposide in combination with intravenous study did not provide further details of the predictive performance cisplatin in extensive-stage small-cell lung cancer: a randomized phase of the model used for dosage adjustment (15). ifi study of the Cancer and Leukemia Group B. J. Clin. Oncol., 13: 1871-1879, 1995. In conclusion, the pharmacodynamic model is statistically 9. Miller, A. A., Rosner, G. L., Ratain, M. J., Hollis, D. R., Green, sound when applied to a population of patients. However, when M. R., and Schilsky, R. L. Pharmacology of 21-day oral etoposide given applied in individual patients for therapeutic drug monitoring, our in combination with iv. cisplatin in patients with extensive-stage small present model lacks precision and accuracy. Increasing the sample cell lung cancer: a Cancer and Leukemia Group B study (CALGB size of the trial is not expected to be helpful. This will improve the 9062). Clin. Cancer Res., 3: 719-725, 1997. PS for the population; however, a review of the large CALGB trial 10. Clark, P. I., Slevin, M. L., Joel, S. P., Osborne, R. J., Talbot, D. I., indicates that great variability existed and that predictions for Johnson, P. W. M., Reznek, R., Masud, T., Gregory, W., and Wrigley, P. F. M. A randomized trial of two etoposide schedules in small-cell individual patients will remain poor (9). Although disappointing, lung cancer: the influence of pharmacokinetics on efficacy and toxicity. this dilemma is not unique in medical oncology. For instance, we J. Clin. Oncol., 12: 1427-1435, 1994. know for a common malignancy and a certain stage what the 11. Grochow, L. B. Individualized dosing of anti-cancer drugs and the median survival time is in the population. However, when faced role of therapeutic monitoring. in: L. B. Grochow and M. M. Ames with an individual patient, the statistically valid figure of a given (eds.), A Clinician’s Guide to Chemotherapy Pharmacokinetics and Phar- macodymamics, pp. 3-16. Baltimore, MD: Williams and Wilkins, 1998. survival time is clinically meaningless. Because of the variability 12. Rataim, M. J. Pharmacokinetics and pharmacodynamics. in: V. T. around the median, which is usually large, even an experienced DeVita, Jr., S. Hellman, S. A. Rosenberg (eds.), Cancer: Principles & clinician cannot tell the individual patient what his/her survival Practice of Oncology, Ed. 5, pp. 375-385. Philadelphia: Lippincott- time will be. We will need to refine the current predictor variables Raven, 1997. and search for other variables that improve the pharmacodynamic 13. McLeod, H. L., and Evans, W. E. Epipodophyllotoxins. in: L. B. model. A study of pharmacodynamic modeling of 21-day, single- Grochow and M. M. Ames (eds.), A Clinician’s Guide to Chemotherapy agent etoposide in patients with relapsed non-Hodgkin’s lym- Pharmacokinetics and Pharmacodynamics, pp. 259-287. Baltimore, MD: Williams and Wilkins, 1998. phoma is presently in progress in CALGB. 14. Ratain, M. J., Mick, R., Schilsky, R. L., Vogelzang, N. J., and Berezin, F. Pharmacologically based dosing of etoposide: a means of safely increasing dose intensity. J. Clin. Oncol., 9: 1480-1486, 1991. ACKNOWLEDGMENTS 15. Madden, T., Crom, W. R., and Mirro, J. Disposition of etoposide We gratefully acknowledge helpful comments by the senior editor, (VP-16) in children with acute nonlymphocytic leukemia (ANLL). Clin. Dr. Bruce A. Chabner. Pharmacol. Ther., 47: 206-210, 1990.

Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 1998 American Association for Cancer Research. Therapeutic drug monitoring of 21-day oral etoposide in patients with advanced non-small cell lung cancer.

A A Miller, E A Tolley and H B Niell

Clin Cancer Res 1998;4:1705-1710.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/4/7/1705

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/4/7/1705. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.