Prospective evaluation of pharmacokinetically guided dosing of carboplatin in Japanese patients with cancer

Abstract

The Calvert formula, that is, carboplatin dose (mg) = target area under the concentration versus time curve (AUC) × (glomerular filtration rate [GFR] + 25), has not been validated in Japanese subjects in whom the GFR was accurately measured. The purpose of this study is to evaluate the validity of this formula for Japanese patients with cancer and modify it for this population. The GFR was measured on the basis of inulin clearance, which is considered to reflect the accurate GFR. Inulin clearance was measured in 28 patients with cancer. The adjusted 24‐h creatinine clearance (24‐h Ccr) was unbiased (mean prediction error [MPE] ± SE = −2.3 ± 4.5%) and acceptably precise (root mean squared error = 23.7%) for GFR assessment. The pharmacokinetics of carboplatin were analyzed in 21 patients with a GFR of 17.2–91.4 mL/min. The original Calvert formula overestimated carboplatin clearance, resulting in an MPE of 14.3%. When we revised the Calvert formula for Japanese patients by substituting a non‐renal clearance of 15 for 25, that is, dose = target AUC × (GFR + 15), the MPE decreased to −0.1% (P < 0.001). We conclude that the adjusted 24‐h Ccr is acceptably precise for GFR assessment, and the non‐renal clearance of carboplatin is suggested to be lower in Japanese patients with cancer than in their Western counterparts. (Cancer Sci 2010; 101: 2601–2605)

Carboplatin, a second‐generation platinum‐containing compound, is widely used to treat solid tumors, especially lung and gynecological cancers. Thrombocytopenia is the major dose‐limiting toxicity and strongly correlates with the area under the plasma concentration versus time curve (AUC) of carboplatin.^{( 1 )} Because carboplatin clearance is closely related to the glomerular filtration rate (GFR),^{( 2 )} Calvert and colleagues developed a formula to estimate the dose of carboplatin required to produce target AUC values in individual patients (the Calvert formula):^{( 3 )}

There are well‐known discrepancies in toxicity of carboplatin‐based chemotherapy between Japanese and Western patients,^{( 4 )} which can be explained in part by the different techniques used to assay creatinine. Because the measurement of the GFR is complex, expensive and impractical in routine clinical practice, creatinine clearance (Ccr) is often substituted for GFR in the Calvert formula.^{( 5 )} The 24‐h creatinine clearance (24‐h Ccr) is the volume of blood plasma that is cleared of creatinine per unit time and is a useful measure for approximating the GFR. Two techniques are commonly used to measure serum creatinine levels: (i) the kinetic Jaffe method; and (ii) the enzymatic peroxidase–antiperoxidase (PAP) method. The Ccr measured by the PAP method overestimates the GFR in subjects with normal renal function,^{( 6 )} and most clinical laboratories in Japan use the PAP method. We previously found that the observed carboplatin clearance is lower than that calculated by the Calvert formula when the 24‐h Ccr estimated by the PAP method is substituted for the GFR.^{( 7 )} To avoid potential overdosing, we proposed an amended formula in which the 24‐h Ccr is adjusted by adding 0.2 mg/dL to the serum creatinine level as measured by the PAP method.^{( 8 )} We then prospectively validated this formula.^{( 9 )} Renal function has also been evaluated by equations using the GFR estimated on the basis of the serum creatinine level.^{( 10} , ¹¹ , ^{12 )} At present, there is no global consensus on the best method for assessing renal function as the basis for determining the dosage of carboplatin.

Ethnic background might be a major determinant of treatment outcomes in patients who receive carboplatin. Interethnic differences have been recognized as important contributory factors to interindividual variability of drug response or toxicity.^{( 13 )} Historically, East Asians are more susceptible to the effects of carboplatin‐based chemotherapy than their Western counterparts.^{( 14 )}

The Calvert formula was developed on the basis of studies in British patients with cancer and has not been validated in Japanese patients in whom the GFR was accurately measured. Since inulin is filtrated freely at the glomerulus and is neither reabsorbed nor secreted via tubules after glomerular filtration, its rate of excretion is identical to the GFR.^{( 15 )} In this study, we prospectively evaluated various methods for assessing the GFR, including inulin clearance, and examined the validity of the Calvert formula in Japanese patients with cancer. We propose a revised version of the Calvert formula for use in Japanese patients.

Patients and Methods

Eligibility.  Patients were eligible for enrollment if they met the following criteria: (i) a diagnosis of solid tumors; (ii) age, >20 years; (iii) scheduled to receive chemotherapy with carboplatin (as monotherapy or in combination with other antineoplastic agents); (iv) Eastern Cooperative Oncology Group (ECOG) performance status, 0–2; (v) no chemotherapy with antineoplastic agents within the past 4 weeks; and (vi) adequate bone marrow function (white‐cell count, >3000/μL; hemoglobin concentration, >9.0 g/dL; platelet count, >100 000/μL). Patients who received cisplatin‐based chemotherapy within the past 12 weeks were ineligible because Ccr does not provide a satisfactory estimate of renal function during this period.^{( 16 )} All patients gave written informed consent to participate in the study, which was approved by the Institutional Review Board of Nagoya University Hospital, and conformed to the provisions of the Declaration of Helsinki in 1995 (as revised in Tokyo 2004).

Assessments of GFR and Ccr.  The GFR was measured on the basis of inulin clearance within a week before the first cycle of chemotherapy. In brief, inulin clearance was calculated from serum and urinary inulin concentrations and urinary flow rate. A solution of inulin (1%) was infused intravenously at 300 mL/h for the first 30 min, followed by 100 mL/h for 90 min. Blood samples were collected at 0 (blank), 45, 75 and 105 min after starting the inulin infusion. Urine was collected between 30 and 60 min, 60 and 90 min, and 90 and 120 min, after the patient completely voided at 30 min. Inulin was assayed by an enzymatic method using a Diacolor Inulin kit (Toyobo Co., Osaka, Japan). The mean of three calculated inulin clearance values was used as the measured GFR for the patient.^{( 17 )}

The accuracies of various methods for assessing renal function on the basis of the GFR were evaluated. Creatinine values were assayed by the PAP method using an Accuras Auto CRE kit (Shino‐Test Co., Tokyo, Japan). The Ccr was measured by 24‐h urine collection. We used data only from patients with a urinary volume of >800 mL/day to obtain accurate estimations of the Ccr within a week before the start of chemotherapy. The mean of two measurements obtained on separate days within a week was used.

The following equations were used to calculate the Ccr:

The Cockcroft–Gault equation was used to estimate the Ccr as follows^{( 18 )}:

where gender is 0 for male subjects and 1 for female subjects.

In addition, the following equation for estimating the GFR in Japan was used^{( 12 )}:

where gender is 0 for male subjects and 1 for female subjects. Because the Japanese equation for estimating the GFR already includes a correction for BSA (units: mL/min per 1.73 m²), it was used after back‐calculation to absolute values (units: mL/min) for carboplatin dosing.

The accuracy of the estimates of the GFR was evaluated on the basis of the mean prediction error (MPE) ± SE and root mean squared error (RMSE).^{( 19 )}

Carboplatin pharmacokinetics.  Carboplatin was administered as a 1‐h intravenous infusion as monotherapy or in combination with other antineoplastic agents. The target AUC of carboplatin was 5 or 6 mg/mL min. The dose of carboplatin was calculated with the following formula^{( 8 )}:

Heparinized blood samples were collected before starting the carboplatin infusion, at the end of the infusion, and 0.25, 0.5, 1, 2, 4, 6, 8 and 24 h after the end of the infusion during the first cycle of chemotherapy. The plasma was immediately separated by centrifugation. Then, the plasma ultrafiltrate was obtained with the use of Amicon Centrifree micropartition units (Millipore, County Cork, Ireland) and stored at −20°C until analysis. The platinum level in the ultrafiltrate was measured by inductively coupled plasma–mass spectrometry (ICP‐MS; Finnigan MAT ELEMENT, Bremen, Germany).^{( 20 )} The carboplatin level was calculated on the basis of a platinum : carboplatin molar ratio of 371.25/195.08.

The pharmacokinetics of carboplatin were analyzed with the use of a two‐compartment model, using WinNonlin version 5.2 software (Pharsight, Mountain View, CA, USA). The volume of distribution at steady state (V _ss) and elimination half‐lives were calculated. The actual AUC of carboplatin was calculated by the trapezoidal method with extrapolation to infinity, and the observed carboplatin clearance was calculated as follows:

Validation and revision of the Calvert formula for Japanese patients.  The observed carboplatin clearance was compared with the estimated clearance, which was calculated as follows: estimated clearance (mL/min) = GFR (mL/min) + 25 (the Calvert formula). The accuracy of the estimated carboplatin clearance was evaluated on the basis of MPE ± SE and RMSE.^{( 19 )} The observed carboplatin clearance and GFR were used to revise the dosage formula, derived with the use of a one‐compartment model as described previously.^{( 3 )} The revised formula was modified by least‐squares analysis to minimize the MPE of the estimated carboplatin clearance and thereby facilitate clinical application.

Subsequently, the accuracy of estimated carboplatin clearance derived by various methods as an index of renal function was assessed. Carboplatin clearance was calculated using the original Calvert formula and the revised formula with crude 24‐h Ccr, adjusted 24‐h Ccr, estimated Ccr and estimated GFR.

Statistical analysis.  Whether different methods to assess renal function were associated with a more accurate estimation of the GFR was analyzed using the Wilcoxon signed‐rank test. Statistical analyses were performed with SPSS version 16.0 software (SPSS Inc., Chicago, IL, USA).

Results

The patients’ characteristics are shown in Table 1. Inulin clearance was measured to evaluate the GFR in 28 Japanese patients. Two patients received cisplatin‐based chemotherapy 18 and 10 months, respectively, before study entry.

Table 1.

 Characteristics of patients

Characteristic	Inulin clearance measured
	All	Carboplatin pharmacokinetic analysis performed
No. patients	28	21
Male/Female	18/10	14/7
Age (years)
Median (range)	68 (54–78)	68 (57–78)
Bodyweight (kg)
Median (range)	53.6 (36.5–72.5)	53.8 (40.3–72.5)
Body surface area (m2)
Median (range)	1.55 (1.25–1.81)	1.55 (1.29–1.81)
Serum creatinine (mg/dL)
Median (range)	0.75 (0.40–1.42)	0.75 (0.56–1.41)
Inulin clearance (mL/min)	17.2–105.4	17.2–91.4
Prior chemotherapy	3	2
Primary tumors
Lung	26	20
Others	2	1
Regimens
Carboplatin‐based combination		20
Paclitaxel		14
Gemcitabine		6
Carboplatin alone		1

Among the 28 patients in whom the GFR was measured, the pharmacokinetics of carboplatin were analyzed in 21 patients whose GFR ranged from 17.2 to 91.4 mL/min. The patient who received cisplatin‐based chemotherapy 10 months before pharmacokinetic analysis of carboplatin was included in the study. Carboplatin monotherapy to achieve a target AUC of 2 mg/mL min plus radiotherapy was given to one patient. The other 20 patients received carboplatin in combination with paclitaxel or gemcitabine.

Comparison of inulin clearance as the GFR with other methods.  Table 2 shows the evaluations of various methods to assess the GFR. Inulin clearance correlated significantly with crude 24‐h Ccr (r = 0.845, P < 0.001), adjusted 24‐h Ccr (r = 0.836, P < 0.001), estimated Ccr (the Cockcroft–Gault equation; r = 0.767, P < 0.001) and estimated GFR (the Japanese equation for estimating GFR; r = 0.798, P < 0.001). The use of the crude 24‐h Ccr overestimated inulin clearance (MPE, 24.2%), but the MPE improved to −2.3% by adjusting the serum creatinine level by adding 0.2 mg/dL (P < 0.001) (Table 2). This adjustment of the serum creatinine level also improved the precision of the estimated GFR: the RMSE decreased by 13.5% points (P = 0.011) (Table 2). When the GFR was assessed by the Japanese equation for estimating the GFR, the assessment was more unbiased and precise than that of the crude 24‐h Ccr and estimated Ccr calculated with the Cockcroft–Gault equation (Table 2).

Table 2.

 Evaluation of various methods to assess GFR

Method	MPE ± SE (%)	RMSE (%)
Crude 24‐h Ccr	24.2 ± 5.4	37.2
Adjusted 24‐h Ccr	−2.3 ± 4.5	23.7
Cockcroft–Gault equation	12.5 ± 5.8	32.7
Japanese equation for estimating GFR	5.2 ± 4.8	25.7

Ccr, creatinine clearance; GFR, glomerular filtration rate; MPE, mean prediction error; RMSE, root mean squared error.

Carboplatin pharmacokinetics and thrombocytopenia.  A profile of the pharmacokinetic data on carboplatin is shown in Table 3. When the target AUC value was 5 mg/mL min, the actual AUC values were slightly higher (112%) than the target value.

Table 3.

 Pharmacokinetics of carboplatin in 21 patients

Target AUC (mg/mL min)	No. patients	Clearance (mL/min)	GFR (mL/min)	T1/2β (h)	V ss (L)	Actual AUC (mg/mL min)	% Target AUC
2	1	21	17	2.6	4.5	5.5	273
5	14	76 ± 18	60 ± 15	2.3 ± 0.6	15.6 ± 3.1	5.6 ± 0.9	112 ± 18
6	6	104 ± 18	76 ± 16	1.8 ± 0.2	15.0 ± 3.7	6.0 ± 0.8	100 ± 13

Mean ± SD. AUC, area under the curve versus time curve; GFR, glomerular filtration rate; T_1/2β, elimination half‐life; V _ss, volume of distribution at steady state.

The patients who received carboplatin in combination with gemcitabine had severe thrombocytopenia. There was no correlation between the grade of thrombocytopenia and the actual AUC values (data not shown).

Validation and revision of the Calvert formula for Japanese patients.  Carboplatin clearance estimated by the original Calvert formula (GFR + 25) clearly overestimated the observed clearance, with a MPE of 14.3% (Fig. 1 and Table 4). Thus, we revised the Calvert formula for Japanese patients. Figure 2 is a plot of the observed carboplatin clearance (dose/AUC) versus GFR for the 21 patients. The following equation was derived from regression analysis: clearance = 1.21 (±0.13) × GFR + 6 (±9) (figures in parentheses are SEM; 95% confidence intervals [CI] of the slope, 0.93–1.48; 95% CI of the y intercept, −13 to 24; R ² = 0.811, P < 0.001). The correlation between the observed carboplatin clearance and BSA was not significant (r = 0.582, data not shown). Since the 95% CI of the slope (i.e. 0.93–1.48) included 1, the slope might be 1; therefore, for convenience, the renal clearance of carboplatin was assumed to be identical to the inulin clearance. The MPE of the estimated carboplatin clearance was lowest (mean ± SE = −0.1 ± 4.1%) on least‐squares analysis when the intercept on the y‐axis was 15. We revised the carboplatin dosage formula for Japanese patients to the following clinically convenient equation:

Figure 1.

 Relation between observed carboplatin clearance and carboplatin clearance predicted by the Calvert formula, that is, glomerular filtration rate (GFR) + 25 (mL/min), where the GFR was measured by inulin clearance. The diagonal line represents the line of identity.

Table 4.

 Evaluation of the original Calvert formula and the revised formula in Japanese patients

Estimated carboplatin clearance	MPE ± SE (%)	RMSE (%)
GFR + 25	14.3 ± 5.7	29.2
GFR + 15	−0.1 ± 4.1	18.2
Crude 24‐h Ccr + 25	35.4 ± 9.2	54.3
Crude 24‐h Ccr + 15	21.0 ± 7.6	39.9
Adjusted 24‐h Ccr + 25	15.0 ± 8.6	41.3
Adjusted 24‐h Ccr + 15	0.6 ± 6.9	30.9
Estimated Ccr + 25	26.1 ± 9.5	49.8
Estimated Ccr + 15	11.6 ± 7.8	36.6
Estimated GFR + 25	20.3 ± 8.9	44.5
Estimated GFR + 15	5.9 ± 7.1	32.3

Ccr, creatinine clearance; Estimated Ccr, creatinine clearance calculated from the Cockcroft–Gault equation; Estimated GFR, GFR calculated from the Japanese equation for estimating GFR; GFR, glomerular filtration rate; MPE, mean prediction error; RMSE, root mean squared error.

Figure 2.

 Correlation of free carboplatin clearance with pretreatment glomerular filtration rate (GFR) in 21 Japanese patients studied during the prospective validation of the Calvert formula. ○, 14 male patients; ×, seven female patients. The solid line is the linear regression line, and the dotted lines are 95% confidence intervals. Mean ± SE for the slope = 1.21 ± 0.13, and for the y intercept = 6 ± 9; r = 0.901; R ² = 0.811, P < 0.001.

Comparison of the original Calvert formula with the revised formula in Japanese patients.

As compared with the original Calvert formula, the MPE obtained with the revised formula in Japanese patients improved from 14.3% to −0.1% (P < 0.001) and was acceptably precise, with an RMSE of 18.2% (Table 4). When renal function was assessed on the basis of crude 24‐h Ccr instead of the GFR, the original Calvert formula overestimated the observed carboplatin clearance, resulting in an MPE of 35.4% (Table 4). The MPE improved to 15.0% by adjusting the serum creatinine level, but the estimated carboplatin clearance was still higher than the observed clearance (Table 4). The revised formula with adjusted 24‐h Ccr decreased the MPE to 0.6% (P < 0.001) (Table 4). When the Japanese equation for estimating the GFR was used, the revised formula was more unbiased than the original Calvert formula, and the MPE improved from 20.3% to 5.9% (P < 0.001) (Table 4).

Discussion

Our study showed that the adjusted 24‐h Ccr could be substituted for the GFR to calculate the dosage of carboplatin and that the Calvert formula using the true GFR overestimates carboplatin clearance in Japanese patients with cancer. We therefore revised the formula by subtracting 10 from the constant 25 to account for the non‐renal clearance of carboplatin: dose (mg) = target AUC × (GFR + 15). Our results suggest that the use of crude Ccr measured by the PAP method combined with the lower non‐renal clearance in Japanese patients can cause an overdose of carboplatin when the dose is determined by the Calvert formula.

Different techniques for the assessment of the GFR would be associated with the overestimation of carboplatin clearance. We measured inulin clearance as an index of the GFR, which is considered the gold standard for GFR assessment, whereas Calvert used ⁵¹Cr‐EDTA clearance, which involves a simpler technique. Because the procedure used by Calvert overestimates inulin clearance in patients with normal GFR,^{( 21 )} the different techniques used to assess GFR could not explain the overestimation of carboplatin clearance in our study.

We reconfirmed the accuracy of adjusting the 24‐h Ccr by adding 0.2 mg/dL to the serum creatinine level measured by the enzymatic PAP method when Ccr was used instead of the GFR, as described previously (Table 2).^{( 8 )} When the GFR was estimated using the equation, the prediction was more unbiased and precise than that obtained with the Cockcroft–Gault equation (Table 2). The Cockcroft–Gault equation was originally designed to predict Ccr,^{( 18 )} and the estimated Ccr derived by this equation must be theoretically higher than the GFR. We therefore recommend the use of the adjusted 24‐h Ccr or the Japanese equation for estimating the GFR in Japanese patients, albeit the latter should be further validated when used to calculate the carboplatin dosage in Japanese patients.

The original Calvert formula includes the constant 25 at the intercept on the y‐axis to account for the non‐renal clearance of carboplatin; we reduced the constant to 15 in the revised formula. The formula might depend on the GFR values in part. However, the GFR in the majority of patients who received carboplatin was generally around 60 mL/min in Japanese patients (Table 3 and Fig. 2); this revised formula would be useful in the clinical setting. Metabolism or biliary excretion does not play a major role in the elimination of carboplatin, and the non‐renal clearance of the drug is mainly due to reactions with tissues and plasma‐protein binding,^{( 22 )} which depend on body size. Newell and colleagues developed a GFR‐based formula for calculating the recommended dosage of carboplatin in children, in which the formula was modified by substituting 0.36 × bodyweight (kg) for 25.^{( 23 )} When this modified formula was used in mice and rats with smaller body sizes, there was an excellent agreement between observed and estimated carboplatin clearance.^{( 23 )} The body sizes of Japanese patients are generally smaller than those of whites, which might explain the lower non‐renal clearance of carboplatin. However, validity of the modified formula should be evaluated prospectively in the future.

In summary, we validated the Calvert formula in Japanese patients with cancer in whom the GFR values were accurately determined on the basis of inulin clearance. Although estimation of carboplatin clearance by the Calvert formula was clinically acceptable and unbiased when the adjusted 24‐h Ccr was used, it still overestimated the carboplatin clearance in Japanese patients. We proposed a revised formula, that is, dose (mg) = target AUC × (GFR + 15), in which the non‐renal clearance is expressed as 15 instead of 25. The clinical usefulness of this new formula should be validated in future studies.

Disclosure Statement

All authors report no conflict of interest.