Sir,
Several chemotherapeutic agents have a narrow therapeutic index, and accurate dosing
becomes essential to avoid toxicity, such as myelosuppression or gastrointestinal
toxicity. Weight-based dosing, body surface area (BSA)-based dosing, and area under
the curve (AUC)-based dosing are commonly employed to estimate doses of anticancer
agents.[1] Of these, AUC-based dosing (area under the plasma concentration multiplied by time)
is the most relevant for drugs which are eliminated by the kidneys. For drugs which
have nonrenal elimination or multiple pathways for elimination, AUC-based dosing is
not useful. In routine oncologic practice, the carboplatin dosage is usually estimated
based on AUC-based dosing since carboplatin clearance closely matches creatinine clearance.
Therein lies the importance of measuring or estimating creatinine clearance.
Certainly, measuring the glomerular filtration rate (GFR) using 51-labeled ethylenediaminetetraacetic
acid would be the most accurate. However, this is not usually available at most centers,
and in actual oncologic practice, the GFR is estimated rather than measuring. Several
equations have been developed in the past for estimating GFR which include Cockcroft–Gault
equation and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation,
among several others. Recently, in an article by Janowitz et al., we came across a newer, more accurate, albeit cumbersome method to estimate creatinine
clearance. The authors used a robust methodology and developed and validated a model
to accurately predict the measured GFR.[2] However, while deciding the dosage of carboplatin (the primary drug for which these
equations are utilized), several factors other than estimated GFR (including but not
limited to performance status [PS]) merit consideration. The key question is whether
we require such a high degree of precision in estimating GFR in real-life settings?
The argument to support this is that, if accuracy was the only essential criteria,
we might be measuring the GFR in all patients rather than estimating. Yet in routine
clinical practice, we avoid doing the former as it is not easily available. The same
argument would probably also hold true for the newer model proposed by Janowitz et al. It is indeed accurate and precise, yet remains complex and inaccessible for a vast
majority. Even though the authors have tried to simplify the calculation by providing
an online tool, the bottom line is that it still remains fairly complex even for people
working in academic settings. The major burden of cancer is currently in the developing
and underdeveloped nations.[3] Oncologists practicing in resource-constrained settings may not have ready access
to an online tool, and the simple desk calculator is often what they have.[4] Further, each equation developed in a specific population needs to be validated
in other ethnic and geographic areas before attempting its application/generalization.
We performed an ad hoc retrospective audit of lung cancer patients undergoing first-line
chemotherapy at our center, with the aim of determining discrepancies between actual
doses of carboplatin administered in the first cycle versus those calculated using
different equations for GFR estimation. The Calvert formula was used to calculate
carboplatin dose, with the estimated GFR being obtained from Cockcroft–Gault equation,
CKD-EPI equation, and the Janowitz et al.'s equation.[2] Carboplatin dose was also calculated using manufacturer's instructions (BSA in kg/m2
× 300 for GFR >60 mL/min; BSA × 250 for GFR = 41–60 mL/min; and BSA × 200 for GFR
≤40 mL/min). Dose derived from GFR estimated using the Janowitz et al.'s online tool was considered as the reference standard. Absolute dosage differences
and percentage errors (PEs) for the above equations were calculated.
From January 1, 2017 till August 31, 2017, 77 patients received carboplatin-based
chemotherapy. Dosage calculated by Cockcroft–Gault-based GFR and manufacturer's recommendation
had significant variation as compared to the authors' new equation-based carboplatin
dose [Table 1]. The dosage calculation based on CKD-EPI equation was largely similar to the latter.
However, the actual administered doses (with reductions being made for PS and vial
package strengths) were lower than both Cockcroft–Gault-based doses and manufacturer's
recommended doses (both of which are routinely used at our center for dose calculations).[5],[6] A significant proportion (n = 48, 62.3%) had >20% absolute PE of carboplatin dose
as compared to the reference standard. Carboplatin dose PEs (actually administered,
calculated as per Cockcroft–Gault equation, manufacturer's recommendation and CKD-EPI
equation) were plotted as a waterfall chart [Figure 1]a, [Figure 1]b, [Figure 1]c, [Figure 1]d. All, except six (7.8%) patients, received doses less than or equal to that calculated
from the reference. None of the above six received carboplatin dose ≥20% than the
predicted reference. Hypothetically, even if the administered carboplatin dose was
exactly as calculated from Cockcroft–Gault equation and manufacturer's recommendation,
majority (83.1% and 81.8%, respectively) would have still received a lower dose compared
to the reference and those receiving ≥20% overdose would have been only two (2.6%)
and three (3.9%) patients, respectively [Figure 1]b and [Figure 1]c.
Table 1
Comparison of carboplatin dose estimation based on different methods and their percentage
errors compared to the reference standard
|
Method of dose calculation
|
Median (IQR) dose in mg
|
Residual dose, median (IQR) mg
|
Median PE (%) IQR
|
Median APE (%) IQR
|
Number of patients with APE >20%, n (%)
|
|
APE – Absolute percentage error; CKD EPI – Chronic Kidney Disease Epidemiology Collaboration;
IQR – Interquartile range; PE – Percentage error
|
|
Cockcroft-Gault based
|
434.1 (382.5-513.9)
|
−50 (−69-14)
|
−12.5 (−17.3-2.79)
|
17.9 (7.2-28.6)
|
18 (23.4)
|
|
CKD EPI based
|
515.5 (434.6-567.4)
|
11.1 (−8.8-23.7)
|
2 (−1.7-4.4)
|
3.5 (2.1-5.0)
|
2 (2.6)
|
|
Manufacturerrecommended doses
|
447.5 (380-501)
|
−59.2 (−123.9-13.6)
|
−13.6 (−28.7-3.3)
|
13.3 (5.2-18.6)
|
36 (46.8)
|
|
Actual dosage given
|
400 (350-450)
|
−106.9 (−160.8-57.5)
|
−26.8 (−45.3-13.5)
|
27.1 (14.7-45.3)
|
48 (62.3)
|
Figure 1: Waterfall plot shows the percentage error in the carboplatin dosage along
Y-axis (each bar represents one patient); (a) dose calculated as per the Chronic Kidney
Disease Epidemiology Collaboration equation, (b) dose estimated by the Cockcroft–Gault
equation, (c) dose estimated as per the carboplatin manufacturer’s recommendations,
and (d) the dose which the patients actually received at our center. The comparator
in all these plots was the dosage calculated as per the reference standard (glomerular
filtration rate based on the new equation proposed by Janowitz et al.). Red line marks
the 20% excess dose from the reference dosage
Thus, the actual dose administered to patients is lower than that predicted in the
majority, often due to PS and vial package strength issues. This is irrespective of
what equation one uses to estimate GFR. Hence, the probability of administering an
unacceptable and potentially toxic (higher) dose of carboplatin, based on an incorrect
GFR estimation, might be much lower in clinical practice than what one would expect.
It is good to be accurate, but it is even better to be safe and simple. While we accept
the inadequacies in estimating GFR by the currently available equations, the actual
administered carboplatin doses that patients generally receive can be safely and conveniently
calculated with these equations without the requirement for having to access a complex
equation online each time – something that is of particular relevance in resource-constrained
settings. Therefore, we believe that using more accurate newer equations may not be
required as multiple factors influence the final administered dose.
