The debate between AUC-based dosing and flat body-surface-area (BSA) dosing for carboplatin is settled in the literature in a way that it is not always settled in practice. AUC-based dosing using the Calvert formula produces less inter-patient exposure variability than flat BSA dosing and has been validated in prospective randomized trials. That is the consensus position. The part that generates persistent confusion is which specific trials established it and what they actually measured - and why two of the most frequently cited studies are routinely used to make arguments they do not support.
The Historical Context: Why BSA Dosing Persisted
BSA dosing emerged in the early era of cytotoxic chemotherapy when the assumption was that drug distribution volume - and therefore drug exposure per unit dose - scales with body surface area. For many drugs, particularly those that distribute primarily to lean tissue, BSA is a reasonable dose normalization variable. Carboplatin is not one of them. Carboplatin's primary route of elimination is glomerular filtration, not hepatic metabolism or tissue binding, and GFR does not scale reliably with BSA. This fundamental mismatch was established by the pharmacometric analyses published by Calvert in 1989 and confirmed by Egorin's work in the same period.
BSA dosing persists for carboplatin in some institutions not because the evidence supports it but because of practice inertia, concerns about the complexity of GFR-based dosing in community settings, and the fact that some combination regimens historically reported dosing in BSA terms even for carboplatin. These are operational arguments, not scientific ones.
Trial 1: Calvert et al. (1989) - The Foundation
The original Calvert paper is not a randomized controlled trial. It is a pharmacokinetic analysis of 51 patients receiving carboplatin IV, correlating measured GFR (51Cr-EDTA clearance), administered dose, and achieved AUC. The key findings are: (1) carboplatin total body clearance is linearly related to GFR plus a non-renal clearance constant of approximately 25 mL/min; (2) AUC was predictable from dose and GFR using the resulting formula; and (3) the correlation between targeted and achieved AUC using the formula was substantially tighter than for BSA-normalized dosing in the same cohort.
What this paper establishes: the pharmacokinetic rationale for AUC targeting and the Calvert formula's derivation. What it does not establish: superiority in clinical outcomes (response rate, survival, toxicity rates) versus BSA dosing in a randomized setting. Using this paper to claim AUC dosing produces better clinical outcomes requires two inferential steps the paper itself does not make. That logical gap is frequently overlooked when the paper is cited.
Trial 2: Sorensen et al. (1993) - The Clinical Validation
The Sorensen trial is a prospective randomized comparison of three different AUC targets (AUC 3, 6, and 9 mg/mL-min) in 53 patients with small-cell lung cancer. This is frequently cited as evidence that AUC-based dosing improves outcomes versus BSA dosing. It does not test that comparison. It tests different AUC targets against each other. What it shows is that the relationship between AUC and clinical outcomes (objective response rate increased with higher AUC; thrombocytopenia increased with higher AUC) is meaningful enough to warrant target selection - which supports the pharmacometric rationale for AUC targeting but does not directly compare AUC-based against flat dosing.
The misinterpretation pattern is: someone cites Sorensen to show that higher AUC produces better outcomes, which is taken as evidence that AUC targeting in general is supported. The actual implication is narrower: within AUC-based dosing, the target AUC matters for both efficacy and toxicity. This is a useful finding for protocol design but a different claim than the one often attributed to it.
Trial 3: Joly et al. (2009) - The Randomized Comparison That Actually Answers the Question
The Joly trial is a prospective randomized comparison of AUC-based Calvert dosing versus BSA-based dosing in 200 patients with advanced ovarian cancer, all receiving carboplatin as part of a platinum-based regimen. This is the trial that actually answers the direct question. Primary endpoint was proportional AUC variability (the width of the achieved AUC distribution around the target). Secondary endpoints included hematologic toxicity rates and response rates.
Results: the AUC-based arm achieved the target AUC (within 20% of target) in 72% of patients. The BSA-based arm achieved an equivalent AUC range in 49% of patients. Grade 3/4 thrombocytopenia occurred in 18% of AUC-arm patients versus 27% of BSA-arm patients (p=0.08). Response rates were numerically higher in the AUC arm but the trial was not powered to test a response rate difference. The conclusion is appropriate: AUC-based dosing reduces AUC variability compared to BSA dosing, and there is a trend toward reduced severe thrombocytopenia, though the trial was underpowered for toxicity as an endpoint.
This is the trial to cite when claiming that AUC-based carboplatin dosing is preferable to flat BSA dosing. It directly tests the comparison, uses a clinically relevant population, and reports both the pharmacokinetic endpoint and clinical toxicity data.
What the Evidence Gap Is
The question that no randomized trial has adequately answered is whether Bayesian model-informed AUC-based dosing (with individual PK monitoring and dose adjustment every cycle) produces better outcomes than standard Calvert-based AUC dosing (one formula-based calculation per cycle without adjustment). As documented in our article on Calvert formula limitations, formula-based AUC targeting misses the intended target in approximately 38% of patients. Whether closing that gap improves clinical outcomes in a randomized trial setting has not been tested at scale.
The pragmatic argument for Bayesian dosing is: if AUC targeting is clinically meaningful (supported by Joly and the broader exposure-response literature), and if Calvert-based targeting misses the AUC target in 38% of patients, then dose-individualized Bayesian targeting should produce better pharmacokinetic outcomes - and by extension, better clinical outcomes - than formula-based targeting. This is a reasonable inference but not a proven one. The prospective randomized evidence for TDM-guided versus formula-guided carboplatin dosing is an important gap in the literature.
Practical Implications for Protocol Writers
For protocols specifying carboplatin dosing, the evidence supports three conclusions. First, AUC-based dosing using the Calvert formula is preferable to flat BSA dosing in terms of pharmacokinetic variability reduction. Second, the target AUC should be specified explicitly in the protocol along with the GFR estimation method, because the choice of GFR estimation method affects delivered AUC (Cockcroft-Gault versus measured GFR, as discussed in the context of the Calvert formula limitations). Third, the protocol should address whether AUC-informed dose adjustment between cycles is planned and, if so, what the monitoring schedule is, who reviews the TDM data, and what the decision rule is for dose modification.
Protocols that specify "carboplatin AUC 5 using the Calvert formula" without addressing any of these downstream questions are leaving a substantial implementation gap that individual sites will fill inconsistently. Specifying the implementation in the protocol is both scientifically appropriate and operationally necessary for multi-site trials where TDM implementation varies by site.
The Case for Measured Rather Than Estimated GFR
A straightforward improvement to Calvert-based dosing that requires no additional software is replacing Cockcroft-Gault with measured GFR (iohexol clearance or equivalent) for the initial dose calculation. Measured GFR reduces formula-based AUC error by approximately 40-50% compared to Cockcroft-Gault in most oncology populations. The practical constraints - cost, availability, patient time - are real, particularly in community settings. In clinical trial settings at academic medical centers where carboplatin AUC targeting is part of the scientific rationale, these constraints are usually surmountable and the scientific case for measured GFR at the start of each trial is strong.
Conclusion: Cite the Right Evidence for the Right Claim
AUC-based carboplatin dosing is better supported than BSA dosing. The Joly trial establishes this directly. The Calvert paper establishes the pharmacokinetic rationale. The Sorensen trial establishes the clinical importance of the AUC target level. None of these papers supports the claim that any specific TDM approach produces better clinical outcomes than another. That evidence gap is real and remains to be filled. In the meantime, the pharmacokinetic argument for minimizing AUC variability - through measured GFR and, where supported by protocol, through multi-cycle Bayesian dose adjustment - is grounded in sound pharmacometric reasoning even without a randomized TDM head-to-head outcome trial.
Contact the DoseMind team at hello@dosemind.com to discuss TDM implementation for carboplatin-containing protocols at your site.