**Assessment of expected drug exposure relative maximum safety limits in early phase studies**

Anders Kristoffersson (1), Daniel Röshammar (1), Elodie Plan (1)

(1) Pharmetheus AB, Uppsala, Sweden

**Introduction:** During early phase drug development, such as in single or multiple ascending dose studies, drug exposure should not exceed the maximum exposure limits (normally established based on the observed toxicity in pre-clinical studies). Pharmacokinetic models are often used when designing such studies for predicting what doses will render exposure below the maximum exposure limits and for assessing whether to progress to the next dose level or not based on interim data analysis. However, there are various approaches to perform such simulations with regard to sources of uncertainty and exposure metrics of interest.

**Objectives: **In this work, we aim to evaluate different simulation strategies and explore the potential impact on decision making.

**Methods: **A single ascending dose study was simulated, using a two-compartment population PK model with first-order absorption and elimination. Doses of 1, 1.5, 3, 6 and 8 mmol were administered to 6 subjects per dose level. Drug concentrations in the central compartment were assessed up to 72h after dose. An interim analysis was performed after the 6 mmol dose level, whereby the model was updated and used to simulate into the next dose level of 8 mmol. The maximum exposure level was arbitrarily set at 1200 nM for the maximum concentration (Cmax) and 5500 nM*h for the area under the concentration to time curve up to 24h after dose (AUC). The study was simulated 300 times under each simulation strategy; 1.) without parameter uncertainty, 2.) with parameter uncertainty from the NONMEM covariance step (COV), and 3.) parameter uncertainty based on Sampling Importance Resampling (SIR) [1]. The expected proportion of study replicates and individuals exceeding the maximum exposure limits at the 8 mmol dose level was assessed for calculating:

- The probability a subject may exceed exposure limits
- The probability a study dose level may have average exposure exceeding limits

The stop criterion for not progressing to the 8 mmol dose level was a >5% risk of exceeding the exposure limit.

All simulations were performed in NONMEM version 7.3.0 installed on an Intel Xeon-based server and PsN 4.8.1. [2]. Post-processing of simulation results was performed using R version 3.3.3.

**Results:** The simulations showed that the choice of exposure metric, simulation strategy, and if to consider the mean exposure or individual exposure per dose level all influenced the decision if to progress to the 8 mmol dose level. There was less variability in the resulting decision between simulation strategies for the AUC than for the Cmax metric. However, whether assessing the exposure limit per subject or the average per dose level had greater impact on decision making for AUC than for Cmax. Based on AUC, progressing to the 8 mmol dose was supported with all simulation strategies when using the average exposure, but not with any simulation strategy when using individual exposure. For Cmax, the proceed/stop decision was the same for the average and individual exposure scenarios; stop at the 6 mmol dose level was only indicated when uncertainty was derived from the NONMEM COV.

In all cases the NONMEM COV was simulating wider distributions compared with SIR, and hence a greater proportion of cases were exceeding the exposure limits. This was most likely an effect of a poor covariance matrix from NONMEM due to the small amount of data simulated (only MATRIX=S successful), which was improved on by SIR.

As expected, the proportion of cases exceeding the exposure limits was much greater for individual subjects compared to for the average exposure per dose level. E.g. for the AUC exposure limit, the fraction of patients exceeding the limit was 13% compared to the fraction of studies where the mean exposure exceeding the limit was 2%.

**Conclusions:** This work shows that the simulation strategy may have impact when deciding whether to proceed to a next higher dose level or not based on available data during early phase trials. We recommend to clearly define upfront how the expected exposure will be assessed relative the maximum exposure limits when exploring the maximum dosing schedule. If of particular importance, different simulation strategies can be applied and subsequently the most conservative approach can be chosen.

**References:**

[1] Dosne AG L, Bergstrand M, Karlsson MO. An automated sampling importance resampling procedure for estimating parameter uncertainty. J Pharmacokinet Pharmacodyn 44(6):509-520, 2017.

[2] Lindbom L, Ribbing J and Jonsson EN. Perl-speaks-NONMEM (PsN) - a Perl module for NONMEM related programming. Comp Meth Prog Biomed 75 (2), 2004.