Results part 1 – Pharmacokinetic model development

The key model building decision points are shown in Fig. 5. Model structures (open and filled circles) were evaluated sequentially based on data (shown in the rectangles). The best structures (shown as filled coloured circles) were carried forward into the next part of model building.

Fig. 5

Iterative model building process. The subset of pharmacokinetic data used are shown in the rectangles. For example, “IV Total” denotes the total concentration profiles of docetaxel after intravenous administration. “PO Total | PO unbound” denotes the total concentration and unbound concentration profiles of oDox + E after administration. “IV Total | PO Total | IV Unbound | PO Unbound” denotes the entire dataset. The circles represent the model feature that was evaluated from the dataset, the bold arrows point to a model feature contained in a coloured circle that best described the data while the thin arrows show the model features contained in the grey circles that were evaluated alongside the chosen model feature. The colour blue is used to represent IV data and models while red represents oDox + E data and models. The combination of colours represents both are present. Abbreviations:IV Total – Dataset of total concentration of docetaxel in plasma over time after IV administration of docetaxel; PO Total – Dataset of total concentration of docetaxel in plasma over time after oDox + E administration; 1CMT – One compartment structural model; 2CMT – Two compartment structural model; 3CMT – Three compartment structural model; IV Total | IV Unbound – Dataset of total and unbound concentration of docetaxel in plasma over time after IV administration of docetaxel; IV Total | PO Total – Dataset of total concentration of docetaxel in plasma over time after IV administration of docetaxel and oDox + E administration; PO Total | PO Unbound – Dataset of total and unbound concentration of docetaxel in plasma over time after oDox + E administration; Time vary Binding – Model with time varying binding mechanics of docetaxel to plasma proteins; 2 site Binding – Model with 2 site binding mechanics of docetaxel to plasma proteins; Michaelis Menten Binding – Model with Michaelis Menten binding mechanics of docetaxel to plasma proteins; Constant Binding – Model with constant binding mechanics of docetaxel to plasma proteins. Constant Bioavailability – Model with a constant bioavailability throughout; Time Varying Bioavailability – Model that allows bioavailability to change with time; IV Total | PO Total | IV Unbound | PO Unbound – Entire dataset containing total and unbound concentration of docetaxel in plasma over time after IV administration of docetaxel and oDox + E administration; Correlated errors – L2 data item used within the model to assess for correlated errors between unbound and total concentration plasma samples; No correlated errors – L2 data item not used within the model

Sub-model 1 (IV docetaxel structural model)

A 3-compartment model best described the total plasma concentration of docetaxel after IV docetaxel administration.

Sub-model 2 (oDox + E structural model)

A 2-compartment model with lag-time best described the total plasma concentration of docetaxel after oDox + E administration.

Sub-model 3 (oDox + E absorption phase)

Time-varying bioavailability was explored in sub-model 3 and did not result in a statistically significant improvement in the objective function; therefore, a constant bioavailability was carried over to subsequent models.

Sub-model 4 (Plasma binding for IV docetaxel)

Different binding mechanics of docetaxel after IV docetaxel administration were explored. For IV docetaxel, binding can be to protein or the excipient Polysorbate 80. Time-varying, Michaelis–Menten and 2-site binding mechanisms did not produce statistically significant improvements in the objective function; therefore, a constant binding mechanism (i.e., Unbound concentration equals total concentration multiplied by fraction unbound) was carried over to subsequent models. The parameters for sub-model 4 is shown in Supplement 1 “IV Docetaxel Model” column.

Sub-model 5 (Plasma binding for oDox + E)

Different binding mechanics of docetaxel after IV docetaxel administration were explored. For oDox + E, binding is to protein only. Time-varying, Michaelis–Menten and 2-site binding mechanisms did not produce statistically significant improvements in the objective function; therefore, a constant binding mechanism was carried over to subsequent models. The parameters for sub-model 5 is shown in Supplement 1 “oDox + E Model” column.

Final model (Sub-model 6)

A 3-compartment model with linear elimination, constant bioavailability, constant binding mechanics, and a combined error model with the L2 data item provided the best fit for the pharmacokinetic data collected which consisted of total and unbound concentrations of docetaxel in plasma after IV docetaxel and oDox + E administration.

The parameters of the final model are shown in Supplement 1. The structure of the final model is shown in Supplement 2.

The parameters in Supplement 1 are reported in the context of unbound docetaxel. The moderate to large standard errors are expected given the dataset size of 9 patients (i.e., proportional to 1/sqrt(n)) and the number of parameters included in the model. Of note, the residual standard errors of Vp1 & ωCL were marginally high, however, this would not invalidate the simulations from the model and the ability to determine whether a feasible regimen exists through the GO / NO-GO framework discussed below remains unchanged.

The parameters can be compared to the total docetaxel pharmacokinetic parameters by an adjustment using the binding constant. For example, the clearance of unbound docetaxel in the final model is estimated at 8570 L h−1 with a fraction unbound of 0.67%. Therefore, the total IV docetaxel clearance is 8570 multiplied by 0.67% which is approximately 57.3 L h−1.

Individual and Goodness of fit plots for final model

Figure 6 shows the individual plots for the total concentration and unbound concentration of docetaxel in plasma after IV docetaxel and oDox + E administration for each patient. The goodness of fit plots are shown in Supplement 3 for total docetaxel and unbound docetaxel samples by route of administration.

Fig. 6

Individual plots of total concentration, unbound concentration of docetaxel after administration of IV docetaxel and oDox + E. The yellow line denotes the lower limit of quantification of 0.084 ng/mL for unbound docetaxel

The individual predictions for the total plasma concentration and unbound plasma concentration of IV docetaxel are shown in Fig. 6A and B, respectively. The predictions closely match the measured plasma concentrations. The individual predictions for the total plasma concentration after oDox + E administration are shown in Fig. 6C and the predictions generally match the measured plasma concentrations, although not as closely when compared to IV docetaxel, suggesting more variability in the oDox + E that has not been explained by the model. The individual predictions for unbound plasma concentration after oDox + E administration are shown in Fig. 6D and appears to only capture the profiles partly, however, this is expected given the very small concentrations that the model is working with for unbound oDox + E (i.e., the peak concentrations are around 1 ng/mL, roughly only 10-times greater that the LLOQ of 0.084 ng/mL). This highlights the importance of modelling the unbound alongside total concentration samples to better estimate the pharmacokinetic model parameters.

The goodness of fit plots for IV docetaxel both bound and unbound is close to linear suggesting an adequate fit (Supplement 3A, right graph and Supplement 3B, right graph). The goodness of fit plot for total concentration after oDox + E has a U-shaped curve which suggests model-misspecification (Supplement 3A, left graph). The goodness of fit for unbound docetaxel after oDox + E (Supplement 3B, left graph) appears to be linear, but this is hard to interpret with confidence due to the small concentrations involved.

Results part 2 – Application of model for oDox + E GO / NO-GO decisionSimulation results

1000 simulations were performed for each oDox + E dose regimen and EC level combination outlined in Table 2. A total of 50 PTA values were calculated for each combination of oDox + E dose and EC included within the simulation space. These PTAs are shown in Figs. 7, 8 and 9f or a single dose, two doses and three doses of oDox + E, respectively.

Fig. 7

Probability target attainment of oDox + E single dose regimens between 400 to 600 mg by dose level. Black line denotes 80% PTA. Dashed line denotes the PTA of a 600 mg dose of oral docetaxel alone (i.e., bioavailability of 8%). PTA – Probability target attainment

Fig. 8

Probability target attainment of oDox + E dose between 400 to 600 mg given twice by dose level. Black line denotes 80% PTA. Dashed line denotes the PTA of a 600 mg dose of oral docetaxel alone (i.e., bioavailability of 8%) given twice. PTA – Probability target attainment

Fig. 9

Probability target attainment of oDox + E dose between 400 to 600 mg given three times by dose level. Black line denotes 80% PTA. Dashed line denotes the PTA of a 600 mg dose of oral docetaxel alone (i.e., bioavailability of 8%) given three times. PTA – Probability target attainment

The PTAs of a 600 mg dose of oral docetaxel alone (with a bioavailability of 8%) was added to each PTA figure as a dashed line. Oral docetaxel alone could not achieve a PTA > 80% for any dose regimen at any EC value.

GO / NO-GO framework 1.

GO / NO-GO recommendation (i.e., existence of a practical regimen of oDox + E?)

No practical regimen of oDox + E existed with a single administration of oDox + E as none of the single dose regimens achieved a PTA of > 80% (Fig. 7). Practical regimens of oDox + E existed when two or three doses are administered (Figs. 8 and 9).

2.

GO conditions

Scenario two of the GO conditions was observed for repeated dosing of oDox + E given either twice or three times, depending on the EC level. For the two-dose regimen of oDox + E (Fig. 8), the PTA > 80% was achieved across the 400-600 mg dose range at an EC of 0.1 ng/mL but could not be achieved for any dose regimen when the EC was greater than 0.5 ng/mL. For the three-dose regimen of oDox + E (Fig. 9), the PTA > 80% was achieved across the 400-600 mg dose range at an EC of less than 0.3 ng/mL but was only achieved by the 550 mg dose level and above when the EC was 0.5 ng/mL. The maximum dose regimen evaluated in this simulation (600 mg oDox + E given three times) could achieve a PTA > 80% when the EC was at or below 0.7 ng/mL. For an EC of 1.0 ng/mL, the maximum dose regimen evaluated produced a PTA of 56%, below the 80% power threshold.

Therefore, a GO recommendation with conditions was proposed for oDox + E. The recommendation prior to commencement of further clinical trials was to better quantify the EC value. If the EC were found to be less than 0.7 ng/mL there would be more confidence in a GO decision and progress to a Phase IIa/b trial, while an EC in excess of 1.0 ng/mL would strongly require consideration of a NO-GO decision and re-evaluation of the development strategy of oDox + E.