Seeking an optimal dosing regimen for OZ439/DSM265 combination therapy for treating uncomplicated falciparum malaria

Abstract

Background

The efficacy of artemisinin-based combination therapies (ACTs), the first-line treatments for uncomplicated falciparum malaria, has been declining in malaria-endemic countries due to the emergence of malaria parasites resistant to these compounds. Novel alternative therapies are needed urgently to prevent the likely surge in morbidity and mortality due to failing ACTs.

Objectives

This study investigates the efficacy of the combination of two novel drugs, OZ439 and DSM265, using a biologically informed within-host mathematical model.

Methods

A within-host model was developed, which accounts for the differential killing of these compounds against different stages of the parasite’s life cycle and accommodates the pharmacodynamic interaction between the drugs. Data of healthy volunteers infected with falciparum malaria collected from four trials (three that administered OZ439 and DSM265 alone, and the fourth a combination of OZ439 and DSM265) were analysed. Model parameters were estimated in a hierarchical Bayesian framework.

Results

The posterior predictive simulations of our model predicted that 800 mg of OZ439 combined with 450 mg of DSM265, which are within the safe and tolerable dose range, can provide above 90% cure rates 42 days after drug administration.

Conclusions

Our results show that the combination of OZ439 and DSM265 can be a promising alternative to replace ACTs. Our model can be used to inform future Phase 2 and 3 clinical trials of OZ439/DSM265, fast-tracking the deployment of this combination therapy in the regions where ACTs are failing. The dosing regimens that are shown to be efficacious and within safe and tolerable limits are suggested for future investigations.

Introduction

Artemisinin-based combination therapies (ACTs) have been the first-line treatment of uncomplicated falciparum malaria in most malaria-endemic countries for more than two decades.¹ During this period, ACTs have played a central role in malaria control and the decline in clinical cases and malaria-attributable deaths. Alarmingly, the efficacy of ACTs has declined below 50% in some regions,² due to the emergence and spread of parasites resistant to the artemisinins across the Greater Mekong Region.^3–5 This worrying trend threatens to reverse the progress against Plasmodium falciparum achieved by widespread availability of ACTs and, of greater concern, highlights the prospect of untreatable falciparum malaria in the absence of efficacious alternative antimalarial treatments.

Various alternative treatments have been suggested, such as combining a failing ACT with an already available partner drug, known as triple artemisinin-based combination therapy (TACT),^2,6 or producing novel synthetic antimalarials.⁷ Key features of a successful treatment include a dosing regimen that is highly effective and easy to adhere to, so that subtherapeutic concentrations are avoided, and combining drugs with different modes of action to prevent the development of resistance to each individual drug.^8,9

OZ439 (also known as artefenomel) is a novel antimalarial drug with a mechanism of action similar to artesunate, i.e. activation of an endoperoxide bond, which in turn damages various proteins of the parasite using free radicals and reactive intermediates.¹⁰ However, unlike artesunate, which has an elimination half-life of ∼1 h (500 mg dose),¹¹ OZ439 has a significantly longer elimination half-life of ∼70 h (500 mg dose). The favourable pharmacokinetic (PK) properties of OZ439, as well as its safety and tolerability at relatively high doses, and in vitro data suggesting that it is active against artemisinin-resistant parasites,¹² make it a potential candidate to replace the artemisinins.

DSM265 is another novel synthetic antimalarial drug with a long elimination half-life (∼100 h) and satisfactory safety and tolerability.^13,14 Similarly to OZ439, the long presence of this drug in the blood plasma allows administration of a single-dose regimen, whereas current dosing regimens for ACTs recommend daily administration (and twice daily for artemether/lumefantrine) for 3 days. DSM265 kills the parasites by inhibiting Plasmodium dihydroorotate dehydrogenase (DHODH), which is a vital enzyme for pyrimidine biosynthesis of the parasite.¹⁵ None of the currently administered antimalarials has this mechanism of action, making DSM265 an attractive candidate for a new antimalarial drug.

The promising pharmacological characteristics of OZ439 and DSM265 described above suggest that these drugs may be suitable candidates as a combination antimalarial treatment. A recent trial evaluating the OZ439/DSM265 combination in healthy volunteers infected with blood-stage falciparum malaria found satisfactory safety and tolerability and promising antimalarial activity.⁷ Relatively low doses were intentionally administered in the trial to allow parasitaemia recrudescence. Subsequent trials are required to investigate the efficacy of higher doses of OZ439 and DSM265 in this combination treatment. A selected efficacious dosing regimen must also satisfy safety and tolerability constraints—both drugs have shown good safety and tolerability profiles up to relatively high doses.^14,16 In addition, the exposure profiles of drugs must overlap to a large extent to reduce the likelihood of resistance selection by the parasites due to their exposure to subtherapeutic levels of only one drug.

This study focuses on the efficacy of the OZ439/DSM265 compound using a biologically informed pharmacodynamic (PD) mathematical model that accounts for the stage-specific killing action of the drugs.^17–20 The PD interaction between OZ439 and DSM265 was determined and accommodated in the model. Data from four separate trials of healthy volunteers inoculated with blood-stage falciparum malaria (OZ439 and DSM265 given alone and the OZ439/DSM265 combination therapy) were analysed. A Bayesian approach was used for parameter estimation and simulations of the fitted model at different OZ439/DSM265 doses were performed to propose regimens required to cure (within 42 days of follow-up) at least 90% of individuals infected with uncomplicated P. falciparum. This work was aimed at informing the selection of dose regimens to investigate in future clinical trials of the OZ439/DSM265 combination.

Materials and methods

Data

Data from four separate studies of volunteers inoculated with P. falciparum malaria were analysed: (i) OZ439 monotherapy (doses: 100, 200 and 500 mg);²¹ (ii) DSM265 monotherapy (dose: 150 mg);¹⁴ (iii) DSM265 monotherapy (dose: 400 mg);¹³ and (iv) OZ439/DSM265 combination therapy (doses: 200 mg of OZ439 combined with 50 and 100 mg of DSM265);⁷ details of these studies are summarized in Table 1. The data from the monotherapy studies were used for constructing the prior distributions of the PD parameters, as detailed in the Model fitting and simulation section.

Table 1.

Dosing regimens of OZ439 and DSM265 administered in four clinical trials of volunteers infected with P. falciparum

Treatment	Cohort (n)	Dose (mg)		Drug administration time (h)	No. parasites/mL at time 0; median (range)
OZ43921	A (8)		100	0	5487 (939–7517)
	B (8)		200	0	5676 (407–24 152)
	C (8)		500	0	4475 (953–10 315)
DSM265a	A (7)		150	0	8442 (1676–13 286)
	B (7)		400	0	8457 (1235–62 044)
OZ439/DSM2657	A (8)	OZ439	200	0	1629 (289–7678)
		DSM265	100	2
	B (5)	OZ439	200	0	12 364 (3495–27 312)
		DSM265	50	2

See the Data section for further information about inoculation of the volunteers with P. falciparum and their treatment.

^aData from two separate trials (cohorts A¹⁴ and B¹³) for DSM265 were used.

Mathematical model

The within-host PD model fitted to the parasitaemia data was based on the models of Saralamba et al.¹⁹ and Zaloumis et al.,²⁰ and includes the stage-specificity of drug action, shown in susceptibility experiments to significantly impact the killing effect of antimalarial drugs.^22,23 Interaction between the PD action of the drugs was also incorporated to capture the combined effect of OZ439 and DSM265.

PK

A two-compartment PK model with first-order absorption for OZ439 and zero-order absorption for DSM265 best described the PK profiles of the volunteers. The PK of each drug were not altered when the drugs were given in combination,⁷ hence no PK drug–drug interaction was considered. A delayed effect of the plasma drug concentration of both OZ439 and DSM265 on parasite killing was incorporated in the model as a transition between two compartments with rate k_e0; see Supplementary data, available at JAC Online. The concentrations at effect site, C_e(t), were substituted into the PD model to derive drug action, as detailed below.

PD

The following PD model was used for the time-evolution of the number of parasites in the body, N:

$$\begin{array}{c}\mathrm{}N(a,t)=\{\begin{array}{cc}N(a-1,t-1)\left(1-E(a-1,t-1)\right),& 1<a\le 48,\\ N(48,t-1)\left(1-E(48,t-1)\right)\times \text{PMF},& a=1,\end{array}\end{array}$$ (1)

where a is parasite age, taking only integer values 1 to 48, t is time, taking only integer values, and PMF is the parasite multiplication factor, which represents the number of merozoites released into the blood by a schizont at the end of its life cycle that successfully invade RBCs. E(a, t) is the killing effect of the drug, taking values between 0 and 1, and dependent upon the age of parasites. The subjects were assumed to be infected with an initial parasite load of $N_0={\sum }_{a=1}^{48}N(a,0)$, which has a discretized normal distribution over age with the mean at μ₀ h and SD σ₀ h and N(1,0) = PMF × N(48,0). To determine N(a, 0) by discretizing a continuous normal distribution, n(a)∼N(μ₀,σ₀), the following formula was used:

$$N(a,0)=N_0\frac{n\left(a\right)}{{\sum }_{a=1}^{48}n\left(a\right)},1\le a\le 48.$$ (2)

The number of detectable parasites circulating in the blood, M(t), was determined by

$$M\left(t\right)={\sum }_{a=1}^{26}N(a,t),$$ (3)

where it is assumed the older parasites sequester in blood capillaries.²⁴ The number of parasites per mL of blood (the unit of parasitaemia in the data) was determined by dividing M(t) by each patient’s blood volume in mL, calculated from 70 mL of blood/kg × patient’s weight.

We assumed Michaelis–Menten kinetics for E:

$$E(a,t)=E_{\mathit{max}}\frac{C_e(t{)}^{\gamma }}{C_e(t{)}^{\gamma }+E{C}_{50}^{\gamma }}q\left(a\right),$$ (4)

where E_max is the maximum killing effect of the antimalarial drug; C_e(t) is the drug concentration at the effect site at time t; EC₅₀ is the concentration at which 50% of the maximum killing effect is obtained; γ is the sigmoidicity (also known as slope) of the concentration–effect curve. The stage-specificity of the killing effect is applied using q(a). A description of all the PD parameters is provided in Table 2.

Table 2.

Description and selected prior bounds (monotherapies) for each parameter of the PD model

Parameter	Drug	Prior bounds	Description
N 0		5.12 × 104–8.02 × 106	initial number of parasites (circulating + sequestered)
		5.18 × 104–1.42 × 107	lower bound: LLOQ × mean blood volume of volunteers in the trial
			upper bound: 2× maximum observed parasitaemia at the first blood measurement × mean blood volume of volunteers in the triala
μ 0		1–48	mean of initial parasite age distribution (h)
			bounds: lowest and highest ages of parasites in a 48  h life cycle
σ 0		4–14	SD of initial age distribution (h)
			bounds: selected to produce a wide range of dispersion (narrow to dispersed) in initial age distribution
PMF		5–50	parasite multiplication factor (/48 h)
			bounds: according to Wockner et al.38
Emax	OZ439	0.05–1	maximum killing effect
	DSM265		bounds: an arbitrarily wide range is selected
EC 50	OZ439	0.5–50	concentration producing Emax/2 effect (ng/mL)
	DSM265	500–5000	bounds: according to McCarthy et al.21 and Phillips et al.15 for OZ439 and DSM265, respectively
γ	OZ439   DSM265	1–10	sigmoidicity of the concentration–effect curves
			bounds: selected to generate a wide range of sigmoidicities
ke 0	OZ439   DSM265	0.01–10	rate of drug transition from blood plasma compartment to the effect-site compartment (h-1)
			bounds: arbitrarily large range was selected
α	OZ439/DSM265	0.2–5	interaction parameter
			bounds: the selected values produce a wide range of interactions, from strong antagonism to strong synergism

^aSee Model fitting and simulation section of Supplementary data for conversion of the observable circulating parasitaemia (no. parasites/mL) to the total parasite biomass (circulating + sequestered).

Previous in vitro experiments showed that OZ439 kills the parasites at all stages of the blood life cycle¹² and DSM265 only kills trophozoites.¹⁵ It has also been shown that OZ439 has maximum activity against trophozoites (reviewed by Phillips et al.²⁵). Therefore, we considered the following step functions for the stage-specificity of the killing action of the drugs:

$$q_D\left(a\right)=\left\{\begin{array}{cc}1,& a\in \left[26,\mathrm{}36\right],\\ 0,& o\mathit{therwise},\end{array}\right.$$ (5)

and

$$q_O\left(a\right)=\left\{\begin{array}{cc}1,& a\in \left[26,\mathrm{}36\right],\\ 0.5,& a\in \left[6,\mathrm{}25\right]\cup \left[37,\mathrm{}44\right],\\ 0,& o\mathit{therwise},\end{array}\right.$$ (6)

where q_D and q_O are the stage-specific functions for DSM265 and OZ439, respectively. Note, early ring and late schizont parasites were considered non-susceptible to OZ439, similar to artemisinin.²⁶

Combined killing effect

OZ439 and DSM265 have different modes of action—the former kills the parasites by activating the endoperoxide bond²¹ and the latter by inhibiting the parasite’s DHODH enzyme.¹⁴ Bliss independence²⁷ was selected as the base model for zero interaction and was then modified to define the combined effect, E_OD:

$$E_{OD}={\left(E_O^{\alpha }+E_D^{\alpha }-E_O^{\alpha }{E}_D^{\alpha }\right)}^{\frac{1}{\alpha }},$$ (7)

where E_O and E_D can be obtained using the Michaelis–Menten function (Equation 4) and α is the interaction parameter. The values of α  =  1, α  >  1 and 0 < α  <  1 correspond to zero interaction, antagonism and synergism, respectively (Figure 1).

Figure 1.

Combined effect of OZ439 and DSM265, E_OD in Equation 7, for three different types of interaction and the following arbitrarily selected set of PD parameter values: E_max,D= E_max,O= 0.4; γ_D= 3; γ_O= 3; EC_50,D= EC_50,O= 100 ng/mL, where O and D in the sub-indices denote correspondence to OZ439 and DSM265, respectively. The (1), (2) and (3) surfaces correspond to α  = 0.3 (synergism), α  = 1 (zero interaction) and α = 7 (antagonism). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Model fitting and simulation

A sequential approach was employed to fit the model to data: the PK parameters were first estimated to simulate drug concentration profiles, which were then substituted into the PD model to estimate the PD parameters. Fitting of the PK models to the data was performed in a non-linear mixed-effects modelling framework using Monolix²⁸ (see Supplementary data).

The PD model was fitted in a Bayesian hierarchical framework that allowed estimation of individual parameters and incorporation of prior information about the parameters. For parasitaemia below the lower limit of quantification (LLOQ), the M3 method was used, by considering the data below the quantification level left-censored and using the cumulative normal distribution in the likelihood function.²⁹ The RStan package³⁰ in the R software³¹ was used to implement the Hamiltonian Monte Carlo (HMC) method; see Supplementary data.

Model simulations for predicting the cure rates were performed using the posterior samples of individual parameters. A total of 20 datasets/cohorts, each including 100 hypothetical patients, were simulated and the cure rate was calculated for each cohort. The cure rate was defined as the proportion of patients whose parasitaemia levels were below the LLOQ (10 parasites/mL) within 42 days. The parasitaemia levels observed in the field³ were used as the baseline parasitaemia in the simulations; see Supplementary data.

Results

Data collected from volunteers infected with P. falciparum in monotherapy trials of OZ439 and DSM265 and the OZ439/DSM265 combination therapy trial were used for model fitting; see Table 1. The measured drug concentrations of the volunteers were first used to estimate the population PK model parameters and between-subject variability. A sequential PK/PD modelling approach was then performed where PK profiles were simulated for each volunteer, based on the post hoc individual PK parameter estimates, which were subsequently substituted into the PD model.

PK

A two-compartment PK model with first- and zero-order absorption rates best described the OZ439 and DSM265 drug concentrations. Table 3 summarizes the estimated PK parameters for the data from the combination therapy trial. Figure 2 shows the observed and simulated drug concentration profiles of the 13 volunteers; see Figure S1 for the PK profiles of the volunteers receiving OZ439 and DSM265 alone. The profiles show significantly higher concentrations of DSM265 in the blood plasma of volunteers compared with OZ439.

Table 3.

Parameter values of the PK models

Parameter	Description	Drug	Estimate (RSE; %)	BSV (RSE; %)
$F$	Relative bioavailability	DSM265	1 (fixed)	0.0897 (44.50)
		OZ439	1 (fixed)	0 (fixed)
$CL/F$	Apparent clearance (L/h)	DSM265	0.514 (3.20)	0.232 (11)
		OZ439	43.2 (9.60)	0.261 (16.40)
$V_c/F$	Apparent central volume (L)	DSM265	12.4 (14)	0.66 (17.20)
		OZ439	148 (12.50)	0 (fixed)
$Q/F$	Apparent intercompartmental clearance (L/h)	DSM265	38.5 (8.24)	0.25 (fixed)
		OZ439	11.1 (8.66)	0 (fixed)
$V_p/F$	Apparent peripheral volume (L)	DSM265	61.8 (4.29)	0.25 (fixed)
		OZ439	903 (12.20)	0.276 (38.80)
$T_{k0}$	Absorption time (h)	DSM265	2.64 (5.86)	0.413 (10.20)
$k_a$	Absorption rate parameter (h-1)	OZ439	0.198 (6.85)	0.106 (58.60)
$T_{\mathit{lag}}$	Absorption lag time (h)	DSM265	0 (fixed)	0 (fixed)
		OZ439	0.418 (2.92)	0.107 (20.70)
${\beta }_F$	Coefficient of dose on F; reference dose: 400 mg	DSM265	−0.114 (22.20)	—
${\mathrm{\beta }}_{V_p}$	Coefficient of weight on F; reference weight: 76.8 kg	DSM265	1.61 (16.90)	—
${\mathrm{\beta }}_{CL}$	Coefficient of dose on CL; reference dose: 400 mg (DSM265) 500 mg (OZ439)	DSM265	0.649 (34.10)	—
		OZ439	−0.473 (18.20)	—

BSV, between-subject variability; RSE, relative standard error, defined as (SD/mean) × 100.

$t$_½ was 63, 66 and 71 h for 100, 200 and 500 mg administered doses of OZ439, respectively, and 101 h for all the administered doses of DSM265. A two-compartment model with zero-order and first-order absorption was used for DSM265 and OZ439, respectively.

Figure 2.

PK profiles of OZ439 and DSM265 for the OZ439/DSM265 combination therapy. The points show the measured drug concentrations and the lines are the generated simulations using the mode of the conditional distribution of the individual PK parameters—the population parameters are listed in Table 3. A two-compartment model with first- and zero-order absorption rates were used for OZ439 and DSM265, respectively. Volunteers 1–8 received 200 mg of OZ439 and 100 mg of DSM265, and volunteers 9–13 received 200 mg of OZ439 and 50 mg of DSM265. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

PD

The antimalarial activities of the drugs were modelled using the mathematical model defined in Equation 1. The model accounts for differential action of the drugs on different stages of the parasite life cycle and PD interaction between the drugs. Initially, the PD model was fitted to the measured parasitaemia of volunteers in the OZ439 and DSM265 monotherapy trials (results shown in Figure S2). The estimated posterior distributions were then used to inform the estimation of E_max, EC₅₀, γ and k_e0 in the OZ439/DSM265 model fitted to the data of the combination therapy trial.

Figure 3 shows the observed parasitaemia profiles for the 13 volunteers (black circles and dots), overlaid with the posterior predictive distributions [red line: median; shaded region: 95% credible interval (CrI)]. The results show that the PK/PD model captures the dynamics of the observed parasitaemia well for all individuals.

Figure 3.

Posterior predictive check of the within-host PK/PD model fitted to the parasitaemia of volunteers in the combination therapy trial. The black circles are the observed parasitaemia; the red line and the shaded area denote the median and 95% CrI (between 2.5^th and 97.5^th percentiles) of 8000 simulations of the model using the posterior samples of the individual PD parameters (see Materials and methods section). Volunteers 1–8 belong to cohort A (OZ439: 200 mg; DSM265: 100 mg) and volunteers 9–13 to cohort B (OZ439: 200 mg; DSM265: 50 mg); see Table 1 for details of the cohorts. The grey vertical line at time = 0 shows the administration time of OZ439; DSM265 was administered 2 h after the OZ439 administration; see Table 1. The LLOQ, shown with the horizontal line, was 10 parasites/mL. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The estimated PD parameters for the combination therapy are summarized in Table 4; Tables S1 and S2 include the details of the posterior distributions of the PD parameters estimated from modelling of the monotherapy trials data. $\widehat{R}$ and $n_{\mathit{eff}}$ metrics (see Table S3) indicate the convergence of the HMC Markov chains: $\widehat{R}$ values are close to 1 and $n_{\mathit{eff}}$ values are large. Other model-fit diagnostics for the four trials are illustrated in Figures S3–S5.

Table 4.

Estimated parameter values of the PD model fitted to the data of OZ439/DSM265 combination therapy

Parametera	Drug	Prior boundsb	Posterior medianc (95% CrI)	BSVd (95% CrI)
N 0		(5.04 × 104–4.00 × 106)	1.91 × 106 (1.16 × 106–2.86 × 106)	1.20 (0.41–2.06)
μ0 (h)		(1–48)	5.55 (2.46–9.5)	0.44 (0.01–1.11)
σ0 (h)		(4–14)	10.13 (8.43–11.72)	0.44 (0.01–1.03)
PMF (/48 h)		(5–50)	10.56 (6.92–17.6)	1.75 (0.94–2.72)
Emax	OZ439	(0.33–0.41)	0.35 (0.34–0.38)	1.08 (0.04–2.38)
	DSM265	(0.49–0.8)	0.63 (0.53–0.75)	1.18 (0.05–2.51)
EC50 (ng/mL)	OZ439	(16.25–28.01)	23.62 (19.02–27.01)	1.07 (0.04–2.30)
	DSM265	(989.1–2085.46)	1354.4 (1088.94–1741.38)	1.91 (0.69–3.15)
γ	OZ439	(1.16–2.84)	1.32 (1.19–1.62)	0.97 (0.03–2.21)
	DSM265	(1.67–7.27)	2.13 (1.77–3)	1.35 (0.07–2.70)
ke0 (h-1)	OZ439	(0.05–0.17)	0.07 (0.05–0.11)	1.13 (0.03–2.52)
	DSM265	(1.7–8.73)	5.11 (2.5–7.84)	1.01 (0.03–2.30)
α	OZ439/DSM265	(0.2–5)	2.25 (0.9–4.06)	1.54 (0.09–2.78)

^aDescription of the parameters and details of the prior bounds are summarized in Table 2.

^bEstimated posterior samples of monotherapy models were used to set the prior boundaries of E_max, EC₅₀, γ and k_e0.

^cThe median and 95% CrIs are drawn from 8000 posterior samples (after burn-in) generated by the HMC algorithm; see Supplementary data for the convergence metrics of HMC chains.

^dBetween-subject variability (BSV) is the SD of individual PD parameters from the population mean PD parameters on the logistic transform scale (see Model fitting and simulation section of Supplementary data).

The results in Table 4 show that the posterior median of the initial parasite load (sum of circulating and sequestered parasites) at the time of first parasitaemia measurement (on average 49.3 h before administration of OZ439/DSM265), N₀, was 1.91 × 10⁶ (95% CrI: 1.16 × 10⁶–2.86 × 10⁶). The estimated initial age distribution indicated that the majority of the parasites at the time of the first measurement were at the ring blood stage of the parasite’s ∼48 h life cycle [the posterior median of μ₀ was 5.55 (95% CrI: 2.46–9.5) h with a spread (σ₀) of 10.13 (95% CrI: 8.43–11.72) h]. The initial decline in parasitaemia of many of the volunteers confirms these results, since ageing of the cluster of ring parasites (observed in the blood) into trophozoites/schizonts (not observed in the blood) would lead to a temporary decline in the observed parasitaemia.

The PMF was estimated to be 10.56 (95% CrI: 6.92–17.6). E_max values for DSM265 and OZ439 were estimated to be 0.63 (95% CrI: 0.53–0.75) and 0.35 (95% CrI: 0.34–0.38), respectively. Note that OZ439 is assumed to kill parasites at all stages²⁵ while DSM265 only kills trophozoites,¹⁵ and the stage-specific action of these drugs was incorporated into the PD model; the killing windows of DSM265 and OZ439 span 11 and 39 h, respectively, of the 48 h life cycle and the killing effect of OZ439 was halved for parasites aged 6–25 and 37–44 h (Equations 5 and 6). Therefore, the lower estimated value of E_max for OZ439 compared with DSM265 does not mean that it is less potent in reducing the parasitaemia. In fact, the average of E_max over all the ages of the parasite’s life cycle is 0.18, which is higher than that of DSM265 (0.14).

The estimated EC₅₀ of OZ439 [23.62 (95% CrI: 19.02–27.01) ng/mL] was significantly lower than that of DSM265 [1354.4 (95% CrI: 1088.94–1741.38) ng/mL]. The average time that drug concentration was above the population average EC₅₀ for OZ439 was 20.2 and 21.6 h for subjects of cohorts A and B (OZ439 dose: 200 mg), respectively, and for DSM265 was 19.9 and 0.3 h for cohorts A (DSM265 dose: 100 mg) and B (DSM265 dose: 50 mg), respectively. The posterior median of the rate of transition between the blood plasma compartment to the effect site (hypothesized) compartment, k_e0, was 0.07 (95% CrI: 0.05–0.11) h⁻¹ for OZ439. For DSM265, the posterior median of k_e0 was 5.11 (95% CrI: 2.5–7.84) h⁻¹. However, the distributions of prior and posterior samples for this parameter were fairly similar [Figure S5(d)], implying that these data were not informative for estimating k_e0 of DSM265.

Prediction of the optimal dosing regimen

To determine a dosing regimen that provides the WHO-recommended 42 day cure rate of at least 90%, parasitaemia profiles post drug administration were simulated using the posterior samples of the individual parameters for different combinations of single doses of OZ439 and DSM265.

Figure 4 shows the mean 42 day cure rates over 20 datasets, each comprising 100 simulated hypothetical patients (2000 patients in total) who received different combinations of OZ439 and DSM265 doses; the patients whose parasitaemia got below the LLOQ (10 parasites/mL) over 42 days of follow-up were considered cured. The lower and upper limits of the 42 day cure rates are shown in Figure S6. The dose combinations that yielded 42 day cure rates above 90% are outlined with black. The selected simulated doses fall within current evidence for safe and tolerable dosing of both drugs. OZ439 has been shown to be safe and well tolerated up to 1200 mg when administered as a capsule and up to 1600 mg when administered as an oral dispersion.¹⁶ The safety profile of DSM265 is seemingly not as good as OZ439 as the number of adverse effects was higher in infected volunteers who received DSM265, compared with those who received placebo,¹⁴ although the number of adverse events were not correlated with the administered dose.

Figure 4.

Expected cure rate within 42 days for different doses of OZ439 and DSM265 combination therapy. For each dose combination, simulations for 20 datasets, each including 100 (new/hypothetical) patients, were generated. The simulations were performed using the individual posterior predictive distributions of the PD parameters (see Table 4 and Model fitting and simulation section) and the obtained 42 day cure rates averaged over the 20 datasets are shown in the grid squares (the lower and upper limits of the 42 day cure rates are shown in Figure S6). The values of parasitaemia at the time of drug administration in the simulations were drawn from a log-normal distribution constructed using the reported values in malaria-endemic countries: median = 52 250 (range: 2560–605 329) parasites/mL.³ Single doses of OZ439 and DSM265 were administered at times 0 and 2 h, respectively. The dose combinations that yielded a 42 day cure rate ≥90% are outlined with black borders. Cure rate: proportion of patients in each dataset who had parasitaemia below the LLOQ (10 parasites/mL) over 42 days of follow-up. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The results in Figure 4 show that the 42 day 90% cure rate cannot be achieved by any dose combinations containing ≤200 mg of OZ439 or ≤100 mg of DSM265. Various dose combinations achieve a ≥90% 42 day cure rate; however, considering the safety and tolerability of each drug, we recommend the combination of 800 mg of OZ439 and 450 mg of DSM265 be investigated in further trials. This dosing regimen also provides high drug concentrations of OZ439 in the blood plasma, covering the exposure duration of DSM265 for up to 42 days.

The initial distribution of the age of parasites for each simulated patient was assumed to follow that estimated from the volunteers. A sensitivity analysis, where the mean of the initial parasite age distribution was assumed to follow a uniform distribution over 0–48 h, found the predicted 42 day cure rate remained above 90% for 800 mg of OZ439 and 450 mg of DSM265, confirming the robustness of this dosing regimen; see Figure S7.

Discussion

We proposed a within-host mathematical PK/PD model for the combined antimalarial activity of two novel drugs: OZ439 and DSM265. Our model incorporated the parasite age-specificity of the killing action of the drugs, parasite sequestration and PD interaction between the drugs. Using a Bayesian hierarchical framework, our model provided a good fit to parasitaemia data collected pre- and post-administration of OZ439 and DSM265 from healthy volunteers inoculated with P. falciparum. Simulating parasitological outcomes using the estimated PD parameters determined that the safe and tolerable dosing regimens of 800 mg for OZ439 and 450 mg for DSM265 can yield the WHO-recommended 42 day cure rates (≥90%).

Our model simulations put forward a set of potential OZ439/DSM265 dose combinations that are predicted to be efficacious and are within the safety and tolerability limits.^14,16 Among these dose combinations, one must be selected for deployment that does not lead to development of parasite resistance to one of the drugs due to long durations of subtherapeutic exposure. The significantly longer exposure time of DSM265 (see Figure 2) reinforces the possibility of resistance development to this drug, if the parasites are not simultaneously exposed to another drug. In fact, a study by Llanos-Cuentas et al.³² found evidence of selection of resistance to DSM265 through a mutation in the DHODH enzyme target in Peruvian patients who were administered a single dose of DSM265. Therefore, the selected dose of OZ439 must be sufficiently high that it exposes the parasites for long enough timespans during which the parasites are exposed to DSM265 as well. As a result, the selection of higher doses of OZ439 must be favoured from the set of all efficacious, safe and tolerable doses (Figure 4), e.g. 600 or 800 mg of OZ439, combined with 400 or 450 mg of DSM265.

We proposed a novel model of the combined action of the drugs based on the Bliss independence concept.²⁷ Fitting the model to the parasitological data showed that OZ439 and DSM265 have a slight antagonistic interaction; however, the antagonistic interaction was not strong enough to significantly nullify their combined effect and the combination compound was still able to produce cure rates above 90%.

The PK/PD model proposed in this work can be used to guide Phase 2 and 3 clinical trials evaluating the efficacy of OZ439/DSM265 regimens, helping to reduce the financial and logistical costs of these trials. Caution must be exercised when extrapolating the findings of this work to patients in the endemic settings where parasitaemia levels are higher and parasites may carry important drug resistance mutations. Our model predictions are based on data of healthy volunteers infected with artemisinin-susceptible P. falciparum, whereas PK and PD responses of patients in the field may differ from those of volunteers. In a previous trial studying the efficacy of a single dose of the combination of OZ439 and piperaquine, the Day 28 efficacy was only 74.5% in patients from Africa and 66.2% in patients from Vietnam.³³ Of note, biomarkers of resistance to artemisinin (which probably implies resistance to OZ439 as well)¹² and piperaquine were found in patients from Vietnam and Africa, respectively.³⁴ This highlights the importance of considering drug resistance as a confounder of the predicted efficacy of dosing regimens.

Our PD model did not include the potential contribution of host immunity to parasite clearance³⁵ since it was validated on volunteers who had not been previously exposed to malaria infection. However, the influence of immunity on parasite clearance would augment drug effect, therefore resulting in an overestimate of the minimal efficacious dose and thereby resulting in a greater safety margin. The efficacy of the suggested dosing regimen in reducing gametocytaemia, and thereby transmission, was not investigated in this work. This will be considered in future work based on a model we have developed for within-host transmission dynamics.³⁶ Further, a more mechanistic model of the combined action of the drugs that accommodates the underlying processes of drug interaction could be used.³⁷ However, to do so would require a greater understanding of the drug–drug interactions, which is yet unavailable, requiring more sophisticated in vitro parasite susceptibility experiments, e.g. chequerboard assays that focus particularly on the combined effect of the drugs. Further validation of our methodology can also be performed by applying our method on PK/PD data of the combination of OZ439 with other partner drugs such as piperaquine³³ and ferroquine (https://clinicaltrials.gov/ct2/show/NCT02497612).

P. falciparum parasites resistant to ACTs are rapidly spreading across South-East Asia, impeding the goal of WHO to achieve malaria elimination by 2030 in this region. The combination of OZ439 and DSM265, administered according to the suggested efficacious and well-tolerated regimens, appears to be a promising alternative treatment to replace the failing ACTs.

Funding

This work was supported in part by the Australian Centre for Research Excellence in Malaria Elimination, funded by the NHMRC (1134989). J.A.S. is funded by an Australian National Health and Medical Research Council of Australia (NHMRC) Investigator Grant (1196068). J.S.M. is funded by an NHMRC Program Grant (1132975) and Practitioner Fellowship (1041802). The clinical trials (NCT02389348, NCT02573857, ACTRN12613000522718, ACTRN12613000527763 and ACTRN12612000814875) from which the data were derived were supported by the Medicines for Malaria Venture (MMV) and funded by the Wellcome Trust (grant reference number: 095909/Z/11/Z), a grant by the Global Health Innovation and Technology Fund (GHIT) (grant no. G2014-108) and by funding from the Bill & Melinda Gates Foundation.

Transparency declarations

N.G., M.C. and J.M. are employed by the Medicines for Malaria Venture, who supported the clinical trials that collected the data used in this study. All other authors: none to declare.

Supplementary data

Tables S1 to S4, Figures S1 to S7 and Supplementary information and references are available as Supplementary data at JAC Online.